module @module {
util.global private @__auto.token_embd.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.token_embd.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"token_embd.weight"> : tensor<128256x4096xf16>
util.global private @__auto.blk.0.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.0.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.0.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.0.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.0.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.1.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.1.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.1.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.1.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.2.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.2.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.2.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.2.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.3.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.3.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.3.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.3.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.4.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.4.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.4.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.4.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.5.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.5.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.5.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.5.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.6.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.6.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.6.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.6.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.7.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.7.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.7.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.7.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.8.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.8.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.8.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.8.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.9.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.9.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.9.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.9.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.10.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.10.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.10.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.10.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.11.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.11.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.11.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.11.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.12.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.12.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.12.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.12.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.13.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.13.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.13.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.13.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.14.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.14.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.14.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.14.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.15.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.15.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.15.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.15.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.16.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.16.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.16.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.16.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.17.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.17.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.17.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.17.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.18.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.18.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.18.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.18.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.19.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.19.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.19.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.19.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.20.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.20.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.20.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.20.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.21.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.21.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.21.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.21.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.22.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.22.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.22.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.22.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.23.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.23.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.23.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.23.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.24.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.24.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.24.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.24.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.25.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.25.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.25.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.25.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.26.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.26.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.26.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.26.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.27.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.27.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.27.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.27.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.28.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.28.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.28.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.28.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.29.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.29.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.29.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.29.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.30.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.30.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.30.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.30.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.attn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.attn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.attn_q.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.0"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.1"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.2"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.3"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.4"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.5"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.6"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_q.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.attn_q.weight.shard.7"> : tensor<512x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_k.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.attn_k.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.0"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.1"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.2"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.3"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.4"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.5"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.6"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_v.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.attn_v.weight.shard.7"> : tensor<128x4096xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.0"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.1"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.2"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.3"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.4"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.5"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.6"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.attn_output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.attn_output.weight.shard.7"> : tensor<4096x512xf16>
util.global private @__auto.blk.31.ffn_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.ffn_norm.weight"> : tensor<4096xf32>
util.global private @__auto.blk.31.ffn_gate.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_gate.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.ffn_gate.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.0"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.1"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.2"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.3"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.4"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.5"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.6"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_up.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.ffn_up.weight.shard.7"> : tensor<1792x4096xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.0"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.1"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.2"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.3"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.4"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.5"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.6"> : tensor<4096x1792xf16>
util.global private @__auto.blk.31.ffn_down.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"blk.31.ffn_down.weight.shard.7"> : tensor<4096x1792xf16>
util.global private @__auto.output_norm.weight {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output_norm.weight$7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"output_norm.weight"> : tensor<4096xf32>
util.global private @__auto.output.weight.shard.0 {iree.abi.affinity = #hal.device.promise<@__device_0>} = #stream.parameter.named<"model"::"output.weight.shard.0"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.1 {iree.abi.affinity = #hal.device.promise<@__device_1>} = #stream.parameter.named<"model"::"output.weight.shard.1"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.2 {iree.abi.affinity = #hal.device.promise<@__device_2>} = #stream.parameter.named<"model"::"output.weight.shard.2"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.3 {iree.abi.affinity = #hal.device.promise<@__device_3>} = #stream.parameter.named<"model"::"output.weight.shard.3"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.4 {iree.abi.affinity = #hal.device.promise<@__device_4>} = #stream.parameter.named<"model"::"output.weight.shard.4"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.5 {iree.abi.affinity = #hal.device.promise<@__device_5>} = #stream.parameter.named<"model"::"output.weight.shard.5"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.6 {iree.abi.affinity = #hal.device.promise<@__device_6>} = #stream.parameter.named<"model"::"output.weight.shard.6"> : tensor<128256x512xf16>
util.global private @__auto.output.weight.shard.7 {iree.abi.affinity = #hal.device.promise<@__device_7>} = #stream.parameter.named<"model"::"output.weight.shard.7"> : tensor<128256x512xf16>
func.func @prefill_bs4(%arg0: !torch.vtensor<[4,?],si64> {iree.abi.affinity = #hal.device.promise<@__device_0>}, %arg1: !torch.vtensor<[4],si64> {iree.abi.affinity = #hal.device.promise<@__device_0>}, %arg2: !torch.vtensor<[4,?],si64> {iree.abi.affinity = #hal.device.promise<@__device_0>}, %arg3: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_0>}, %arg4: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_1>}, %arg5: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_2>}, %arg6: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_3>}, %arg7: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_4>}, %arg8: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_5>}, %arg9: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_6>}, %arg10: !torch.tensor<[?,131072],f16> {iree.abi.affinity = #hal.device.promise<@__device_7>}) -> !torch.vtensor<[4,?,128256],f16> attributes {torch.assume_strict_symbolic_shapes} {
%__auto.token_embd.weight = util.global.load @__auto.token_embd.weight : tensor<128256x4096xf16>
%0 = torch_c.from_builtin_tensor %__auto.token_embd.weight : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$1 = util.global.load @__auto.token_embd.weight$1 : tensor<128256x4096xf16>
%1 = torch_c.from_builtin_tensor %__auto.token_embd.weight$1 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$2 = util.global.load @__auto.token_embd.weight$2 : tensor<128256x4096xf16>
%2 = torch_c.from_builtin_tensor %__auto.token_embd.weight$2 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$3 = util.global.load @__auto.token_embd.weight$3 : tensor<128256x4096xf16>
%3 = torch_c.from_builtin_tensor %__auto.token_embd.weight$3 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$4 = util.global.load @__auto.token_embd.weight$4 : tensor<128256x4096xf16>
%4 = torch_c.from_builtin_tensor %__auto.token_embd.weight$4 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$5 = util.global.load @__auto.token_embd.weight$5 : tensor<128256x4096xf16>
%5 = torch_c.from_builtin_tensor %__auto.token_embd.weight$5 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$6 = util.global.load @__auto.token_embd.weight$6 : tensor<128256x4096xf16>
%6 = torch_c.from_builtin_tensor %__auto.token_embd.weight$6 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.token_embd.weight$7 = util.global.load @__auto.token_embd.weight$7 : tensor<128256x4096xf16>
%7 = torch_c.from_builtin_tensor %__auto.token_embd.weight$7 : tensor<128256x4096xf16> -> !torch.vtensor<[128256,4096],f16>
%__auto.blk.0.attn_norm.weight = util.global.load @__auto.blk.0.attn_norm.weight : tensor<4096xf32>
%8 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$1 = util.global.load @__auto.blk.0.attn_norm.weight$1 : tensor<4096xf32>
%9 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$2 = util.global.load @__auto.blk.0.attn_norm.weight$2 : tensor<4096xf32>
%10 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$3 = util.global.load @__auto.blk.0.attn_norm.weight$3 : tensor<4096xf32>
%11 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$4 = util.global.load @__auto.blk.0.attn_norm.weight$4 : tensor<4096xf32>
%12 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$5 = util.global.load @__auto.blk.0.attn_norm.weight$5 : tensor<4096xf32>
%13 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$6 = util.global.load @__auto.blk.0.attn_norm.weight$6 : tensor<4096xf32>
%14 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_norm.weight$7 = util.global.load @__auto.blk.0.attn_norm.weight$7 : tensor<4096xf32>
%15 = torch_c.from_builtin_tensor %__auto.blk.0.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.attn_q.weight.shard.0 = util.global.load @__auto.blk.0.attn_q.weight.shard.0 : tensor<512x4096xf16>
%16 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.1 = util.global.load @__auto.blk.0.attn_q.weight.shard.1 : tensor<512x4096xf16>
%17 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.2 = util.global.load @__auto.blk.0.attn_q.weight.shard.2 : tensor<512x4096xf16>
%18 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.3 = util.global.load @__auto.blk.0.attn_q.weight.shard.3 : tensor<512x4096xf16>
%19 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.4 = util.global.load @__auto.blk.0.attn_q.weight.shard.4 : tensor<512x4096xf16>
%20 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.5 = util.global.load @__auto.blk.0.attn_q.weight.shard.5 : tensor<512x4096xf16>
%21 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.6 = util.global.load @__auto.blk.0.attn_q.weight.shard.6 : tensor<512x4096xf16>
%22 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_q.weight.shard.7 = util.global.load @__auto.blk.0.attn_q.weight.shard.7 : tensor<512x4096xf16>
%23 = torch_c.from_builtin_tensor %__auto.blk.0.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.0.attn_k.weight.shard.0 = util.global.load @__auto.blk.0.attn_k.weight.shard.0 : tensor<128x4096xf16>
%24 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.1 = util.global.load @__auto.blk.0.attn_k.weight.shard.1 : tensor<128x4096xf16>
%25 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.2 = util.global.load @__auto.blk.0.attn_k.weight.shard.2 : tensor<128x4096xf16>
%26 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.3 = util.global.load @__auto.blk.0.attn_k.weight.shard.3 : tensor<128x4096xf16>
%27 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.4 = util.global.load @__auto.blk.0.attn_k.weight.shard.4 : tensor<128x4096xf16>
%28 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.5 = util.global.load @__auto.blk.0.attn_k.weight.shard.5 : tensor<128x4096xf16>
%29 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.6 = util.global.load @__auto.blk.0.attn_k.weight.shard.6 : tensor<128x4096xf16>
%30 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_k.weight.shard.7 = util.global.load @__auto.blk.0.attn_k.weight.shard.7 : tensor<128x4096xf16>
%31 = torch_c.from_builtin_tensor %__auto.blk.0.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.0 = util.global.load @__auto.blk.0.attn_v.weight.shard.0 : tensor<128x4096xf16>
%32 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.1 = util.global.load @__auto.blk.0.attn_v.weight.shard.1 : tensor<128x4096xf16>
%33 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.2 = util.global.load @__auto.blk.0.attn_v.weight.shard.2 : tensor<128x4096xf16>
%34 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.3 = util.global.load @__auto.blk.0.attn_v.weight.shard.3 : tensor<128x4096xf16>
%35 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.4 = util.global.load @__auto.blk.0.attn_v.weight.shard.4 : tensor<128x4096xf16>
%36 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.5 = util.global.load @__auto.blk.0.attn_v.weight.shard.5 : tensor<128x4096xf16>
%37 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.6 = util.global.load @__auto.blk.0.attn_v.weight.shard.6 : tensor<128x4096xf16>
%38 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_v.weight.shard.7 = util.global.load @__auto.blk.0.attn_v.weight.shard.7 : tensor<128x4096xf16>
%39 = torch_c.from_builtin_tensor %__auto.blk.0.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.0.attn_output.weight.shard.0 = util.global.load @__auto.blk.0.attn_output.weight.shard.0 : tensor<4096x512xf16>
%40 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.1 = util.global.load @__auto.blk.0.attn_output.weight.shard.1 : tensor<4096x512xf16>
%41 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.2 = util.global.load @__auto.blk.0.attn_output.weight.shard.2 : tensor<4096x512xf16>
%42 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.3 = util.global.load @__auto.blk.0.attn_output.weight.shard.3 : tensor<4096x512xf16>
%43 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.4 = util.global.load @__auto.blk.0.attn_output.weight.shard.4 : tensor<4096x512xf16>
%44 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.5 = util.global.load @__auto.blk.0.attn_output.weight.shard.5 : tensor<4096x512xf16>
%45 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.6 = util.global.load @__auto.blk.0.attn_output.weight.shard.6 : tensor<4096x512xf16>
%46 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.attn_output.weight.shard.7 = util.global.load @__auto.blk.0.attn_output.weight.shard.7 : tensor<4096x512xf16>
%47 = torch_c.from_builtin_tensor %__auto.blk.0.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.0.ffn_norm.weight = util.global.load @__auto.blk.0.ffn_norm.weight : tensor<4096xf32>
%48 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$1 = util.global.load @__auto.blk.0.ffn_norm.weight$1 : tensor<4096xf32>
%49 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$2 = util.global.load @__auto.blk.0.ffn_norm.weight$2 : tensor<4096xf32>
%50 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$3 = util.global.load @__auto.blk.0.ffn_norm.weight$3 : tensor<4096xf32>
%51 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$4 = util.global.load @__auto.blk.0.ffn_norm.weight$4 : tensor<4096xf32>
%52 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$5 = util.global.load @__auto.blk.0.ffn_norm.weight$5 : tensor<4096xf32>
%53 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$6 = util.global.load @__auto.blk.0.ffn_norm.weight$6 : tensor<4096xf32>
%54 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_norm.weight$7 = util.global.load @__auto.blk.0.ffn_norm.weight$7 : tensor<4096xf32>
%55 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.0.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%56 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%57 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%58 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%59 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%60 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%61 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%62 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.0.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%63 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.0 = util.global.load @__auto.blk.0.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%64 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.1 = util.global.load @__auto.blk.0.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%65 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.2 = util.global.load @__auto.blk.0.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%66 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.3 = util.global.load @__auto.blk.0.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%67 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.4 = util.global.load @__auto.blk.0.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%68 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.5 = util.global.load @__auto.blk.0.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%69 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.6 = util.global.load @__auto.blk.0.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%70 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_up.weight.shard.7 = util.global.load @__auto.blk.0.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%71 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.0.ffn_down.weight.shard.0 = util.global.load @__auto.blk.0.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%72 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.1 = util.global.load @__auto.blk.0.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%73 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.2 = util.global.load @__auto.blk.0.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%74 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.3 = util.global.load @__auto.blk.0.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%75 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.4 = util.global.load @__auto.blk.0.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%76 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.5 = util.global.load @__auto.blk.0.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%77 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.6 = util.global.load @__auto.blk.0.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%78 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.0.ffn_down.weight.shard.7 = util.global.load @__auto.blk.0.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%79 = torch_c.from_builtin_tensor %__auto.blk.0.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.attn_norm.weight = util.global.load @__auto.blk.1.attn_norm.weight : tensor<4096xf32>
%80 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$1 = util.global.load @__auto.blk.1.attn_norm.weight$1 : tensor<4096xf32>
%81 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$2 = util.global.load @__auto.blk.1.attn_norm.weight$2 : tensor<4096xf32>
%82 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$3 = util.global.load @__auto.blk.1.attn_norm.weight$3 : tensor<4096xf32>
%83 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$4 = util.global.load @__auto.blk.1.attn_norm.weight$4 : tensor<4096xf32>
%84 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$5 = util.global.load @__auto.blk.1.attn_norm.weight$5 : tensor<4096xf32>
%85 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$6 = util.global.load @__auto.blk.1.attn_norm.weight$6 : tensor<4096xf32>
%86 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_norm.weight$7 = util.global.load @__auto.blk.1.attn_norm.weight$7 : tensor<4096xf32>
%87 = torch_c.from_builtin_tensor %__auto.blk.1.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.attn_q.weight.shard.0 = util.global.load @__auto.blk.1.attn_q.weight.shard.0 : tensor<512x4096xf16>
%88 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.1 = util.global.load @__auto.blk.1.attn_q.weight.shard.1 : tensor<512x4096xf16>
%89 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.2 = util.global.load @__auto.blk.1.attn_q.weight.shard.2 : tensor<512x4096xf16>
%90 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.3 = util.global.load @__auto.blk.1.attn_q.weight.shard.3 : tensor<512x4096xf16>
%91 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.4 = util.global.load @__auto.blk.1.attn_q.weight.shard.4 : tensor<512x4096xf16>
%92 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.5 = util.global.load @__auto.blk.1.attn_q.weight.shard.5 : tensor<512x4096xf16>
%93 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.6 = util.global.load @__auto.blk.1.attn_q.weight.shard.6 : tensor<512x4096xf16>
%94 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_q.weight.shard.7 = util.global.load @__auto.blk.1.attn_q.weight.shard.7 : tensor<512x4096xf16>
%95 = torch_c.from_builtin_tensor %__auto.blk.1.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.1.attn_k.weight.shard.0 = util.global.load @__auto.blk.1.attn_k.weight.shard.0 : tensor<128x4096xf16>
%96 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.1 = util.global.load @__auto.blk.1.attn_k.weight.shard.1 : tensor<128x4096xf16>
%97 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.2 = util.global.load @__auto.blk.1.attn_k.weight.shard.2 : tensor<128x4096xf16>
%98 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.3 = util.global.load @__auto.blk.1.attn_k.weight.shard.3 : tensor<128x4096xf16>
%99 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.4 = util.global.load @__auto.blk.1.attn_k.weight.shard.4 : tensor<128x4096xf16>
%100 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.5 = util.global.load @__auto.blk.1.attn_k.weight.shard.5 : tensor<128x4096xf16>
%101 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.6 = util.global.load @__auto.blk.1.attn_k.weight.shard.6 : tensor<128x4096xf16>
%102 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_k.weight.shard.7 = util.global.load @__auto.blk.1.attn_k.weight.shard.7 : tensor<128x4096xf16>
%103 = torch_c.from_builtin_tensor %__auto.blk.1.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.0 = util.global.load @__auto.blk.1.attn_v.weight.shard.0 : tensor<128x4096xf16>
%104 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.1 = util.global.load @__auto.blk.1.attn_v.weight.shard.1 : tensor<128x4096xf16>
%105 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.2 = util.global.load @__auto.blk.1.attn_v.weight.shard.2 : tensor<128x4096xf16>
%106 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.3 = util.global.load @__auto.blk.1.attn_v.weight.shard.3 : tensor<128x4096xf16>
%107 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.4 = util.global.load @__auto.blk.1.attn_v.weight.shard.4 : tensor<128x4096xf16>
%108 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.5 = util.global.load @__auto.blk.1.attn_v.weight.shard.5 : tensor<128x4096xf16>
%109 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.6 = util.global.load @__auto.blk.1.attn_v.weight.shard.6 : tensor<128x4096xf16>
%110 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_v.weight.shard.7 = util.global.load @__auto.blk.1.attn_v.weight.shard.7 : tensor<128x4096xf16>
%111 = torch_c.from_builtin_tensor %__auto.blk.1.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.1.attn_output.weight.shard.0 = util.global.load @__auto.blk.1.attn_output.weight.shard.0 : tensor<4096x512xf16>
%112 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.1 = util.global.load @__auto.blk.1.attn_output.weight.shard.1 : tensor<4096x512xf16>
%113 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.2 = util.global.load @__auto.blk.1.attn_output.weight.shard.2 : tensor<4096x512xf16>
%114 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.3 = util.global.load @__auto.blk.1.attn_output.weight.shard.3 : tensor<4096x512xf16>
%115 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.4 = util.global.load @__auto.blk.1.attn_output.weight.shard.4 : tensor<4096x512xf16>
%116 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.5 = util.global.load @__auto.blk.1.attn_output.weight.shard.5 : tensor<4096x512xf16>
%117 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.6 = util.global.load @__auto.blk.1.attn_output.weight.shard.6 : tensor<4096x512xf16>
%118 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.attn_output.weight.shard.7 = util.global.load @__auto.blk.1.attn_output.weight.shard.7 : tensor<4096x512xf16>
%119 = torch_c.from_builtin_tensor %__auto.blk.1.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.1.ffn_norm.weight = util.global.load @__auto.blk.1.ffn_norm.weight : tensor<4096xf32>
%120 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$1 = util.global.load @__auto.blk.1.ffn_norm.weight$1 : tensor<4096xf32>
%121 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$2 = util.global.load @__auto.blk.1.ffn_norm.weight$2 : tensor<4096xf32>
%122 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$3 = util.global.load @__auto.blk.1.ffn_norm.weight$3 : tensor<4096xf32>
%123 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$4 = util.global.load @__auto.blk.1.ffn_norm.weight$4 : tensor<4096xf32>
%124 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$5 = util.global.load @__auto.blk.1.ffn_norm.weight$5 : tensor<4096xf32>
%125 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$6 = util.global.load @__auto.blk.1.ffn_norm.weight$6 : tensor<4096xf32>
%126 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_norm.weight$7 = util.global.load @__auto.blk.1.ffn_norm.weight$7 : tensor<4096xf32>
%127 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.1.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%128 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%129 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%130 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%131 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%132 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%133 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%134 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.1.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%135 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.0 = util.global.load @__auto.blk.1.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%136 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.1 = util.global.load @__auto.blk.1.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%137 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.2 = util.global.load @__auto.blk.1.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%138 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.3 = util.global.load @__auto.blk.1.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%139 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.4 = util.global.load @__auto.blk.1.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%140 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.5 = util.global.load @__auto.blk.1.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%141 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.6 = util.global.load @__auto.blk.1.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%142 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_up.weight.shard.7 = util.global.load @__auto.blk.1.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%143 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.1.ffn_down.weight.shard.0 = util.global.load @__auto.blk.1.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%144 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.1 = util.global.load @__auto.blk.1.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%145 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.2 = util.global.load @__auto.blk.1.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%146 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.3 = util.global.load @__auto.blk.1.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%147 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.4 = util.global.load @__auto.blk.1.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%148 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.5 = util.global.load @__auto.blk.1.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%149 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.6 = util.global.load @__auto.blk.1.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%150 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.1.ffn_down.weight.shard.7 = util.global.load @__auto.blk.1.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%151 = torch_c.from_builtin_tensor %__auto.blk.1.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.attn_norm.weight = util.global.load @__auto.blk.2.attn_norm.weight : tensor<4096xf32>
%152 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$1 = util.global.load @__auto.blk.2.attn_norm.weight$1 : tensor<4096xf32>
%153 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$2 = util.global.load @__auto.blk.2.attn_norm.weight$2 : tensor<4096xf32>
%154 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$3 = util.global.load @__auto.blk.2.attn_norm.weight$3 : tensor<4096xf32>
%155 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$4 = util.global.load @__auto.blk.2.attn_norm.weight$4 : tensor<4096xf32>
%156 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$5 = util.global.load @__auto.blk.2.attn_norm.weight$5 : tensor<4096xf32>
%157 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$6 = util.global.load @__auto.blk.2.attn_norm.weight$6 : tensor<4096xf32>
%158 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_norm.weight$7 = util.global.load @__auto.blk.2.attn_norm.weight$7 : tensor<4096xf32>
%159 = torch_c.from_builtin_tensor %__auto.blk.2.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.attn_q.weight.shard.0 = util.global.load @__auto.blk.2.attn_q.weight.shard.0 : tensor<512x4096xf16>
%160 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.1 = util.global.load @__auto.blk.2.attn_q.weight.shard.1 : tensor<512x4096xf16>
%161 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.2 = util.global.load @__auto.blk.2.attn_q.weight.shard.2 : tensor<512x4096xf16>
%162 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.3 = util.global.load @__auto.blk.2.attn_q.weight.shard.3 : tensor<512x4096xf16>
%163 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.4 = util.global.load @__auto.blk.2.attn_q.weight.shard.4 : tensor<512x4096xf16>
%164 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.5 = util.global.load @__auto.blk.2.attn_q.weight.shard.5 : tensor<512x4096xf16>
%165 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.6 = util.global.load @__auto.blk.2.attn_q.weight.shard.6 : tensor<512x4096xf16>
%166 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_q.weight.shard.7 = util.global.load @__auto.blk.2.attn_q.weight.shard.7 : tensor<512x4096xf16>
%167 = torch_c.from_builtin_tensor %__auto.blk.2.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.2.attn_k.weight.shard.0 = util.global.load @__auto.blk.2.attn_k.weight.shard.0 : tensor<128x4096xf16>
%168 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.1 = util.global.load @__auto.blk.2.attn_k.weight.shard.1 : tensor<128x4096xf16>
%169 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.2 = util.global.load @__auto.blk.2.attn_k.weight.shard.2 : tensor<128x4096xf16>
%170 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.3 = util.global.load @__auto.blk.2.attn_k.weight.shard.3 : tensor<128x4096xf16>
%171 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.4 = util.global.load @__auto.blk.2.attn_k.weight.shard.4 : tensor<128x4096xf16>
%172 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.5 = util.global.load @__auto.blk.2.attn_k.weight.shard.5 : tensor<128x4096xf16>
%173 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.6 = util.global.load @__auto.blk.2.attn_k.weight.shard.6 : tensor<128x4096xf16>
%174 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_k.weight.shard.7 = util.global.load @__auto.blk.2.attn_k.weight.shard.7 : tensor<128x4096xf16>
%175 = torch_c.from_builtin_tensor %__auto.blk.2.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.0 = util.global.load @__auto.blk.2.attn_v.weight.shard.0 : tensor<128x4096xf16>
%176 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.1 = util.global.load @__auto.blk.2.attn_v.weight.shard.1 : tensor<128x4096xf16>
%177 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.2 = util.global.load @__auto.blk.2.attn_v.weight.shard.2 : tensor<128x4096xf16>
%178 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.3 = util.global.load @__auto.blk.2.attn_v.weight.shard.3 : tensor<128x4096xf16>
%179 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.4 = util.global.load @__auto.blk.2.attn_v.weight.shard.4 : tensor<128x4096xf16>
%180 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.5 = util.global.load @__auto.blk.2.attn_v.weight.shard.5 : tensor<128x4096xf16>
%181 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.6 = util.global.load @__auto.blk.2.attn_v.weight.shard.6 : tensor<128x4096xf16>
%182 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_v.weight.shard.7 = util.global.load @__auto.blk.2.attn_v.weight.shard.7 : tensor<128x4096xf16>
%183 = torch_c.from_builtin_tensor %__auto.blk.2.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.2.attn_output.weight.shard.0 = util.global.load @__auto.blk.2.attn_output.weight.shard.0 : tensor<4096x512xf16>
%184 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.1 = util.global.load @__auto.blk.2.attn_output.weight.shard.1 : tensor<4096x512xf16>
%185 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.2 = util.global.load @__auto.blk.2.attn_output.weight.shard.2 : tensor<4096x512xf16>
%186 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.3 = util.global.load @__auto.blk.2.attn_output.weight.shard.3 : tensor<4096x512xf16>
%187 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.4 = util.global.load @__auto.blk.2.attn_output.weight.shard.4 : tensor<4096x512xf16>
%188 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.5 = util.global.load @__auto.blk.2.attn_output.weight.shard.5 : tensor<4096x512xf16>
%189 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.6 = util.global.load @__auto.blk.2.attn_output.weight.shard.6 : tensor<4096x512xf16>
%190 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.attn_output.weight.shard.7 = util.global.load @__auto.blk.2.attn_output.weight.shard.7 : tensor<4096x512xf16>
%191 = torch_c.from_builtin_tensor %__auto.blk.2.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.2.ffn_norm.weight = util.global.load @__auto.blk.2.ffn_norm.weight : tensor<4096xf32>
%192 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$1 = util.global.load @__auto.blk.2.ffn_norm.weight$1 : tensor<4096xf32>
%193 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$2 = util.global.load @__auto.blk.2.ffn_norm.weight$2 : tensor<4096xf32>
%194 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$3 = util.global.load @__auto.blk.2.ffn_norm.weight$3 : tensor<4096xf32>
%195 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$4 = util.global.load @__auto.blk.2.ffn_norm.weight$4 : tensor<4096xf32>
%196 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$5 = util.global.load @__auto.blk.2.ffn_norm.weight$5 : tensor<4096xf32>
%197 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$6 = util.global.load @__auto.blk.2.ffn_norm.weight$6 : tensor<4096xf32>
%198 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_norm.weight$7 = util.global.load @__auto.blk.2.ffn_norm.weight$7 : tensor<4096xf32>
%199 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.2.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%200 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%201 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%202 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%203 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%204 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%205 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%206 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.2.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%207 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.0 = util.global.load @__auto.blk.2.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%208 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.1 = util.global.load @__auto.blk.2.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%209 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.2 = util.global.load @__auto.blk.2.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%210 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.3 = util.global.load @__auto.blk.2.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%211 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.4 = util.global.load @__auto.blk.2.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%212 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.5 = util.global.load @__auto.blk.2.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%213 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.6 = util.global.load @__auto.blk.2.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%214 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_up.weight.shard.7 = util.global.load @__auto.blk.2.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%215 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.2.ffn_down.weight.shard.0 = util.global.load @__auto.blk.2.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%216 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.1 = util.global.load @__auto.blk.2.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%217 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.2 = util.global.load @__auto.blk.2.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%218 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.3 = util.global.load @__auto.blk.2.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%219 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.4 = util.global.load @__auto.blk.2.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%220 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.5 = util.global.load @__auto.blk.2.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%221 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.6 = util.global.load @__auto.blk.2.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%222 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.2.ffn_down.weight.shard.7 = util.global.load @__auto.blk.2.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%223 = torch_c.from_builtin_tensor %__auto.blk.2.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.attn_norm.weight = util.global.load @__auto.blk.3.attn_norm.weight : tensor<4096xf32>
%224 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$1 = util.global.load @__auto.blk.3.attn_norm.weight$1 : tensor<4096xf32>
%225 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$2 = util.global.load @__auto.blk.3.attn_norm.weight$2 : tensor<4096xf32>
%226 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$3 = util.global.load @__auto.blk.3.attn_norm.weight$3 : tensor<4096xf32>
%227 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$4 = util.global.load @__auto.blk.3.attn_norm.weight$4 : tensor<4096xf32>
%228 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$5 = util.global.load @__auto.blk.3.attn_norm.weight$5 : tensor<4096xf32>
%229 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$6 = util.global.load @__auto.blk.3.attn_norm.weight$6 : tensor<4096xf32>
%230 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_norm.weight$7 = util.global.load @__auto.blk.3.attn_norm.weight$7 : tensor<4096xf32>
%231 = torch_c.from_builtin_tensor %__auto.blk.3.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.attn_q.weight.shard.0 = util.global.load @__auto.blk.3.attn_q.weight.shard.0 : tensor<512x4096xf16>
%232 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.1 = util.global.load @__auto.blk.3.attn_q.weight.shard.1 : tensor<512x4096xf16>
%233 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.2 = util.global.load @__auto.blk.3.attn_q.weight.shard.2 : tensor<512x4096xf16>
%234 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.3 = util.global.load @__auto.blk.3.attn_q.weight.shard.3 : tensor<512x4096xf16>
%235 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.4 = util.global.load @__auto.blk.3.attn_q.weight.shard.4 : tensor<512x4096xf16>
%236 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.5 = util.global.load @__auto.blk.3.attn_q.weight.shard.5 : tensor<512x4096xf16>
%237 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.6 = util.global.load @__auto.blk.3.attn_q.weight.shard.6 : tensor<512x4096xf16>
%238 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_q.weight.shard.7 = util.global.load @__auto.blk.3.attn_q.weight.shard.7 : tensor<512x4096xf16>
%239 = torch_c.from_builtin_tensor %__auto.blk.3.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.3.attn_k.weight.shard.0 = util.global.load @__auto.blk.3.attn_k.weight.shard.0 : tensor<128x4096xf16>
%240 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.1 = util.global.load @__auto.blk.3.attn_k.weight.shard.1 : tensor<128x4096xf16>
%241 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.2 = util.global.load @__auto.blk.3.attn_k.weight.shard.2 : tensor<128x4096xf16>
%242 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.3 = util.global.load @__auto.blk.3.attn_k.weight.shard.3 : tensor<128x4096xf16>
%243 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.4 = util.global.load @__auto.blk.3.attn_k.weight.shard.4 : tensor<128x4096xf16>
%244 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.5 = util.global.load @__auto.blk.3.attn_k.weight.shard.5 : tensor<128x4096xf16>
%245 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.6 = util.global.load @__auto.blk.3.attn_k.weight.shard.6 : tensor<128x4096xf16>
%246 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_k.weight.shard.7 = util.global.load @__auto.blk.3.attn_k.weight.shard.7 : tensor<128x4096xf16>
%247 = torch_c.from_builtin_tensor %__auto.blk.3.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.0 = util.global.load @__auto.blk.3.attn_v.weight.shard.0 : tensor<128x4096xf16>
%248 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.1 = util.global.load @__auto.blk.3.attn_v.weight.shard.1 : tensor<128x4096xf16>
%249 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.2 = util.global.load @__auto.blk.3.attn_v.weight.shard.2 : tensor<128x4096xf16>
%250 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.3 = util.global.load @__auto.blk.3.attn_v.weight.shard.3 : tensor<128x4096xf16>
%251 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.4 = util.global.load @__auto.blk.3.attn_v.weight.shard.4 : tensor<128x4096xf16>
%252 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.5 = util.global.load @__auto.blk.3.attn_v.weight.shard.5 : tensor<128x4096xf16>
%253 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.6 = util.global.load @__auto.blk.3.attn_v.weight.shard.6 : tensor<128x4096xf16>
%254 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_v.weight.shard.7 = util.global.load @__auto.blk.3.attn_v.weight.shard.7 : tensor<128x4096xf16>
%255 = torch_c.from_builtin_tensor %__auto.blk.3.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.3.attn_output.weight.shard.0 = util.global.load @__auto.blk.3.attn_output.weight.shard.0 : tensor<4096x512xf16>
%256 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.1 = util.global.load @__auto.blk.3.attn_output.weight.shard.1 : tensor<4096x512xf16>
%257 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.2 = util.global.load @__auto.blk.3.attn_output.weight.shard.2 : tensor<4096x512xf16>
%258 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.3 = util.global.load @__auto.blk.3.attn_output.weight.shard.3 : tensor<4096x512xf16>
%259 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.4 = util.global.load @__auto.blk.3.attn_output.weight.shard.4 : tensor<4096x512xf16>
%260 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.5 = util.global.load @__auto.blk.3.attn_output.weight.shard.5 : tensor<4096x512xf16>
%261 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.6 = util.global.load @__auto.blk.3.attn_output.weight.shard.6 : tensor<4096x512xf16>
%262 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.attn_output.weight.shard.7 = util.global.load @__auto.blk.3.attn_output.weight.shard.7 : tensor<4096x512xf16>
%263 = torch_c.from_builtin_tensor %__auto.blk.3.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.3.ffn_norm.weight = util.global.load @__auto.blk.3.ffn_norm.weight : tensor<4096xf32>
%264 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$1 = util.global.load @__auto.blk.3.ffn_norm.weight$1 : tensor<4096xf32>
%265 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$2 = util.global.load @__auto.blk.3.ffn_norm.weight$2 : tensor<4096xf32>
%266 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$3 = util.global.load @__auto.blk.3.ffn_norm.weight$3 : tensor<4096xf32>
%267 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$4 = util.global.load @__auto.blk.3.ffn_norm.weight$4 : tensor<4096xf32>
%268 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$5 = util.global.load @__auto.blk.3.ffn_norm.weight$5 : tensor<4096xf32>
%269 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$6 = util.global.load @__auto.blk.3.ffn_norm.weight$6 : tensor<4096xf32>
%270 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_norm.weight$7 = util.global.load @__auto.blk.3.ffn_norm.weight$7 : tensor<4096xf32>
%271 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.3.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%272 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%273 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%274 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%275 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%276 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%277 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%278 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.3.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%279 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.0 = util.global.load @__auto.blk.3.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%280 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.1 = util.global.load @__auto.blk.3.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%281 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.2 = util.global.load @__auto.blk.3.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%282 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.3 = util.global.load @__auto.blk.3.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%283 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.4 = util.global.load @__auto.blk.3.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%284 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.5 = util.global.load @__auto.blk.3.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%285 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.6 = util.global.load @__auto.blk.3.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%286 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_up.weight.shard.7 = util.global.load @__auto.blk.3.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%287 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.3.ffn_down.weight.shard.0 = util.global.load @__auto.blk.3.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%288 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.1 = util.global.load @__auto.blk.3.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%289 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.2 = util.global.load @__auto.blk.3.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%290 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.3 = util.global.load @__auto.blk.3.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%291 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.4 = util.global.load @__auto.blk.3.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%292 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.5 = util.global.load @__auto.blk.3.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%293 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.6 = util.global.load @__auto.blk.3.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%294 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.3.ffn_down.weight.shard.7 = util.global.load @__auto.blk.3.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%295 = torch_c.from_builtin_tensor %__auto.blk.3.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.attn_norm.weight = util.global.load @__auto.blk.4.attn_norm.weight : tensor<4096xf32>
%296 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$1 = util.global.load @__auto.blk.4.attn_norm.weight$1 : tensor<4096xf32>
%297 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$2 = util.global.load @__auto.blk.4.attn_norm.weight$2 : tensor<4096xf32>
%298 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$3 = util.global.load @__auto.blk.4.attn_norm.weight$3 : tensor<4096xf32>
%299 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$4 = util.global.load @__auto.blk.4.attn_norm.weight$4 : tensor<4096xf32>
%300 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$5 = util.global.load @__auto.blk.4.attn_norm.weight$5 : tensor<4096xf32>
%301 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$6 = util.global.load @__auto.blk.4.attn_norm.weight$6 : tensor<4096xf32>
%302 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_norm.weight$7 = util.global.load @__auto.blk.4.attn_norm.weight$7 : tensor<4096xf32>
%303 = torch_c.from_builtin_tensor %__auto.blk.4.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.attn_q.weight.shard.0 = util.global.load @__auto.blk.4.attn_q.weight.shard.0 : tensor<512x4096xf16>
%304 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.1 = util.global.load @__auto.blk.4.attn_q.weight.shard.1 : tensor<512x4096xf16>
%305 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.2 = util.global.load @__auto.blk.4.attn_q.weight.shard.2 : tensor<512x4096xf16>
%306 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.3 = util.global.load @__auto.blk.4.attn_q.weight.shard.3 : tensor<512x4096xf16>
%307 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.4 = util.global.load @__auto.blk.4.attn_q.weight.shard.4 : tensor<512x4096xf16>
%308 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.5 = util.global.load @__auto.blk.4.attn_q.weight.shard.5 : tensor<512x4096xf16>
%309 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.6 = util.global.load @__auto.blk.4.attn_q.weight.shard.6 : tensor<512x4096xf16>
%310 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_q.weight.shard.7 = util.global.load @__auto.blk.4.attn_q.weight.shard.7 : tensor<512x4096xf16>
%311 = torch_c.from_builtin_tensor %__auto.blk.4.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.4.attn_k.weight.shard.0 = util.global.load @__auto.blk.4.attn_k.weight.shard.0 : tensor<128x4096xf16>
%312 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.1 = util.global.load @__auto.blk.4.attn_k.weight.shard.1 : tensor<128x4096xf16>
%313 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.2 = util.global.load @__auto.blk.4.attn_k.weight.shard.2 : tensor<128x4096xf16>
%314 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.3 = util.global.load @__auto.blk.4.attn_k.weight.shard.3 : tensor<128x4096xf16>
%315 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.4 = util.global.load @__auto.blk.4.attn_k.weight.shard.4 : tensor<128x4096xf16>
%316 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.5 = util.global.load @__auto.blk.4.attn_k.weight.shard.5 : tensor<128x4096xf16>
%317 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.6 = util.global.load @__auto.blk.4.attn_k.weight.shard.6 : tensor<128x4096xf16>
%318 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_k.weight.shard.7 = util.global.load @__auto.blk.4.attn_k.weight.shard.7 : tensor<128x4096xf16>
%319 = torch_c.from_builtin_tensor %__auto.blk.4.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.0 = util.global.load @__auto.blk.4.attn_v.weight.shard.0 : tensor<128x4096xf16>
%320 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.1 = util.global.load @__auto.blk.4.attn_v.weight.shard.1 : tensor<128x4096xf16>
%321 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.2 = util.global.load @__auto.blk.4.attn_v.weight.shard.2 : tensor<128x4096xf16>
%322 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.3 = util.global.load @__auto.blk.4.attn_v.weight.shard.3 : tensor<128x4096xf16>
%323 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.4 = util.global.load @__auto.blk.4.attn_v.weight.shard.4 : tensor<128x4096xf16>
%324 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.5 = util.global.load @__auto.blk.4.attn_v.weight.shard.5 : tensor<128x4096xf16>
%325 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.6 = util.global.load @__auto.blk.4.attn_v.weight.shard.6 : tensor<128x4096xf16>
%326 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_v.weight.shard.7 = util.global.load @__auto.blk.4.attn_v.weight.shard.7 : tensor<128x4096xf16>
%327 = torch_c.from_builtin_tensor %__auto.blk.4.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.4.attn_output.weight.shard.0 = util.global.load @__auto.blk.4.attn_output.weight.shard.0 : tensor<4096x512xf16>
%328 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.1 = util.global.load @__auto.blk.4.attn_output.weight.shard.1 : tensor<4096x512xf16>
%329 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.2 = util.global.load @__auto.blk.4.attn_output.weight.shard.2 : tensor<4096x512xf16>
%330 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.3 = util.global.load @__auto.blk.4.attn_output.weight.shard.3 : tensor<4096x512xf16>
%331 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.4 = util.global.load @__auto.blk.4.attn_output.weight.shard.4 : tensor<4096x512xf16>
%332 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.5 = util.global.load @__auto.blk.4.attn_output.weight.shard.5 : tensor<4096x512xf16>
%333 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.6 = util.global.load @__auto.blk.4.attn_output.weight.shard.6 : tensor<4096x512xf16>
%334 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.attn_output.weight.shard.7 = util.global.load @__auto.blk.4.attn_output.weight.shard.7 : tensor<4096x512xf16>
%335 = torch_c.from_builtin_tensor %__auto.blk.4.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.4.ffn_norm.weight = util.global.load @__auto.blk.4.ffn_norm.weight : tensor<4096xf32>
%336 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$1 = util.global.load @__auto.blk.4.ffn_norm.weight$1 : tensor<4096xf32>
%337 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$2 = util.global.load @__auto.blk.4.ffn_norm.weight$2 : tensor<4096xf32>
%338 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$3 = util.global.load @__auto.blk.4.ffn_norm.weight$3 : tensor<4096xf32>
%339 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$4 = util.global.load @__auto.blk.4.ffn_norm.weight$4 : tensor<4096xf32>
%340 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$5 = util.global.load @__auto.blk.4.ffn_norm.weight$5 : tensor<4096xf32>
%341 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$6 = util.global.load @__auto.blk.4.ffn_norm.weight$6 : tensor<4096xf32>
%342 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_norm.weight$7 = util.global.load @__auto.blk.4.ffn_norm.weight$7 : tensor<4096xf32>
%343 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.4.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%344 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%345 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%346 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%347 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%348 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%349 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%350 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.4.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%351 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.0 = util.global.load @__auto.blk.4.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%352 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.1 = util.global.load @__auto.blk.4.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%353 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.2 = util.global.load @__auto.blk.4.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%354 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.3 = util.global.load @__auto.blk.4.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%355 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.4 = util.global.load @__auto.blk.4.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%356 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.5 = util.global.load @__auto.blk.4.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%357 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.6 = util.global.load @__auto.blk.4.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%358 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_up.weight.shard.7 = util.global.load @__auto.blk.4.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%359 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.4.ffn_down.weight.shard.0 = util.global.load @__auto.blk.4.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%360 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.1 = util.global.load @__auto.blk.4.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%361 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.2 = util.global.load @__auto.blk.4.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%362 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.3 = util.global.load @__auto.blk.4.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%363 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.4 = util.global.load @__auto.blk.4.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%364 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.5 = util.global.load @__auto.blk.4.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%365 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.6 = util.global.load @__auto.blk.4.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%366 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.4.ffn_down.weight.shard.7 = util.global.load @__auto.blk.4.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%367 = torch_c.from_builtin_tensor %__auto.blk.4.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.attn_norm.weight = util.global.load @__auto.blk.5.attn_norm.weight : tensor<4096xf32>
%368 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$1 = util.global.load @__auto.blk.5.attn_norm.weight$1 : tensor<4096xf32>
%369 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$2 = util.global.load @__auto.blk.5.attn_norm.weight$2 : tensor<4096xf32>
%370 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$3 = util.global.load @__auto.blk.5.attn_norm.weight$3 : tensor<4096xf32>
%371 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$4 = util.global.load @__auto.blk.5.attn_norm.weight$4 : tensor<4096xf32>
%372 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$5 = util.global.load @__auto.blk.5.attn_norm.weight$5 : tensor<4096xf32>
%373 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$6 = util.global.load @__auto.blk.5.attn_norm.weight$6 : tensor<4096xf32>
%374 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_norm.weight$7 = util.global.load @__auto.blk.5.attn_norm.weight$7 : tensor<4096xf32>
%375 = torch_c.from_builtin_tensor %__auto.blk.5.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.attn_q.weight.shard.0 = util.global.load @__auto.blk.5.attn_q.weight.shard.0 : tensor<512x4096xf16>
%376 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.1 = util.global.load @__auto.blk.5.attn_q.weight.shard.1 : tensor<512x4096xf16>
%377 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.2 = util.global.load @__auto.blk.5.attn_q.weight.shard.2 : tensor<512x4096xf16>
%378 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.3 = util.global.load @__auto.blk.5.attn_q.weight.shard.3 : tensor<512x4096xf16>
%379 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.4 = util.global.load @__auto.blk.5.attn_q.weight.shard.4 : tensor<512x4096xf16>
%380 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.5 = util.global.load @__auto.blk.5.attn_q.weight.shard.5 : tensor<512x4096xf16>
%381 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.6 = util.global.load @__auto.blk.5.attn_q.weight.shard.6 : tensor<512x4096xf16>
%382 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_q.weight.shard.7 = util.global.load @__auto.blk.5.attn_q.weight.shard.7 : tensor<512x4096xf16>
%383 = torch_c.from_builtin_tensor %__auto.blk.5.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.5.attn_k.weight.shard.0 = util.global.load @__auto.blk.5.attn_k.weight.shard.0 : tensor<128x4096xf16>
%384 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.1 = util.global.load @__auto.blk.5.attn_k.weight.shard.1 : tensor<128x4096xf16>
%385 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.2 = util.global.load @__auto.blk.5.attn_k.weight.shard.2 : tensor<128x4096xf16>
%386 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.3 = util.global.load @__auto.blk.5.attn_k.weight.shard.3 : tensor<128x4096xf16>
%387 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.4 = util.global.load @__auto.blk.5.attn_k.weight.shard.4 : tensor<128x4096xf16>
%388 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.5 = util.global.load @__auto.blk.5.attn_k.weight.shard.5 : tensor<128x4096xf16>
%389 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.6 = util.global.load @__auto.blk.5.attn_k.weight.shard.6 : tensor<128x4096xf16>
%390 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_k.weight.shard.7 = util.global.load @__auto.blk.5.attn_k.weight.shard.7 : tensor<128x4096xf16>
%391 = torch_c.from_builtin_tensor %__auto.blk.5.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.0 = util.global.load @__auto.blk.5.attn_v.weight.shard.0 : tensor<128x4096xf16>
%392 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.1 = util.global.load @__auto.blk.5.attn_v.weight.shard.1 : tensor<128x4096xf16>
%393 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.2 = util.global.load @__auto.blk.5.attn_v.weight.shard.2 : tensor<128x4096xf16>
%394 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.3 = util.global.load @__auto.blk.5.attn_v.weight.shard.3 : tensor<128x4096xf16>
%395 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.4 = util.global.load @__auto.blk.5.attn_v.weight.shard.4 : tensor<128x4096xf16>
%396 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.5 = util.global.load @__auto.blk.5.attn_v.weight.shard.5 : tensor<128x4096xf16>
%397 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.6 = util.global.load @__auto.blk.5.attn_v.weight.shard.6 : tensor<128x4096xf16>
%398 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_v.weight.shard.7 = util.global.load @__auto.blk.5.attn_v.weight.shard.7 : tensor<128x4096xf16>
%399 = torch_c.from_builtin_tensor %__auto.blk.5.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.5.attn_output.weight.shard.0 = util.global.load @__auto.blk.5.attn_output.weight.shard.0 : tensor<4096x512xf16>
%400 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.1 = util.global.load @__auto.blk.5.attn_output.weight.shard.1 : tensor<4096x512xf16>
%401 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.2 = util.global.load @__auto.blk.5.attn_output.weight.shard.2 : tensor<4096x512xf16>
%402 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.3 = util.global.load @__auto.blk.5.attn_output.weight.shard.3 : tensor<4096x512xf16>
%403 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.4 = util.global.load @__auto.blk.5.attn_output.weight.shard.4 : tensor<4096x512xf16>
%404 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.5 = util.global.load @__auto.blk.5.attn_output.weight.shard.5 : tensor<4096x512xf16>
%405 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.6 = util.global.load @__auto.blk.5.attn_output.weight.shard.6 : tensor<4096x512xf16>
%406 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.attn_output.weight.shard.7 = util.global.load @__auto.blk.5.attn_output.weight.shard.7 : tensor<4096x512xf16>
%407 = torch_c.from_builtin_tensor %__auto.blk.5.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.5.ffn_norm.weight = util.global.load @__auto.blk.5.ffn_norm.weight : tensor<4096xf32>
%408 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$1 = util.global.load @__auto.blk.5.ffn_norm.weight$1 : tensor<4096xf32>
%409 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$2 = util.global.load @__auto.blk.5.ffn_norm.weight$2 : tensor<4096xf32>
%410 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$3 = util.global.load @__auto.blk.5.ffn_norm.weight$3 : tensor<4096xf32>
%411 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$4 = util.global.load @__auto.blk.5.ffn_norm.weight$4 : tensor<4096xf32>
%412 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$5 = util.global.load @__auto.blk.5.ffn_norm.weight$5 : tensor<4096xf32>
%413 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$6 = util.global.load @__auto.blk.5.ffn_norm.weight$6 : tensor<4096xf32>
%414 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_norm.weight$7 = util.global.load @__auto.blk.5.ffn_norm.weight$7 : tensor<4096xf32>
%415 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.5.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%416 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%417 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%418 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%419 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%420 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%421 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%422 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.5.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%423 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.0 = util.global.load @__auto.blk.5.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%424 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.1 = util.global.load @__auto.blk.5.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%425 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.2 = util.global.load @__auto.blk.5.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%426 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.3 = util.global.load @__auto.blk.5.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%427 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.4 = util.global.load @__auto.blk.5.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%428 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.5 = util.global.load @__auto.blk.5.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%429 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.6 = util.global.load @__auto.blk.5.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%430 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_up.weight.shard.7 = util.global.load @__auto.blk.5.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%431 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.5.ffn_down.weight.shard.0 = util.global.load @__auto.blk.5.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%432 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.1 = util.global.load @__auto.blk.5.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%433 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.2 = util.global.load @__auto.blk.5.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%434 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.3 = util.global.load @__auto.blk.5.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%435 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.4 = util.global.load @__auto.blk.5.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%436 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.5 = util.global.load @__auto.blk.5.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%437 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.6 = util.global.load @__auto.blk.5.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%438 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.5.ffn_down.weight.shard.7 = util.global.load @__auto.blk.5.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%439 = torch_c.from_builtin_tensor %__auto.blk.5.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.attn_norm.weight = util.global.load @__auto.blk.6.attn_norm.weight : tensor<4096xf32>
%440 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$1 = util.global.load @__auto.blk.6.attn_norm.weight$1 : tensor<4096xf32>
%441 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$2 = util.global.load @__auto.blk.6.attn_norm.weight$2 : tensor<4096xf32>
%442 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$3 = util.global.load @__auto.blk.6.attn_norm.weight$3 : tensor<4096xf32>
%443 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$4 = util.global.load @__auto.blk.6.attn_norm.weight$4 : tensor<4096xf32>
%444 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$5 = util.global.load @__auto.blk.6.attn_norm.weight$5 : tensor<4096xf32>
%445 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$6 = util.global.load @__auto.blk.6.attn_norm.weight$6 : tensor<4096xf32>
%446 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_norm.weight$7 = util.global.load @__auto.blk.6.attn_norm.weight$7 : tensor<4096xf32>
%447 = torch_c.from_builtin_tensor %__auto.blk.6.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.attn_q.weight.shard.0 = util.global.load @__auto.blk.6.attn_q.weight.shard.0 : tensor<512x4096xf16>
%448 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.1 = util.global.load @__auto.blk.6.attn_q.weight.shard.1 : tensor<512x4096xf16>
%449 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.2 = util.global.load @__auto.blk.6.attn_q.weight.shard.2 : tensor<512x4096xf16>
%450 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.3 = util.global.load @__auto.blk.6.attn_q.weight.shard.3 : tensor<512x4096xf16>
%451 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.4 = util.global.load @__auto.blk.6.attn_q.weight.shard.4 : tensor<512x4096xf16>
%452 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.5 = util.global.load @__auto.blk.6.attn_q.weight.shard.5 : tensor<512x4096xf16>
%453 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.6 = util.global.load @__auto.blk.6.attn_q.weight.shard.6 : tensor<512x4096xf16>
%454 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_q.weight.shard.7 = util.global.load @__auto.blk.6.attn_q.weight.shard.7 : tensor<512x4096xf16>
%455 = torch_c.from_builtin_tensor %__auto.blk.6.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.6.attn_k.weight.shard.0 = util.global.load @__auto.blk.6.attn_k.weight.shard.0 : tensor<128x4096xf16>
%456 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.1 = util.global.load @__auto.blk.6.attn_k.weight.shard.1 : tensor<128x4096xf16>
%457 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.2 = util.global.load @__auto.blk.6.attn_k.weight.shard.2 : tensor<128x4096xf16>
%458 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.3 = util.global.load @__auto.blk.6.attn_k.weight.shard.3 : tensor<128x4096xf16>
%459 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.4 = util.global.load @__auto.blk.6.attn_k.weight.shard.4 : tensor<128x4096xf16>
%460 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.5 = util.global.load @__auto.blk.6.attn_k.weight.shard.5 : tensor<128x4096xf16>
%461 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.6 = util.global.load @__auto.blk.6.attn_k.weight.shard.6 : tensor<128x4096xf16>
%462 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_k.weight.shard.7 = util.global.load @__auto.blk.6.attn_k.weight.shard.7 : tensor<128x4096xf16>
%463 = torch_c.from_builtin_tensor %__auto.blk.6.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.0 = util.global.load @__auto.blk.6.attn_v.weight.shard.0 : tensor<128x4096xf16>
%464 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.1 = util.global.load @__auto.blk.6.attn_v.weight.shard.1 : tensor<128x4096xf16>
%465 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.2 = util.global.load @__auto.blk.6.attn_v.weight.shard.2 : tensor<128x4096xf16>
%466 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.3 = util.global.load @__auto.blk.6.attn_v.weight.shard.3 : tensor<128x4096xf16>
%467 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.4 = util.global.load @__auto.blk.6.attn_v.weight.shard.4 : tensor<128x4096xf16>
%468 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.5 = util.global.load @__auto.blk.6.attn_v.weight.shard.5 : tensor<128x4096xf16>
%469 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.6 = util.global.load @__auto.blk.6.attn_v.weight.shard.6 : tensor<128x4096xf16>
%470 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_v.weight.shard.7 = util.global.load @__auto.blk.6.attn_v.weight.shard.7 : tensor<128x4096xf16>
%471 = torch_c.from_builtin_tensor %__auto.blk.6.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.6.attn_output.weight.shard.0 = util.global.load @__auto.blk.6.attn_output.weight.shard.0 : tensor<4096x512xf16>
%472 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.1 = util.global.load @__auto.blk.6.attn_output.weight.shard.1 : tensor<4096x512xf16>
%473 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.2 = util.global.load @__auto.blk.6.attn_output.weight.shard.2 : tensor<4096x512xf16>
%474 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.3 = util.global.load @__auto.blk.6.attn_output.weight.shard.3 : tensor<4096x512xf16>
%475 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.4 = util.global.load @__auto.blk.6.attn_output.weight.shard.4 : tensor<4096x512xf16>
%476 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.5 = util.global.load @__auto.blk.6.attn_output.weight.shard.5 : tensor<4096x512xf16>
%477 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.6 = util.global.load @__auto.blk.6.attn_output.weight.shard.6 : tensor<4096x512xf16>
%478 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.attn_output.weight.shard.7 = util.global.load @__auto.blk.6.attn_output.weight.shard.7 : tensor<4096x512xf16>
%479 = torch_c.from_builtin_tensor %__auto.blk.6.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.6.ffn_norm.weight = util.global.load @__auto.blk.6.ffn_norm.weight : tensor<4096xf32>
%480 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$1 = util.global.load @__auto.blk.6.ffn_norm.weight$1 : tensor<4096xf32>
%481 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$2 = util.global.load @__auto.blk.6.ffn_norm.weight$2 : tensor<4096xf32>
%482 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$3 = util.global.load @__auto.blk.6.ffn_norm.weight$3 : tensor<4096xf32>
%483 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$4 = util.global.load @__auto.blk.6.ffn_norm.weight$4 : tensor<4096xf32>
%484 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$5 = util.global.load @__auto.blk.6.ffn_norm.weight$5 : tensor<4096xf32>
%485 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$6 = util.global.load @__auto.blk.6.ffn_norm.weight$6 : tensor<4096xf32>
%486 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_norm.weight$7 = util.global.load @__auto.blk.6.ffn_norm.weight$7 : tensor<4096xf32>
%487 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.6.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%488 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%489 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%490 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%491 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%492 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%493 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%494 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.6.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%495 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.0 = util.global.load @__auto.blk.6.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%496 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.1 = util.global.load @__auto.blk.6.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%497 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.2 = util.global.load @__auto.blk.6.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%498 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.3 = util.global.load @__auto.blk.6.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%499 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.4 = util.global.load @__auto.blk.6.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%500 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.5 = util.global.load @__auto.blk.6.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%501 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.6 = util.global.load @__auto.blk.6.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%502 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_up.weight.shard.7 = util.global.load @__auto.blk.6.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%503 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.6.ffn_down.weight.shard.0 = util.global.load @__auto.blk.6.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%504 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.1 = util.global.load @__auto.blk.6.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%505 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.2 = util.global.load @__auto.blk.6.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%506 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.3 = util.global.load @__auto.blk.6.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%507 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.4 = util.global.load @__auto.blk.6.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%508 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.5 = util.global.load @__auto.blk.6.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%509 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.6 = util.global.load @__auto.blk.6.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%510 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.6.ffn_down.weight.shard.7 = util.global.load @__auto.blk.6.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%511 = torch_c.from_builtin_tensor %__auto.blk.6.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.attn_norm.weight = util.global.load @__auto.blk.7.attn_norm.weight : tensor<4096xf32>
%512 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$1 = util.global.load @__auto.blk.7.attn_norm.weight$1 : tensor<4096xf32>
%513 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$2 = util.global.load @__auto.blk.7.attn_norm.weight$2 : tensor<4096xf32>
%514 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$3 = util.global.load @__auto.blk.7.attn_norm.weight$3 : tensor<4096xf32>
%515 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$4 = util.global.load @__auto.blk.7.attn_norm.weight$4 : tensor<4096xf32>
%516 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$5 = util.global.load @__auto.blk.7.attn_norm.weight$5 : tensor<4096xf32>
%517 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$6 = util.global.load @__auto.blk.7.attn_norm.weight$6 : tensor<4096xf32>
%518 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_norm.weight$7 = util.global.load @__auto.blk.7.attn_norm.weight$7 : tensor<4096xf32>
%519 = torch_c.from_builtin_tensor %__auto.blk.7.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.attn_q.weight.shard.0 = util.global.load @__auto.blk.7.attn_q.weight.shard.0 : tensor<512x4096xf16>
%520 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.1 = util.global.load @__auto.blk.7.attn_q.weight.shard.1 : tensor<512x4096xf16>
%521 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.2 = util.global.load @__auto.blk.7.attn_q.weight.shard.2 : tensor<512x4096xf16>
%522 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.3 = util.global.load @__auto.blk.7.attn_q.weight.shard.3 : tensor<512x4096xf16>
%523 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.4 = util.global.load @__auto.blk.7.attn_q.weight.shard.4 : tensor<512x4096xf16>
%524 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.5 = util.global.load @__auto.blk.7.attn_q.weight.shard.5 : tensor<512x4096xf16>
%525 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.6 = util.global.load @__auto.blk.7.attn_q.weight.shard.6 : tensor<512x4096xf16>
%526 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_q.weight.shard.7 = util.global.load @__auto.blk.7.attn_q.weight.shard.7 : tensor<512x4096xf16>
%527 = torch_c.from_builtin_tensor %__auto.blk.7.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.7.attn_k.weight.shard.0 = util.global.load @__auto.blk.7.attn_k.weight.shard.0 : tensor<128x4096xf16>
%528 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.1 = util.global.load @__auto.blk.7.attn_k.weight.shard.1 : tensor<128x4096xf16>
%529 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.2 = util.global.load @__auto.blk.7.attn_k.weight.shard.2 : tensor<128x4096xf16>
%530 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.3 = util.global.load @__auto.blk.7.attn_k.weight.shard.3 : tensor<128x4096xf16>
%531 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.4 = util.global.load @__auto.blk.7.attn_k.weight.shard.4 : tensor<128x4096xf16>
%532 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.5 = util.global.load @__auto.blk.7.attn_k.weight.shard.5 : tensor<128x4096xf16>
%533 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.6 = util.global.load @__auto.blk.7.attn_k.weight.shard.6 : tensor<128x4096xf16>
%534 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_k.weight.shard.7 = util.global.load @__auto.blk.7.attn_k.weight.shard.7 : tensor<128x4096xf16>
%535 = torch_c.from_builtin_tensor %__auto.blk.7.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.0 = util.global.load @__auto.blk.7.attn_v.weight.shard.0 : tensor<128x4096xf16>
%536 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.1 = util.global.load @__auto.blk.7.attn_v.weight.shard.1 : tensor<128x4096xf16>
%537 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.2 = util.global.load @__auto.blk.7.attn_v.weight.shard.2 : tensor<128x4096xf16>
%538 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.3 = util.global.load @__auto.blk.7.attn_v.weight.shard.3 : tensor<128x4096xf16>
%539 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.4 = util.global.load @__auto.blk.7.attn_v.weight.shard.4 : tensor<128x4096xf16>
%540 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.5 = util.global.load @__auto.blk.7.attn_v.weight.shard.5 : tensor<128x4096xf16>
%541 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.6 = util.global.load @__auto.blk.7.attn_v.weight.shard.6 : tensor<128x4096xf16>
%542 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_v.weight.shard.7 = util.global.load @__auto.blk.7.attn_v.weight.shard.7 : tensor<128x4096xf16>
%543 = torch_c.from_builtin_tensor %__auto.blk.7.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.7.attn_output.weight.shard.0 = util.global.load @__auto.blk.7.attn_output.weight.shard.0 : tensor<4096x512xf16>
%544 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.1 = util.global.load @__auto.blk.7.attn_output.weight.shard.1 : tensor<4096x512xf16>
%545 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.2 = util.global.load @__auto.blk.7.attn_output.weight.shard.2 : tensor<4096x512xf16>
%546 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.3 = util.global.load @__auto.blk.7.attn_output.weight.shard.3 : tensor<4096x512xf16>
%547 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.4 = util.global.load @__auto.blk.7.attn_output.weight.shard.4 : tensor<4096x512xf16>
%548 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.5 = util.global.load @__auto.blk.7.attn_output.weight.shard.5 : tensor<4096x512xf16>
%549 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.6 = util.global.load @__auto.blk.7.attn_output.weight.shard.6 : tensor<4096x512xf16>
%550 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.attn_output.weight.shard.7 = util.global.load @__auto.blk.7.attn_output.weight.shard.7 : tensor<4096x512xf16>
%551 = torch_c.from_builtin_tensor %__auto.blk.7.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.7.ffn_norm.weight = util.global.load @__auto.blk.7.ffn_norm.weight : tensor<4096xf32>
%552 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$1 = util.global.load @__auto.blk.7.ffn_norm.weight$1 : tensor<4096xf32>
%553 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$2 = util.global.load @__auto.blk.7.ffn_norm.weight$2 : tensor<4096xf32>
%554 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$3 = util.global.load @__auto.blk.7.ffn_norm.weight$3 : tensor<4096xf32>
%555 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$4 = util.global.load @__auto.blk.7.ffn_norm.weight$4 : tensor<4096xf32>
%556 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$5 = util.global.load @__auto.blk.7.ffn_norm.weight$5 : tensor<4096xf32>
%557 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$6 = util.global.load @__auto.blk.7.ffn_norm.weight$6 : tensor<4096xf32>
%558 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_norm.weight$7 = util.global.load @__auto.blk.7.ffn_norm.weight$7 : tensor<4096xf32>
%559 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.7.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%560 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%561 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%562 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%563 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%564 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%565 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%566 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.7.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%567 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.0 = util.global.load @__auto.blk.7.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%568 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.1 = util.global.load @__auto.blk.7.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%569 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.2 = util.global.load @__auto.blk.7.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%570 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.3 = util.global.load @__auto.blk.7.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%571 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.4 = util.global.load @__auto.blk.7.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%572 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.5 = util.global.load @__auto.blk.7.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%573 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.6 = util.global.load @__auto.blk.7.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%574 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_up.weight.shard.7 = util.global.load @__auto.blk.7.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%575 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.7.ffn_down.weight.shard.0 = util.global.load @__auto.blk.7.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%576 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.1 = util.global.load @__auto.blk.7.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%577 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.2 = util.global.load @__auto.blk.7.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%578 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.3 = util.global.load @__auto.blk.7.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%579 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.4 = util.global.load @__auto.blk.7.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%580 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.5 = util.global.load @__auto.blk.7.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%581 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.6 = util.global.load @__auto.blk.7.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%582 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.7.ffn_down.weight.shard.7 = util.global.load @__auto.blk.7.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%583 = torch_c.from_builtin_tensor %__auto.blk.7.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.attn_norm.weight = util.global.load @__auto.blk.8.attn_norm.weight : tensor<4096xf32>
%584 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$1 = util.global.load @__auto.blk.8.attn_norm.weight$1 : tensor<4096xf32>
%585 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$2 = util.global.load @__auto.blk.8.attn_norm.weight$2 : tensor<4096xf32>
%586 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$3 = util.global.load @__auto.blk.8.attn_norm.weight$3 : tensor<4096xf32>
%587 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$4 = util.global.load @__auto.blk.8.attn_norm.weight$4 : tensor<4096xf32>
%588 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$5 = util.global.load @__auto.blk.8.attn_norm.weight$5 : tensor<4096xf32>
%589 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$6 = util.global.load @__auto.blk.8.attn_norm.weight$6 : tensor<4096xf32>
%590 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_norm.weight$7 = util.global.load @__auto.blk.8.attn_norm.weight$7 : tensor<4096xf32>
%591 = torch_c.from_builtin_tensor %__auto.blk.8.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.attn_q.weight.shard.0 = util.global.load @__auto.blk.8.attn_q.weight.shard.0 : tensor<512x4096xf16>
%592 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.1 = util.global.load @__auto.blk.8.attn_q.weight.shard.1 : tensor<512x4096xf16>
%593 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.2 = util.global.load @__auto.blk.8.attn_q.weight.shard.2 : tensor<512x4096xf16>
%594 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.3 = util.global.load @__auto.blk.8.attn_q.weight.shard.3 : tensor<512x4096xf16>
%595 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.4 = util.global.load @__auto.blk.8.attn_q.weight.shard.4 : tensor<512x4096xf16>
%596 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.5 = util.global.load @__auto.blk.8.attn_q.weight.shard.5 : tensor<512x4096xf16>
%597 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.6 = util.global.load @__auto.blk.8.attn_q.weight.shard.6 : tensor<512x4096xf16>
%598 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_q.weight.shard.7 = util.global.load @__auto.blk.8.attn_q.weight.shard.7 : tensor<512x4096xf16>
%599 = torch_c.from_builtin_tensor %__auto.blk.8.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.8.attn_k.weight.shard.0 = util.global.load @__auto.blk.8.attn_k.weight.shard.0 : tensor<128x4096xf16>
%600 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.1 = util.global.load @__auto.blk.8.attn_k.weight.shard.1 : tensor<128x4096xf16>
%601 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.2 = util.global.load @__auto.blk.8.attn_k.weight.shard.2 : tensor<128x4096xf16>
%602 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.3 = util.global.load @__auto.blk.8.attn_k.weight.shard.3 : tensor<128x4096xf16>
%603 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.4 = util.global.load @__auto.blk.8.attn_k.weight.shard.4 : tensor<128x4096xf16>
%604 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.5 = util.global.load @__auto.blk.8.attn_k.weight.shard.5 : tensor<128x4096xf16>
%605 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.6 = util.global.load @__auto.blk.8.attn_k.weight.shard.6 : tensor<128x4096xf16>
%606 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_k.weight.shard.7 = util.global.load @__auto.blk.8.attn_k.weight.shard.7 : tensor<128x4096xf16>
%607 = torch_c.from_builtin_tensor %__auto.blk.8.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.0 = util.global.load @__auto.blk.8.attn_v.weight.shard.0 : tensor<128x4096xf16>
%608 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.1 = util.global.load @__auto.blk.8.attn_v.weight.shard.1 : tensor<128x4096xf16>
%609 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.2 = util.global.load @__auto.blk.8.attn_v.weight.shard.2 : tensor<128x4096xf16>
%610 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.3 = util.global.load @__auto.blk.8.attn_v.weight.shard.3 : tensor<128x4096xf16>
%611 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.4 = util.global.load @__auto.blk.8.attn_v.weight.shard.4 : tensor<128x4096xf16>
%612 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.5 = util.global.load @__auto.blk.8.attn_v.weight.shard.5 : tensor<128x4096xf16>
%613 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.6 = util.global.load @__auto.blk.8.attn_v.weight.shard.6 : tensor<128x4096xf16>
%614 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_v.weight.shard.7 = util.global.load @__auto.blk.8.attn_v.weight.shard.7 : tensor<128x4096xf16>
%615 = torch_c.from_builtin_tensor %__auto.blk.8.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.8.attn_output.weight.shard.0 = util.global.load @__auto.blk.8.attn_output.weight.shard.0 : tensor<4096x512xf16>
%616 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.1 = util.global.load @__auto.blk.8.attn_output.weight.shard.1 : tensor<4096x512xf16>
%617 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.2 = util.global.load @__auto.blk.8.attn_output.weight.shard.2 : tensor<4096x512xf16>
%618 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.3 = util.global.load @__auto.blk.8.attn_output.weight.shard.3 : tensor<4096x512xf16>
%619 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.4 = util.global.load @__auto.blk.8.attn_output.weight.shard.4 : tensor<4096x512xf16>
%620 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.5 = util.global.load @__auto.blk.8.attn_output.weight.shard.5 : tensor<4096x512xf16>
%621 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.6 = util.global.load @__auto.blk.8.attn_output.weight.shard.6 : tensor<4096x512xf16>
%622 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.attn_output.weight.shard.7 = util.global.load @__auto.blk.8.attn_output.weight.shard.7 : tensor<4096x512xf16>
%623 = torch_c.from_builtin_tensor %__auto.blk.8.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.8.ffn_norm.weight = util.global.load @__auto.blk.8.ffn_norm.weight : tensor<4096xf32>
%624 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$1 = util.global.load @__auto.blk.8.ffn_norm.weight$1 : tensor<4096xf32>
%625 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$2 = util.global.load @__auto.blk.8.ffn_norm.weight$2 : tensor<4096xf32>
%626 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$3 = util.global.load @__auto.blk.8.ffn_norm.weight$3 : tensor<4096xf32>
%627 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$4 = util.global.load @__auto.blk.8.ffn_norm.weight$4 : tensor<4096xf32>
%628 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$5 = util.global.load @__auto.blk.8.ffn_norm.weight$5 : tensor<4096xf32>
%629 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$6 = util.global.load @__auto.blk.8.ffn_norm.weight$6 : tensor<4096xf32>
%630 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_norm.weight$7 = util.global.load @__auto.blk.8.ffn_norm.weight$7 : tensor<4096xf32>
%631 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.8.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%632 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%633 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%634 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%635 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%636 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%637 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%638 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.8.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%639 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.0 = util.global.load @__auto.blk.8.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%640 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.1 = util.global.load @__auto.blk.8.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%641 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.2 = util.global.load @__auto.blk.8.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%642 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.3 = util.global.load @__auto.blk.8.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%643 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.4 = util.global.load @__auto.blk.8.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%644 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.5 = util.global.load @__auto.blk.8.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%645 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.6 = util.global.load @__auto.blk.8.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%646 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_up.weight.shard.7 = util.global.load @__auto.blk.8.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%647 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.8.ffn_down.weight.shard.0 = util.global.load @__auto.blk.8.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%648 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.1 = util.global.load @__auto.blk.8.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%649 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.2 = util.global.load @__auto.blk.8.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%650 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.3 = util.global.load @__auto.blk.8.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%651 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.4 = util.global.load @__auto.blk.8.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%652 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.5 = util.global.load @__auto.blk.8.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%653 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.6 = util.global.load @__auto.blk.8.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%654 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.8.ffn_down.weight.shard.7 = util.global.load @__auto.blk.8.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%655 = torch_c.from_builtin_tensor %__auto.blk.8.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.attn_norm.weight = util.global.load @__auto.blk.9.attn_norm.weight : tensor<4096xf32>
%656 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$1 = util.global.load @__auto.blk.9.attn_norm.weight$1 : tensor<4096xf32>
%657 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$2 = util.global.load @__auto.blk.9.attn_norm.weight$2 : tensor<4096xf32>
%658 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$3 = util.global.load @__auto.blk.9.attn_norm.weight$3 : tensor<4096xf32>
%659 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$4 = util.global.load @__auto.blk.9.attn_norm.weight$4 : tensor<4096xf32>
%660 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$5 = util.global.load @__auto.blk.9.attn_norm.weight$5 : tensor<4096xf32>
%661 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$6 = util.global.load @__auto.blk.9.attn_norm.weight$6 : tensor<4096xf32>
%662 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_norm.weight$7 = util.global.load @__auto.blk.9.attn_norm.weight$7 : tensor<4096xf32>
%663 = torch_c.from_builtin_tensor %__auto.blk.9.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.attn_q.weight.shard.0 = util.global.load @__auto.blk.9.attn_q.weight.shard.0 : tensor<512x4096xf16>
%664 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.1 = util.global.load @__auto.blk.9.attn_q.weight.shard.1 : tensor<512x4096xf16>
%665 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.2 = util.global.load @__auto.blk.9.attn_q.weight.shard.2 : tensor<512x4096xf16>
%666 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.3 = util.global.load @__auto.blk.9.attn_q.weight.shard.3 : tensor<512x4096xf16>
%667 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.4 = util.global.load @__auto.blk.9.attn_q.weight.shard.4 : tensor<512x4096xf16>
%668 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.5 = util.global.load @__auto.blk.9.attn_q.weight.shard.5 : tensor<512x4096xf16>
%669 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.6 = util.global.load @__auto.blk.9.attn_q.weight.shard.6 : tensor<512x4096xf16>
%670 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_q.weight.shard.7 = util.global.load @__auto.blk.9.attn_q.weight.shard.7 : tensor<512x4096xf16>
%671 = torch_c.from_builtin_tensor %__auto.blk.9.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.9.attn_k.weight.shard.0 = util.global.load @__auto.blk.9.attn_k.weight.shard.0 : tensor<128x4096xf16>
%672 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.1 = util.global.load @__auto.blk.9.attn_k.weight.shard.1 : tensor<128x4096xf16>
%673 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.2 = util.global.load @__auto.blk.9.attn_k.weight.shard.2 : tensor<128x4096xf16>
%674 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.3 = util.global.load @__auto.blk.9.attn_k.weight.shard.3 : tensor<128x4096xf16>
%675 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.4 = util.global.load @__auto.blk.9.attn_k.weight.shard.4 : tensor<128x4096xf16>
%676 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.5 = util.global.load @__auto.blk.9.attn_k.weight.shard.5 : tensor<128x4096xf16>
%677 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.6 = util.global.load @__auto.blk.9.attn_k.weight.shard.6 : tensor<128x4096xf16>
%678 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_k.weight.shard.7 = util.global.load @__auto.blk.9.attn_k.weight.shard.7 : tensor<128x4096xf16>
%679 = torch_c.from_builtin_tensor %__auto.blk.9.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.0 = util.global.load @__auto.blk.9.attn_v.weight.shard.0 : tensor<128x4096xf16>
%680 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.1 = util.global.load @__auto.blk.9.attn_v.weight.shard.1 : tensor<128x4096xf16>
%681 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.2 = util.global.load @__auto.blk.9.attn_v.weight.shard.2 : tensor<128x4096xf16>
%682 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.3 = util.global.load @__auto.blk.9.attn_v.weight.shard.3 : tensor<128x4096xf16>
%683 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.4 = util.global.load @__auto.blk.9.attn_v.weight.shard.4 : tensor<128x4096xf16>
%684 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.5 = util.global.load @__auto.blk.9.attn_v.weight.shard.5 : tensor<128x4096xf16>
%685 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.6 = util.global.load @__auto.blk.9.attn_v.weight.shard.6 : tensor<128x4096xf16>
%686 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_v.weight.shard.7 = util.global.load @__auto.blk.9.attn_v.weight.shard.7 : tensor<128x4096xf16>
%687 = torch_c.from_builtin_tensor %__auto.blk.9.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.9.attn_output.weight.shard.0 = util.global.load @__auto.blk.9.attn_output.weight.shard.0 : tensor<4096x512xf16>
%688 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.1 = util.global.load @__auto.blk.9.attn_output.weight.shard.1 : tensor<4096x512xf16>
%689 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.2 = util.global.load @__auto.blk.9.attn_output.weight.shard.2 : tensor<4096x512xf16>
%690 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.3 = util.global.load @__auto.blk.9.attn_output.weight.shard.3 : tensor<4096x512xf16>
%691 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.4 = util.global.load @__auto.blk.9.attn_output.weight.shard.4 : tensor<4096x512xf16>
%692 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.5 = util.global.load @__auto.blk.9.attn_output.weight.shard.5 : tensor<4096x512xf16>
%693 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.6 = util.global.load @__auto.blk.9.attn_output.weight.shard.6 : tensor<4096x512xf16>
%694 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.attn_output.weight.shard.7 = util.global.load @__auto.blk.9.attn_output.weight.shard.7 : tensor<4096x512xf16>
%695 = torch_c.from_builtin_tensor %__auto.blk.9.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.9.ffn_norm.weight = util.global.load @__auto.blk.9.ffn_norm.weight : tensor<4096xf32>
%696 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$1 = util.global.load @__auto.blk.9.ffn_norm.weight$1 : tensor<4096xf32>
%697 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$2 = util.global.load @__auto.blk.9.ffn_norm.weight$2 : tensor<4096xf32>
%698 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$3 = util.global.load @__auto.blk.9.ffn_norm.weight$3 : tensor<4096xf32>
%699 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$4 = util.global.load @__auto.blk.9.ffn_norm.weight$4 : tensor<4096xf32>
%700 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$5 = util.global.load @__auto.blk.9.ffn_norm.weight$5 : tensor<4096xf32>
%701 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$6 = util.global.load @__auto.blk.9.ffn_norm.weight$6 : tensor<4096xf32>
%702 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_norm.weight$7 = util.global.load @__auto.blk.9.ffn_norm.weight$7 : tensor<4096xf32>
%703 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.9.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%704 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%705 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%706 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%707 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%708 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%709 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%710 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.9.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%711 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.0 = util.global.load @__auto.blk.9.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%712 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.1 = util.global.load @__auto.blk.9.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%713 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.2 = util.global.load @__auto.blk.9.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%714 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.3 = util.global.load @__auto.blk.9.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%715 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.4 = util.global.load @__auto.blk.9.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%716 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.5 = util.global.load @__auto.blk.9.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%717 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.6 = util.global.load @__auto.blk.9.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%718 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_up.weight.shard.7 = util.global.load @__auto.blk.9.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%719 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.9.ffn_down.weight.shard.0 = util.global.load @__auto.blk.9.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%720 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.1 = util.global.load @__auto.blk.9.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%721 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.2 = util.global.load @__auto.blk.9.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%722 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.3 = util.global.load @__auto.blk.9.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%723 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.4 = util.global.load @__auto.blk.9.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%724 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.5 = util.global.load @__auto.blk.9.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%725 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.6 = util.global.load @__auto.blk.9.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%726 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.9.ffn_down.weight.shard.7 = util.global.load @__auto.blk.9.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%727 = torch_c.from_builtin_tensor %__auto.blk.9.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.attn_norm.weight = util.global.load @__auto.blk.10.attn_norm.weight : tensor<4096xf32>
%728 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$1 = util.global.load @__auto.blk.10.attn_norm.weight$1 : tensor<4096xf32>
%729 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$2 = util.global.load @__auto.blk.10.attn_norm.weight$2 : tensor<4096xf32>
%730 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$3 = util.global.load @__auto.blk.10.attn_norm.weight$3 : tensor<4096xf32>
%731 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$4 = util.global.load @__auto.blk.10.attn_norm.weight$4 : tensor<4096xf32>
%732 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$5 = util.global.load @__auto.blk.10.attn_norm.weight$5 : tensor<4096xf32>
%733 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$6 = util.global.load @__auto.blk.10.attn_norm.weight$6 : tensor<4096xf32>
%734 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_norm.weight$7 = util.global.load @__auto.blk.10.attn_norm.weight$7 : tensor<4096xf32>
%735 = torch_c.from_builtin_tensor %__auto.blk.10.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.attn_q.weight.shard.0 = util.global.load @__auto.blk.10.attn_q.weight.shard.0 : tensor<512x4096xf16>
%736 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.1 = util.global.load @__auto.blk.10.attn_q.weight.shard.1 : tensor<512x4096xf16>
%737 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.2 = util.global.load @__auto.blk.10.attn_q.weight.shard.2 : tensor<512x4096xf16>
%738 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.3 = util.global.load @__auto.blk.10.attn_q.weight.shard.3 : tensor<512x4096xf16>
%739 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.4 = util.global.load @__auto.blk.10.attn_q.weight.shard.4 : tensor<512x4096xf16>
%740 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.5 = util.global.load @__auto.blk.10.attn_q.weight.shard.5 : tensor<512x4096xf16>
%741 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.6 = util.global.load @__auto.blk.10.attn_q.weight.shard.6 : tensor<512x4096xf16>
%742 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_q.weight.shard.7 = util.global.load @__auto.blk.10.attn_q.weight.shard.7 : tensor<512x4096xf16>
%743 = torch_c.from_builtin_tensor %__auto.blk.10.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.10.attn_k.weight.shard.0 = util.global.load @__auto.blk.10.attn_k.weight.shard.0 : tensor<128x4096xf16>
%744 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.1 = util.global.load @__auto.blk.10.attn_k.weight.shard.1 : tensor<128x4096xf16>
%745 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.2 = util.global.load @__auto.blk.10.attn_k.weight.shard.2 : tensor<128x4096xf16>
%746 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.3 = util.global.load @__auto.blk.10.attn_k.weight.shard.3 : tensor<128x4096xf16>
%747 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.4 = util.global.load @__auto.blk.10.attn_k.weight.shard.4 : tensor<128x4096xf16>
%748 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.5 = util.global.load @__auto.blk.10.attn_k.weight.shard.5 : tensor<128x4096xf16>
%749 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.6 = util.global.load @__auto.blk.10.attn_k.weight.shard.6 : tensor<128x4096xf16>
%750 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_k.weight.shard.7 = util.global.load @__auto.blk.10.attn_k.weight.shard.7 : tensor<128x4096xf16>
%751 = torch_c.from_builtin_tensor %__auto.blk.10.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.0 = util.global.load @__auto.blk.10.attn_v.weight.shard.0 : tensor<128x4096xf16>
%752 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.1 = util.global.load @__auto.blk.10.attn_v.weight.shard.1 : tensor<128x4096xf16>
%753 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.2 = util.global.load @__auto.blk.10.attn_v.weight.shard.2 : tensor<128x4096xf16>
%754 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.3 = util.global.load @__auto.blk.10.attn_v.weight.shard.3 : tensor<128x4096xf16>
%755 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.4 = util.global.load @__auto.blk.10.attn_v.weight.shard.4 : tensor<128x4096xf16>
%756 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.5 = util.global.load @__auto.blk.10.attn_v.weight.shard.5 : tensor<128x4096xf16>
%757 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.6 = util.global.load @__auto.blk.10.attn_v.weight.shard.6 : tensor<128x4096xf16>
%758 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_v.weight.shard.7 = util.global.load @__auto.blk.10.attn_v.weight.shard.7 : tensor<128x4096xf16>
%759 = torch_c.from_builtin_tensor %__auto.blk.10.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.10.attn_output.weight.shard.0 = util.global.load @__auto.blk.10.attn_output.weight.shard.0 : tensor<4096x512xf16>
%760 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.1 = util.global.load @__auto.blk.10.attn_output.weight.shard.1 : tensor<4096x512xf16>
%761 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.2 = util.global.load @__auto.blk.10.attn_output.weight.shard.2 : tensor<4096x512xf16>
%762 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.3 = util.global.load @__auto.blk.10.attn_output.weight.shard.3 : tensor<4096x512xf16>
%763 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.4 = util.global.load @__auto.blk.10.attn_output.weight.shard.4 : tensor<4096x512xf16>
%764 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.5 = util.global.load @__auto.blk.10.attn_output.weight.shard.5 : tensor<4096x512xf16>
%765 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.6 = util.global.load @__auto.blk.10.attn_output.weight.shard.6 : tensor<4096x512xf16>
%766 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.attn_output.weight.shard.7 = util.global.load @__auto.blk.10.attn_output.weight.shard.7 : tensor<4096x512xf16>
%767 = torch_c.from_builtin_tensor %__auto.blk.10.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.10.ffn_norm.weight = util.global.load @__auto.blk.10.ffn_norm.weight : tensor<4096xf32>
%768 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$1 = util.global.load @__auto.blk.10.ffn_norm.weight$1 : tensor<4096xf32>
%769 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$2 = util.global.load @__auto.blk.10.ffn_norm.weight$2 : tensor<4096xf32>
%770 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$3 = util.global.load @__auto.blk.10.ffn_norm.weight$3 : tensor<4096xf32>
%771 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$4 = util.global.load @__auto.blk.10.ffn_norm.weight$4 : tensor<4096xf32>
%772 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$5 = util.global.load @__auto.blk.10.ffn_norm.weight$5 : tensor<4096xf32>
%773 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$6 = util.global.load @__auto.blk.10.ffn_norm.weight$6 : tensor<4096xf32>
%774 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_norm.weight$7 = util.global.load @__auto.blk.10.ffn_norm.weight$7 : tensor<4096xf32>
%775 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.10.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%776 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%777 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%778 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%779 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%780 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%781 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%782 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.10.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%783 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.0 = util.global.load @__auto.blk.10.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%784 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.1 = util.global.load @__auto.blk.10.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%785 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.2 = util.global.load @__auto.blk.10.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%786 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.3 = util.global.load @__auto.blk.10.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%787 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.4 = util.global.load @__auto.blk.10.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%788 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.5 = util.global.load @__auto.blk.10.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%789 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.6 = util.global.load @__auto.blk.10.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%790 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_up.weight.shard.7 = util.global.load @__auto.blk.10.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%791 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.10.ffn_down.weight.shard.0 = util.global.load @__auto.blk.10.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%792 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.1 = util.global.load @__auto.blk.10.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%793 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.2 = util.global.load @__auto.blk.10.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%794 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.3 = util.global.load @__auto.blk.10.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%795 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.4 = util.global.load @__auto.blk.10.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%796 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.5 = util.global.load @__auto.blk.10.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%797 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.6 = util.global.load @__auto.blk.10.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%798 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.10.ffn_down.weight.shard.7 = util.global.load @__auto.blk.10.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%799 = torch_c.from_builtin_tensor %__auto.blk.10.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.attn_norm.weight = util.global.load @__auto.blk.11.attn_norm.weight : tensor<4096xf32>
%800 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$1 = util.global.load @__auto.blk.11.attn_norm.weight$1 : tensor<4096xf32>
%801 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$2 = util.global.load @__auto.blk.11.attn_norm.weight$2 : tensor<4096xf32>
%802 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$3 = util.global.load @__auto.blk.11.attn_norm.weight$3 : tensor<4096xf32>
%803 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$4 = util.global.load @__auto.blk.11.attn_norm.weight$4 : tensor<4096xf32>
%804 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$5 = util.global.load @__auto.blk.11.attn_norm.weight$5 : tensor<4096xf32>
%805 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$6 = util.global.load @__auto.blk.11.attn_norm.weight$6 : tensor<4096xf32>
%806 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_norm.weight$7 = util.global.load @__auto.blk.11.attn_norm.weight$7 : tensor<4096xf32>
%807 = torch_c.from_builtin_tensor %__auto.blk.11.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.attn_q.weight.shard.0 = util.global.load @__auto.blk.11.attn_q.weight.shard.0 : tensor<512x4096xf16>
%808 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.1 = util.global.load @__auto.blk.11.attn_q.weight.shard.1 : tensor<512x4096xf16>
%809 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.2 = util.global.load @__auto.blk.11.attn_q.weight.shard.2 : tensor<512x4096xf16>
%810 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.3 = util.global.load @__auto.blk.11.attn_q.weight.shard.3 : tensor<512x4096xf16>
%811 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.4 = util.global.load @__auto.blk.11.attn_q.weight.shard.4 : tensor<512x4096xf16>
%812 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.5 = util.global.load @__auto.blk.11.attn_q.weight.shard.5 : tensor<512x4096xf16>
%813 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.6 = util.global.load @__auto.blk.11.attn_q.weight.shard.6 : tensor<512x4096xf16>
%814 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_q.weight.shard.7 = util.global.load @__auto.blk.11.attn_q.weight.shard.7 : tensor<512x4096xf16>
%815 = torch_c.from_builtin_tensor %__auto.blk.11.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.11.attn_k.weight.shard.0 = util.global.load @__auto.blk.11.attn_k.weight.shard.0 : tensor<128x4096xf16>
%816 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.1 = util.global.load @__auto.blk.11.attn_k.weight.shard.1 : tensor<128x4096xf16>
%817 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.2 = util.global.load @__auto.blk.11.attn_k.weight.shard.2 : tensor<128x4096xf16>
%818 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.3 = util.global.load @__auto.blk.11.attn_k.weight.shard.3 : tensor<128x4096xf16>
%819 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.4 = util.global.load @__auto.blk.11.attn_k.weight.shard.4 : tensor<128x4096xf16>
%820 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.5 = util.global.load @__auto.blk.11.attn_k.weight.shard.5 : tensor<128x4096xf16>
%821 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.6 = util.global.load @__auto.blk.11.attn_k.weight.shard.6 : tensor<128x4096xf16>
%822 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_k.weight.shard.7 = util.global.load @__auto.blk.11.attn_k.weight.shard.7 : tensor<128x4096xf16>
%823 = torch_c.from_builtin_tensor %__auto.blk.11.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.0 = util.global.load @__auto.blk.11.attn_v.weight.shard.0 : tensor<128x4096xf16>
%824 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.1 = util.global.load @__auto.blk.11.attn_v.weight.shard.1 : tensor<128x4096xf16>
%825 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.2 = util.global.load @__auto.blk.11.attn_v.weight.shard.2 : tensor<128x4096xf16>
%826 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.3 = util.global.load @__auto.blk.11.attn_v.weight.shard.3 : tensor<128x4096xf16>
%827 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.4 = util.global.load @__auto.blk.11.attn_v.weight.shard.4 : tensor<128x4096xf16>
%828 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.5 = util.global.load @__auto.blk.11.attn_v.weight.shard.5 : tensor<128x4096xf16>
%829 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.6 = util.global.load @__auto.blk.11.attn_v.weight.shard.6 : tensor<128x4096xf16>
%830 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_v.weight.shard.7 = util.global.load @__auto.blk.11.attn_v.weight.shard.7 : tensor<128x4096xf16>
%831 = torch_c.from_builtin_tensor %__auto.blk.11.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.11.attn_output.weight.shard.0 = util.global.load @__auto.blk.11.attn_output.weight.shard.0 : tensor<4096x512xf16>
%832 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.1 = util.global.load @__auto.blk.11.attn_output.weight.shard.1 : tensor<4096x512xf16>
%833 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.2 = util.global.load @__auto.blk.11.attn_output.weight.shard.2 : tensor<4096x512xf16>
%834 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.3 = util.global.load @__auto.blk.11.attn_output.weight.shard.3 : tensor<4096x512xf16>
%835 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.4 = util.global.load @__auto.blk.11.attn_output.weight.shard.4 : tensor<4096x512xf16>
%836 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.5 = util.global.load @__auto.blk.11.attn_output.weight.shard.5 : tensor<4096x512xf16>
%837 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.6 = util.global.load @__auto.blk.11.attn_output.weight.shard.6 : tensor<4096x512xf16>
%838 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.attn_output.weight.shard.7 = util.global.load @__auto.blk.11.attn_output.weight.shard.7 : tensor<4096x512xf16>
%839 = torch_c.from_builtin_tensor %__auto.blk.11.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.11.ffn_norm.weight = util.global.load @__auto.blk.11.ffn_norm.weight : tensor<4096xf32>
%840 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$1 = util.global.load @__auto.blk.11.ffn_norm.weight$1 : tensor<4096xf32>
%841 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$2 = util.global.load @__auto.blk.11.ffn_norm.weight$2 : tensor<4096xf32>
%842 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$3 = util.global.load @__auto.blk.11.ffn_norm.weight$3 : tensor<4096xf32>
%843 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$4 = util.global.load @__auto.blk.11.ffn_norm.weight$4 : tensor<4096xf32>
%844 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$5 = util.global.load @__auto.blk.11.ffn_norm.weight$5 : tensor<4096xf32>
%845 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$6 = util.global.load @__auto.blk.11.ffn_norm.weight$6 : tensor<4096xf32>
%846 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_norm.weight$7 = util.global.load @__auto.blk.11.ffn_norm.weight$7 : tensor<4096xf32>
%847 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.11.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%848 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%849 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%850 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%851 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%852 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%853 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%854 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.11.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%855 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.0 = util.global.load @__auto.blk.11.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%856 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.1 = util.global.load @__auto.blk.11.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%857 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.2 = util.global.load @__auto.blk.11.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%858 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.3 = util.global.load @__auto.blk.11.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%859 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.4 = util.global.load @__auto.blk.11.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%860 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.5 = util.global.load @__auto.blk.11.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%861 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.6 = util.global.load @__auto.blk.11.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%862 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_up.weight.shard.7 = util.global.load @__auto.blk.11.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%863 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.11.ffn_down.weight.shard.0 = util.global.load @__auto.blk.11.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%864 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.1 = util.global.load @__auto.blk.11.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%865 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.2 = util.global.load @__auto.blk.11.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%866 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.3 = util.global.load @__auto.blk.11.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%867 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.4 = util.global.load @__auto.blk.11.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%868 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.5 = util.global.load @__auto.blk.11.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%869 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.6 = util.global.load @__auto.blk.11.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%870 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.11.ffn_down.weight.shard.7 = util.global.load @__auto.blk.11.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%871 = torch_c.from_builtin_tensor %__auto.blk.11.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.attn_norm.weight = util.global.load @__auto.blk.12.attn_norm.weight : tensor<4096xf32>
%872 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$1 = util.global.load @__auto.blk.12.attn_norm.weight$1 : tensor<4096xf32>
%873 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$2 = util.global.load @__auto.blk.12.attn_norm.weight$2 : tensor<4096xf32>
%874 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$3 = util.global.load @__auto.blk.12.attn_norm.weight$3 : tensor<4096xf32>
%875 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$4 = util.global.load @__auto.blk.12.attn_norm.weight$4 : tensor<4096xf32>
%876 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$5 = util.global.load @__auto.blk.12.attn_norm.weight$5 : tensor<4096xf32>
%877 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$6 = util.global.load @__auto.blk.12.attn_norm.weight$6 : tensor<4096xf32>
%878 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_norm.weight$7 = util.global.load @__auto.blk.12.attn_norm.weight$7 : tensor<4096xf32>
%879 = torch_c.from_builtin_tensor %__auto.blk.12.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.attn_q.weight.shard.0 = util.global.load @__auto.blk.12.attn_q.weight.shard.0 : tensor<512x4096xf16>
%880 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.1 = util.global.load @__auto.blk.12.attn_q.weight.shard.1 : tensor<512x4096xf16>
%881 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.2 = util.global.load @__auto.blk.12.attn_q.weight.shard.2 : tensor<512x4096xf16>
%882 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.3 = util.global.load @__auto.blk.12.attn_q.weight.shard.3 : tensor<512x4096xf16>
%883 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.4 = util.global.load @__auto.blk.12.attn_q.weight.shard.4 : tensor<512x4096xf16>
%884 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.5 = util.global.load @__auto.blk.12.attn_q.weight.shard.5 : tensor<512x4096xf16>
%885 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.6 = util.global.load @__auto.blk.12.attn_q.weight.shard.6 : tensor<512x4096xf16>
%886 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_q.weight.shard.7 = util.global.load @__auto.blk.12.attn_q.weight.shard.7 : tensor<512x4096xf16>
%887 = torch_c.from_builtin_tensor %__auto.blk.12.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.12.attn_k.weight.shard.0 = util.global.load @__auto.blk.12.attn_k.weight.shard.0 : tensor<128x4096xf16>
%888 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.1 = util.global.load @__auto.blk.12.attn_k.weight.shard.1 : tensor<128x4096xf16>
%889 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.2 = util.global.load @__auto.blk.12.attn_k.weight.shard.2 : tensor<128x4096xf16>
%890 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.3 = util.global.load @__auto.blk.12.attn_k.weight.shard.3 : tensor<128x4096xf16>
%891 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.4 = util.global.load @__auto.blk.12.attn_k.weight.shard.4 : tensor<128x4096xf16>
%892 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.5 = util.global.load @__auto.blk.12.attn_k.weight.shard.5 : tensor<128x4096xf16>
%893 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.6 = util.global.load @__auto.blk.12.attn_k.weight.shard.6 : tensor<128x4096xf16>
%894 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_k.weight.shard.7 = util.global.load @__auto.blk.12.attn_k.weight.shard.7 : tensor<128x4096xf16>
%895 = torch_c.from_builtin_tensor %__auto.blk.12.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.0 = util.global.load @__auto.blk.12.attn_v.weight.shard.0 : tensor<128x4096xf16>
%896 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.1 = util.global.load @__auto.blk.12.attn_v.weight.shard.1 : tensor<128x4096xf16>
%897 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.2 = util.global.load @__auto.blk.12.attn_v.weight.shard.2 : tensor<128x4096xf16>
%898 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.3 = util.global.load @__auto.blk.12.attn_v.weight.shard.3 : tensor<128x4096xf16>
%899 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.4 = util.global.load @__auto.blk.12.attn_v.weight.shard.4 : tensor<128x4096xf16>
%900 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.5 = util.global.load @__auto.blk.12.attn_v.weight.shard.5 : tensor<128x4096xf16>
%901 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.6 = util.global.load @__auto.blk.12.attn_v.weight.shard.6 : tensor<128x4096xf16>
%902 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_v.weight.shard.7 = util.global.load @__auto.blk.12.attn_v.weight.shard.7 : tensor<128x4096xf16>
%903 = torch_c.from_builtin_tensor %__auto.blk.12.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.12.attn_output.weight.shard.0 = util.global.load @__auto.blk.12.attn_output.weight.shard.0 : tensor<4096x512xf16>
%904 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.1 = util.global.load @__auto.blk.12.attn_output.weight.shard.1 : tensor<4096x512xf16>
%905 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.2 = util.global.load @__auto.blk.12.attn_output.weight.shard.2 : tensor<4096x512xf16>
%906 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.3 = util.global.load @__auto.blk.12.attn_output.weight.shard.3 : tensor<4096x512xf16>
%907 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.4 = util.global.load @__auto.blk.12.attn_output.weight.shard.4 : tensor<4096x512xf16>
%908 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.5 = util.global.load @__auto.blk.12.attn_output.weight.shard.5 : tensor<4096x512xf16>
%909 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.6 = util.global.load @__auto.blk.12.attn_output.weight.shard.6 : tensor<4096x512xf16>
%910 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.attn_output.weight.shard.7 = util.global.load @__auto.blk.12.attn_output.weight.shard.7 : tensor<4096x512xf16>
%911 = torch_c.from_builtin_tensor %__auto.blk.12.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.12.ffn_norm.weight = util.global.load @__auto.blk.12.ffn_norm.weight : tensor<4096xf32>
%912 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$1 = util.global.load @__auto.blk.12.ffn_norm.weight$1 : tensor<4096xf32>
%913 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$2 = util.global.load @__auto.blk.12.ffn_norm.weight$2 : tensor<4096xf32>
%914 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$3 = util.global.load @__auto.blk.12.ffn_norm.weight$3 : tensor<4096xf32>
%915 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$4 = util.global.load @__auto.blk.12.ffn_norm.weight$4 : tensor<4096xf32>
%916 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$5 = util.global.load @__auto.blk.12.ffn_norm.weight$5 : tensor<4096xf32>
%917 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$6 = util.global.load @__auto.blk.12.ffn_norm.weight$6 : tensor<4096xf32>
%918 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_norm.weight$7 = util.global.load @__auto.blk.12.ffn_norm.weight$7 : tensor<4096xf32>
%919 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.12.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%920 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%921 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%922 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%923 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%924 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%925 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%926 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.12.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%927 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.0 = util.global.load @__auto.blk.12.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%928 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.1 = util.global.load @__auto.blk.12.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%929 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.2 = util.global.load @__auto.blk.12.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%930 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.3 = util.global.load @__auto.blk.12.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%931 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.4 = util.global.load @__auto.blk.12.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%932 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.5 = util.global.load @__auto.blk.12.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%933 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.6 = util.global.load @__auto.blk.12.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%934 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_up.weight.shard.7 = util.global.load @__auto.blk.12.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%935 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.12.ffn_down.weight.shard.0 = util.global.load @__auto.blk.12.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%936 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.1 = util.global.load @__auto.blk.12.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%937 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.2 = util.global.load @__auto.blk.12.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%938 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.3 = util.global.load @__auto.blk.12.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%939 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.4 = util.global.load @__auto.blk.12.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%940 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.5 = util.global.load @__auto.blk.12.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%941 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.6 = util.global.load @__auto.blk.12.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%942 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.12.ffn_down.weight.shard.7 = util.global.load @__auto.blk.12.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%943 = torch_c.from_builtin_tensor %__auto.blk.12.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.attn_norm.weight = util.global.load @__auto.blk.13.attn_norm.weight : tensor<4096xf32>
%944 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$1 = util.global.load @__auto.blk.13.attn_norm.weight$1 : tensor<4096xf32>
%945 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$2 = util.global.load @__auto.blk.13.attn_norm.weight$2 : tensor<4096xf32>
%946 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$3 = util.global.load @__auto.blk.13.attn_norm.weight$3 : tensor<4096xf32>
%947 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$4 = util.global.load @__auto.blk.13.attn_norm.weight$4 : tensor<4096xf32>
%948 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$5 = util.global.load @__auto.blk.13.attn_norm.weight$5 : tensor<4096xf32>
%949 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$6 = util.global.load @__auto.blk.13.attn_norm.weight$6 : tensor<4096xf32>
%950 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_norm.weight$7 = util.global.load @__auto.blk.13.attn_norm.weight$7 : tensor<4096xf32>
%951 = torch_c.from_builtin_tensor %__auto.blk.13.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.attn_q.weight.shard.0 = util.global.load @__auto.blk.13.attn_q.weight.shard.0 : tensor<512x4096xf16>
%952 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.1 = util.global.load @__auto.blk.13.attn_q.weight.shard.1 : tensor<512x4096xf16>
%953 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.2 = util.global.load @__auto.blk.13.attn_q.weight.shard.2 : tensor<512x4096xf16>
%954 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.3 = util.global.load @__auto.blk.13.attn_q.weight.shard.3 : tensor<512x4096xf16>
%955 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.4 = util.global.load @__auto.blk.13.attn_q.weight.shard.4 : tensor<512x4096xf16>
%956 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.5 = util.global.load @__auto.blk.13.attn_q.weight.shard.5 : tensor<512x4096xf16>
%957 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.6 = util.global.load @__auto.blk.13.attn_q.weight.shard.6 : tensor<512x4096xf16>
%958 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_q.weight.shard.7 = util.global.load @__auto.blk.13.attn_q.weight.shard.7 : tensor<512x4096xf16>
%959 = torch_c.from_builtin_tensor %__auto.blk.13.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.13.attn_k.weight.shard.0 = util.global.load @__auto.blk.13.attn_k.weight.shard.0 : tensor<128x4096xf16>
%960 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.1 = util.global.load @__auto.blk.13.attn_k.weight.shard.1 : tensor<128x4096xf16>
%961 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.2 = util.global.load @__auto.blk.13.attn_k.weight.shard.2 : tensor<128x4096xf16>
%962 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.3 = util.global.load @__auto.blk.13.attn_k.weight.shard.3 : tensor<128x4096xf16>
%963 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.4 = util.global.load @__auto.blk.13.attn_k.weight.shard.4 : tensor<128x4096xf16>
%964 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.5 = util.global.load @__auto.blk.13.attn_k.weight.shard.5 : tensor<128x4096xf16>
%965 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.6 = util.global.load @__auto.blk.13.attn_k.weight.shard.6 : tensor<128x4096xf16>
%966 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_k.weight.shard.7 = util.global.load @__auto.blk.13.attn_k.weight.shard.7 : tensor<128x4096xf16>
%967 = torch_c.from_builtin_tensor %__auto.blk.13.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.0 = util.global.load @__auto.blk.13.attn_v.weight.shard.0 : tensor<128x4096xf16>
%968 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.1 = util.global.load @__auto.blk.13.attn_v.weight.shard.1 : tensor<128x4096xf16>
%969 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.2 = util.global.load @__auto.blk.13.attn_v.weight.shard.2 : tensor<128x4096xf16>
%970 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.3 = util.global.load @__auto.blk.13.attn_v.weight.shard.3 : tensor<128x4096xf16>
%971 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.4 = util.global.load @__auto.blk.13.attn_v.weight.shard.4 : tensor<128x4096xf16>
%972 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.5 = util.global.load @__auto.blk.13.attn_v.weight.shard.5 : tensor<128x4096xf16>
%973 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.6 = util.global.load @__auto.blk.13.attn_v.weight.shard.6 : tensor<128x4096xf16>
%974 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_v.weight.shard.7 = util.global.load @__auto.blk.13.attn_v.weight.shard.7 : tensor<128x4096xf16>
%975 = torch_c.from_builtin_tensor %__auto.blk.13.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.13.attn_output.weight.shard.0 = util.global.load @__auto.blk.13.attn_output.weight.shard.0 : tensor<4096x512xf16>
%976 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.1 = util.global.load @__auto.blk.13.attn_output.weight.shard.1 : tensor<4096x512xf16>
%977 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.2 = util.global.load @__auto.blk.13.attn_output.weight.shard.2 : tensor<4096x512xf16>
%978 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.3 = util.global.load @__auto.blk.13.attn_output.weight.shard.3 : tensor<4096x512xf16>
%979 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.4 = util.global.load @__auto.blk.13.attn_output.weight.shard.4 : tensor<4096x512xf16>
%980 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.5 = util.global.load @__auto.blk.13.attn_output.weight.shard.5 : tensor<4096x512xf16>
%981 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.6 = util.global.load @__auto.blk.13.attn_output.weight.shard.6 : tensor<4096x512xf16>
%982 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.attn_output.weight.shard.7 = util.global.load @__auto.blk.13.attn_output.weight.shard.7 : tensor<4096x512xf16>
%983 = torch_c.from_builtin_tensor %__auto.blk.13.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.13.ffn_norm.weight = util.global.load @__auto.blk.13.ffn_norm.weight : tensor<4096xf32>
%984 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$1 = util.global.load @__auto.blk.13.ffn_norm.weight$1 : tensor<4096xf32>
%985 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$2 = util.global.load @__auto.blk.13.ffn_norm.weight$2 : tensor<4096xf32>
%986 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$3 = util.global.load @__auto.blk.13.ffn_norm.weight$3 : tensor<4096xf32>
%987 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$4 = util.global.load @__auto.blk.13.ffn_norm.weight$4 : tensor<4096xf32>
%988 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$5 = util.global.load @__auto.blk.13.ffn_norm.weight$5 : tensor<4096xf32>
%989 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$6 = util.global.load @__auto.blk.13.ffn_norm.weight$6 : tensor<4096xf32>
%990 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_norm.weight$7 = util.global.load @__auto.blk.13.ffn_norm.weight$7 : tensor<4096xf32>
%991 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.13.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%992 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%993 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%994 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%995 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%996 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%997 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%998 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.13.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%999 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.0 = util.global.load @__auto.blk.13.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1000 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.1 = util.global.load @__auto.blk.13.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1001 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.2 = util.global.load @__auto.blk.13.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1002 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.3 = util.global.load @__auto.blk.13.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1003 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.4 = util.global.load @__auto.blk.13.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1004 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.5 = util.global.load @__auto.blk.13.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1005 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.6 = util.global.load @__auto.blk.13.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1006 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_up.weight.shard.7 = util.global.load @__auto.blk.13.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1007 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.13.ffn_down.weight.shard.0 = util.global.load @__auto.blk.13.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1008 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.1 = util.global.load @__auto.blk.13.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1009 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.2 = util.global.load @__auto.blk.13.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1010 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.3 = util.global.load @__auto.blk.13.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1011 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.4 = util.global.load @__auto.blk.13.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1012 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.5 = util.global.load @__auto.blk.13.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1013 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.6 = util.global.load @__auto.blk.13.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1014 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.13.ffn_down.weight.shard.7 = util.global.load @__auto.blk.13.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1015 = torch_c.from_builtin_tensor %__auto.blk.13.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.attn_norm.weight = util.global.load @__auto.blk.14.attn_norm.weight : tensor<4096xf32>
%1016 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$1 = util.global.load @__auto.blk.14.attn_norm.weight$1 : tensor<4096xf32>
%1017 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$2 = util.global.load @__auto.blk.14.attn_norm.weight$2 : tensor<4096xf32>
%1018 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$3 = util.global.load @__auto.blk.14.attn_norm.weight$3 : tensor<4096xf32>
%1019 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$4 = util.global.load @__auto.blk.14.attn_norm.weight$4 : tensor<4096xf32>
%1020 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$5 = util.global.load @__auto.blk.14.attn_norm.weight$5 : tensor<4096xf32>
%1021 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$6 = util.global.load @__auto.blk.14.attn_norm.weight$6 : tensor<4096xf32>
%1022 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_norm.weight$7 = util.global.load @__auto.blk.14.attn_norm.weight$7 : tensor<4096xf32>
%1023 = torch_c.from_builtin_tensor %__auto.blk.14.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.attn_q.weight.shard.0 = util.global.load @__auto.blk.14.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1024 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.1 = util.global.load @__auto.blk.14.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1025 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.2 = util.global.load @__auto.blk.14.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1026 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.3 = util.global.load @__auto.blk.14.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1027 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.4 = util.global.load @__auto.blk.14.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1028 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.5 = util.global.load @__auto.blk.14.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1029 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.6 = util.global.load @__auto.blk.14.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1030 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_q.weight.shard.7 = util.global.load @__auto.blk.14.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1031 = torch_c.from_builtin_tensor %__auto.blk.14.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.14.attn_k.weight.shard.0 = util.global.load @__auto.blk.14.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1032 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.1 = util.global.load @__auto.blk.14.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1033 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.2 = util.global.load @__auto.blk.14.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1034 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.3 = util.global.load @__auto.blk.14.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1035 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.4 = util.global.load @__auto.blk.14.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1036 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.5 = util.global.load @__auto.blk.14.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1037 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.6 = util.global.load @__auto.blk.14.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1038 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_k.weight.shard.7 = util.global.load @__auto.blk.14.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1039 = torch_c.from_builtin_tensor %__auto.blk.14.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.0 = util.global.load @__auto.blk.14.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1040 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.1 = util.global.load @__auto.blk.14.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1041 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.2 = util.global.load @__auto.blk.14.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1042 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.3 = util.global.load @__auto.blk.14.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1043 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.4 = util.global.load @__auto.blk.14.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1044 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.5 = util.global.load @__auto.blk.14.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1045 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.6 = util.global.load @__auto.blk.14.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1046 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_v.weight.shard.7 = util.global.load @__auto.blk.14.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1047 = torch_c.from_builtin_tensor %__auto.blk.14.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.14.attn_output.weight.shard.0 = util.global.load @__auto.blk.14.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1048 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.1 = util.global.load @__auto.blk.14.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1049 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.2 = util.global.load @__auto.blk.14.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1050 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.3 = util.global.load @__auto.blk.14.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1051 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.4 = util.global.load @__auto.blk.14.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1052 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.5 = util.global.load @__auto.blk.14.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1053 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.6 = util.global.load @__auto.blk.14.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1054 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.attn_output.weight.shard.7 = util.global.load @__auto.blk.14.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1055 = torch_c.from_builtin_tensor %__auto.blk.14.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.14.ffn_norm.weight = util.global.load @__auto.blk.14.ffn_norm.weight : tensor<4096xf32>
%1056 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$1 = util.global.load @__auto.blk.14.ffn_norm.weight$1 : tensor<4096xf32>
%1057 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$2 = util.global.load @__auto.blk.14.ffn_norm.weight$2 : tensor<4096xf32>
%1058 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$3 = util.global.load @__auto.blk.14.ffn_norm.weight$3 : tensor<4096xf32>
%1059 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$4 = util.global.load @__auto.blk.14.ffn_norm.weight$4 : tensor<4096xf32>
%1060 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$5 = util.global.load @__auto.blk.14.ffn_norm.weight$5 : tensor<4096xf32>
%1061 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$6 = util.global.load @__auto.blk.14.ffn_norm.weight$6 : tensor<4096xf32>
%1062 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_norm.weight$7 = util.global.load @__auto.blk.14.ffn_norm.weight$7 : tensor<4096xf32>
%1063 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.14.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1064 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1065 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1066 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1067 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1068 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1069 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1070 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.14.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1071 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.0 = util.global.load @__auto.blk.14.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1072 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.1 = util.global.load @__auto.blk.14.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1073 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.2 = util.global.load @__auto.blk.14.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1074 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.3 = util.global.load @__auto.blk.14.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1075 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.4 = util.global.load @__auto.blk.14.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1076 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.5 = util.global.load @__auto.blk.14.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1077 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.6 = util.global.load @__auto.blk.14.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1078 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_up.weight.shard.7 = util.global.load @__auto.blk.14.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1079 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.14.ffn_down.weight.shard.0 = util.global.load @__auto.blk.14.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1080 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.1 = util.global.load @__auto.blk.14.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1081 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.2 = util.global.load @__auto.blk.14.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1082 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.3 = util.global.load @__auto.blk.14.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1083 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.4 = util.global.load @__auto.blk.14.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1084 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.5 = util.global.load @__auto.blk.14.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1085 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.6 = util.global.load @__auto.blk.14.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1086 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.14.ffn_down.weight.shard.7 = util.global.load @__auto.blk.14.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1087 = torch_c.from_builtin_tensor %__auto.blk.14.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.attn_norm.weight = util.global.load @__auto.blk.15.attn_norm.weight : tensor<4096xf32>
%1088 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$1 = util.global.load @__auto.blk.15.attn_norm.weight$1 : tensor<4096xf32>
%1089 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$2 = util.global.load @__auto.blk.15.attn_norm.weight$2 : tensor<4096xf32>
%1090 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$3 = util.global.load @__auto.blk.15.attn_norm.weight$3 : tensor<4096xf32>
%1091 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$4 = util.global.load @__auto.blk.15.attn_norm.weight$4 : tensor<4096xf32>
%1092 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$5 = util.global.load @__auto.blk.15.attn_norm.weight$5 : tensor<4096xf32>
%1093 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$6 = util.global.load @__auto.blk.15.attn_norm.weight$6 : tensor<4096xf32>
%1094 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_norm.weight$7 = util.global.load @__auto.blk.15.attn_norm.weight$7 : tensor<4096xf32>
%1095 = torch_c.from_builtin_tensor %__auto.blk.15.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.attn_q.weight.shard.0 = util.global.load @__auto.blk.15.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1096 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.1 = util.global.load @__auto.blk.15.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1097 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.2 = util.global.load @__auto.blk.15.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1098 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.3 = util.global.load @__auto.blk.15.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1099 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.4 = util.global.load @__auto.blk.15.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1100 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.5 = util.global.load @__auto.blk.15.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1101 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.6 = util.global.load @__auto.blk.15.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1102 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_q.weight.shard.7 = util.global.load @__auto.blk.15.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1103 = torch_c.from_builtin_tensor %__auto.blk.15.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.15.attn_k.weight.shard.0 = util.global.load @__auto.blk.15.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1104 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.1 = util.global.load @__auto.blk.15.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1105 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.2 = util.global.load @__auto.blk.15.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1106 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.3 = util.global.load @__auto.blk.15.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1107 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.4 = util.global.load @__auto.blk.15.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1108 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.5 = util.global.load @__auto.blk.15.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1109 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.6 = util.global.load @__auto.blk.15.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1110 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_k.weight.shard.7 = util.global.load @__auto.blk.15.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1111 = torch_c.from_builtin_tensor %__auto.blk.15.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.0 = util.global.load @__auto.blk.15.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1112 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.1 = util.global.load @__auto.blk.15.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1113 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.2 = util.global.load @__auto.blk.15.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1114 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.3 = util.global.load @__auto.blk.15.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1115 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.4 = util.global.load @__auto.blk.15.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1116 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.5 = util.global.load @__auto.blk.15.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1117 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.6 = util.global.load @__auto.blk.15.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1118 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_v.weight.shard.7 = util.global.load @__auto.blk.15.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1119 = torch_c.from_builtin_tensor %__auto.blk.15.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.15.attn_output.weight.shard.0 = util.global.load @__auto.blk.15.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1120 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.1 = util.global.load @__auto.blk.15.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1121 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.2 = util.global.load @__auto.blk.15.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1122 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.3 = util.global.load @__auto.blk.15.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1123 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.4 = util.global.load @__auto.blk.15.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1124 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.5 = util.global.load @__auto.blk.15.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1125 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.6 = util.global.load @__auto.blk.15.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1126 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.attn_output.weight.shard.7 = util.global.load @__auto.blk.15.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1127 = torch_c.from_builtin_tensor %__auto.blk.15.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.15.ffn_norm.weight = util.global.load @__auto.blk.15.ffn_norm.weight : tensor<4096xf32>
%1128 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$1 = util.global.load @__auto.blk.15.ffn_norm.weight$1 : tensor<4096xf32>
%1129 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$2 = util.global.load @__auto.blk.15.ffn_norm.weight$2 : tensor<4096xf32>
%1130 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$3 = util.global.load @__auto.blk.15.ffn_norm.weight$3 : tensor<4096xf32>
%1131 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$4 = util.global.load @__auto.blk.15.ffn_norm.weight$4 : tensor<4096xf32>
%1132 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$5 = util.global.load @__auto.blk.15.ffn_norm.weight$5 : tensor<4096xf32>
%1133 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$6 = util.global.load @__auto.blk.15.ffn_norm.weight$6 : tensor<4096xf32>
%1134 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_norm.weight$7 = util.global.load @__auto.blk.15.ffn_norm.weight$7 : tensor<4096xf32>
%1135 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.15.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1136 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1137 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1138 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1139 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1140 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1141 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1142 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.15.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1143 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.0 = util.global.load @__auto.blk.15.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1144 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.1 = util.global.load @__auto.blk.15.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1145 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.2 = util.global.load @__auto.blk.15.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1146 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.3 = util.global.load @__auto.blk.15.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1147 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.4 = util.global.load @__auto.blk.15.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1148 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.5 = util.global.load @__auto.blk.15.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1149 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.6 = util.global.load @__auto.blk.15.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1150 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_up.weight.shard.7 = util.global.load @__auto.blk.15.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1151 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.15.ffn_down.weight.shard.0 = util.global.load @__auto.blk.15.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1152 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.1 = util.global.load @__auto.blk.15.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1153 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.2 = util.global.load @__auto.blk.15.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1154 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.3 = util.global.load @__auto.blk.15.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1155 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.4 = util.global.load @__auto.blk.15.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1156 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.5 = util.global.load @__auto.blk.15.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1157 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.6 = util.global.load @__auto.blk.15.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1158 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.15.ffn_down.weight.shard.7 = util.global.load @__auto.blk.15.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1159 = torch_c.from_builtin_tensor %__auto.blk.15.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.attn_norm.weight = util.global.load @__auto.blk.16.attn_norm.weight : tensor<4096xf32>
%1160 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$1 = util.global.load @__auto.blk.16.attn_norm.weight$1 : tensor<4096xf32>
%1161 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$2 = util.global.load @__auto.blk.16.attn_norm.weight$2 : tensor<4096xf32>
%1162 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$3 = util.global.load @__auto.blk.16.attn_norm.weight$3 : tensor<4096xf32>
%1163 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$4 = util.global.load @__auto.blk.16.attn_norm.weight$4 : tensor<4096xf32>
%1164 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$5 = util.global.load @__auto.blk.16.attn_norm.weight$5 : tensor<4096xf32>
%1165 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$6 = util.global.load @__auto.blk.16.attn_norm.weight$6 : tensor<4096xf32>
%1166 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_norm.weight$7 = util.global.load @__auto.blk.16.attn_norm.weight$7 : tensor<4096xf32>
%1167 = torch_c.from_builtin_tensor %__auto.blk.16.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.attn_q.weight.shard.0 = util.global.load @__auto.blk.16.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1168 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.1 = util.global.load @__auto.blk.16.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1169 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.2 = util.global.load @__auto.blk.16.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1170 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.3 = util.global.load @__auto.blk.16.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1171 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.4 = util.global.load @__auto.blk.16.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1172 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.5 = util.global.load @__auto.blk.16.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1173 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.6 = util.global.load @__auto.blk.16.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1174 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_q.weight.shard.7 = util.global.load @__auto.blk.16.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1175 = torch_c.from_builtin_tensor %__auto.blk.16.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.16.attn_k.weight.shard.0 = util.global.load @__auto.blk.16.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1176 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.1 = util.global.load @__auto.blk.16.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1177 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.2 = util.global.load @__auto.blk.16.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1178 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.3 = util.global.load @__auto.blk.16.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1179 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.4 = util.global.load @__auto.blk.16.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1180 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.5 = util.global.load @__auto.blk.16.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1181 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.6 = util.global.load @__auto.blk.16.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1182 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_k.weight.shard.7 = util.global.load @__auto.blk.16.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1183 = torch_c.from_builtin_tensor %__auto.blk.16.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.0 = util.global.load @__auto.blk.16.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1184 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.1 = util.global.load @__auto.blk.16.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1185 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.2 = util.global.load @__auto.blk.16.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1186 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.3 = util.global.load @__auto.blk.16.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1187 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.4 = util.global.load @__auto.blk.16.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1188 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.5 = util.global.load @__auto.blk.16.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1189 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.6 = util.global.load @__auto.blk.16.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1190 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_v.weight.shard.7 = util.global.load @__auto.blk.16.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1191 = torch_c.from_builtin_tensor %__auto.blk.16.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.16.attn_output.weight.shard.0 = util.global.load @__auto.blk.16.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1192 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.1 = util.global.load @__auto.blk.16.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1193 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.2 = util.global.load @__auto.blk.16.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1194 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.3 = util.global.load @__auto.blk.16.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1195 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.4 = util.global.load @__auto.blk.16.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1196 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.5 = util.global.load @__auto.blk.16.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1197 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.6 = util.global.load @__auto.blk.16.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1198 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.attn_output.weight.shard.7 = util.global.load @__auto.blk.16.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1199 = torch_c.from_builtin_tensor %__auto.blk.16.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.16.ffn_norm.weight = util.global.load @__auto.blk.16.ffn_norm.weight : tensor<4096xf32>
%1200 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$1 = util.global.load @__auto.blk.16.ffn_norm.weight$1 : tensor<4096xf32>
%1201 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$2 = util.global.load @__auto.blk.16.ffn_norm.weight$2 : tensor<4096xf32>
%1202 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$3 = util.global.load @__auto.blk.16.ffn_norm.weight$3 : tensor<4096xf32>
%1203 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$4 = util.global.load @__auto.blk.16.ffn_norm.weight$4 : tensor<4096xf32>
%1204 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$5 = util.global.load @__auto.blk.16.ffn_norm.weight$5 : tensor<4096xf32>
%1205 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$6 = util.global.load @__auto.blk.16.ffn_norm.weight$6 : tensor<4096xf32>
%1206 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_norm.weight$7 = util.global.load @__auto.blk.16.ffn_norm.weight$7 : tensor<4096xf32>
%1207 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.16.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1208 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1209 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1210 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1211 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1212 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1213 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1214 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.16.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1215 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.0 = util.global.load @__auto.blk.16.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1216 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.1 = util.global.load @__auto.blk.16.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1217 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.2 = util.global.load @__auto.blk.16.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1218 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.3 = util.global.load @__auto.blk.16.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1219 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.4 = util.global.load @__auto.blk.16.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1220 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.5 = util.global.load @__auto.blk.16.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1221 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.6 = util.global.load @__auto.blk.16.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1222 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_up.weight.shard.7 = util.global.load @__auto.blk.16.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1223 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.16.ffn_down.weight.shard.0 = util.global.load @__auto.blk.16.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1224 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.1 = util.global.load @__auto.blk.16.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1225 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.2 = util.global.load @__auto.blk.16.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1226 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.3 = util.global.load @__auto.blk.16.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1227 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.4 = util.global.load @__auto.blk.16.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1228 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.5 = util.global.load @__auto.blk.16.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1229 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.6 = util.global.load @__auto.blk.16.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1230 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.16.ffn_down.weight.shard.7 = util.global.load @__auto.blk.16.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1231 = torch_c.from_builtin_tensor %__auto.blk.16.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.attn_norm.weight = util.global.load @__auto.blk.17.attn_norm.weight : tensor<4096xf32>
%1232 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$1 = util.global.load @__auto.blk.17.attn_norm.weight$1 : tensor<4096xf32>
%1233 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$2 = util.global.load @__auto.blk.17.attn_norm.weight$2 : tensor<4096xf32>
%1234 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$3 = util.global.load @__auto.blk.17.attn_norm.weight$3 : tensor<4096xf32>
%1235 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$4 = util.global.load @__auto.blk.17.attn_norm.weight$4 : tensor<4096xf32>
%1236 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$5 = util.global.load @__auto.blk.17.attn_norm.weight$5 : tensor<4096xf32>
%1237 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$6 = util.global.load @__auto.blk.17.attn_norm.weight$6 : tensor<4096xf32>
%1238 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_norm.weight$7 = util.global.load @__auto.blk.17.attn_norm.weight$7 : tensor<4096xf32>
%1239 = torch_c.from_builtin_tensor %__auto.blk.17.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.attn_q.weight.shard.0 = util.global.load @__auto.blk.17.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1240 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.1 = util.global.load @__auto.blk.17.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1241 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.2 = util.global.load @__auto.blk.17.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1242 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.3 = util.global.load @__auto.blk.17.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1243 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.4 = util.global.load @__auto.blk.17.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1244 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.5 = util.global.load @__auto.blk.17.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1245 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.6 = util.global.load @__auto.blk.17.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1246 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_q.weight.shard.7 = util.global.load @__auto.blk.17.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1247 = torch_c.from_builtin_tensor %__auto.blk.17.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.17.attn_k.weight.shard.0 = util.global.load @__auto.blk.17.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1248 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.1 = util.global.load @__auto.blk.17.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1249 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.2 = util.global.load @__auto.blk.17.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1250 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.3 = util.global.load @__auto.blk.17.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1251 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.4 = util.global.load @__auto.blk.17.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1252 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.5 = util.global.load @__auto.blk.17.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1253 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.6 = util.global.load @__auto.blk.17.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1254 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_k.weight.shard.7 = util.global.load @__auto.blk.17.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1255 = torch_c.from_builtin_tensor %__auto.blk.17.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.0 = util.global.load @__auto.blk.17.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1256 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.1 = util.global.load @__auto.blk.17.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1257 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.2 = util.global.load @__auto.blk.17.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1258 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.3 = util.global.load @__auto.blk.17.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1259 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.4 = util.global.load @__auto.blk.17.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1260 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.5 = util.global.load @__auto.blk.17.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1261 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.6 = util.global.load @__auto.blk.17.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1262 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_v.weight.shard.7 = util.global.load @__auto.blk.17.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1263 = torch_c.from_builtin_tensor %__auto.blk.17.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.17.attn_output.weight.shard.0 = util.global.load @__auto.blk.17.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1264 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.1 = util.global.load @__auto.blk.17.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1265 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.2 = util.global.load @__auto.blk.17.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1266 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.3 = util.global.load @__auto.blk.17.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1267 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.4 = util.global.load @__auto.blk.17.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1268 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.5 = util.global.load @__auto.blk.17.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1269 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.6 = util.global.load @__auto.blk.17.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1270 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.attn_output.weight.shard.7 = util.global.load @__auto.blk.17.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1271 = torch_c.from_builtin_tensor %__auto.blk.17.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.17.ffn_norm.weight = util.global.load @__auto.blk.17.ffn_norm.weight : tensor<4096xf32>
%1272 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$1 = util.global.load @__auto.blk.17.ffn_norm.weight$1 : tensor<4096xf32>
%1273 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$2 = util.global.load @__auto.blk.17.ffn_norm.weight$2 : tensor<4096xf32>
%1274 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$3 = util.global.load @__auto.blk.17.ffn_norm.weight$3 : tensor<4096xf32>
%1275 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$4 = util.global.load @__auto.blk.17.ffn_norm.weight$4 : tensor<4096xf32>
%1276 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$5 = util.global.load @__auto.blk.17.ffn_norm.weight$5 : tensor<4096xf32>
%1277 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$6 = util.global.load @__auto.blk.17.ffn_norm.weight$6 : tensor<4096xf32>
%1278 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_norm.weight$7 = util.global.load @__auto.blk.17.ffn_norm.weight$7 : tensor<4096xf32>
%1279 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.17.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1280 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1281 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1282 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1283 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1284 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1285 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1286 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.17.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1287 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.0 = util.global.load @__auto.blk.17.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1288 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.1 = util.global.load @__auto.blk.17.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1289 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.2 = util.global.load @__auto.blk.17.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1290 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.3 = util.global.load @__auto.blk.17.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1291 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.4 = util.global.load @__auto.blk.17.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1292 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.5 = util.global.load @__auto.blk.17.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1293 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.6 = util.global.load @__auto.blk.17.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1294 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_up.weight.shard.7 = util.global.load @__auto.blk.17.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1295 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.17.ffn_down.weight.shard.0 = util.global.load @__auto.blk.17.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1296 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.1 = util.global.load @__auto.blk.17.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1297 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.2 = util.global.load @__auto.blk.17.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1298 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.3 = util.global.load @__auto.blk.17.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1299 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.4 = util.global.load @__auto.blk.17.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1300 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.5 = util.global.load @__auto.blk.17.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1301 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.6 = util.global.load @__auto.blk.17.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1302 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.17.ffn_down.weight.shard.7 = util.global.load @__auto.blk.17.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1303 = torch_c.from_builtin_tensor %__auto.blk.17.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.attn_norm.weight = util.global.load @__auto.blk.18.attn_norm.weight : tensor<4096xf32>
%1304 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$1 = util.global.load @__auto.blk.18.attn_norm.weight$1 : tensor<4096xf32>
%1305 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$2 = util.global.load @__auto.blk.18.attn_norm.weight$2 : tensor<4096xf32>
%1306 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$3 = util.global.load @__auto.blk.18.attn_norm.weight$3 : tensor<4096xf32>
%1307 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$4 = util.global.load @__auto.blk.18.attn_norm.weight$4 : tensor<4096xf32>
%1308 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$5 = util.global.load @__auto.blk.18.attn_norm.weight$5 : tensor<4096xf32>
%1309 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$6 = util.global.load @__auto.blk.18.attn_norm.weight$6 : tensor<4096xf32>
%1310 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_norm.weight$7 = util.global.load @__auto.blk.18.attn_norm.weight$7 : tensor<4096xf32>
%1311 = torch_c.from_builtin_tensor %__auto.blk.18.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.attn_q.weight.shard.0 = util.global.load @__auto.blk.18.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1312 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.1 = util.global.load @__auto.blk.18.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1313 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.2 = util.global.load @__auto.blk.18.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1314 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.3 = util.global.load @__auto.blk.18.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1315 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.4 = util.global.load @__auto.blk.18.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1316 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.5 = util.global.load @__auto.blk.18.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1317 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.6 = util.global.load @__auto.blk.18.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1318 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_q.weight.shard.7 = util.global.load @__auto.blk.18.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1319 = torch_c.from_builtin_tensor %__auto.blk.18.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.18.attn_k.weight.shard.0 = util.global.load @__auto.blk.18.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1320 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.1 = util.global.load @__auto.blk.18.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1321 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.2 = util.global.load @__auto.blk.18.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1322 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.3 = util.global.load @__auto.blk.18.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1323 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.4 = util.global.load @__auto.blk.18.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1324 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.5 = util.global.load @__auto.blk.18.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1325 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.6 = util.global.load @__auto.blk.18.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1326 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_k.weight.shard.7 = util.global.load @__auto.blk.18.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1327 = torch_c.from_builtin_tensor %__auto.blk.18.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.0 = util.global.load @__auto.blk.18.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1328 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.1 = util.global.load @__auto.blk.18.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1329 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.2 = util.global.load @__auto.blk.18.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1330 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.3 = util.global.load @__auto.blk.18.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1331 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.4 = util.global.load @__auto.blk.18.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1332 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.5 = util.global.load @__auto.blk.18.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1333 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.6 = util.global.load @__auto.blk.18.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1334 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_v.weight.shard.7 = util.global.load @__auto.blk.18.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1335 = torch_c.from_builtin_tensor %__auto.blk.18.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.18.attn_output.weight.shard.0 = util.global.load @__auto.blk.18.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1336 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.1 = util.global.load @__auto.blk.18.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1337 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.2 = util.global.load @__auto.blk.18.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1338 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.3 = util.global.load @__auto.blk.18.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1339 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.4 = util.global.load @__auto.blk.18.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1340 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.5 = util.global.load @__auto.blk.18.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1341 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.6 = util.global.load @__auto.blk.18.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1342 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.attn_output.weight.shard.7 = util.global.load @__auto.blk.18.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1343 = torch_c.from_builtin_tensor %__auto.blk.18.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.18.ffn_norm.weight = util.global.load @__auto.blk.18.ffn_norm.weight : tensor<4096xf32>
%1344 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$1 = util.global.load @__auto.blk.18.ffn_norm.weight$1 : tensor<4096xf32>
%1345 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$2 = util.global.load @__auto.blk.18.ffn_norm.weight$2 : tensor<4096xf32>
%1346 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$3 = util.global.load @__auto.blk.18.ffn_norm.weight$3 : tensor<4096xf32>
%1347 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$4 = util.global.load @__auto.blk.18.ffn_norm.weight$4 : tensor<4096xf32>
%1348 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$5 = util.global.load @__auto.blk.18.ffn_norm.weight$5 : tensor<4096xf32>
%1349 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$6 = util.global.load @__auto.blk.18.ffn_norm.weight$6 : tensor<4096xf32>
%1350 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_norm.weight$7 = util.global.load @__auto.blk.18.ffn_norm.weight$7 : tensor<4096xf32>
%1351 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.18.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1352 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1353 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1354 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1355 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1356 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1357 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1358 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.18.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1359 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.0 = util.global.load @__auto.blk.18.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1360 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.1 = util.global.load @__auto.blk.18.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1361 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.2 = util.global.load @__auto.blk.18.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1362 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.3 = util.global.load @__auto.blk.18.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1363 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.4 = util.global.load @__auto.blk.18.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1364 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.5 = util.global.load @__auto.blk.18.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1365 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.6 = util.global.load @__auto.blk.18.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1366 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_up.weight.shard.7 = util.global.load @__auto.blk.18.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1367 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.18.ffn_down.weight.shard.0 = util.global.load @__auto.blk.18.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1368 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.1 = util.global.load @__auto.blk.18.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1369 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.2 = util.global.load @__auto.blk.18.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1370 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.3 = util.global.load @__auto.blk.18.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1371 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.4 = util.global.load @__auto.blk.18.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1372 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.5 = util.global.load @__auto.blk.18.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1373 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.6 = util.global.load @__auto.blk.18.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1374 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.18.ffn_down.weight.shard.7 = util.global.load @__auto.blk.18.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1375 = torch_c.from_builtin_tensor %__auto.blk.18.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.attn_norm.weight = util.global.load @__auto.blk.19.attn_norm.weight : tensor<4096xf32>
%1376 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$1 = util.global.load @__auto.blk.19.attn_norm.weight$1 : tensor<4096xf32>
%1377 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$2 = util.global.load @__auto.blk.19.attn_norm.weight$2 : tensor<4096xf32>
%1378 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$3 = util.global.load @__auto.blk.19.attn_norm.weight$3 : tensor<4096xf32>
%1379 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$4 = util.global.load @__auto.blk.19.attn_norm.weight$4 : tensor<4096xf32>
%1380 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$5 = util.global.load @__auto.blk.19.attn_norm.weight$5 : tensor<4096xf32>
%1381 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$6 = util.global.load @__auto.blk.19.attn_norm.weight$6 : tensor<4096xf32>
%1382 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_norm.weight$7 = util.global.load @__auto.blk.19.attn_norm.weight$7 : tensor<4096xf32>
%1383 = torch_c.from_builtin_tensor %__auto.blk.19.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.attn_q.weight.shard.0 = util.global.load @__auto.blk.19.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1384 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.1 = util.global.load @__auto.blk.19.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1385 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.2 = util.global.load @__auto.blk.19.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1386 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.3 = util.global.load @__auto.blk.19.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1387 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.4 = util.global.load @__auto.blk.19.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1388 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.5 = util.global.load @__auto.blk.19.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1389 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.6 = util.global.load @__auto.blk.19.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1390 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_q.weight.shard.7 = util.global.load @__auto.blk.19.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1391 = torch_c.from_builtin_tensor %__auto.blk.19.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.19.attn_k.weight.shard.0 = util.global.load @__auto.blk.19.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1392 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.1 = util.global.load @__auto.blk.19.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1393 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.2 = util.global.load @__auto.blk.19.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1394 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.3 = util.global.load @__auto.blk.19.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1395 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.4 = util.global.load @__auto.blk.19.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1396 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.5 = util.global.load @__auto.blk.19.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1397 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.6 = util.global.load @__auto.blk.19.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1398 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_k.weight.shard.7 = util.global.load @__auto.blk.19.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1399 = torch_c.from_builtin_tensor %__auto.blk.19.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.0 = util.global.load @__auto.blk.19.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1400 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.1 = util.global.load @__auto.blk.19.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1401 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.2 = util.global.load @__auto.blk.19.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1402 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.3 = util.global.load @__auto.blk.19.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1403 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.4 = util.global.load @__auto.blk.19.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1404 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.5 = util.global.load @__auto.blk.19.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1405 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.6 = util.global.load @__auto.blk.19.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1406 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_v.weight.shard.7 = util.global.load @__auto.blk.19.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1407 = torch_c.from_builtin_tensor %__auto.blk.19.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.19.attn_output.weight.shard.0 = util.global.load @__auto.blk.19.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1408 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.1 = util.global.load @__auto.blk.19.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1409 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.2 = util.global.load @__auto.blk.19.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1410 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.3 = util.global.load @__auto.blk.19.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1411 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.4 = util.global.load @__auto.blk.19.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1412 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.5 = util.global.load @__auto.blk.19.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1413 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.6 = util.global.load @__auto.blk.19.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1414 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.attn_output.weight.shard.7 = util.global.load @__auto.blk.19.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1415 = torch_c.from_builtin_tensor %__auto.blk.19.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.19.ffn_norm.weight = util.global.load @__auto.blk.19.ffn_norm.weight : tensor<4096xf32>
%1416 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$1 = util.global.load @__auto.blk.19.ffn_norm.weight$1 : tensor<4096xf32>
%1417 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$2 = util.global.load @__auto.blk.19.ffn_norm.weight$2 : tensor<4096xf32>
%1418 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$3 = util.global.load @__auto.blk.19.ffn_norm.weight$3 : tensor<4096xf32>
%1419 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$4 = util.global.load @__auto.blk.19.ffn_norm.weight$4 : tensor<4096xf32>
%1420 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$5 = util.global.load @__auto.blk.19.ffn_norm.weight$5 : tensor<4096xf32>
%1421 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$6 = util.global.load @__auto.blk.19.ffn_norm.weight$6 : tensor<4096xf32>
%1422 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_norm.weight$7 = util.global.load @__auto.blk.19.ffn_norm.weight$7 : tensor<4096xf32>
%1423 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.19.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1424 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1425 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1426 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1427 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1428 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1429 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1430 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.19.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1431 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.0 = util.global.load @__auto.blk.19.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1432 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.1 = util.global.load @__auto.blk.19.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1433 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.2 = util.global.load @__auto.blk.19.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1434 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.3 = util.global.load @__auto.blk.19.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1435 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.4 = util.global.load @__auto.blk.19.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1436 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.5 = util.global.load @__auto.blk.19.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1437 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.6 = util.global.load @__auto.blk.19.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1438 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_up.weight.shard.7 = util.global.load @__auto.blk.19.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1439 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.19.ffn_down.weight.shard.0 = util.global.load @__auto.blk.19.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1440 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.1 = util.global.load @__auto.blk.19.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1441 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.2 = util.global.load @__auto.blk.19.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1442 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.3 = util.global.load @__auto.blk.19.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1443 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.4 = util.global.load @__auto.blk.19.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1444 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.5 = util.global.load @__auto.blk.19.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1445 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.6 = util.global.load @__auto.blk.19.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1446 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.19.ffn_down.weight.shard.7 = util.global.load @__auto.blk.19.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1447 = torch_c.from_builtin_tensor %__auto.blk.19.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.attn_norm.weight = util.global.load @__auto.blk.20.attn_norm.weight : tensor<4096xf32>
%1448 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$1 = util.global.load @__auto.blk.20.attn_norm.weight$1 : tensor<4096xf32>
%1449 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$2 = util.global.load @__auto.blk.20.attn_norm.weight$2 : tensor<4096xf32>
%1450 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$3 = util.global.load @__auto.blk.20.attn_norm.weight$3 : tensor<4096xf32>
%1451 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$4 = util.global.load @__auto.blk.20.attn_norm.weight$4 : tensor<4096xf32>
%1452 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$5 = util.global.load @__auto.blk.20.attn_norm.weight$5 : tensor<4096xf32>
%1453 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$6 = util.global.load @__auto.blk.20.attn_norm.weight$6 : tensor<4096xf32>
%1454 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_norm.weight$7 = util.global.load @__auto.blk.20.attn_norm.weight$7 : tensor<4096xf32>
%1455 = torch_c.from_builtin_tensor %__auto.blk.20.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.attn_q.weight.shard.0 = util.global.load @__auto.blk.20.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1456 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.1 = util.global.load @__auto.blk.20.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1457 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.2 = util.global.load @__auto.blk.20.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1458 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.3 = util.global.load @__auto.blk.20.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1459 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.4 = util.global.load @__auto.blk.20.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1460 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.5 = util.global.load @__auto.blk.20.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1461 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.6 = util.global.load @__auto.blk.20.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1462 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_q.weight.shard.7 = util.global.load @__auto.blk.20.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1463 = torch_c.from_builtin_tensor %__auto.blk.20.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.20.attn_k.weight.shard.0 = util.global.load @__auto.blk.20.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1464 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.1 = util.global.load @__auto.blk.20.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1465 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.2 = util.global.load @__auto.blk.20.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1466 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.3 = util.global.load @__auto.blk.20.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1467 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.4 = util.global.load @__auto.blk.20.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1468 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.5 = util.global.load @__auto.blk.20.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1469 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.6 = util.global.load @__auto.blk.20.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1470 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_k.weight.shard.7 = util.global.load @__auto.blk.20.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1471 = torch_c.from_builtin_tensor %__auto.blk.20.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.0 = util.global.load @__auto.blk.20.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1472 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.1 = util.global.load @__auto.blk.20.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1473 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.2 = util.global.load @__auto.blk.20.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1474 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.3 = util.global.load @__auto.blk.20.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1475 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.4 = util.global.load @__auto.blk.20.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1476 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.5 = util.global.load @__auto.blk.20.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1477 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.6 = util.global.load @__auto.blk.20.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1478 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_v.weight.shard.7 = util.global.load @__auto.blk.20.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1479 = torch_c.from_builtin_tensor %__auto.blk.20.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.20.attn_output.weight.shard.0 = util.global.load @__auto.blk.20.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1480 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.1 = util.global.load @__auto.blk.20.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1481 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.2 = util.global.load @__auto.blk.20.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1482 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.3 = util.global.load @__auto.blk.20.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1483 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.4 = util.global.load @__auto.blk.20.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1484 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.5 = util.global.load @__auto.blk.20.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1485 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.6 = util.global.load @__auto.blk.20.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1486 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.attn_output.weight.shard.7 = util.global.load @__auto.blk.20.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1487 = torch_c.from_builtin_tensor %__auto.blk.20.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.20.ffn_norm.weight = util.global.load @__auto.blk.20.ffn_norm.weight : tensor<4096xf32>
%1488 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$1 = util.global.load @__auto.blk.20.ffn_norm.weight$1 : tensor<4096xf32>
%1489 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$2 = util.global.load @__auto.blk.20.ffn_norm.weight$2 : tensor<4096xf32>
%1490 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$3 = util.global.load @__auto.blk.20.ffn_norm.weight$3 : tensor<4096xf32>
%1491 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$4 = util.global.load @__auto.blk.20.ffn_norm.weight$4 : tensor<4096xf32>
%1492 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$5 = util.global.load @__auto.blk.20.ffn_norm.weight$5 : tensor<4096xf32>
%1493 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$6 = util.global.load @__auto.blk.20.ffn_norm.weight$6 : tensor<4096xf32>
%1494 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_norm.weight$7 = util.global.load @__auto.blk.20.ffn_norm.weight$7 : tensor<4096xf32>
%1495 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.20.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1496 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1497 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1498 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1499 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1500 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1501 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1502 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.20.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1503 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.0 = util.global.load @__auto.blk.20.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1504 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.1 = util.global.load @__auto.blk.20.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1505 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.2 = util.global.load @__auto.blk.20.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1506 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.3 = util.global.load @__auto.blk.20.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1507 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.4 = util.global.load @__auto.blk.20.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1508 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.5 = util.global.load @__auto.blk.20.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1509 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.6 = util.global.load @__auto.blk.20.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1510 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_up.weight.shard.7 = util.global.load @__auto.blk.20.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1511 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.20.ffn_down.weight.shard.0 = util.global.load @__auto.blk.20.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1512 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.1 = util.global.load @__auto.blk.20.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1513 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.2 = util.global.load @__auto.blk.20.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1514 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.3 = util.global.load @__auto.blk.20.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1515 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.4 = util.global.load @__auto.blk.20.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1516 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.5 = util.global.load @__auto.blk.20.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1517 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.6 = util.global.load @__auto.blk.20.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1518 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.20.ffn_down.weight.shard.7 = util.global.load @__auto.blk.20.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1519 = torch_c.from_builtin_tensor %__auto.blk.20.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.attn_norm.weight = util.global.load @__auto.blk.21.attn_norm.weight : tensor<4096xf32>
%1520 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$1 = util.global.load @__auto.blk.21.attn_norm.weight$1 : tensor<4096xf32>
%1521 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$2 = util.global.load @__auto.blk.21.attn_norm.weight$2 : tensor<4096xf32>
%1522 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$3 = util.global.load @__auto.blk.21.attn_norm.weight$3 : tensor<4096xf32>
%1523 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$4 = util.global.load @__auto.blk.21.attn_norm.weight$4 : tensor<4096xf32>
%1524 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$5 = util.global.load @__auto.blk.21.attn_norm.weight$5 : tensor<4096xf32>
%1525 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$6 = util.global.load @__auto.blk.21.attn_norm.weight$6 : tensor<4096xf32>
%1526 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_norm.weight$7 = util.global.load @__auto.blk.21.attn_norm.weight$7 : tensor<4096xf32>
%1527 = torch_c.from_builtin_tensor %__auto.blk.21.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.attn_q.weight.shard.0 = util.global.load @__auto.blk.21.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1528 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.1 = util.global.load @__auto.blk.21.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1529 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.2 = util.global.load @__auto.blk.21.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1530 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.3 = util.global.load @__auto.blk.21.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1531 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.4 = util.global.load @__auto.blk.21.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1532 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.5 = util.global.load @__auto.blk.21.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1533 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.6 = util.global.load @__auto.blk.21.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1534 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_q.weight.shard.7 = util.global.load @__auto.blk.21.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1535 = torch_c.from_builtin_tensor %__auto.blk.21.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.21.attn_k.weight.shard.0 = util.global.load @__auto.blk.21.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1536 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.1 = util.global.load @__auto.blk.21.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1537 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.2 = util.global.load @__auto.blk.21.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1538 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.3 = util.global.load @__auto.blk.21.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1539 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.4 = util.global.load @__auto.blk.21.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1540 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.5 = util.global.load @__auto.blk.21.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1541 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.6 = util.global.load @__auto.blk.21.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1542 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_k.weight.shard.7 = util.global.load @__auto.blk.21.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1543 = torch_c.from_builtin_tensor %__auto.blk.21.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.0 = util.global.load @__auto.blk.21.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1544 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.1 = util.global.load @__auto.blk.21.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1545 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.2 = util.global.load @__auto.blk.21.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1546 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.3 = util.global.load @__auto.blk.21.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1547 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.4 = util.global.load @__auto.blk.21.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1548 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.5 = util.global.load @__auto.blk.21.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1549 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.6 = util.global.load @__auto.blk.21.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1550 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_v.weight.shard.7 = util.global.load @__auto.blk.21.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1551 = torch_c.from_builtin_tensor %__auto.blk.21.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.21.attn_output.weight.shard.0 = util.global.load @__auto.blk.21.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1552 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.1 = util.global.load @__auto.blk.21.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1553 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.2 = util.global.load @__auto.blk.21.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1554 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.3 = util.global.load @__auto.blk.21.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1555 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.4 = util.global.load @__auto.blk.21.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1556 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.5 = util.global.load @__auto.blk.21.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1557 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.6 = util.global.load @__auto.blk.21.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1558 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.attn_output.weight.shard.7 = util.global.load @__auto.blk.21.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1559 = torch_c.from_builtin_tensor %__auto.blk.21.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.21.ffn_norm.weight = util.global.load @__auto.blk.21.ffn_norm.weight : tensor<4096xf32>
%1560 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$1 = util.global.load @__auto.blk.21.ffn_norm.weight$1 : tensor<4096xf32>
%1561 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$2 = util.global.load @__auto.blk.21.ffn_norm.weight$2 : tensor<4096xf32>
%1562 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$3 = util.global.load @__auto.blk.21.ffn_norm.weight$3 : tensor<4096xf32>
%1563 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$4 = util.global.load @__auto.blk.21.ffn_norm.weight$4 : tensor<4096xf32>
%1564 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$5 = util.global.load @__auto.blk.21.ffn_norm.weight$5 : tensor<4096xf32>
%1565 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$6 = util.global.load @__auto.blk.21.ffn_norm.weight$6 : tensor<4096xf32>
%1566 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_norm.weight$7 = util.global.load @__auto.blk.21.ffn_norm.weight$7 : tensor<4096xf32>
%1567 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.21.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1568 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1569 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1570 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1571 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1572 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1573 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1574 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.21.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1575 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.0 = util.global.load @__auto.blk.21.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1576 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.1 = util.global.load @__auto.blk.21.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1577 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.2 = util.global.load @__auto.blk.21.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1578 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.3 = util.global.load @__auto.blk.21.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1579 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.4 = util.global.load @__auto.blk.21.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1580 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.5 = util.global.load @__auto.blk.21.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1581 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.6 = util.global.load @__auto.blk.21.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1582 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_up.weight.shard.7 = util.global.load @__auto.blk.21.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1583 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.21.ffn_down.weight.shard.0 = util.global.load @__auto.blk.21.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1584 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.1 = util.global.load @__auto.blk.21.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1585 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.2 = util.global.load @__auto.blk.21.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1586 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.3 = util.global.load @__auto.blk.21.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1587 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.4 = util.global.load @__auto.blk.21.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1588 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.5 = util.global.load @__auto.blk.21.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1589 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.6 = util.global.load @__auto.blk.21.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1590 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.21.ffn_down.weight.shard.7 = util.global.load @__auto.blk.21.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1591 = torch_c.from_builtin_tensor %__auto.blk.21.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.attn_norm.weight = util.global.load @__auto.blk.22.attn_norm.weight : tensor<4096xf32>
%1592 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$1 = util.global.load @__auto.blk.22.attn_norm.weight$1 : tensor<4096xf32>
%1593 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$2 = util.global.load @__auto.blk.22.attn_norm.weight$2 : tensor<4096xf32>
%1594 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$3 = util.global.load @__auto.blk.22.attn_norm.weight$3 : tensor<4096xf32>
%1595 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$4 = util.global.load @__auto.blk.22.attn_norm.weight$4 : tensor<4096xf32>
%1596 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$5 = util.global.load @__auto.blk.22.attn_norm.weight$5 : tensor<4096xf32>
%1597 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$6 = util.global.load @__auto.blk.22.attn_norm.weight$6 : tensor<4096xf32>
%1598 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_norm.weight$7 = util.global.load @__auto.blk.22.attn_norm.weight$7 : tensor<4096xf32>
%1599 = torch_c.from_builtin_tensor %__auto.blk.22.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.attn_q.weight.shard.0 = util.global.load @__auto.blk.22.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1600 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.1 = util.global.load @__auto.blk.22.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1601 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.2 = util.global.load @__auto.blk.22.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1602 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.3 = util.global.load @__auto.blk.22.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1603 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.4 = util.global.load @__auto.blk.22.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1604 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.5 = util.global.load @__auto.blk.22.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1605 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.6 = util.global.load @__auto.blk.22.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1606 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_q.weight.shard.7 = util.global.load @__auto.blk.22.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1607 = torch_c.from_builtin_tensor %__auto.blk.22.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.22.attn_k.weight.shard.0 = util.global.load @__auto.blk.22.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1608 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.1 = util.global.load @__auto.blk.22.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1609 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.2 = util.global.load @__auto.blk.22.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1610 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.3 = util.global.load @__auto.blk.22.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1611 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.4 = util.global.load @__auto.blk.22.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1612 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.5 = util.global.load @__auto.blk.22.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1613 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.6 = util.global.load @__auto.blk.22.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1614 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_k.weight.shard.7 = util.global.load @__auto.blk.22.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1615 = torch_c.from_builtin_tensor %__auto.blk.22.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.0 = util.global.load @__auto.blk.22.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1616 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.1 = util.global.load @__auto.blk.22.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1617 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.2 = util.global.load @__auto.blk.22.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1618 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.3 = util.global.load @__auto.blk.22.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1619 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.4 = util.global.load @__auto.blk.22.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1620 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.5 = util.global.load @__auto.blk.22.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1621 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.6 = util.global.load @__auto.blk.22.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1622 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_v.weight.shard.7 = util.global.load @__auto.blk.22.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1623 = torch_c.from_builtin_tensor %__auto.blk.22.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.22.attn_output.weight.shard.0 = util.global.load @__auto.blk.22.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1624 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.1 = util.global.load @__auto.blk.22.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1625 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.2 = util.global.load @__auto.blk.22.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1626 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.3 = util.global.load @__auto.blk.22.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1627 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.4 = util.global.load @__auto.blk.22.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1628 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.5 = util.global.load @__auto.blk.22.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1629 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.6 = util.global.load @__auto.blk.22.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1630 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.attn_output.weight.shard.7 = util.global.load @__auto.blk.22.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1631 = torch_c.from_builtin_tensor %__auto.blk.22.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.22.ffn_norm.weight = util.global.load @__auto.blk.22.ffn_norm.weight : tensor<4096xf32>
%1632 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$1 = util.global.load @__auto.blk.22.ffn_norm.weight$1 : tensor<4096xf32>
%1633 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$2 = util.global.load @__auto.blk.22.ffn_norm.weight$2 : tensor<4096xf32>
%1634 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$3 = util.global.load @__auto.blk.22.ffn_norm.weight$3 : tensor<4096xf32>
%1635 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$4 = util.global.load @__auto.blk.22.ffn_norm.weight$4 : tensor<4096xf32>
%1636 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$5 = util.global.load @__auto.blk.22.ffn_norm.weight$5 : tensor<4096xf32>
%1637 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$6 = util.global.load @__auto.blk.22.ffn_norm.weight$6 : tensor<4096xf32>
%1638 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_norm.weight$7 = util.global.load @__auto.blk.22.ffn_norm.weight$7 : tensor<4096xf32>
%1639 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.22.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1640 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1641 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1642 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1643 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1644 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1645 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1646 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.22.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1647 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.0 = util.global.load @__auto.blk.22.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1648 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.1 = util.global.load @__auto.blk.22.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1649 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.2 = util.global.load @__auto.blk.22.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1650 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.3 = util.global.load @__auto.blk.22.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1651 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.4 = util.global.load @__auto.blk.22.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1652 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.5 = util.global.load @__auto.blk.22.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1653 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.6 = util.global.load @__auto.blk.22.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1654 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_up.weight.shard.7 = util.global.load @__auto.blk.22.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1655 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.22.ffn_down.weight.shard.0 = util.global.load @__auto.blk.22.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1656 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.1 = util.global.load @__auto.blk.22.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1657 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.2 = util.global.load @__auto.blk.22.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1658 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.3 = util.global.load @__auto.blk.22.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1659 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.4 = util.global.load @__auto.blk.22.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1660 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.5 = util.global.load @__auto.blk.22.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1661 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.6 = util.global.load @__auto.blk.22.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1662 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.22.ffn_down.weight.shard.7 = util.global.load @__auto.blk.22.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1663 = torch_c.from_builtin_tensor %__auto.blk.22.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.attn_norm.weight = util.global.load @__auto.blk.23.attn_norm.weight : tensor<4096xf32>
%1664 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$1 = util.global.load @__auto.blk.23.attn_norm.weight$1 : tensor<4096xf32>
%1665 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$2 = util.global.load @__auto.blk.23.attn_norm.weight$2 : tensor<4096xf32>
%1666 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$3 = util.global.load @__auto.blk.23.attn_norm.weight$3 : tensor<4096xf32>
%1667 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$4 = util.global.load @__auto.blk.23.attn_norm.weight$4 : tensor<4096xf32>
%1668 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$5 = util.global.load @__auto.blk.23.attn_norm.weight$5 : tensor<4096xf32>
%1669 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$6 = util.global.load @__auto.blk.23.attn_norm.weight$6 : tensor<4096xf32>
%1670 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_norm.weight$7 = util.global.load @__auto.blk.23.attn_norm.weight$7 : tensor<4096xf32>
%1671 = torch_c.from_builtin_tensor %__auto.blk.23.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.attn_q.weight.shard.0 = util.global.load @__auto.blk.23.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1672 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.1 = util.global.load @__auto.blk.23.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1673 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.2 = util.global.load @__auto.blk.23.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1674 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.3 = util.global.load @__auto.blk.23.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1675 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.4 = util.global.load @__auto.blk.23.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1676 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.5 = util.global.load @__auto.blk.23.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1677 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.6 = util.global.load @__auto.blk.23.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1678 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_q.weight.shard.7 = util.global.load @__auto.blk.23.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1679 = torch_c.from_builtin_tensor %__auto.blk.23.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.23.attn_k.weight.shard.0 = util.global.load @__auto.blk.23.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1680 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.1 = util.global.load @__auto.blk.23.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1681 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.2 = util.global.load @__auto.blk.23.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1682 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.3 = util.global.load @__auto.blk.23.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1683 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.4 = util.global.load @__auto.blk.23.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1684 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.5 = util.global.load @__auto.blk.23.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1685 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.6 = util.global.load @__auto.blk.23.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1686 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_k.weight.shard.7 = util.global.load @__auto.blk.23.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1687 = torch_c.from_builtin_tensor %__auto.blk.23.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.0 = util.global.load @__auto.blk.23.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1688 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.1 = util.global.load @__auto.blk.23.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1689 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.2 = util.global.load @__auto.blk.23.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1690 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.3 = util.global.load @__auto.blk.23.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1691 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.4 = util.global.load @__auto.blk.23.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1692 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.5 = util.global.load @__auto.blk.23.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1693 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.6 = util.global.load @__auto.blk.23.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1694 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_v.weight.shard.7 = util.global.load @__auto.blk.23.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1695 = torch_c.from_builtin_tensor %__auto.blk.23.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.23.attn_output.weight.shard.0 = util.global.load @__auto.blk.23.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1696 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.1 = util.global.load @__auto.blk.23.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1697 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.2 = util.global.load @__auto.blk.23.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1698 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.3 = util.global.load @__auto.blk.23.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1699 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.4 = util.global.load @__auto.blk.23.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1700 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.5 = util.global.load @__auto.blk.23.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1701 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.6 = util.global.load @__auto.blk.23.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1702 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.attn_output.weight.shard.7 = util.global.load @__auto.blk.23.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1703 = torch_c.from_builtin_tensor %__auto.blk.23.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.23.ffn_norm.weight = util.global.load @__auto.blk.23.ffn_norm.weight : tensor<4096xf32>
%1704 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$1 = util.global.load @__auto.blk.23.ffn_norm.weight$1 : tensor<4096xf32>
%1705 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$2 = util.global.load @__auto.blk.23.ffn_norm.weight$2 : tensor<4096xf32>
%1706 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$3 = util.global.load @__auto.blk.23.ffn_norm.weight$3 : tensor<4096xf32>
%1707 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$4 = util.global.load @__auto.blk.23.ffn_norm.weight$4 : tensor<4096xf32>
%1708 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$5 = util.global.load @__auto.blk.23.ffn_norm.weight$5 : tensor<4096xf32>
%1709 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$6 = util.global.load @__auto.blk.23.ffn_norm.weight$6 : tensor<4096xf32>
%1710 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_norm.weight$7 = util.global.load @__auto.blk.23.ffn_norm.weight$7 : tensor<4096xf32>
%1711 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.23.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1712 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1713 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1714 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1715 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1716 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1717 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1718 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.23.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1719 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.0 = util.global.load @__auto.blk.23.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1720 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.1 = util.global.load @__auto.blk.23.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1721 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.2 = util.global.load @__auto.blk.23.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1722 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.3 = util.global.load @__auto.blk.23.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1723 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.4 = util.global.load @__auto.blk.23.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1724 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.5 = util.global.load @__auto.blk.23.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1725 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.6 = util.global.load @__auto.blk.23.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1726 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_up.weight.shard.7 = util.global.load @__auto.blk.23.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1727 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.23.ffn_down.weight.shard.0 = util.global.load @__auto.blk.23.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1728 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.1 = util.global.load @__auto.blk.23.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1729 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.2 = util.global.load @__auto.blk.23.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1730 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.3 = util.global.load @__auto.blk.23.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1731 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.4 = util.global.load @__auto.blk.23.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1732 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.5 = util.global.load @__auto.blk.23.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1733 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.6 = util.global.load @__auto.blk.23.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1734 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.23.ffn_down.weight.shard.7 = util.global.load @__auto.blk.23.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1735 = torch_c.from_builtin_tensor %__auto.blk.23.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.attn_norm.weight = util.global.load @__auto.blk.24.attn_norm.weight : tensor<4096xf32>
%1736 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$1 = util.global.load @__auto.blk.24.attn_norm.weight$1 : tensor<4096xf32>
%1737 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$2 = util.global.load @__auto.blk.24.attn_norm.weight$2 : tensor<4096xf32>
%1738 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$3 = util.global.load @__auto.blk.24.attn_norm.weight$3 : tensor<4096xf32>
%1739 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$4 = util.global.load @__auto.blk.24.attn_norm.weight$4 : tensor<4096xf32>
%1740 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$5 = util.global.load @__auto.blk.24.attn_norm.weight$5 : tensor<4096xf32>
%1741 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$6 = util.global.load @__auto.blk.24.attn_norm.weight$6 : tensor<4096xf32>
%1742 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_norm.weight$7 = util.global.load @__auto.blk.24.attn_norm.weight$7 : tensor<4096xf32>
%1743 = torch_c.from_builtin_tensor %__auto.blk.24.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.attn_q.weight.shard.0 = util.global.load @__auto.blk.24.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1744 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.1 = util.global.load @__auto.blk.24.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1745 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.2 = util.global.load @__auto.blk.24.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1746 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.3 = util.global.load @__auto.blk.24.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1747 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.4 = util.global.load @__auto.blk.24.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1748 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.5 = util.global.load @__auto.blk.24.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1749 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.6 = util.global.load @__auto.blk.24.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1750 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_q.weight.shard.7 = util.global.load @__auto.blk.24.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1751 = torch_c.from_builtin_tensor %__auto.blk.24.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.24.attn_k.weight.shard.0 = util.global.load @__auto.blk.24.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1752 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.1 = util.global.load @__auto.blk.24.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1753 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.2 = util.global.load @__auto.blk.24.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1754 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.3 = util.global.load @__auto.blk.24.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1755 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.4 = util.global.load @__auto.blk.24.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1756 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.5 = util.global.load @__auto.blk.24.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1757 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.6 = util.global.load @__auto.blk.24.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1758 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_k.weight.shard.7 = util.global.load @__auto.blk.24.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1759 = torch_c.from_builtin_tensor %__auto.blk.24.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.0 = util.global.load @__auto.blk.24.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1760 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.1 = util.global.load @__auto.blk.24.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1761 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.2 = util.global.load @__auto.blk.24.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1762 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.3 = util.global.load @__auto.blk.24.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1763 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.4 = util.global.load @__auto.blk.24.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1764 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.5 = util.global.load @__auto.blk.24.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1765 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.6 = util.global.load @__auto.blk.24.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1766 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_v.weight.shard.7 = util.global.load @__auto.blk.24.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1767 = torch_c.from_builtin_tensor %__auto.blk.24.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.24.attn_output.weight.shard.0 = util.global.load @__auto.blk.24.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1768 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.1 = util.global.load @__auto.blk.24.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1769 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.2 = util.global.load @__auto.blk.24.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1770 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.3 = util.global.load @__auto.blk.24.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1771 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.4 = util.global.load @__auto.blk.24.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1772 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.5 = util.global.load @__auto.blk.24.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1773 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.6 = util.global.load @__auto.blk.24.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1774 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.attn_output.weight.shard.7 = util.global.load @__auto.blk.24.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1775 = torch_c.from_builtin_tensor %__auto.blk.24.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.24.ffn_norm.weight = util.global.load @__auto.blk.24.ffn_norm.weight : tensor<4096xf32>
%1776 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$1 = util.global.load @__auto.blk.24.ffn_norm.weight$1 : tensor<4096xf32>
%1777 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$2 = util.global.load @__auto.blk.24.ffn_norm.weight$2 : tensor<4096xf32>
%1778 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$3 = util.global.load @__auto.blk.24.ffn_norm.weight$3 : tensor<4096xf32>
%1779 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$4 = util.global.load @__auto.blk.24.ffn_norm.weight$4 : tensor<4096xf32>
%1780 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$5 = util.global.load @__auto.blk.24.ffn_norm.weight$5 : tensor<4096xf32>
%1781 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$6 = util.global.load @__auto.blk.24.ffn_norm.weight$6 : tensor<4096xf32>
%1782 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_norm.weight$7 = util.global.load @__auto.blk.24.ffn_norm.weight$7 : tensor<4096xf32>
%1783 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.24.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1784 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1785 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1786 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1787 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1788 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1789 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1790 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.24.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1791 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.0 = util.global.load @__auto.blk.24.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1792 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.1 = util.global.load @__auto.blk.24.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1793 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.2 = util.global.load @__auto.blk.24.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1794 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.3 = util.global.load @__auto.blk.24.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1795 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.4 = util.global.load @__auto.blk.24.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1796 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.5 = util.global.load @__auto.blk.24.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1797 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.6 = util.global.load @__auto.blk.24.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1798 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_up.weight.shard.7 = util.global.load @__auto.blk.24.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1799 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.24.ffn_down.weight.shard.0 = util.global.load @__auto.blk.24.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1800 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.1 = util.global.load @__auto.blk.24.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1801 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.2 = util.global.load @__auto.blk.24.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1802 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.3 = util.global.load @__auto.blk.24.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1803 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.4 = util.global.load @__auto.blk.24.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1804 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.5 = util.global.load @__auto.blk.24.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1805 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.6 = util.global.load @__auto.blk.24.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1806 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.24.ffn_down.weight.shard.7 = util.global.load @__auto.blk.24.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1807 = torch_c.from_builtin_tensor %__auto.blk.24.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.attn_norm.weight = util.global.load @__auto.blk.25.attn_norm.weight : tensor<4096xf32>
%1808 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$1 = util.global.load @__auto.blk.25.attn_norm.weight$1 : tensor<4096xf32>
%1809 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$2 = util.global.load @__auto.blk.25.attn_norm.weight$2 : tensor<4096xf32>
%1810 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$3 = util.global.load @__auto.blk.25.attn_norm.weight$3 : tensor<4096xf32>
%1811 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$4 = util.global.load @__auto.blk.25.attn_norm.weight$4 : tensor<4096xf32>
%1812 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$5 = util.global.load @__auto.blk.25.attn_norm.weight$5 : tensor<4096xf32>
%1813 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$6 = util.global.load @__auto.blk.25.attn_norm.weight$6 : tensor<4096xf32>
%1814 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_norm.weight$7 = util.global.load @__auto.blk.25.attn_norm.weight$7 : tensor<4096xf32>
%1815 = torch_c.from_builtin_tensor %__auto.blk.25.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.attn_q.weight.shard.0 = util.global.load @__auto.blk.25.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1816 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.1 = util.global.load @__auto.blk.25.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1817 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.2 = util.global.load @__auto.blk.25.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1818 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.3 = util.global.load @__auto.blk.25.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1819 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.4 = util.global.load @__auto.blk.25.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1820 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.5 = util.global.load @__auto.blk.25.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1821 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.6 = util.global.load @__auto.blk.25.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1822 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_q.weight.shard.7 = util.global.load @__auto.blk.25.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1823 = torch_c.from_builtin_tensor %__auto.blk.25.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.25.attn_k.weight.shard.0 = util.global.load @__auto.blk.25.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1824 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.1 = util.global.load @__auto.blk.25.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1825 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.2 = util.global.load @__auto.blk.25.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1826 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.3 = util.global.load @__auto.blk.25.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1827 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.4 = util.global.load @__auto.blk.25.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1828 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.5 = util.global.load @__auto.blk.25.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1829 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.6 = util.global.load @__auto.blk.25.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1830 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_k.weight.shard.7 = util.global.load @__auto.blk.25.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1831 = torch_c.from_builtin_tensor %__auto.blk.25.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.0 = util.global.load @__auto.blk.25.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1832 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.1 = util.global.load @__auto.blk.25.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1833 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.2 = util.global.load @__auto.blk.25.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1834 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.3 = util.global.load @__auto.blk.25.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1835 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.4 = util.global.load @__auto.blk.25.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1836 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.5 = util.global.load @__auto.blk.25.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1837 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.6 = util.global.load @__auto.blk.25.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1838 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_v.weight.shard.7 = util.global.load @__auto.blk.25.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1839 = torch_c.from_builtin_tensor %__auto.blk.25.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.25.attn_output.weight.shard.0 = util.global.load @__auto.blk.25.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1840 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.1 = util.global.load @__auto.blk.25.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1841 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.2 = util.global.load @__auto.blk.25.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1842 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.3 = util.global.load @__auto.blk.25.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1843 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.4 = util.global.load @__auto.blk.25.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1844 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.5 = util.global.load @__auto.blk.25.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1845 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.6 = util.global.load @__auto.blk.25.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1846 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.attn_output.weight.shard.7 = util.global.load @__auto.blk.25.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1847 = torch_c.from_builtin_tensor %__auto.blk.25.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.25.ffn_norm.weight = util.global.load @__auto.blk.25.ffn_norm.weight : tensor<4096xf32>
%1848 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$1 = util.global.load @__auto.blk.25.ffn_norm.weight$1 : tensor<4096xf32>
%1849 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$2 = util.global.load @__auto.blk.25.ffn_norm.weight$2 : tensor<4096xf32>
%1850 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$3 = util.global.load @__auto.blk.25.ffn_norm.weight$3 : tensor<4096xf32>
%1851 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$4 = util.global.load @__auto.blk.25.ffn_norm.weight$4 : tensor<4096xf32>
%1852 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$5 = util.global.load @__auto.blk.25.ffn_norm.weight$5 : tensor<4096xf32>
%1853 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$6 = util.global.load @__auto.blk.25.ffn_norm.weight$6 : tensor<4096xf32>
%1854 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_norm.weight$7 = util.global.load @__auto.blk.25.ffn_norm.weight$7 : tensor<4096xf32>
%1855 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.25.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1856 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1857 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1858 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1859 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1860 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1861 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1862 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.25.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1863 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.0 = util.global.load @__auto.blk.25.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1864 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.1 = util.global.load @__auto.blk.25.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1865 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.2 = util.global.load @__auto.blk.25.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1866 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.3 = util.global.load @__auto.blk.25.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1867 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.4 = util.global.load @__auto.blk.25.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1868 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.5 = util.global.load @__auto.blk.25.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1869 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.6 = util.global.load @__auto.blk.25.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1870 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_up.weight.shard.7 = util.global.load @__auto.blk.25.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1871 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.25.ffn_down.weight.shard.0 = util.global.load @__auto.blk.25.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1872 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.1 = util.global.load @__auto.blk.25.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1873 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.2 = util.global.load @__auto.blk.25.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1874 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.3 = util.global.load @__auto.blk.25.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1875 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.4 = util.global.load @__auto.blk.25.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1876 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.5 = util.global.load @__auto.blk.25.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1877 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.6 = util.global.load @__auto.blk.25.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1878 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.25.ffn_down.weight.shard.7 = util.global.load @__auto.blk.25.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1879 = torch_c.from_builtin_tensor %__auto.blk.25.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.attn_norm.weight = util.global.load @__auto.blk.26.attn_norm.weight : tensor<4096xf32>
%1880 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$1 = util.global.load @__auto.blk.26.attn_norm.weight$1 : tensor<4096xf32>
%1881 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$2 = util.global.load @__auto.blk.26.attn_norm.weight$2 : tensor<4096xf32>
%1882 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$3 = util.global.load @__auto.blk.26.attn_norm.weight$3 : tensor<4096xf32>
%1883 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$4 = util.global.load @__auto.blk.26.attn_norm.weight$4 : tensor<4096xf32>
%1884 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$5 = util.global.load @__auto.blk.26.attn_norm.weight$5 : tensor<4096xf32>
%1885 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$6 = util.global.load @__auto.blk.26.attn_norm.weight$6 : tensor<4096xf32>
%1886 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_norm.weight$7 = util.global.load @__auto.blk.26.attn_norm.weight$7 : tensor<4096xf32>
%1887 = torch_c.from_builtin_tensor %__auto.blk.26.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.attn_q.weight.shard.0 = util.global.load @__auto.blk.26.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1888 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.1 = util.global.load @__auto.blk.26.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1889 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.2 = util.global.load @__auto.blk.26.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1890 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.3 = util.global.load @__auto.blk.26.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1891 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.4 = util.global.load @__auto.blk.26.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1892 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.5 = util.global.load @__auto.blk.26.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1893 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.6 = util.global.load @__auto.blk.26.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1894 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_q.weight.shard.7 = util.global.load @__auto.blk.26.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1895 = torch_c.from_builtin_tensor %__auto.blk.26.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.26.attn_k.weight.shard.0 = util.global.load @__auto.blk.26.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1896 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.1 = util.global.load @__auto.blk.26.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1897 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.2 = util.global.load @__auto.blk.26.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1898 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.3 = util.global.load @__auto.blk.26.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1899 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.4 = util.global.load @__auto.blk.26.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1900 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.5 = util.global.load @__auto.blk.26.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1901 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.6 = util.global.load @__auto.blk.26.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1902 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_k.weight.shard.7 = util.global.load @__auto.blk.26.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1903 = torch_c.from_builtin_tensor %__auto.blk.26.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.0 = util.global.load @__auto.blk.26.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1904 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.1 = util.global.load @__auto.blk.26.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1905 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.2 = util.global.load @__auto.blk.26.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1906 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.3 = util.global.load @__auto.blk.26.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1907 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.4 = util.global.load @__auto.blk.26.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1908 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.5 = util.global.load @__auto.blk.26.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1909 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.6 = util.global.load @__auto.blk.26.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1910 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_v.weight.shard.7 = util.global.load @__auto.blk.26.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1911 = torch_c.from_builtin_tensor %__auto.blk.26.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.26.attn_output.weight.shard.0 = util.global.load @__auto.blk.26.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1912 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.1 = util.global.load @__auto.blk.26.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1913 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.2 = util.global.load @__auto.blk.26.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1914 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.3 = util.global.load @__auto.blk.26.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1915 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.4 = util.global.load @__auto.blk.26.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1916 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.5 = util.global.load @__auto.blk.26.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1917 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.6 = util.global.load @__auto.blk.26.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1918 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.attn_output.weight.shard.7 = util.global.load @__auto.blk.26.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1919 = torch_c.from_builtin_tensor %__auto.blk.26.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.26.ffn_norm.weight = util.global.load @__auto.blk.26.ffn_norm.weight : tensor<4096xf32>
%1920 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$1 = util.global.load @__auto.blk.26.ffn_norm.weight$1 : tensor<4096xf32>
%1921 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$2 = util.global.load @__auto.blk.26.ffn_norm.weight$2 : tensor<4096xf32>
%1922 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$3 = util.global.load @__auto.blk.26.ffn_norm.weight$3 : tensor<4096xf32>
%1923 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$4 = util.global.load @__auto.blk.26.ffn_norm.weight$4 : tensor<4096xf32>
%1924 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$5 = util.global.load @__auto.blk.26.ffn_norm.weight$5 : tensor<4096xf32>
%1925 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$6 = util.global.load @__auto.blk.26.ffn_norm.weight$6 : tensor<4096xf32>
%1926 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_norm.weight$7 = util.global.load @__auto.blk.26.ffn_norm.weight$7 : tensor<4096xf32>
%1927 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.26.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%1928 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%1929 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%1930 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%1931 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%1932 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%1933 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%1934 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.26.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%1935 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.0 = util.global.load @__auto.blk.26.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%1936 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.1 = util.global.load @__auto.blk.26.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%1937 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.2 = util.global.load @__auto.blk.26.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%1938 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.3 = util.global.load @__auto.blk.26.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%1939 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.4 = util.global.load @__auto.blk.26.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%1940 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.5 = util.global.load @__auto.blk.26.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%1941 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.6 = util.global.load @__auto.blk.26.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%1942 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_up.weight.shard.7 = util.global.load @__auto.blk.26.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%1943 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.26.ffn_down.weight.shard.0 = util.global.load @__auto.blk.26.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%1944 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.1 = util.global.load @__auto.blk.26.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%1945 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.2 = util.global.load @__auto.blk.26.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%1946 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.3 = util.global.load @__auto.blk.26.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%1947 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.4 = util.global.load @__auto.blk.26.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%1948 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.5 = util.global.load @__auto.blk.26.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%1949 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.6 = util.global.load @__auto.blk.26.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%1950 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.26.ffn_down.weight.shard.7 = util.global.load @__auto.blk.26.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%1951 = torch_c.from_builtin_tensor %__auto.blk.26.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.attn_norm.weight = util.global.load @__auto.blk.27.attn_norm.weight : tensor<4096xf32>
%1952 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$1 = util.global.load @__auto.blk.27.attn_norm.weight$1 : tensor<4096xf32>
%1953 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$2 = util.global.load @__auto.blk.27.attn_norm.weight$2 : tensor<4096xf32>
%1954 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$3 = util.global.load @__auto.blk.27.attn_norm.weight$3 : tensor<4096xf32>
%1955 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$4 = util.global.load @__auto.blk.27.attn_norm.weight$4 : tensor<4096xf32>
%1956 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$5 = util.global.load @__auto.blk.27.attn_norm.weight$5 : tensor<4096xf32>
%1957 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$6 = util.global.load @__auto.blk.27.attn_norm.weight$6 : tensor<4096xf32>
%1958 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_norm.weight$7 = util.global.load @__auto.blk.27.attn_norm.weight$7 : tensor<4096xf32>
%1959 = torch_c.from_builtin_tensor %__auto.blk.27.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.attn_q.weight.shard.0 = util.global.load @__auto.blk.27.attn_q.weight.shard.0 : tensor<512x4096xf16>
%1960 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.1 = util.global.load @__auto.blk.27.attn_q.weight.shard.1 : tensor<512x4096xf16>
%1961 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.2 = util.global.load @__auto.blk.27.attn_q.weight.shard.2 : tensor<512x4096xf16>
%1962 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.3 = util.global.load @__auto.blk.27.attn_q.weight.shard.3 : tensor<512x4096xf16>
%1963 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.4 = util.global.load @__auto.blk.27.attn_q.weight.shard.4 : tensor<512x4096xf16>
%1964 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.5 = util.global.load @__auto.blk.27.attn_q.weight.shard.5 : tensor<512x4096xf16>
%1965 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.6 = util.global.load @__auto.blk.27.attn_q.weight.shard.6 : tensor<512x4096xf16>
%1966 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_q.weight.shard.7 = util.global.load @__auto.blk.27.attn_q.weight.shard.7 : tensor<512x4096xf16>
%1967 = torch_c.from_builtin_tensor %__auto.blk.27.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.27.attn_k.weight.shard.0 = util.global.load @__auto.blk.27.attn_k.weight.shard.0 : tensor<128x4096xf16>
%1968 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.1 = util.global.load @__auto.blk.27.attn_k.weight.shard.1 : tensor<128x4096xf16>
%1969 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.2 = util.global.load @__auto.blk.27.attn_k.weight.shard.2 : tensor<128x4096xf16>
%1970 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.3 = util.global.load @__auto.blk.27.attn_k.weight.shard.3 : tensor<128x4096xf16>
%1971 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.4 = util.global.load @__auto.blk.27.attn_k.weight.shard.4 : tensor<128x4096xf16>
%1972 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.5 = util.global.load @__auto.blk.27.attn_k.weight.shard.5 : tensor<128x4096xf16>
%1973 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.6 = util.global.load @__auto.blk.27.attn_k.weight.shard.6 : tensor<128x4096xf16>
%1974 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_k.weight.shard.7 = util.global.load @__auto.blk.27.attn_k.weight.shard.7 : tensor<128x4096xf16>
%1975 = torch_c.from_builtin_tensor %__auto.blk.27.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.0 = util.global.load @__auto.blk.27.attn_v.weight.shard.0 : tensor<128x4096xf16>
%1976 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.1 = util.global.load @__auto.blk.27.attn_v.weight.shard.1 : tensor<128x4096xf16>
%1977 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.2 = util.global.load @__auto.blk.27.attn_v.weight.shard.2 : tensor<128x4096xf16>
%1978 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.3 = util.global.load @__auto.blk.27.attn_v.weight.shard.3 : tensor<128x4096xf16>
%1979 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.4 = util.global.load @__auto.blk.27.attn_v.weight.shard.4 : tensor<128x4096xf16>
%1980 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.5 = util.global.load @__auto.blk.27.attn_v.weight.shard.5 : tensor<128x4096xf16>
%1981 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.6 = util.global.load @__auto.blk.27.attn_v.weight.shard.6 : tensor<128x4096xf16>
%1982 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_v.weight.shard.7 = util.global.load @__auto.blk.27.attn_v.weight.shard.7 : tensor<128x4096xf16>
%1983 = torch_c.from_builtin_tensor %__auto.blk.27.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.27.attn_output.weight.shard.0 = util.global.load @__auto.blk.27.attn_output.weight.shard.0 : tensor<4096x512xf16>
%1984 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.1 = util.global.load @__auto.blk.27.attn_output.weight.shard.1 : tensor<4096x512xf16>
%1985 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.2 = util.global.load @__auto.blk.27.attn_output.weight.shard.2 : tensor<4096x512xf16>
%1986 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.3 = util.global.load @__auto.blk.27.attn_output.weight.shard.3 : tensor<4096x512xf16>
%1987 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.4 = util.global.load @__auto.blk.27.attn_output.weight.shard.4 : tensor<4096x512xf16>
%1988 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.5 = util.global.load @__auto.blk.27.attn_output.weight.shard.5 : tensor<4096x512xf16>
%1989 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.6 = util.global.load @__auto.blk.27.attn_output.weight.shard.6 : tensor<4096x512xf16>
%1990 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.attn_output.weight.shard.7 = util.global.load @__auto.blk.27.attn_output.weight.shard.7 : tensor<4096x512xf16>
%1991 = torch_c.from_builtin_tensor %__auto.blk.27.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.27.ffn_norm.weight = util.global.load @__auto.blk.27.ffn_norm.weight : tensor<4096xf32>
%1992 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$1 = util.global.load @__auto.blk.27.ffn_norm.weight$1 : tensor<4096xf32>
%1993 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$2 = util.global.load @__auto.blk.27.ffn_norm.weight$2 : tensor<4096xf32>
%1994 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$3 = util.global.load @__auto.blk.27.ffn_norm.weight$3 : tensor<4096xf32>
%1995 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$4 = util.global.load @__auto.blk.27.ffn_norm.weight$4 : tensor<4096xf32>
%1996 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$5 = util.global.load @__auto.blk.27.ffn_norm.weight$5 : tensor<4096xf32>
%1997 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$6 = util.global.load @__auto.blk.27.ffn_norm.weight$6 : tensor<4096xf32>
%1998 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_norm.weight$7 = util.global.load @__auto.blk.27.ffn_norm.weight$7 : tensor<4096xf32>
%1999 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.27.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%2000 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%2001 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%2002 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%2003 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%2004 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%2005 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%2006 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.27.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%2007 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.0 = util.global.load @__auto.blk.27.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%2008 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.1 = util.global.load @__auto.blk.27.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%2009 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.2 = util.global.load @__auto.blk.27.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%2010 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.3 = util.global.load @__auto.blk.27.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%2011 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.4 = util.global.load @__auto.blk.27.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%2012 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.5 = util.global.load @__auto.blk.27.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%2013 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.6 = util.global.load @__auto.blk.27.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%2014 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_up.weight.shard.7 = util.global.load @__auto.blk.27.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%2015 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.27.ffn_down.weight.shard.0 = util.global.load @__auto.blk.27.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%2016 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.1 = util.global.load @__auto.blk.27.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%2017 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.2 = util.global.load @__auto.blk.27.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%2018 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.3 = util.global.load @__auto.blk.27.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%2019 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.4 = util.global.load @__auto.blk.27.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%2020 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.5 = util.global.load @__auto.blk.27.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%2021 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.6 = util.global.load @__auto.blk.27.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%2022 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.27.ffn_down.weight.shard.7 = util.global.load @__auto.blk.27.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%2023 = torch_c.from_builtin_tensor %__auto.blk.27.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.attn_norm.weight = util.global.load @__auto.blk.28.attn_norm.weight : tensor<4096xf32>
%2024 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$1 = util.global.load @__auto.blk.28.attn_norm.weight$1 : tensor<4096xf32>
%2025 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$2 = util.global.load @__auto.blk.28.attn_norm.weight$2 : tensor<4096xf32>
%2026 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$3 = util.global.load @__auto.blk.28.attn_norm.weight$3 : tensor<4096xf32>
%2027 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$4 = util.global.load @__auto.blk.28.attn_norm.weight$4 : tensor<4096xf32>
%2028 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$5 = util.global.load @__auto.blk.28.attn_norm.weight$5 : tensor<4096xf32>
%2029 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$6 = util.global.load @__auto.blk.28.attn_norm.weight$6 : tensor<4096xf32>
%2030 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_norm.weight$7 = util.global.load @__auto.blk.28.attn_norm.weight$7 : tensor<4096xf32>
%2031 = torch_c.from_builtin_tensor %__auto.blk.28.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.attn_q.weight.shard.0 = util.global.load @__auto.blk.28.attn_q.weight.shard.0 : tensor<512x4096xf16>
%2032 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.1 = util.global.load @__auto.blk.28.attn_q.weight.shard.1 : tensor<512x4096xf16>
%2033 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.2 = util.global.load @__auto.blk.28.attn_q.weight.shard.2 : tensor<512x4096xf16>
%2034 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.3 = util.global.load @__auto.blk.28.attn_q.weight.shard.3 : tensor<512x4096xf16>
%2035 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.4 = util.global.load @__auto.blk.28.attn_q.weight.shard.4 : tensor<512x4096xf16>
%2036 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.5 = util.global.load @__auto.blk.28.attn_q.weight.shard.5 : tensor<512x4096xf16>
%2037 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.6 = util.global.load @__auto.blk.28.attn_q.weight.shard.6 : tensor<512x4096xf16>
%2038 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_q.weight.shard.7 = util.global.load @__auto.blk.28.attn_q.weight.shard.7 : tensor<512x4096xf16>
%2039 = torch_c.from_builtin_tensor %__auto.blk.28.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.28.attn_k.weight.shard.0 = util.global.load @__auto.blk.28.attn_k.weight.shard.0 : tensor<128x4096xf16>
%2040 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.1 = util.global.load @__auto.blk.28.attn_k.weight.shard.1 : tensor<128x4096xf16>
%2041 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.2 = util.global.load @__auto.blk.28.attn_k.weight.shard.2 : tensor<128x4096xf16>
%2042 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.3 = util.global.load @__auto.blk.28.attn_k.weight.shard.3 : tensor<128x4096xf16>
%2043 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.4 = util.global.load @__auto.blk.28.attn_k.weight.shard.4 : tensor<128x4096xf16>
%2044 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.5 = util.global.load @__auto.blk.28.attn_k.weight.shard.5 : tensor<128x4096xf16>
%2045 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.6 = util.global.load @__auto.blk.28.attn_k.weight.shard.6 : tensor<128x4096xf16>
%2046 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_k.weight.shard.7 = util.global.load @__auto.blk.28.attn_k.weight.shard.7 : tensor<128x4096xf16>
%2047 = torch_c.from_builtin_tensor %__auto.blk.28.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.0 = util.global.load @__auto.blk.28.attn_v.weight.shard.0 : tensor<128x4096xf16>
%2048 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.1 = util.global.load @__auto.blk.28.attn_v.weight.shard.1 : tensor<128x4096xf16>
%2049 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.2 = util.global.load @__auto.blk.28.attn_v.weight.shard.2 : tensor<128x4096xf16>
%2050 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.3 = util.global.load @__auto.blk.28.attn_v.weight.shard.3 : tensor<128x4096xf16>
%2051 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.4 = util.global.load @__auto.blk.28.attn_v.weight.shard.4 : tensor<128x4096xf16>
%2052 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.5 = util.global.load @__auto.blk.28.attn_v.weight.shard.5 : tensor<128x4096xf16>
%2053 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.6 = util.global.load @__auto.blk.28.attn_v.weight.shard.6 : tensor<128x4096xf16>
%2054 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_v.weight.shard.7 = util.global.load @__auto.blk.28.attn_v.weight.shard.7 : tensor<128x4096xf16>
%2055 = torch_c.from_builtin_tensor %__auto.blk.28.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.28.attn_output.weight.shard.0 = util.global.load @__auto.blk.28.attn_output.weight.shard.0 : tensor<4096x512xf16>
%2056 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.1 = util.global.load @__auto.blk.28.attn_output.weight.shard.1 : tensor<4096x512xf16>
%2057 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.2 = util.global.load @__auto.blk.28.attn_output.weight.shard.2 : tensor<4096x512xf16>
%2058 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.3 = util.global.load @__auto.blk.28.attn_output.weight.shard.3 : tensor<4096x512xf16>
%2059 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.4 = util.global.load @__auto.blk.28.attn_output.weight.shard.4 : tensor<4096x512xf16>
%2060 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.5 = util.global.load @__auto.blk.28.attn_output.weight.shard.5 : tensor<4096x512xf16>
%2061 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.6 = util.global.load @__auto.blk.28.attn_output.weight.shard.6 : tensor<4096x512xf16>
%2062 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.attn_output.weight.shard.7 = util.global.load @__auto.blk.28.attn_output.weight.shard.7 : tensor<4096x512xf16>
%2063 = torch_c.from_builtin_tensor %__auto.blk.28.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.28.ffn_norm.weight = util.global.load @__auto.blk.28.ffn_norm.weight : tensor<4096xf32>
%2064 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$1 = util.global.load @__auto.blk.28.ffn_norm.weight$1 : tensor<4096xf32>
%2065 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$2 = util.global.load @__auto.blk.28.ffn_norm.weight$2 : tensor<4096xf32>
%2066 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$3 = util.global.load @__auto.blk.28.ffn_norm.weight$3 : tensor<4096xf32>
%2067 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$4 = util.global.load @__auto.blk.28.ffn_norm.weight$4 : tensor<4096xf32>
%2068 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$5 = util.global.load @__auto.blk.28.ffn_norm.weight$5 : tensor<4096xf32>
%2069 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$6 = util.global.load @__auto.blk.28.ffn_norm.weight$6 : tensor<4096xf32>
%2070 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_norm.weight$7 = util.global.load @__auto.blk.28.ffn_norm.weight$7 : tensor<4096xf32>
%2071 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.28.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%2072 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%2073 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%2074 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%2075 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%2076 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%2077 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%2078 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.28.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%2079 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.0 = util.global.load @__auto.blk.28.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%2080 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.1 = util.global.load @__auto.blk.28.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%2081 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.2 = util.global.load @__auto.blk.28.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%2082 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.3 = util.global.load @__auto.blk.28.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%2083 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.4 = util.global.load @__auto.blk.28.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%2084 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.5 = util.global.load @__auto.blk.28.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%2085 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.6 = util.global.load @__auto.blk.28.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%2086 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_up.weight.shard.7 = util.global.load @__auto.blk.28.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%2087 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.28.ffn_down.weight.shard.0 = util.global.load @__auto.blk.28.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%2088 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.1 = util.global.load @__auto.blk.28.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%2089 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.2 = util.global.load @__auto.blk.28.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%2090 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.3 = util.global.load @__auto.blk.28.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%2091 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.4 = util.global.load @__auto.blk.28.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%2092 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.5 = util.global.load @__auto.blk.28.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%2093 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.6 = util.global.load @__auto.blk.28.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%2094 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.28.ffn_down.weight.shard.7 = util.global.load @__auto.blk.28.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%2095 = torch_c.from_builtin_tensor %__auto.blk.28.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.attn_norm.weight = util.global.load @__auto.blk.29.attn_norm.weight : tensor<4096xf32>
%2096 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$1 = util.global.load @__auto.blk.29.attn_norm.weight$1 : tensor<4096xf32>
%2097 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$2 = util.global.load @__auto.blk.29.attn_norm.weight$2 : tensor<4096xf32>
%2098 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$3 = util.global.load @__auto.blk.29.attn_norm.weight$3 : tensor<4096xf32>
%2099 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$4 = util.global.load @__auto.blk.29.attn_norm.weight$4 : tensor<4096xf32>
%2100 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$5 = util.global.load @__auto.blk.29.attn_norm.weight$5 : tensor<4096xf32>
%2101 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$6 = util.global.load @__auto.blk.29.attn_norm.weight$6 : tensor<4096xf32>
%2102 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_norm.weight$7 = util.global.load @__auto.blk.29.attn_norm.weight$7 : tensor<4096xf32>
%2103 = torch_c.from_builtin_tensor %__auto.blk.29.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.attn_q.weight.shard.0 = util.global.load @__auto.blk.29.attn_q.weight.shard.0 : tensor<512x4096xf16>
%2104 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.1 = util.global.load @__auto.blk.29.attn_q.weight.shard.1 : tensor<512x4096xf16>
%2105 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.2 = util.global.load @__auto.blk.29.attn_q.weight.shard.2 : tensor<512x4096xf16>
%2106 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.3 = util.global.load @__auto.blk.29.attn_q.weight.shard.3 : tensor<512x4096xf16>
%2107 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.4 = util.global.load @__auto.blk.29.attn_q.weight.shard.4 : tensor<512x4096xf16>
%2108 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.5 = util.global.load @__auto.blk.29.attn_q.weight.shard.5 : tensor<512x4096xf16>
%2109 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.6 = util.global.load @__auto.blk.29.attn_q.weight.shard.6 : tensor<512x4096xf16>
%2110 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_q.weight.shard.7 = util.global.load @__auto.blk.29.attn_q.weight.shard.7 : tensor<512x4096xf16>
%2111 = torch_c.from_builtin_tensor %__auto.blk.29.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.29.attn_k.weight.shard.0 = util.global.load @__auto.blk.29.attn_k.weight.shard.0 : tensor<128x4096xf16>
%2112 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.1 = util.global.load @__auto.blk.29.attn_k.weight.shard.1 : tensor<128x4096xf16>
%2113 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.2 = util.global.load @__auto.blk.29.attn_k.weight.shard.2 : tensor<128x4096xf16>
%2114 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.3 = util.global.load @__auto.blk.29.attn_k.weight.shard.3 : tensor<128x4096xf16>
%2115 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.4 = util.global.load @__auto.blk.29.attn_k.weight.shard.4 : tensor<128x4096xf16>
%2116 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.5 = util.global.load @__auto.blk.29.attn_k.weight.shard.5 : tensor<128x4096xf16>
%2117 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.6 = util.global.load @__auto.blk.29.attn_k.weight.shard.6 : tensor<128x4096xf16>
%2118 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_k.weight.shard.7 = util.global.load @__auto.blk.29.attn_k.weight.shard.7 : tensor<128x4096xf16>
%2119 = torch_c.from_builtin_tensor %__auto.blk.29.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.0 = util.global.load @__auto.blk.29.attn_v.weight.shard.0 : tensor<128x4096xf16>
%2120 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.1 = util.global.load @__auto.blk.29.attn_v.weight.shard.1 : tensor<128x4096xf16>
%2121 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.2 = util.global.load @__auto.blk.29.attn_v.weight.shard.2 : tensor<128x4096xf16>
%2122 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.3 = util.global.load @__auto.blk.29.attn_v.weight.shard.3 : tensor<128x4096xf16>
%2123 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.4 = util.global.load @__auto.blk.29.attn_v.weight.shard.4 : tensor<128x4096xf16>
%2124 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.5 = util.global.load @__auto.blk.29.attn_v.weight.shard.5 : tensor<128x4096xf16>
%2125 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.6 = util.global.load @__auto.blk.29.attn_v.weight.shard.6 : tensor<128x4096xf16>
%2126 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_v.weight.shard.7 = util.global.load @__auto.blk.29.attn_v.weight.shard.7 : tensor<128x4096xf16>
%2127 = torch_c.from_builtin_tensor %__auto.blk.29.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.29.attn_output.weight.shard.0 = util.global.load @__auto.blk.29.attn_output.weight.shard.0 : tensor<4096x512xf16>
%2128 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.1 = util.global.load @__auto.blk.29.attn_output.weight.shard.1 : tensor<4096x512xf16>
%2129 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.2 = util.global.load @__auto.blk.29.attn_output.weight.shard.2 : tensor<4096x512xf16>
%2130 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.3 = util.global.load @__auto.blk.29.attn_output.weight.shard.3 : tensor<4096x512xf16>
%2131 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.4 = util.global.load @__auto.blk.29.attn_output.weight.shard.4 : tensor<4096x512xf16>
%2132 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.5 = util.global.load @__auto.blk.29.attn_output.weight.shard.5 : tensor<4096x512xf16>
%2133 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.6 = util.global.load @__auto.blk.29.attn_output.weight.shard.6 : tensor<4096x512xf16>
%2134 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.attn_output.weight.shard.7 = util.global.load @__auto.blk.29.attn_output.weight.shard.7 : tensor<4096x512xf16>
%2135 = torch_c.from_builtin_tensor %__auto.blk.29.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.29.ffn_norm.weight = util.global.load @__auto.blk.29.ffn_norm.weight : tensor<4096xf32>
%2136 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$1 = util.global.load @__auto.blk.29.ffn_norm.weight$1 : tensor<4096xf32>
%2137 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$2 = util.global.load @__auto.blk.29.ffn_norm.weight$2 : tensor<4096xf32>
%2138 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$3 = util.global.load @__auto.blk.29.ffn_norm.weight$3 : tensor<4096xf32>
%2139 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$4 = util.global.load @__auto.blk.29.ffn_norm.weight$4 : tensor<4096xf32>
%2140 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$5 = util.global.load @__auto.blk.29.ffn_norm.weight$5 : tensor<4096xf32>
%2141 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$6 = util.global.load @__auto.blk.29.ffn_norm.weight$6 : tensor<4096xf32>
%2142 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_norm.weight$7 = util.global.load @__auto.blk.29.ffn_norm.weight$7 : tensor<4096xf32>
%2143 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.29.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%2144 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%2145 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%2146 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%2147 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%2148 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%2149 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%2150 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.29.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%2151 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.0 = util.global.load @__auto.blk.29.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%2152 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.1 = util.global.load @__auto.blk.29.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%2153 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.2 = util.global.load @__auto.blk.29.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%2154 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.3 = util.global.load @__auto.blk.29.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%2155 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.4 = util.global.load @__auto.blk.29.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%2156 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.5 = util.global.load @__auto.blk.29.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%2157 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.6 = util.global.load @__auto.blk.29.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%2158 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_up.weight.shard.7 = util.global.load @__auto.blk.29.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%2159 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.29.ffn_down.weight.shard.0 = util.global.load @__auto.blk.29.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%2160 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.1 = util.global.load @__auto.blk.29.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%2161 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.2 = util.global.load @__auto.blk.29.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%2162 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.3 = util.global.load @__auto.blk.29.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%2163 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.4 = util.global.load @__auto.blk.29.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%2164 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.5 = util.global.load @__auto.blk.29.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%2165 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.6 = util.global.load @__auto.blk.29.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%2166 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.29.ffn_down.weight.shard.7 = util.global.load @__auto.blk.29.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%2167 = torch_c.from_builtin_tensor %__auto.blk.29.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.attn_norm.weight = util.global.load @__auto.blk.30.attn_norm.weight : tensor<4096xf32>
%2168 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$1 = util.global.load @__auto.blk.30.attn_norm.weight$1 : tensor<4096xf32>
%2169 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$2 = util.global.load @__auto.blk.30.attn_norm.weight$2 : tensor<4096xf32>
%2170 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$3 = util.global.load @__auto.blk.30.attn_norm.weight$3 : tensor<4096xf32>
%2171 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$4 = util.global.load @__auto.blk.30.attn_norm.weight$4 : tensor<4096xf32>
%2172 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$5 = util.global.load @__auto.blk.30.attn_norm.weight$5 : tensor<4096xf32>
%2173 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$6 = util.global.load @__auto.blk.30.attn_norm.weight$6 : tensor<4096xf32>
%2174 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_norm.weight$7 = util.global.load @__auto.blk.30.attn_norm.weight$7 : tensor<4096xf32>
%2175 = torch_c.from_builtin_tensor %__auto.blk.30.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.attn_q.weight.shard.0 = util.global.load @__auto.blk.30.attn_q.weight.shard.0 : tensor<512x4096xf16>
%2176 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.1 = util.global.load @__auto.blk.30.attn_q.weight.shard.1 : tensor<512x4096xf16>
%2177 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.2 = util.global.load @__auto.blk.30.attn_q.weight.shard.2 : tensor<512x4096xf16>
%2178 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.3 = util.global.load @__auto.blk.30.attn_q.weight.shard.3 : tensor<512x4096xf16>
%2179 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.4 = util.global.load @__auto.blk.30.attn_q.weight.shard.4 : tensor<512x4096xf16>
%2180 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.5 = util.global.load @__auto.blk.30.attn_q.weight.shard.5 : tensor<512x4096xf16>
%2181 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.6 = util.global.load @__auto.blk.30.attn_q.weight.shard.6 : tensor<512x4096xf16>
%2182 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_q.weight.shard.7 = util.global.load @__auto.blk.30.attn_q.weight.shard.7 : tensor<512x4096xf16>
%2183 = torch_c.from_builtin_tensor %__auto.blk.30.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.30.attn_k.weight.shard.0 = util.global.load @__auto.blk.30.attn_k.weight.shard.0 : tensor<128x4096xf16>
%2184 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.1 = util.global.load @__auto.blk.30.attn_k.weight.shard.1 : tensor<128x4096xf16>
%2185 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.2 = util.global.load @__auto.blk.30.attn_k.weight.shard.2 : tensor<128x4096xf16>
%2186 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.3 = util.global.load @__auto.blk.30.attn_k.weight.shard.3 : tensor<128x4096xf16>
%2187 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.4 = util.global.load @__auto.blk.30.attn_k.weight.shard.4 : tensor<128x4096xf16>
%2188 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.5 = util.global.load @__auto.blk.30.attn_k.weight.shard.5 : tensor<128x4096xf16>
%2189 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.6 = util.global.load @__auto.blk.30.attn_k.weight.shard.6 : tensor<128x4096xf16>
%2190 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_k.weight.shard.7 = util.global.load @__auto.blk.30.attn_k.weight.shard.7 : tensor<128x4096xf16>
%2191 = torch_c.from_builtin_tensor %__auto.blk.30.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.0 = util.global.load @__auto.blk.30.attn_v.weight.shard.0 : tensor<128x4096xf16>
%2192 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.1 = util.global.load @__auto.blk.30.attn_v.weight.shard.1 : tensor<128x4096xf16>
%2193 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.2 = util.global.load @__auto.blk.30.attn_v.weight.shard.2 : tensor<128x4096xf16>
%2194 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.3 = util.global.load @__auto.blk.30.attn_v.weight.shard.3 : tensor<128x4096xf16>
%2195 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.4 = util.global.load @__auto.blk.30.attn_v.weight.shard.4 : tensor<128x4096xf16>
%2196 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.5 = util.global.load @__auto.blk.30.attn_v.weight.shard.5 : tensor<128x4096xf16>
%2197 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.6 = util.global.load @__auto.blk.30.attn_v.weight.shard.6 : tensor<128x4096xf16>
%2198 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_v.weight.shard.7 = util.global.load @__auto.blk.30.attn_v.weight.shard.7 : tensor<128x4096xf16>
%2199 = torch_c.from_builtin_tensor %__auto.blk.30.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.30.attn_output.weight.shard.0 = util.global.load @__auto.blk.30.attn_output.weight.shard.0 : tensor<4096x512xf16>
%2200 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.1 = util.global.load @__auto.blk.30.attn_output.weight.shard.1 : tensor<4096x512xf16>
%2201 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.2 = util.global.load @__auto.blk.30.attn_output.weight.shard.2 : tensor<4096x512xf16>
%2202 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.3 = util.global.load @__auto.blk.30.attn_output.weight.shard.3 : tensor<4096x512xf16>
%2203 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.4 = util.global.load @__auto.blk.30.attn_output.weight.shard.4 : tensor<4096x512xf16>
%2204 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.5 = util.global.load @__auto.blk.30.attn_output.weight.shard.5 : tensor<4096x512xf16>
%2205 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.6 = util.global.load @__auto.blk.30.attn_output.weight.shard.6 : tensor<4096x512xf16>
%2206 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.attn_output.weight.shard.7 = util.global.load @__auto.blk.30.attn_output.weight.shard.7 : tensor<4096x512xf16>
%2207 = torch_c.from_builtin_tensor %__auto.blk.30.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.30.ffn_norm.weight = util.global.load @__auto.blk.30.ffn_norm.weight : tensor<4096xf32>
%2208 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$1 = util.global.load @__auto.blk.30.ffn_norm.weight$1 : tensor<4096xf32>
%2209 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$2 = util.global.load @__auto.blk.30.ffn_norm.weight$2 : tensor<4096xf32>
%2210 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$3 = util.global.load @__auto.blk.30.ffn_norm.weight$3 : tensor<4096xf32>
%2211 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$4 = util.global.load @__auto.blk.30.ffn_norm.weight$4 : tensor<4096xf32>
%2212 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$5 = util.global.load @__auto.blk.30.ffn_norm.weight$5 : tensor<4096xf32>
%2213 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$6 = util.global.load @__auto.blk.30.ffn_norm.weight$6 : tensor<4096xf32>
%2214 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_norm.weight$7 = util.global.load @__auto.blk.30.ffn_norm.weight$7 : tensor<4096xf32>
%2215 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.30.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%2216 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%2217 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%2218 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%2219 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%2220 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%2221 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%2222 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.30.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%2223 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.0 = util.global.load @__auto.blk.30.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%2224 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.1 = util.global.load @__auto.blk.30.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%2225 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.2 = util.global.load @__auto.blk.30.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%2226 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.3 = util.global.load @__auto.blk.30.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%2227 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.4 = util.global.load @__auto.blk.30.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%2228 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.5 = util.global.load @__auto.blk.30.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%2229 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.6 = util.global.load @__auto.blk.30.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%2230 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_up.weight.shard.7 = util.global.load @__auto.blk.30.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%2231 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.30.ffn_down.weight.shard.0 = util.global.load @__auto.blk.30.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%2232 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.1 = util.global.load @__auto.blk.30.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%2233 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.2 = util.global.load @__auto.blk.30.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%2234 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.3 = util.global.load @__auto.blk.30.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%2235 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.4 = util.global.load @__auto.blk.30.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%2236 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.5 = util.global.load @__auto.blk.30.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%2237 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.6 = util.global.load @__auto.blk.30.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%2238 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.30.ffn_down.weight.shard.7 = util.global.load @__auto.blk.30.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%2239 = torch_c.from_builtin_tensor %__auto.blk.30.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.attn_norm.weight = util.global.load @__auto.blk.31.attn_norm.weight : tensor<4096xf32>
%2240 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$1 = util.global.load @__auto.blk.31.attn_norm.weight$1 : tensor<4096xf32>
%2241 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$2 = util.global.load @__auto.blk.31.attn_norm.weight$2 : tensor<4096xf32>
%2242 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$3 = util.global.load @__auto.blk.31.attn_norm.weight$3 : tensor<4096xf32>
%2243 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$4 = util.global.load @__auto.blk.31.attn_norm.weight$4 : tensor<4096xf32>
%2244 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$5 = util.global.load @__auto.blk.31.attn_norm.weight$5 : tensor<4096xf32>
%2245 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$6 = util.global.load @__auto.blk.31.attn_norm.weight$6 : tensor<4096xf32>
%2246 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_norm.weight$7 = util.global.load @__auto.blk.31.attn_norm.weight$7 : tensor<4096xf32>
%2247 = torch_c.from_builtin_tensor %__auto.blk.31.attn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.attn_q.weight.shard.0 = util.global.load @__auto.blk.31.attn_q.weight.shard.0 : tensor<512x4096xf16>
%2248 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.0 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.1 = util.global.load @__auto.blk.31.attn_q.weight.shard.1 : tensor<512x4096xf16>
%2249 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.1 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.2 = util.global.load @__auto.blk.31.attn_q.weight.shard.2 : tensor<512x4096xf16>
%2250 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.2 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.3 = util.global.load @__auto.blk.31.attn_q.weight.shard.3 : tensor<512x4096xf16>
%2251 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.3 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.4 = util.global.load @__auto.blk.31.attn_q.weight.shard.4 : tensor<512x4096xf16>
%2252 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.4 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.5 = util.global.load @__auto.blk.31.attn_q.weight.shard.5 : tensor<512x4096xf16>
%2253 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.5 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.6 = util.global.load @__auto.blk.31.attn_q.weight.shard.6 : tensor<512x4096xf16>
%2254 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.6 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_q.weight.shard.7 = util.global.load @__auto.blk.31.attn_q.weight.shard.7 : tensor<512x4096xf16>
%2255 = torch_c.from_builtin_tensor %__auto.blk.31.attn_q.weight.shard.7 : tensor<512x4096xf16> -> !torch.vtensor<[512,4096],f16>
%__auto.blk.31.attn_k.weight.shard.0 = util.global.load @__auto.blk.31.attn_k.weight.shard.0 : tensor<128x4096xf16>
%2256 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.1 = util.global.load @__auto.blk.31.attn_k.weight.shard.1 : tensor<128x4096xf16>
%2257 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.2 = util.global.load @__auto.blk.31.attn_k.weight.shard.2 : tensor<128x4096xf16>
%2258 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.3 = util.global.load @__auto.blk.31.attn_k.weight.shard.3 : tensor<128x4096xf16>
%2259 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.4 = util.global.load @__auto.blk.31.attn_k.weight.shard.4 : tensor<128x4096xf16>
%2260 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.5 = util.global.load @__auto.blk.31.attn_k.weight.shard.5 : tensor<128x4096xf16>
%2261 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.6 = util.global.load @__auto.blk.31.attn_k.weight.shard.6 : tensor<128x4096xf16>
%2262 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_k.weight.shard.7 = util.global.load @__auto.blk.31.attn_k.weight.shard.7 : tensor<128x4096xf16>
%2263 = torch_c.from_builtin_tensor %__auto.blk.31.attn_k.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.0 = util.global.load @__auto.blk.31.attn_v.weight.shard.0 : tensor<128x4096xf16>
%2264 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.0 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.1 = util.global.load @__auto.blk.31.attn_v.weight.shard.1 : tensor<128x4096xf16>
%2265 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.1 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.2 = util.global.load @__auto.blk.31.attn_v.weight.shard.2 : tensor<128x4096xf16>
%2266 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.2 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.3 = util.global.load @__auto.blk.31.attn_v.weight.shard.3 : tensor<128x4096xf16>
%2267 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.3 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.4 = util.global.load @__auto.blk.31.attn_v.weight.shard.4 : tensor<128x4096xf16>
%2268 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.4 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.5 = util.global.load @__auto.blk.31.attn_v.weight.shard.5 : tensor<128x4096xf16>
%2269 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.5 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.6 = util.global.load @__auto.blk.31.attn_v.weight.shard.6 : tensor<128x4096xf16>
%2270 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.6 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_v.weight.shard.7 = util.global.load @__auto.blk.31.attn_v.weight.shard.7 : tensor<128x4096xf16>
%2271 = torch_c.from_builtin_tensor %__auto.blk.31.attn_v.weight.shard.7 : tensor<128x4096xf16> -> !torch.vtensor<[128,4096],f16>
%__auto.blk.31.attn_output.weight.shard.0 = util.global.load @__auto.blk.31.attn_output.weight.shard.0 : tensor<4096x512xf16>
%2272 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.0 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.1 = util.global.load @__auto.blk.31.attn_output.weight.shard.1 : tensor<4096x512xf16>
%2273 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.1 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.2 = util.global.load @__auto.blk.31.attn_output.weight.shard.2 : tensor<4096x512xf16>
%2274 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.2 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.3 = util.global.load @__auto.blk.31.attn_output.weight.shard.3 : tensor<4096x512xf16>
%2275 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.3 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.4 = util.global.load @__auto.blk.31.attn_output.weight.shard.4 : tensor<4096x512xf16>
%2276 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.4 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.5 = util.global.load @__auto.blk.31.attn_output.weight.shard.5 : tensor<4096x512xf16>
%2277 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.5 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.6 = util.global.load @__auto.blk.31.attn_output.weight.shard.6 : tensor<4096x512xf16>
%2278 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.6 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.attn_output.weight.shard.7 = util.global.load @__auto.blk.31.attn_output.weight.shard.7 : tensor<4096x512xf16>
%2279 = torch_c.from_builtin_tensor %__auto.blk.31.attn_output.weight.shard.7 : tensor<4096x512xf16> -> !torch.vtensor<[4096,512],f16>
%__auto.blk.31.ffn_norm.weight = util.global.load @__auto.blk.31.ffn_norm.weight : tensor<4096xf32>
%2280 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$1 = util.global.load @__auto.blk.31.ffn_norm.weight$1 : tensor<4096xf32>
%2281 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$2 = util.global.load @__auto.blk.31.ffn_norm.weight$2 : tensor<4096xf32>
%2282 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$3 = util.global.load @__auto.blk.31.ffn_norm.weight$3 : tensor<4096xf32>
%2283 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$4 = util.global.load @__auto.blk.31.ffn_norm.weight$4 : tensor<4096xf32>
%2284 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$5 = util.global.load @__auto.blk.31.ffn_norm.weight$5 : tensor<4096xf32>
%2285 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$6 = util.global.load @__auto.blk.31.ffn_norm.weight$6 : tensor<4096xf32>
%2286 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_norm.weight$7 = util.global.load @__auto.blk.31.ffn_norm.weight$7 : tensor<4096xf32>
%2287 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.blk.31.ffn_gate.weight.shard.0 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.0 : tensor<1792x4096xf16>
%2288 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.1 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.1 : tensor<1792x4096xf16>
%2289 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.2 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.2 : tensor<1792x4096xf16>
%2290 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.3 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.3 : tensor<1792x4096xf16>
%2291 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.4 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.4 : tensor<1792x4096xf16>
%2292 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.5 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.5 : tensor<1792x4096xf16>
%2293 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.6 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.6 : tensor<1792x4096xf16>
%2294 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_gate.weight.shard.7 = util.global.load @__auto.blk.31.ffn_gate.weight.shard.7 : tensor<1792x4096xf16>
%2295 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_gate.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.0 = util.global.load @__auto.blk.31.ffn_up.weight.shard.0 : tensor<1792x4096xf16>
%2296 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.0 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.1 = util.global.load @__auto.blk.31.ffn_up.weight.shard.1 : tensor<1792x4096xf16>
%2297 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.1 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.2 = util.global.load @__auto.blk.31.ffn_up.weight.shard.2 : tensor<1792x4096xf16>
%2298 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.2 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.3 = util.global.load @__auto.blk.31.ffn_up.weight.shard.3 : tensor<1792x4096xf16>
%2299 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.3 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.4 = util.global.load @__auto.blk.31.ffn_up.weight.shard.4 : tensor<1792x4096xf16>
%2300 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.4 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.5 = util.global.load @__auto.blk.31.ffn_up.weight.shard.5 : tensor<1792x4096xf16>
%2301 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.5 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.6 = util.global.load @__auto.blk.31.ffn_up.weight.shard.6 : tensor<1792x4096xf16>
%2302 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.6 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_up.weight.shard.7 = util.global.load @__auto.blk.31.ffn_up.weight.shard.7 : tensor<1792x4096xf16>
%2303 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_up.weight.shard.7 : tensor<1792x4096xf16> -> !torch.vtensor<[1792,4096],f16>
%__auto.blk.31.ffn_down.weight.shard.0 = util.global.load @__auto.blk.31.ffn_down.weight.shard.0 : tensor<4096x1792xf16>
%2304 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.0 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.1 = util.global.load @__auto.blk.31.ffn_down.weight.shard.1 : tensor<4096x1792xf16>
%2305 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.1 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.2 = util.global.load @__auto.blk.31.ffn_down.weight.shard.2 : tensor<4096x1792xf16>
%2306 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.2 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.3 = util.global.load @__auto.blk.31.ffn_down.weight.shard.3 : tensor<4096x1792xf16>
%2307 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.3 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.4 = util.global.load @__auto.blk.31.ffn_down.weight.shard.4 : tensor<4096x1792xf16>
%2308 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.4 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.5 = util.global.load @__auto.blk.31.ffn_down.weight.shard.5 : tensor<4096x1792xf16>
%2309 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.5 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.6 = util.global.load @__auto.blk.31.ffn_down.weight.shard.6 : tensor<4096x1792xf16>
%2310 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.6 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.blk.31.ffn_down.weight.shard.7 = util.global.load @__auto.blk.31.ffn_down.weight.shard.7 : tensor<4096x1792xf16>
%2311 = torch_c.from_builtin_tensor %__auto.blk.31.ffn_down.weight.shard.7 : tensor<4096x1792xf16> -> !torch.vtensor<[4096,1792],f16>
%__auto.output_norm.weight = util.global.load @__auto.output_norm.weight : tensor<4096xf32>
%2312 = torch_c.from_builtin_tensor %__auto.output_norm.weight : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$1 = util.global.load @__auto.output_norm.weight$1 : tensor<4096xf32>
%2313 = torch_c.from_builtin_tensor %__auto.output_norm.weight$1 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$2 = util.global.load @__auto.output_norm.weight$2 : tensor<4096xf32>
%2314 = torch_c.from_builtin_tensor %__auto.output_norm.weight$2 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$3 = util.global.load @__auto.output_norm.weight$3 : tensor<4096xf32>
%2315 = torch_c.from_builtin_tensor %__auto.output_norm.weight$3 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$4 = util.global.load @__auto.output_norm.weight$4 : tensor<4096xf32>
%2316 = torch_c.from_builtin_tensor %__auto.output_norm.weight$4 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$5 = util.global.load @__auto.output_norm.weight$5 : tensor<4096xf32>
%2317 = torch_c.from_builtin_tensor %__auto.output_norm.weight$5 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$6 = util.global.load @__auto.output_norm.weight$6 : tensor<4096xf32>
%2318 = torch_c.from_builtin_tensor %__auto.output_norm.weight$6 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output_norm.weight$7 = util.global.load @__auto.output_norm.weight$7 : tensor<4096xf32>
%2319 = torch_c.from_builtin_tensor %__auto.output_norm.weight$7 : tensor<4096xf32> -> !torch.vtensor<[4096],f32>
%__auto.output.weight.shard.0 = util.global.load @__auto.output.weight.shard.0 : tensor<128256x512xf16>
%2320 = torch_c.from_builtin_tensor %__auto.output.weight.shard.0 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.1 = util.global.load @__auto.output.weight.shard.1 : tensor<128256x512xf16>
%2321 = torch_c.from_builtin_tensor %__auto.output.weight.shard.1 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.2 = util.global.load @__auto.output.weight.shard.2 : tensor<128256x512xf16>
%2322 = torch_c.from_builtin_tensor %__auto.output.weight.shard.2 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.3 = util.global.load @__auto.output.weight.shard.3 : tensor<128256x512xf16>
%2323 = torch_c.from_builtin_tensor %__auto.output.weight.shard.3 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.4 = util.global.load @__auto.output.weight.shard.4 : tensor<128256x512xf16>
%2324 = torch_c.from_builtin_tensor %__auto.output.weight.shard.4 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.5 = util.global.load @__auto.output.weight.shard.5 : tensor<128256x512xf16>
%2325 = torch_c.from_builtin_tensor %__auto.output.weight.shard.5 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.6 = util.global.load @__auto.output.weight.shard.6 : tensor<128256x512xf16>
%2326 = torch_c.from_builtin_tensor %__auto.output.weight.shard.6 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%__auto.output.weight.shard.7 = util.global.load @__auto.output.weight.shard.7 : tensor<128256x512xf16>
%2327 = torch_c.from_builtin_tensor %__auto.output.weight.shard.7 : tensor<128256x512xf16> -> !torch.vtensor<[128256,512],f16>
%2328 = torch.copy.to_vtensor %arg3 : !torch.vtensor<[?,131072],f16>
%2329 = torch.copy.to_vtensor %arg4 : !torch.vtensor<[?,131072],f16>
%2330 = torch.copy.to_vtensor %arg5 : !torch.vtensor<[?,131072],f16>
%2331 = torch.copy.to_vtensor %arg6 : !torch.vtensor<[?,131072],f16>
%2332 = torch.copy.to_vtensor %arg7 : !torch.vtensor<[?,131072],f16>
%2333 = torch.copy.to_vtensor %arg8 : !torch.vtensor<[?,131072],f16>
%2334 = torch.copy.to_vtensor %arg9 : !torch.vtensor<[?,131072],f16>
%2335 = torch.copy.to_vtensor %arg10 : !torch.vtensor<[?,131072],f16>
%2336 = torch.symbolic_int "s1" {min_val = 2, max_val = 8191} : !torch.int
%2337 = torch.symbolic_int "s2" {min_val = 2, max_val = 9223372036854775806} : !torch.int
torch.bind_symbolic_shape %arg0, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %arg2, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2328, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2329, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2330, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2331, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2332, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2333, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2334, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %2335, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%2338 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1 = arith.constant 1 : index
%dim = tensor.dim %2338, %c1 : tensor<4x?xi64>
%2339 = flow.tensor.transfer %2338 : tensor<4x?xi64>{%dim} to #hal.device.promise<@__device_0>
%2340 = torch_c.from_builtin_tensor %2339 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2340, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2341 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_0 = arith.constant 1 : index
%dim_1 = tensor.dim %2341, %c1_0 : tensor<4x?xi64>
%2342 = flow.tensor.transfer %2341 : tensor<4x?xi64>{%dim_1} to #hal.device.promise<@__device_1>
%2343 = torch_c.from_builtin_tensor %2342 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2343, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2344 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_2 = arith.constant 1 : index
%dim_3 = tensor.dim %2344, %c1_2 : tensor<4x?xi64>
%2345 = flow.tensor.transfer %2344 : tensor<4x?xi64>{%dim_3} to #hal.device.promise<@__device_2>
%2346 = torch_c.from_builtin_tensor %2345 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2346, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2347 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_4 = arith.constant 1 : index
%dim_5 = tensor.dim %2347, %c1_4 : tensor<4x?xi64>
%2348 = flow.tensor.transfer %2347 : tensor<4x?xi64>{%dim_5} to #hal.device.promise<@__device_3>
%2349 = torch_c.from_builtin_tensor %2348 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2349, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2350 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_6 = arith.constant 1 : index
%dim_7 = tensor.dim %2350, %c1_6 : tensor<4x?xi64>
%2351 = flow.tensor.transfer %2350 : tensor<4x?xi64>{%dim_7} to #hal.device.promise<@__device_4>
%2352 = torch_c.from_builtin_tensor %2351 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2352, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2353 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_8 = arith.constant 1 : index
%dim_9 = tensor.dim %2353, %c1_8 : tensor<4x?xi64>
%2354 = flow.tensor.transfer %2353 : tensor<4x?xi64>{%dim_9} to #hal.device.promise<@__device_5>
%2355 = torch_c.from_builtin_tensor %2354 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2355, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2356 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_10 = arith.constant 1 : index
%dim_11 = tensor.dim %2356, %c1_10 : tensor<4x?xi64>
%2357 = flow.tensor.transfer %2356 : tensor<4x?xi64>{%dim_11} to #hal.device.promise<@__device_6>
%2358 = torch_c.from_builtin_tensor %2357 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2358, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2359 = torch_c.to_builtin_tensor %arg0 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_12 = arith.constant 1 : index
%dim_13 = tensor.dim %2359, %c1_12 : tensor<4x?xi64>
%2360 = flow.tensor.transfer %2359 : tensor<4x?xi64>{%dim_13} to #hal.device.promise<@__device_7>
%2361 = torch_c.from_builtin_tensor %2360 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2361, [%2336], affine_map<()[s0] -> (4, s0 * 16)> : !torch.vtensor<[4,?],si64>
%2362 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_14 = arith.constant 1 : index
%dim_15 = tensor.dim %2362, %c1_14 : tensor<4x?xi64>
%2363 = flow.tensor.transfer %2362 : tensor<4x?xi64>{%dim_15} to #hal.device.promise<@__device_0>
%2364 = torch_c.from_builtin_tensor %2363 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2364, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2365 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_16 = arith.constant 1 : index
%dim_17 = tensor.dim %2365, %c1_16 : tensor<4x?xi64>
%2366 = flow.tensor.transfer %2365 : tensor<4x?xi64>{%dim_17} to #hal.device.promise<@__device_1>
%2367 = torch_c.from_builtin_tensor %2366 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2367, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2368 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_18 = arith.constant 1 : index
%dim_19 = tensor.dim %2368, %c1_18 : tensor<4x?xi64>
%2369 = flow.tensor.transfer %2368 : tensor<4x?xi64>{%dim_19} to #hal.device.promise<@__device_2>
%2370 = torch_c.from_builtin_tensor %2369 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2370, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2371 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_20 = arith.constant 1 : index
%dim_21 = tensor.dim %2371, %c1_20 : tensor<4x?xi64>
%2372 = flow.tensor.transfer %2371 : tensor<4x?xi64>{%dim_21} to #hal.device.promise<@__device_3>
%2373 = torch_c.from_builtin_tensor %2372 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2373, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2374 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_22 = arith.constant 1 : index
%dim_23 = tensor.dim %2374, %c1_22 : tensor<4x?xi64>
%2375 = flow.tensor.transfer %2374 : tensor<4x?xi64>{%dim_23} to #hal.device.promise<@__device_4>
%2376 = torch_c.from_builtin_tensor %2375 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2376, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2377 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_24 = arith.constant 1 : index
%dim_25 = tensor.dim %2377, %c1_24 : tensor<4x?xi64>
%2378 = flow.tensor.transfer %2377 : tensor<4x?xi64>{%dim_25} to #hal.device.promise<@__device_5>
%2379 = torch_c.from_builtin_tensor %2378 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2379, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2380 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_26 = arith.constant 1 : index
%dim_27 = tensor.dim %2380, %c1_26 : tensor<4x?xi64>
%2381 = flow.tensor.transfer %2380 : tensor<4x?xi64>{%dim_27} to #hal.device.promise<@__device_6>
%2382 = torch_c.from_builtin_tensor %2381 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2382, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%2383 = torch_c.to_builtin_tensor %arg2 : !torch.vtensor<[4,?],si64> -> tensor<4x?xi64>
%c1_28 = arith.constant 1 : index
%dim_29 = tensor.dim %2383, %c1_28 : tensor<4x?xi64>
%2384 = flow.tensor.transfer %2383 : tensor<4x?xi64>{%dim_29} to #hal.device.promise<@__device_7>
%2385 = torch_c.from_builtin_tensor %2384 : tensor<4x?xi64> -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %2385, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int-1 = torch.constant.int -1
%false = torch.constant.bool false
%false_30 = torch.constant.bool false
%2386 = torch.aten.embedding %0, %2340, %int-1, %false, %false_30 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_31 = torch.constant.int -1
%false_32 = torch.constant.bool false
%false_33 = torch.constant.bool false
%2387 = torch.aten.embedding %1, %2343, %int-1_31, %false_32, %false_33 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_34 = torch.constant.int -1
%false_35 = torch.constant.bool false
%false_36 = torch.constant.bool false
%2388 = torch.aten.embedding %2, %2346, %int-1_34, %false_35, %false_36 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_37 = torch.constant.int -1
%false_38 = torch.constant.bool false
%false_39 = torch.constant.bool false
%2389 = torch.aten.embedding %3, %2349, %int-1_37, %false_38, %false_39 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_40 = torch.constant.int -1
%false_41 = torch.constant.bool false
%false_42 = torch.constant.bool false
%2390 = torch.aten.embedding %4, %2352, %int-1_40, %false_41, %false_42 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_43 = torch.constant.int -1
%false_44 = torch.constant.bool false
%false_45 = torch.constant.bool false
%2391 = torch.aten.embedding %5, %2355, %int-1_43, %false_44, %false_45 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_46 = torch.constant.int -1
%false_47 = torch.constant.bool false
%false_48 = torch.constant.bool false
%2392 = torch.aten.embedding %6, %2358, %int-1_46, %false_47, %false_48 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int-1_49 = torch.constant.int -1
%false_50 = torch.constant.bool false
%false_51 = torch.constant.bool false
%2393 = torch.aten.embedding %7, %2361, %int-1_49, %false_50, %false_51 : !torch.vtensor<[128256,4096],f16>, !torch.vtensor<[4,?],si64>, !torch.int, !torch.bool, !torch.bool -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6 = torch.constant.int 6
%2394 = torch.prims.convert_element_type %2386, %int6 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_52 = torch.constant.int 6
%2395 = torch.prims.convert_element_type %2387, %int6_52 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_53 = torch.constant.int 6
%2396 = torch.prims.convert_element_type %2388, %int6_53 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_54 = torch.constant.int 6
%2397 = torch.prims.convert_element_type %2389, %int6_54 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_55 = torch.constant.int 6
%2398 = torch.prims.convert_element_type %2390, %int6_55 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_56 = torch.constant.int 6
%2399 = torch.prims.convert_element_type %2391, %int6_56 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_57 = torch.constant.int 6
%2400 = torch.prims.convert_element_type %2392, %int6_57 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_58 = torch.constant.int 6
%2401 = torch.prims.convert_element_type %2393, %int6_58 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2 = torch.constant.int 2
%2402 = torch.aten.pow.Tensor_Scalar %2394, %int2 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2402, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_59 = torch.constant.int 2
%2403 = torch.aten.pow.Tensor_Scalar %2395, %int2_59 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_60 = torch.constant.int 2
%2404 = torch.aten.pow.Tensor_Scalar %2396, %int2_60 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_61 = torch.constant.int 2
%2405 = torch.aten.pow.Tensor_Scalar %2397, %int2_61 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_62 = torch.constant.int 2
%2406 = torch.aten.pow.Tensor_Scalar %2398, %int2_62 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_63 = torch.constant.int 2
%2407 = torch.aten.pow.Tensor_Scalar %2399, %int2_63 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_64 = torch.constant.int 2
%2408 = torch.aten.pow.Tensor_Scalar %2400, %int2_64 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_65 = torch.constant.int 2
%2409 = torch.aten.pow.Tensor_Scalar %2401, %int2_65 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_66 = torch.constant.int -1
%2410 = torch.prim.ListConstruct %int-1_66 : (!torch.int) -> !torch.list<int>
%true = torch.constant.bool true
%none = torch.constant.none
%2411 = torch.aten.mean.dim %2402, %2410, %true, %none : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_67 = torch.constant.int -1
%2412 = torch.prim.ListConstruct %int-1_67 : (!torch.int) -> !torch.list<int>
%true_68 = torch.constant.bool true
%none_69 = torch.constant.none
%2413 = torch.aten.mean.dim %2403, %2412, %true_68, %none_69 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_70 = torch.constant.int -1
%2414 = torch.prim.ListConstruct %int-1_70 : (!torch.int) -> !torch.list<int>
%true_71 = torch.constant.bool true
%none_72 = torch.constant.none
%2415 = torch.aten.mean.dim %2404, %2414, %true_71, %none_72 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_73 = torch.constant.int -1
%2416 = torch.prim.ListConstruct %int-1_73 : (!torch.int) -> !torch.list<int>
%true_74 = torch.constant.bool true
%none_75 = torch.constant.none
%2417 = torch.aten.mean.dim %2405, %2416, %true_74, %none_75 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_76 = torch.constant.int -1
%2418 = torch.prim.ListConstruct %int-1_76 : (!torch.int) -> !torch.list<int>
%true_77 = torch.constant.bool true
%none_78 = torch.constant.none
%2419 = torch.aten.mean.dim %2406, %2418, %true_77, %none_78 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_79 = torch.constant.int -1
%2420 = torch.prim.ListConstruct %int-1_79 : (!torch.int) -> !torch.list<int>
%true_80 = torch.constant.bool true
%none_81 = torch.constant.none
%2421 = torch.aten.mean.dim %2407, %2420, %true_80, %none_81 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_82 = torch.constant.int -1
%2422 = torch.prim.ListConstruct %int-1_82 : (!torch.int) -> !torch.list<int>
%true_83 = torch.constant.bool true
%none_84 = torch.constant.none
%2423 = torch.aten.mean.dim %2408, %2422, %true_83, %none_84 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_85 = torch.constant.int -1
%2424 = torch.prim.ListConstruct %int-1_85 : (!torch.int) -> !torch.list<int>
%true_86 = torch.constant.bool true
%none_87 = torch.constant.none
%2425 = torch.aten.mean.dim %2409, %2424, %true_86, %none_87 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06 = torch.constant.float 9.9999997473787516E-6
%int1 = torch.constant.int 1
%2426 = torch.aten.add.Scalar %2411, %float9.999990e-06, %int1 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_88 = torch.constant.float 9.9999997473787516E-6
%int1_89 = torch.constant.int 1
%2427 = torch.aten.add.Scalar %2413, %float9.999990e-06_88, %int1_89 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_90 = torch.constant.float 9.9999997473787516E-6
%int1_91 = torch.constant.int 1
%2428 = torch.aten.add.Scalar %2415, %float9.999990e-06_90, %int1_91 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_92 = torch.constant.float 9.9999997473787516E-6
%int1_93 = torch.constant.int 1
%2429 = torch.aten.add.Scalar %2417, %float9.999990e-06_92, %int1_93 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_94 = torch.constant.float 9.9999997473787516E-6
%int1_95 = torch.constant.int 1
%2430 = torch.aten.add.Scalar %2419, %float9.999990e-06_94, %int1_95 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_96 = torch.constant.float 9.9999997473787516E-6
%int1_97 = torch.constant.int 1
%2431 = torch.aten.add.Scalar %2421, %float9.999990e-06_96, %int1_97 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_98 = torch.constant.float 9.9999997473787516E-6
%int1_99 = torch.constant.int 1
%2432 = torch.aten.add.Scalar %2423, %float9.999990e-06_98, %int1_99 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_100 = torch.constant.float 9.9999997473787516E-6
%int1_101 = torch.constant.int 1
%2433 = torch.aten.add.Scalar %2425, %float9.999990e-06_100, %int1_101 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2434 = torch.aten.rsqrt %2426 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2435 = torch.aten.rsqrt %2427 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2436 = torch.aten.rsqrt %2428 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2437 = torch.aten.rsqrt %2429 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2438 = torch.aten.rsqrt %2430 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2439 = torch.aten.rsqrt %2431 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2440 = torch.aten.rsqrt %2432 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2441 = torch.aten.rsqrt %2433 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %2441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%2442 = torch.aten.mul.Tensor %2394, %2434 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2443 = torch.aten.mul.Tensor %2395, %2435 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2444 = torch.aten.mul.Tensor %2396, %2436 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2445 = torch.aten.mul.Tensor %2397, %2437 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2446 = torch.aten.mul.Tensor %2398, %2438 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2447 = torch.aten.mul.Tensor %2399, %2439 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2448 = torch.aten.mul.Tensor %2400, %2440 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2449 = torch.aten.mul.Tensor %2401, %2441 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2450 = torch.aten.mul.Tensor %8, %2442 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2451 = torch.aten.mul.Tensor %9, %2443 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2452 = torch.aten.mul.Tensor %10, %2444 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2453 = torch.aten.mul.Tensor %11, %2445 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2454 = torch.aten.mul.Tensor %12, %2446 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2455 = torch.aten.mul.Tensor %13, %2447 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2456 = torch.aten.mul.Tensor %14, %2448 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%2457 = torch.aten.mul.Tensor %15, %2449 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %2457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5 = torch.constant.int 5
%2458 = torch.prims.convert_element_type %2450, %int5 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_102 = torch.constant.int 5
%2459 = torch.prims.convert_element_type %2451, %int5_102 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_103 = torch.constant.int 5
%2460 = torch.prims.convert_element_type %2452, %int5_103 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_104 = torch.constant.int 5
%2461 = torch.prims.convert_element_type %2453, %int5_104 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_105 = torch.constant.int 5
%2462 = torch.prims.convert_element_type %2454, %int5_105 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_106 = torch.constant.int 5
%2463 = torch.prims.convert_element_type %2455, %int5_106 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_107 = torch.constant.int 5
%2464 = torch.prims.convert_element_type %2456, %int5_107 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_108 = torch.constant.int 5
%2465 = torch.prims.convert_element_type %2457, %int5_108 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %2465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_109 = torch.constant.int 1
%int0 = torch.constant.int 0
%2466 = torch.prim.ListConstruct %int1_109, %int0 : (!torch.int, !torch.int) -> !torch.list<int>
%2467 = torch.aten.permute %16, %2466 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_110 = torch.constant.int 1
%int0_111 = torch.constant.int 0
%2468 = torch.prim.ListConstruct %int1_110, %int0_111 : (!torch.int, !torch.int) -> !torch.list<int>
%2469 = torch.aten.permute %17, %2468 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_112 = torch.constant.int 1
%int0_113 = torch.constant.int 0
%2470 = torch.prim.ListConstruct %int1_112, %int0_113 : (!torch.int, !torch.int) -> !torch.list<int>
%2471 = torch.aten.permute %18, %2470 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_114 = torch.constant.int 1
%int0_115 = torch.constant.int 0
%2472 = torch.prim.ListConstruct %int1_114, %int0_115 : (!torch.int, !torch.int) -> !torch.list<int>
%2473 = torch.aten.permute %19, %2472 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_116 = torch.constant.int 1
%int0_117 = torch.constant.int 0
%2474 = torch.prim.ListConstruct %int1_116, %int0_117 : (!torch.int, !torch.int) -> !torch.list<int>
%2475 = torch.aten.permute %20, %2474 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_118 = torch.constant.int 1
%int0_119 = torch.constant.int 0
%2476 = torch.prim.ListConstruct %int1_118, %int0_119 : (!torch.int, !torch.int) -> !torch.list<int>
%2477 = torch.aten.permute %21, %2476 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_120 = torch.constant.int 1
%int0_121 = torch.constant.int 0
%2478 = torch.prim.ListConstruct %int1_120, %int0_121 : (!torch.int, !torch.int) -> !torch.list<int>
%2479 = torch.aten.permute %22, %2478 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_122 = torch.constant.int 1
%int0_123 = torch.constant.int 0
%2480 = torch.prim.ListConstruct %int1_122, %int0_123 : (!torch.int, !torch.int) -> !torch.list<int>
%2481 = torch.aten.permute %23, %2480 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_124 = torch.constant.int 1
%2482 = torch.aten.size.int %arg0, %int1_124 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.int
%int4 = torch.constant.int 4
%2483 = torch.aten.mul.int %int4, %2482 : !torch.int, !torch.int -> !torch.int
%int4096 = torch.constant.int 4096
%2484 = torch.prim.ListConstruct %2483, %int4096 : (!torch.int, !torch.int) -> !torch.list<int>
%2485 = torch.aten.view %2458, %2484 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2485, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2486 = torch.aten.mm %2485, %2467 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2486, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_125 = torch.constant.int 4
%int512 = torch.constant.int 512
%2487 = torch.prim.ListConstruct %int4_125, %2482, %int512 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2488 = torch.aten.view %2486, %2487 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_126 = torch.constant.int 4
%2489 = torch.aten.mul.int %int4_126, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_127 = torch.constant.int 4096
%2490 = torch.prim.ListConstruct %2489, %int4096_127 : (!torch.int, !torch.int) -> !torch.list<int>
%2491 = torch.aten.view %2459, %2490 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2491, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2492 = torch.aten.mm %2491, %2469 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2492, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_128 = torch.constant.int 4
%int512_129 = torch.constant.int 512
%2493 = torch.prim.ListConstruct %int4_128, %2482, %int512_129 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2494 = torch.aten.view %2492, %2493 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_130 = torch.constant.int 4
%2495 = torch.aten.mul.int %int4_130, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_131 = torch.constant.int 4096
%2496 = torch.prim.ListConstruct %2495, %int4096_131 : (!torch.int, !torch.int) -> !torch.list<int>
%2497 = torch.aten.view %2460, %2496 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2497, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2498 = torch.aten.mm %2497, %2471 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2498, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_132 = torch.constant.int 4
%int512_133 = torch.constant.int 512
%2499 = torch.prim.ListConstruct %int4_132, %2482, %int512_133 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2500 = torch.aten.view %2498, %2499 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_134 = torch.constant.int 4
%2501 = torch.aten.mul.int %int4_134, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_135 = torch.constant.int 4096
%2502 = torch.prim.ListConstruct %2501, %int4096_135 : (!torch.int, !torch.int) -> !torch.list<int>
%2503 = torch.aten.view %2461, %2502 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2503, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2504 = torch.aten.mm %2503, %2473 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2504, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_136 = torch.constant.int 4
%int512_137 = torch.constant.int 512
%2505 = torch.prim.ListConstruct %int4_136, %2482, %int512_137 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2506 = torch.aten.view %2504, %2505 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_138 = torch.constant.int 4
%2507 = torch.aten.mul.int %int4_138, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_139 = torch.constant.int 4096
%2508 = torch.prim.ListConstruct %2507, %int4096_139 : (!torch.int, !torch.int) -> !torch.list<int>
%2509 = torch.aten.view %2462, %2508 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2509, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2510 = torch.aten.mm %2509, %2475 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2510, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_140 = torch.constant.int 4
%int512_141 = torch.constant.int 512
%2511 = torch.prim.ListConstruct %int4_140, %2482, %int512_141 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2512 = torch.aten.view %2510, %2511 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_142 = torch.constant.int 4
%2513 = torch.aten.mul.int %int4_142, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_143 = torch.constant.int 4096
%2514 = torch.prim.ListConstruct %2513, %int4096_143 : (!torch.int, !torch.int) -> !torch.list<int>
%2515 = torch.aten.view %2463, %2514 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2515, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2516 = torch.aten.mm %2515, %2477 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2516, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_144 = torch.constant.int 4
%int512_145 = torch.constant.int 512
%2517 = torch.prim.ListConstruct %int4_144, %2482, %int512_145 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2518 = torch.aten.view %2516, %2517 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_146 = torch.constant.int 4
%2519 = torch.aten.mul.int %int4_146, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_147 = torch.constant.int 4096
%2520 = torch.prim.ListConstruct %2519, %int4096_147 : (!torch.int, !torch.int) -> !torch.list<int>
%2521 = torch.aten.view %2464, %2520 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2521, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2522 = torch.aten.mm %2521, %2479 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2522, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_148 = torch.constant.int 4
%int512_149 = torch.constant.int 512
%2523 = torch.prim.ListConstruct %int4_148, %2482, %int512_149 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2524 = torch.aten.view %2522, %2523 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_150 = torch.constant.int 4
%2525 = torch.aten.mul.int %int4_150, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_151 = torch.constant.int 4096
%2526 = torch.prim.ListConstruct %2525, %int4096_151 : (!torch.int, !torch.int) -> !torch.list<int>
%2527 = torch.aten.view %2465, %2526 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2527, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2528 = torch.aten.mm %2527, %2481 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %2528, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_152 = torch.constant.int 4
%int512_153 = torch.constant.int 512
%2529 = torch.prim.ListConstruct %int4_152, %2482, %int512_153 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2530 = torch.aten.view %2528, %2529 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %2530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_154 = torch.constant.int 1
%int0_155 = torch.constant.int 0
%2531 = torch.prim.ListConstruct %int1_154, %int0_155 : (!torch.int, !torch.int) -> !torch.list<int>
%2532 = torch.aten.permute %24, %2531 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_156 = torch.constant.int 1
%int0_157 = torch.constant.int 0
%2533 = torch.prim.ListConstruct %int1_156, %int0_157 : (!torch.int, !torch.int) -> !torch.list<int>
%2534 = torch.aten.permute %25, %2533 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_158 = torch.constant.int 1
%int0_159 = torch.constant.int 0
%2535 = torch.prim.ListConstruct %int1_158, %int0_159 : (!torch.int, !torch.int) -> !torch.list<int>
%2536 = torch.aten.permute %26, %2535 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_160 = torch.constant.int 1
%int0_161 = torch.constant.int 0
%2537 = torch.prim.ListConstruct %int1_160, %int0_161 : (!torch.int, !torch.int) -> !torch.list<int>
%2538 = torch.aten.permute %27, %2537 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_162 = torch.constant.int 1
%int0_163 = torch.constant.int 0
%2539 = torch.prim.ListConstruct %int1_162, %int0_163 : (!torch.int, !torch.int) -> !torch.list<int>
%2540 = torch.aten.permute %28, %2539 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_164 = torch.constant.int 1
%int0_165 = torch.constant.int 0
%2541 = torch.prim.ListConstruct %int1_164, %int0_165 : (!torch.int, !torch.int) -> !torch.list<int>
%2542 = torch.aten.permute %29, %2541 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_166 = torch.constant.int 1
%int0_167 = torch.constant.int 0
%2543 = torch.prim.ListConstruct %int1_166, %int0_167 : (!torch.int, !torch.int) -> !torch.list<int>
%2544 = torch.aten.permute %30, %2543 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_168 = torch.constant.int 1
%int0_169 = torch.constant.int 0
%2545 = torch.prim.ListConstruct %int1_168, %int0_169 : (!torch.int, !torch.int) -> !torch.list<int>
%2546 = torch.aten.permute %31, %2545 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_170 = torch.constant.int 4
%2547 = torch.aten.mul.int %int4_170, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_171 = torch.constant.int 4096
%2548 = torch.prim.ListConstruct %2547, %int4096_171 : (!torch.int, !torch.int) -> !torch.list<int>
%2549 = torch.aten.view %2458, %2548 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2549, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2550 = torch.aten.mm %2549, %2532 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2550, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_172 = torch.constant.int 4
%int128 = torch.constant.int 128
%2551 = torch.prim.ListConstruct %int4_172, %2482, %int128 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2552 = torch.aten.view %2550, %2551 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_173 = torch.constant.int 4
%2553 = torch.aten.mul.int %int4_173, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_174 = torch.constant.int 4096
%2554 = torch.prim.ListConstruct %2553, %int4096_174 : (!torch.int, !torch.int) -> !torch.list<int>
%2555 = torch.aten.view %2459, %2554 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2555, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2556 = torch.aten.mm %2555, %2534 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2556, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_175 = torch.constant.int 4
%int128_176 = torch.constant.int 128
%2557 = torch.prim.ListConstruct %int4_175, %2482, %int128_176 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2558 = torch.aten.view %2556, %2557 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_177 = torch.constant.int 4
%2559 = torch.aten.mul.int %int4_177, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_178 = torch.constant.int 4096
%2560 = torch.prim.ListConstruct %2559, %int4096_178 : (!torch.int, !torch.int) -> !torch.list<int>
%2561 = torch.aten.view %2460, %2560 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2561, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2562 = torch.aten.mm %2561, %2536 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2562, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_179 = torch.constant.int 4
%int128_180 = torch.constant.int 128
%2563 = torch.prim.ListConstruct %int4_179, %2482, %int128_180 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2564 = torch.aten.view %2562, %2563 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_181 = torch.constant.int 4
%2565 = torch.aten.mul.int %int4_181, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_182 = torch.constant.int 4096
%2566 = torch.prim.ListConstruct %2565, %int4096_182 : (!torch.int, !torch.int) -> !torch.list<int>
%2567 = torch.aten.view %2461, %2566 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2567, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2568 = torch.aten.mm %2567, %2538 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2568, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_183 = torch.constant.int 4
%int128_184 = torch.constant.int 128
%2569 = torch.prim.ListConstruct %int4_183, %2482, %int128_184 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2570 = torch.aten.view %2568, %2569 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_185 = torch.constant.int 4
%2571 = torch.aten.mul.int %int4_185, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_186 = torch.constant.int 4096
%2572 = torch.prim.ListConstruct %2571, %int4096_186 : (!torch.int, !torch.int) -> !torch.list<int>
%2573 = torch.aten.view %2462, %2572 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2573, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2574 = torch.aten.mm %2573, %2540 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2574, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_187 = torch.constant.int 4
%int128_188 = torch.constant.int 128
%2575 = torch.prim.ListConstruct %int4_187, %2482, %int128_188 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2576 = torch.aten.view %2574, %2575 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_189 = torch.constant.int 4
%2577 = torch.aten.mul.int %int4_189, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_190 = torch.constant.int 4096
%2578 = torch.prim.ListConstruct %2577, %int4096_190 : (!torch.int, !torch.int) -> !torch.list<int>
%2579 = torch.aten.view %2463, %2578 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2579, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2580 = torch.aten.mm %2579, %2542 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2580, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_191 = torch.constant.int 4
%int128_192 = torch.constant.int 128
%2581 = torch.prim.ListConstruct %int4_191, %2482, %int128_192 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2582 = torch.aten.view %2580, %2581 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_193 = torch.constant.int 4
%2583 = torch.aten.mul.int %int4_193, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_194 = torch.constant.int 4096
%2584 = torch.prim.ListConstruct %2583, %int4096_194 : (!torch.int, !torch.int) -> !torch.list<int>
%2585 = torch.aten.view %2464, %2584 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2585, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2586 = torch.aten.mm %2585, %2544 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2586, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_195 = torch.constant.int 4
%int128_196 = torch.constant.int 128
%2587 = torch.prim.ListConstruct %int4_195, %2482, %int128_196 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2588 = torch.aten.view %2586, %2587 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_197 = torch.constant.int 4
%2589 = torch.aten.mul.int %int4_197, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_198 = torch.constant.int 4096
%2590 = torch.prim.ListConstruct %2589, %int4096_198 : (!torch.int, !torch.int) -> !torch.list<int>
%2591 = torch.aten.view %2465, %2590 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2591, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2592 = torch.aten.mm %2591, %2546 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2592, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_199 = torch.constant.int 4
%int128_200 = torch.constant.int 128
%2593 = torch.prim.ListConstruct %int4_199, %2482, %int128_200 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2594 = torch.aten.view %2592, %2593 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_201 = torch.constant.int 1
%int0_202 = torch.constant.int 0
%2595 = torch.prim.ListConstruct %int1_201, %int0_202 : (!torch.int, !torch.int) -> !torch.list<int>
%2596 = torch.aten.permute %32, %2595 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_203 = torch.constant.int 1
%int0_204 = torch.constant.int 0
%2597 = torch.prim.ListConstruct %int1_203, %int0_204 : (!torch.int, !torch.int) -> !torch.list<int>
%2598 = torch.aten.permute %33, %2597 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_205 = torch.constant.int 1
%int0_206 = torch.constant.int 0
%2599 = torch.prim.ListConstruct %int1_205, %int0_206 : (!torch.int, !torch.int) -> !torch.list<int>
%2600 = torch.aten.permute %34, %2599 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_207 = torch.constant.int 1
%int0_208 = torch.constant.int 0
%2601 = torch.prim.ListConstruct %int1_207, %int0_208 : (!torch.int, !torch.int) -> !torch.list<int>
%2602 = torch.aten.permute %35, %2601 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_209 = torch.constant.int 1
%int0_210 = torch.constant.int 0
%2603 = torch.prim.ListConstruct %int1_209, %int0_210 : (!torch.int, !torch.int) -> !torch.list<int>
%2604 = torch.aten.permute %36, %2603 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_211 = torch.constant.int 1
%int0_212 = torch.constant.int 0
%2605 = torch.prim.ListConstruct %int1_211, %int0_212 : (!torch.int, !torch.int) -> !torch.list<int>
%2606 = torch.aten.permute %37, %2605 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_213 = torch.constant.int 1
%int0_214 = torch.constant.int 0
%2607 = torch.prim.ListConstruct %int1_213, %int0_214 : (!torch.int, !torch.int) -> !torch.list<int>
%2608 = torch.aten.permute %38, %2607 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_215 = torch.constant.int 1
%int0_216 = torch.constant.int 0
%2609 = torch.prim.ListConstruct %int1_215, %int0_216 : (!torch.int, !torch.int) -> !torch.list<int>
%2610 = torch.aten.permute %39, %2609 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_217 = torch.constant.int 4
%2611 = torch.aten.mul.int %int4_217, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_218 = torch.constant.int 4096
%2612 = torch.prim.ListConstruct %2611, %int4096_218 : (!torch.int, !torch.int) -> !torch.list<int>
%2613 = torch.aten.view %2458, %2612 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2613, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2614 = torch.aten.mm %2613, %2596 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2614, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_219 = torch.constant.int 4
%int128_220 = torch.constant.int 128
%2615 = torch.prim.ListConstruct %int4_219, %2482, %int128_220 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2616 = torch.aten.view %2614, %2615 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_221 = torch.constant.int 4
%2617 = torch.aten.mul.int %int4_221, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_222 = torch.constant.int 4096
%2618 = torch.prim.ListConstruct %2617, %int4096_222 : (!torch.int, !torch.int) -> !torch.list<int>
%2619 = torch.aten.view %2459, %2618 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2619, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2620 = torch.aten.mm %2619, %2598 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2620, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_223 = torch.constant.int 4
%int128_224 = torch.constant.int 128
%2621 = torch.prim.ListConstruct %int4_223, %2482, %int128_224 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2622 = torch.aten.view %2620, %2621 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_225 = torch.constant.int 4
%2623 = torch.aten.mul.int %int4_225, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_226 = torch.constant.int 4096
%2624 = torch.prim.ListConstruct %2623, %int4096_226 : (!torch.int, !torch.int) -> !torch.list<int>
%2625 = torch.aten.view %2460, %2624 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2625, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2626 = torch.aten.mm %2625, %2600 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2626, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_227 = torch.constant.int 4
%int128_228 = torch.constant.int 128
%2627 = torch.prim.ListConstruct %int4_227, %2482, %int128_228 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2628 = torch.aten.view %2626, %2627 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_229 = torch.constant.int 4
%2629 = torch.aten.mul.int %int4_229, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_230 = torch.constant.int 4096
%2630 = torch.prim.ListConstruct %2629, %int4096_230 : (!torch.int, !torch.int) -> !torch.list<int>
%2631 = torch.aten.view %2461, %2630 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2631, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2632 = torch.aten.mm %2631, %2602 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2632, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_231 = torch.constant.int 4
%int128_232 = torch.constant.int 128
%2633 = torch.prim.ListConstruct %int4_231, %2482, %int128_232 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2634 = torch.aten.view %2632, %2633 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_233 = torch.constant.int 4
%2635 = torch.aten.mul.int %int4_233, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_234 = torch.constant.int 4096
%2636 = torch.prim.ListConstruct %2635, %int4096_234 : (!torch.int, !torch.int) -> !torch.list<int>
%2637 = torch.aten.view %2462, %2636 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2637, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2638 = torch.aten.mm %2637, %2604 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2638, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_235 = torch.constant.int 4
%int128_236 = torch.constant.int 128
%2639 = torch.prim.ListConstruct %int4_235, %2482, %int128_236 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2640 = torch.aten.view %2638, %2639 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_237 = torch.constant.int 4
%2641 = torch.aten.mul.int %int4_237, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_238 = torch.constant.int 4096
%2642 = torch.prim.ListConstruct %2641, %int4096_238 : (!torch.int, !torch.int) -> !torch.list<int>
%2643 = torch.aten.view %2463, %2642 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2643, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2644 = torch.aten.mm %2643, %2606 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2644, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_239 = torch.constant.int 4
%int128_240 = torch.constant.int 128
%2645 = torch.prim.ListConstruct %int4_239, %2482, %int128_240 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2646 = torch.aten.view %2644, %2645 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_241 = torch.constant.int 4
%2647 = torch.aten.mul.int %int4_241, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_242 = torch.constant.int 4096
%2648 = torch.prim.ListConstruct %2647, %int4096_242 : (!torch.int, !torch.int) -> !torch.list<int>
%2649 = torch.aten.view %2464, %2648 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2649, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2650 = torch.aten.mm %2649, %2608 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2650, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_243 = torch.constant.int 4
%int128_244 = torch.constant.int 128
%2651 = torch.prim.ListConstruct %int4_243, %2482, %int128_244 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2652 = torch.aten.view %2650, %2651 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_245 = torch.constant.int 4
%2653 = torch.aten.mul.int %int4_245, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_246 = torch.constant.int 4096
%2654 = torch.prim.ListConstruct %2653, %int4096_246 : (!torch.int, !torch.int) -> !torch.list<int>
%2655 = torch.aten.view %2465, %2654 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %2655, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%2656 = torch.aten.mm %2655, %2610 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %2656, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_247 = torch.constant.int 4
%int128_248 = torch.constant.int 128
%2657 = torch.prim.ListConstruct %int4_247, %2482, %int128_248 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2658 = torch.aten.view %2656, %2657 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %2658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_249 = torch.constant.int 4
%int4_250 = torch.constant.int 4
%int128_251 = torch.constant.int 128
%2659 = torch.prim.ListConstruct %int4_249, %2482, %int4_250, %int128_251 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2660 = torch.aten.view %2488, %2659 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_252 = torch.constant.int 4
%int4_253 = torch.constant.int 4
%int128_254 = torch.constant.int 128
%2661 = torch.prim.ListConstruct %int4_252, %2482, %int4_253, %int128_254 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2662 = torch.aten.view %2494, %2661 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_255 = torch.constant.int 4
%int4_256 = torch.constant.int 4
%int128_257 = torch.constant.int 128
%2663 = torch.prim.ListConstruct %int4_255, %2482, %int4_256, %int128_257 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2664 = torch.aten.view %2500, %2663 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_258 = torch.constant.int 4
%int4_259 = torch.constant.int 4
%int128_260 = torch.constant.int 128
%2665 = torch.prim.ListConstruct %int4_258, %2482, %int4_259, %int128_260 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2666 = torch.aten.view %2506, %2665 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_261 = torch.constant.int 4
%int4_262 = torch.constant.int 4
%int128_263 = torch.constant.int 128
%2667 = torch.prim.ListConstruct %int4_261, %2482, %int4_262, %int128_263 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2668 = torch.aten.view %2512, %2667 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_264 = torch.constant.int 4
%int4_265 = torch.constant.int 4
%int128_266 = torch.constant.int 128
%2669 = torch.prim.ListConstruct %int4_264, %2482, %int4_265, %int128_266 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2670 = torch.aten.view %2518, %2669 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_267 = torch.constant.int 4
%int4_268 = torch.constant.int 4
%int128_269 = torch.constant.int 128
%2671 = torch.prim.ListConstruct %int4_267, %2482, %int4_268, %int128_269 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2672 = torch.aten.view %2524, %2671 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_270 = torch.constant.int 4
%int4_271 = torch.constant.int 4
%int128_272 = torch.constant.int 128
%2673 = torch.prim.ListConstruct %int4_270, %2482, %int4_271, %int128_272 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2674 = torch.aten.view %2530, %2673 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_273 = torch.constant.int 4
%int1_274 = torch.constant.int 1
%int128_275 = torch.constant.int 128
%2675 = torch.prim.ListConstruct %int4_273, %2482, %int1_274, %int128_275 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2676 = torch.aten.view %2552, %2675 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_276 = torch.constant.int 4
%int1_277 = torch.constant.int 1
%int128_278 = torch.constant.int 128
%2677 = torch.prim.ListConstruct %int4_276, %2482, %int1_277, %int128_278 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2678 = torch.aten.view %2558, %2677 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_279 = torch.constant.int 4
%int1_280 = torch.constant.int 1
%int128_281 = torch.constant.int 128
%2679 = torch.prim.ListConstruct %int4_279, %2482, %int1_280, %int128_281 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2680 = torch.aten.view %2564, %2679 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_282 = torch.constant.int 4
%int1_283 = torch.constant.int 1
%int128_284 = torch.constant.int 128
%2681 = torch.prim.ListConstruct %int4_282, %2482, %int1_283, %int128_284 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2682 = torch.aten.view %2570, %2681 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_285 = torch.constant.int 4
%int1_286 = torch.constant.int 1
%int128_287 = torch.constant.int 128
%2683 = torch.prim.ListConstruct %int4_285, %2482, %int1_286, %int128_287 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2684 = torch.aten.view %2576, %2683 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_288 = torch.constant.int 4
%int1_289 = torch.constant.int 1
%int128_290 = torch.constant.int 128
%2685 = torch.prim.ListConstruct %int4_288, %2482, %int1_289, %int128_290 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2686 = torch.aten.view %2582, %2685 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_291 = torch.constant.int 4
%int1_292 = torch.constant.int 1
%int128_293 = torch.constant.int 128
%2687 = torch.prim.ListConstruct %int4_291, %2482, %int1_292, %int128_293 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2688 = torch.aten.view %2588, %2687 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_294 = torch.constant.int 4
%int1_295 = torch.constant.int 1
%int128_296 = torch.constant.int 128
%2689 = torch.prim.ListConstruct %int4_294, %2482, %int1_295, %int128_296 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2690 = torch.aten.view %2594, %2689 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_297 = torch.constant.int 4
%int1_298 = torch.constant.int 1
%int128_299 = torch.constant.int 128
%2691 = torch.prim.ListConstruct %int4_297, %2482, %int1_298, %int128_299 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2692 = torch.aten.view %2616, %2691 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_300 = torch.constant.int 4
%int1_301 = torch.constant.int 1
%int128_302 = torch.constant.int 128
%2693 = torch.prim.ListConstruct %int4_300, %2482, %int1_301, %int128_302 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2694 = torch.aten.view %2622, %2693 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_303 = torch.constant.int 4
%int1_304 = torch.constant.int 1
%int128_305 = torch.constant.int 128
%2695 = torch.prim.ListConstruct %int4_303, %2482, %int1_304, %int128_305 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2696 = torch.aten.view %2628, %2695 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_306 = torch.constant.int 4
%int1_307 = torch.constant.int 1
%int128_308 = torch.constant.int 128
%2697 = torch.prim.ListConstruct %int4_306, %2482, %int1_307, %int128_308 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2698 = torch.aten.view %2634, %2697 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_309 = torch.constant.int 4
%int1_310 = torch.constant.int 1
%int128_311 = torch.constant.int 128
%2699 = torch.prim.ListConstruct %int4_309, %2482, %int1_310, %int128_311 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2700 = torch.aten.view %2640, %2699 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_312 = torch.constant.int 4
%int1_313 = torch.constant.int 1
%int128_314 = torch.constant.int 128
%2701 = torch.prim.ListConstruct %int4_312, %2482, %int1_313, %int128_314 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2702 = torch.aten.view %2646, %2701 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_315 = torch.constant.int 4
%int1_316 = torch.constant.int 1
%int128_317 = torch.constant.int 128
%2703 = torch.prim.ListConstruct %int4_315, %2482, %int1_316, %int128_317 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2704 = torch.aten.view %2652, %2703 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_318 = torch.constant.int 4
%int1_319 = torch.constant.int 1
%int128_320 = torch.constant.int 128
%2705 = torch.prim.ListConstruct %int4_318, %2482, %int1_319, %int128_320 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%2706 = torch.aten.view %2658, %2705 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072 = torch.constant.int 131072
%none_321 = torch.constant.none
%none_322 = torch.constant.none
%cpu = torch.constant.device "cpu"
%false_323 = torch.constant.bool false
%2707 = torch.aten.arange %int131072, %none_321, %none_322, %cpu, %false_323 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_324 = torch.constant.int 0
%int128_325 = torch.constant.int 128
%int2_326 = torch.constant.int 2
%none_327 = torch.constant.none
%none_328 = torch.constant.none
%cpu_329 = torch.constant.device "cpu"
%false_330 = torch.constant.bool false
%2708 = torch.aten.arange.start_step %int0_324, %int128_325, %int2_326, %none_327, %none_328, %cpu_329, %false_330 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_331 = torch.constant.int 0
%int0_332 = torch.constant.int 0
%int64 = torch.constant.int 64
%int1_333 = torch.constant.int 1
%2709 = torch.aten.slice.Tensor %2708, %int0_331, %int0_332, %int64, %int1_333 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_334 = torch.constant.int 6
%2710 = torch.prims.convert_element_type %2709, %int6_334 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_335 = torch.constant.int 128
%2711 = torch.aten.div.Scalar %2710, %int128_335 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05 = torch.constant.float 5.000000e+05
%2712 = torch.aten.pow.Scalar %float5.000000e05, %2711 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%2713 = torch.aten.reciprocal %2712 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00 = torch.constant.float 1.000000e+00
%2714 = torch.aten.mul.Scalar %2713, %float1.000000e00 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_336 = torch.constant.int 131072
%int1_337 = torch.constant.int 1
%2715 = torch.prim.ListConstruct %int131072_336, %int1_337 : (!torch.int, !torch.int) -> !torch.list<int>
%2716 = torch.aten.view %2707, %2715 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%2717 = torch.aten.mul.Tensor %2716, %2714 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%2718 = torch.aten.cos %2717 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%2719 = torch.aten.sin %2717 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%2720 = torch.aten.complex %2718, %2719 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%2721 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2722 = flow.tensor.transfer %2721 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%2723 = torch_c.from_builtin_tensor %2722 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2724 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2725 = flow.tensor.transfer %2724 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%2726 = torch_c.from_builtin_tensor %2725 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2727 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2728 = flow.tensor.transfer %2727 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%2729 = torch_c.from_builtin_tensor %2728 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2730 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2731 = flow.tensor.transfer %2730 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%2732 = torch_c.from_builtin_tensor %2731 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2733 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2734 = flow.tensor.transfer %2733 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%2735 = torch_c.from_builtin_tensor %2734 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2736 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2737 = flow.tensor.transfer %2736 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%2738 = torch_c.from_builtin_tensor %2737 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2739 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2740 = flow.tensor.transfer %2739 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%2741 = torch_c.from_builtin_tensor %2740 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2742 = torch_c.to_builtin_tensor %2720 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2743 = flow.tensor.transfer %2742 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%2744 = torch_c.from_builtin_tensor %2743 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_338 = torch.constant.int 1
%2745 = torch.aten.size.int %2488, %int1_338 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_339 = torch.constant.int 0
%2746 = torch.aten.add.int %int0_339, %2745 : !torch.int, !torch.int -> !torch.int
%int0_340 = torch.constant.int 0
%int0_341 = torch.constant.int 0
%int1_342 = torch.constant.int 1
%2747 = torch.aten.slice.Tensor %2723, %int0_340, %int0_341, %2746, %int1_342 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2747, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_343 = torch.constant.int 1
%int0_344 = torch.constant.int 0
%int9223372036854775807 = torch.constant.int 9223372036854775807
%int1_345 = torch.constant.int 1
%2748 = torch.aten.slice.Tensor %2747, %int1_343, %int0_344, %int9223372036854775807, %int1_345 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2748, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_346 = torch.constant.int 0
%2749 = torch.aten.unsqueeze %2748, %int0_346 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2749, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_347 = torch.constant.int 2
%2750 = torch.aten.unsqueeze %2749, %int2_347 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2750, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3 = torch.constant.int 3
%int0_348 = torch.constant.int 0
%int9223372036854775807_349 = torch.constant.int 9223372036854775807
%int1_350 = torch.constant.int 1
%2751 = torch.aten.slice.Tensor %2750, %int3, %int0_348, %int9223372036854775807_349, %int1_350 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2751, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2752 = torch_c.to_builtin_tensor %2660 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_351 = arith.constant 1 : index
%dim_352 = tensor.dim %2752, %c1_351 : tensor<4x?x4x128xf16>
%2753 = flow.tensor.bitcast %2752 : tensor<4x?x4x128xf16>{%dim_352} -> tensor<4x?x4x64xcomplex<f16>>{%dim_352}
%2754 = torch_c.from_builtin_tensor %2753 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2755 = torch.aten.mul.Tensor %2754, %2751 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2756 = torch_c.to_builtin_tensor %2755 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_353 = arith.constant 1 : index
%dim_354 = tensor.dim %2756, %c1_353 : tensor<4x?x4x64xcomplex<f32>>
%2757 = flow.tensor.bitcast %2756 : tensor<4x?x4x64xcomplex<f32>>{%dim_354} -> tensor<4x?x4x128xf32>{%dim_354}
%2758 = torch_c.from_builtin_tensor %2757 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_355 = torch.constant.int 5
%2759 = torch.prims.convert_element_type %2758, %int5_355 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_356 = torch.constant.int 1
%2760 = torch.aten.size.int %2494, %int1_356 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_357 = torch.constant.int 0
%2761 = torch.aten.add.int %int0_357, %2760 : !torch.int, !torch.int -> !torch.int
%int0_358 = torch.constant.int 0
%int0_359 = torch.constant.int 0
%int1_360 = torch.constant.int 1
%2762 = torch.aten.slice.Tensor %2726, %int0_358, %int0_359, %2761, %int1_360 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2762, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_361 = torch.constant.int 1
%int0_362 = torch.constant.int 0
%int9223372036854775807_363 = torch.constant.int 9223372036854775807
%int1_364 = torch.constant.int 1
%2763 = torch.aten.slice.Tensor %2762, %int1_361, %int0_362, %int9223372036854775807_363, %int1_364 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2763, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_365 = torch.constant.int 0
%2764 = torch.aten.unsqueeze %2763, %int0_365 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2764, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_366 = torch.constant.int 2
%2765 = torch.aten.unsqueeze %2764, %int2_366 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2765, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_367 = torch.constant.int 3
%int0_368 = torch.constant.int 0
%int9223372036854775807_369 = torch.constant.int 9223372036854775807
%int1_370 = torch.constant.int 1
%2766 = torch.aten.slice.Tensor %2765, %int3_367, %int0_368, %int9223372036854775807_369, %int1_370 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2766, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2767 = torch_c.to_builtin_tensor %2662 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_371 = arith.constant 1 : index
%dim_372 = tensor.dim %2767, %c1_371 : tensor<4x?x4x128xf16>
%2768 = flow.tensor.bitcast %2767 : tensor<4x?x4x128xf16>{%dim_372} -> tensor<4x?x4x64xcomplex<f16>>{%dim_372}
%2769 = torch_c.from_builtin_tensor %2768 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2770 = torch.aten.mul.Tensor %2769, %2766 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2771 = torch_c.to_builtin_tensor %2770 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_373 = arith.constant 1 : index
%dim_374 = tensor.dim %2771, %c1_373 : tensor<4x?x4x64xcomplex<f32>>
%2772 = flow.tensor.bitcast %2771 : tensor<4x?x4x64xcomplex<f32>>{%dim_374} -> tensor<4x?x4x128xf32>{%dim_374}
%2773 = torch_c.from_builtin_tensor %2772 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_375 = torch.constant.int 5
%2774 = torch.prims.convert_element_type %2773, %int5_375 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_376 = torch.constant.int 1
%2775 = torch.aten.size.int %2500, %int1_376 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_377 = torch.constant.int 0
%2776 = torch.aten.add.int %int0_377, %2775 : !torch.int, !torch.int -> !torch.int
%int0_378 = torch.constant.int 0
%int0_379 = torch.constant.int 0
%int1_380 = torch.constant.int 1
%2777 = torch.aten.slice.Tensor %2729, %int0_378, %int0_379, %2776, %int1_380 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2777, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_381 = torch.constant.int 1
%int0_382 = torch.constant.int 0
%int9223372036854775807_383 = torch.constant.int 9223372036854775807
%int1_384 = torch.constant.int 1
%2778 = torch.aten.slice.Tensor %2777, %int1_381, %int0_382, %int9223372036854775807_383, %int1_384 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2778, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_385 = torch.constant.int 0
%2779 = torch.aten.unsqueeze %2778, %int0_385 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2779, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_386 = torch.constant.int 2
%2780 = torch.aten.unsqueeze %2779, %int2_386 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2780, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_387 = torch.constant.int 3
%int0_388 = torch.constant.int 0
%int9223372036854775807_389 = torch.constant.int 9223372036854775807
%int1_390 = torch.constant.int 1
%2781 = torch.aten.slice.Tensor %2780, %int3_387, %int0_388, %int9223372036854775807_389, %int1_390 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2781, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2782 = torch_c.to_builtin_tensor %2664 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_391 = arith.constant 1 : index
%dim_392 = tensor.dim %2782, %c1_391 : tensor<4x?x4x128xf16>
%2783 = flow.tensor.bitcast %2782 : tensor<4x?x4x128xf16>{%dim_392} -> tensor<4x?x4x64xcomplex<f16>>{%dim_392}
%2784 = torch_c.from_builtin_tensor %2783 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2785 = torch.aten.mul.Tensor %2784, %2781 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2786 = torch_c.to_builtin_tensor %2785 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_393 = arith.constant 1 : index
%dim_394 = tensor.dim %2786, %c1_393 : tensor<4x?x4x64xcomplex<f32>>
%2787 = flow.tensor.bitcast %2786 : tensor<4x?x4x64xcomplex<f32>>{%dim_394} -> tensor<4x?x4x128xf32>{%dim_394}
%2788 = torch_c.from_builtin_tensor %2787 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_395 = torch.constant.int 5
%2789 = torch.prims.convert_element_type %2788, %int5_395 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_396 = torch.constant.int 1
%2790 = torch.aten.size.int %2506, %int1_396 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_397 = torch.constant.int 0
%2791 = torch.aten.add.int %int0_397, %2790 : !torch.int, !torch.int -> !torch.int
%int0_398 = torch.constant.int 0
%int0_399 = torch.constant.int 0
%int1_400 = torch.constant.int 1
%2792 = torch.aten.slice.Tensor %2732, %int0_398, %int0_399, %2791, %int1_400 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2792, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_401 = torch.constant.int 1
%int0_402 = torch.constant.int 0
%int9223372036854775807_403 = torch.constant.int 9223372036854775807
%int1_404 = torch.constant.int 1
%2793 = torch.aten.slice.Tensor %2792, %int1_401, %int0_402, %int9223372036854775807_403, %int1_404 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2793, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_405 = torch.constant.int 0
%2794 = torch.aten.unsqueeze %2793, %int0_405 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2794, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_406 = torch.constant.int 2
%2795 = torch.aten.unsqueeze %2794, %int2_406 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2795, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_407 = torch.constant.int 3
%int0_408 = torch.constant.int 0
%int9223372036854775807_409 = torch.constant.int 9223372036854775807
%int1_410 = torch.constant.int 1
%2796 = torch.aten.slice.Tensor %2795, %int3_407, %int0_408, %int9223372036854775807_409, %int1_410 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2796, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2797 = torch_c.to_builtin_tensor %2666 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_411 = arith.constant 1 : index
%dim_412 = tensor.dim %2797, %c1_411 : tensor<4x?x4x128xf16>
%2798 = flow.tensor.bitcast %2797 : tensor<4x?x4x128xf16>{%dim_412} -> tensor<4x?x4x64xcomplex<f16>>{%dim_412}
%2799 = torch_c.from_builtin_tensor %2798 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2800 = torch.aten.mul.Tensor %2799, %2796 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2801 = torch_c.to_builtin_tensor %2800 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_413 = arith.constant 1 : index
%dim_414 = tensor.dim %2801, %c1_413 : tensor<4x?x4x64xcomplex<f32>>
%2802 = flow.tensor.bitcast %2801 : tensor<4x?x4x64xcomplex<f32>>{%dim_414} -> tensor<4x?x4x128xf32>{%dim_414}
%2803 = torch_c.from_builtin_tensor %2802 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_415 = torch.constant.int 5
%2804 = torch.prims.convert_element_type %2803, %int5_415 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_416 = torch.constant.int 1
%2805 = torch.aten.size.int %2512, %int1_416 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_417 = torch.constant.int 0
%2806 = torch.aten.add.int %int0_417, %2805 : !torch.int, !torch.int -> !torch.int
%int0_418 = torch.constant.int 0
%int0_419 = torch.constant.int 0
%int1_420 = torch.constant.int 1
%2807 = torch.aten.slice.Tensor %2735, %int0_418, %int0_419, %2806, %int1_420 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2807, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_421 = torch.constant.int 1
%int0_422 = torch.constant.int 0
%int9223372036854775807_423 = torch.constant.int 9223372036854775807
%int1_424 = torch.constant.int 1
%2808 = torch.aten.slice.Tensor %2807, %int1_421, %int0_422, %int9223372036854775807_423, %int1_424 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2808, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_425 = torch.constant.int 0
%2809 = torch.aten.unsqueeze %2808, %int0_425 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2809, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_426 = torch.constant.int 2
%2810 = torch.aten.unsqueeze %2809, %int2_426 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2810, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_427 = torch.constant.int 3
%int0_428 = torch.constant.int 0
%int9223372036854775807_429 = torch.constant.int 9223372036854775807
%int1_430 = torch.constant.int 1
%2811 = torch.aten.slice.Tensor %2810, %int3_427, %int0_428, %int9223372036854775807_429, %int1_430 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2811, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2812 = torch_c.to_builtin_tensor %2668 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_431 = arith.constant 1 : index
%dim_432 = tensor.dim %2812, %c1_431 : tensor<4x?x4x128xf16>
%2813 = flow.tensor.bitcast %2812 : tensor<4x?x4x128xf16>{%dim_432} -> tensor<4x?x4x64xcomplex<f16>>{%dim_432}
%2814 = torch_c.from_builtin_tensor %2813 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2815 = torch.aten.mul.Tensor %2814, %2811 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2816 = torch_c.to_builtin_tensor %2815 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_433 = arith.constant 1 : index
%dim_434 = tensor.dim %2816, %c1_433 : tensor<4x?x4x64xcomplex<f32>>
%2817 = flow.tensor.bitcast %2816 : tensor<4x?x4x64xcomplex<f32>>{%dim_434} -> tensor<4x?x4x128xf32>{%dim_434}
%2818 = torch_c.from_builtin_tensor %2817 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_435 = torch.constant.int 5
%2819 = torch.prims.convert_element_type %2818, %int5_435 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_436 = torch.constant.int 1
%2820 = torch.aten.size.int %2518, %int1_436 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_437 = torch.constant.int 0
%2821 = torch.aten.add.int %int0_437, %2820 : !torch.int, !torch.int -> !torch.int
%int0_438 = torch.constant.int 0
%int0_439 = torch.constant.int 0
%int1_440 = torch.constant.int 1
%2822 = torch.aten.slice.Tensor %2738, %int0_438, %int0_439, %2821, %int1_440 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2822, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_441 = torch.constant.int 1
%int0_442 = torch.constant.int 0
%int9223372036854775807_443 = torch.constant.int 9223372036854775807
%int1_444 = torch.constant.int 1
%2823 = torch.aten.slice.Tensor %2822, %int1_441, %int0_442, %int9223372036854775807_443, %int1_444 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2823, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_445 = torch.constant.int 0
%2824 = torch.aten.unsqueeze %2823, %int0_445 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2824, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_446 = torch.constant.int 2
%2825 = torch.aten.unsqueeze %2824, %int2_446 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2825, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_447 = torch.constant.int 3
%int0_448 = torch.constant.int 0
%int9223372036854775807_449 = torch.constant.int 9223372036854775807
%int1_450 = torch.constant.int 1
%2826 = torch.aten.slice.Tensor %2825, %int3_447, %int0_448, %int9223372036854775807_449, %int1_450 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2826, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2827 = torch_c.to_builtin_tensor %2670 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_451 = arith.constant 1 : index
%dim_452 = tensor.dim %2827, %c1_451 : tensor<4x?x4x128xf16>
%2828 = flow.tensor.bitcast %2827 : tensor<4x?x4x128xf16>{%dim_452} -> tensor<4x?x4x64xcomplex<f16>>{%dim_452}
%2829 = torch_c.from_builtin_tensor %2828 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2830 = torch.aten.mul.Tensor %2829, %2826 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2831 = torch_c.to_builtin_tensor %2830 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_453 = arith.constant 1 : index
%dim_454 = tensor.dim %2831, %c1_453 : tensor<4x?x4x64xcomplex<f32>>
%2832 = flow.tensor.bitcast %2831 : tensor<4x?x4x64xcomplex<f32>>{%dim_454} -> tensor<4x?x4x128xf32>{%dim_454}
%2833 = torch_c.from_builtin_tensor %2832 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_455 = torch.constant.int 5
%2834 = torch.prims.convert_element_type %2833, %int5_455 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_456 = torch.constant.int 1
%2835 = torch.aten.size.int %2524, %int1_456 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_457 = torch.constant.int 0
%2836 = torch.aten.add.int %int0_457, %2835 : !torch.int, !torch.int -> !torch.int
%int0_458 = torch.constant.int 0
%int0_459 = torch.constant.int 0
%int1_460 = torch.constant.int 1
%2837 = torch.aten.slice.Tensor %2741, %int0_458, %int0_459, %2836, %int1_460 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2837, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_461 = torch.constant.int 1
%int0_462 = torch.constant.int 0
%int9223372036854775807_463 = torch.constant.int 9223372036854775807
%int1_464 = torch.constant.int 1
%2838 = torch.aten.slice.Tensor %2837, %int1_461, %int0_462, %int9223372036854775807_463, %int1_464 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2838, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_465 = torch.constant.int 0
%2839 = torch.aten.unsqueeze %2838, %int0_465 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2839, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_466 = torch.constant.int 2
%2840 = torch.aten.unsqueeze %2839, %int2_466 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2840, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_467 = torch.constant.int 3
%int0_468 = torch.constant.int 0
%int9223372036854775807_469 = torch.constant.int 9223372036854775807
%int1_470 = torch.constant.int 1
%2841 = torch.aten.slice.Tensor %2840, %int3_467, %int0_468, %int9223372036854775807_469, %int1_470 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2841, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2842 = torch_c.to_builtin_tensor %2672 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_471 = arith.constant 1 : index
%dim_472 = tensor.dim %2842, %c1_471 : tensor<4x?x4x128xf16>
%2843 = flow.tensor.bitcast %2842 : tensor<4x?x4x128xf16>{%dim_472} -> tensor<4x?x4x64xcomplex<f16>>{%dim_472}
%2844 = torch_c.from_builtin_tensor %2843 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2845 = torch.aten.mul.Tensor %2844, %2841 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2846 = torch_c.to_builtin_tensor %2845 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_473 = arith.constant 1 : index
%dim_474 = tensor.dim %2846, %c1_473 : tensor<4x?x4x64xcomplex<f32>>
%2847 = flow.tensor.bitcast %2846 : tensor<4x?x4x64xcomplex<f32>>{%dim_474} -> tensor<4x?x4x128xf32>{%dim_474}
%2848 = torch_c.from_builtin_tensor %2847 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_475 = torch.constant.int 5
%2849 = torch.prims.convert_element_type %2848, %int5_475 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_476 = torch.constant.int 1
%2850 = torch.aten.size.int %2530, %int1_476 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_477 = torch.constant.int 0
%2851 = torch.aten.add.int %int0_477, %2850 : !torch.int, !torch.int -> !torch.int
%int0_478 = torch.constant.int 0
%int0_479 = torch.constant.int 0
%int1_480 = torch.constant.int 1
%2852 = torch.aten.slice.Tensor %2744, %int0_478, %int0_479, %2851, %int1_480 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2852, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_481 = torch.constant.int 1
%int0_482 = torch.constant.int 0
%int9223372036854775807_483 = torch.constant.int 9223372036854775807
%int1_484 = torch.constant.int 1
%2853 = torch.aten.slice.Tensor %2852, %int1_481, %int0_482, %int9223372036854775807_483, %int1_484 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2853, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_485 = torch.constant.int 0
%2854 = torch.aten.unsqueeze %2853, %int0_485 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2854, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_486 = torch.constant.int 2
%2855 = torch.aten.unsqueeze %2854, %int2_486 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2855, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_487 = torch.constant.int 3
%int0_488 = torch.constant.int 0
%int9223372036854775807_489 = torch.constant.int 9223372036854775807
%int1_490 = torch.constant.int 1
%2856 = torch.aten.slice.Tensor %2855, %int3_487, %int0_488, %int9223372036854775807_489, %int1_490 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2856, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2857 = torch_c.to_builtin_tensor %2674 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_491 = arith.constant 1 : index
%dim_492 = tensor.dim %2857, %c1_491 : tensor<4x?x4x128xf16>
%2858 = flow.tensor.bitcast %2857 : tensor<4x?x4x128xf16>{%dim_492} -> tensor<4x?x4x64xcomplex<f16>>{%dim_492}
%2859 = torch_c.from_builtin_tensor %2858 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %2859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%2860 = torch.aten.mul.Tensor %2859, %2856 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %2860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%2861 = torch_c.to_builtin_tensor %2860 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_493 = arith.constant 1 : index
%dim_494 = tensor.dim %2861, %c1_493 : tensor<4x?x4x64xcomplex<f32>>
%2862 = flow.tensor.bitcast %2861 : tensor<4x?x4x64xcomplex<f32>>{%dim_494} -> tensor<4x?x4x128xf32>{%dim_494}
%2863 = torch_c.from_builtin_tensor %2862 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %2863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_495 = torch.constant.int 5
%2864 = torch.prims.convert_element_type %2863, %int5_495 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %2864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_496 = torch.constant.int 131072
%none_497 = torch.constant.none
%none_498 = torch.constant.none
%cpu_499 = torch.constant.device "cpu"
%false_500 = torch.constant.bool false
%2865 = torch.aten.arange %int131072_496, %none_497, %none_498, %cpu_499, %false_500 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_501 = torch.constant.int 0
%int128_502 = torch.constant.int 128
%int2_503 = torch.constant.int 2
%none_504 = torch.constant.none
%none_505 = torch.constant.none
%cpu_506 = torch.constant.device "cpu"
%false_507 = torch.constant.bool false
%2866 = torch.aten.arange.start_step %int0_501, %int128_502, %int2_503, %none_504, %none_505, %cpu_506, %false_507 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_508 = torch.constant.int 0
%int0_509 = torch.constant.int 0
%int64_510 = torch.constant.int 64
%int1_511 = torch.constant.int 1
%2867 = torch.aten.slice.Tensor %2866, %int0_508, %int0_509, %int64_510, %int1_511 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_512 = torch.constant.int 6
%2868 = torch.prims.convert_element_type %2867, %int6_512 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_513 = torch.constant.int 128
%2869 = torch.aten.div.Scalar %2868, %int128_513 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_514 = torch.constant.float 5.000000e+05
%2870 = torch.aten.pow.Scalar %float5.000000e05_514, %2869 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%2871 = torch.aten.reciprocal %2870 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_515 = torch.constant.float 1.000000e+00
%2872 = torch.aten.mul.Scalar %2871, %float1.000000e00_515 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_516 = torch.constant.int 131072
%int1_517 = torch.constant.int 1
%2873 = torch.prim.ListConstruct %int131072_516, %int1_517 : (!torch.int, !torch.int) -> !torch.list<int>
%2874 = torch.aten.view %2865, %2873 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%2875 = torch.aten.mul.Tensor %2874, %2872 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%2876 = torch.aten.cos %2875 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%2877 = torch.aten.sin %2875 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%2878 = torch.aten.complex %2876, %2877 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%2879 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2880 = flow.tensor.transfer %2879 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%2881 = torch_c.from_builtin_tensor %2880 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2882 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2883 = flow.tensor.transfer %2882 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%2884 = torch_c.from_builtin_tensor %2883 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2885 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2886 = flow.tensor.transfer %2885 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%2887 = torch_c.from_builtin_tensor %2886 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2888 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2889 = flow.tensor.transfer %2888 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%2890 = torch_c.from_builtin_tensor %2889 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2891 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2892 = flow.tensor.transfer %2891 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%2893 = torch_c.from_builtin_tensor %2892 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2894 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2895 = flow.tensor.transfer %2894 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%2896 = torch_c.from_builtin_tensor %2895 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2897 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2898 = flow.tensor.transfer %2897 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%2899 = torch_c.from_builtin_tensor %2898 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%2900 = torch_c.to_builtin_tensor %2878 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%2901 = flow.tensor.transfer %2900 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%2902 = torch_c.from_builtin_tensor %2901 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_518 = torch.constant.int 1
%2903 = torch.aten.size.int %2552, %int1_518 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_519 = torch.constant.int 0
%2904 = torch.aten.add.int %int0_519, %2903 : !torch.int, !torch.int -> !torch.int
%int0_520 = torch.constant.int 0
%int0_521 = torch.constant.int 0
%int1_522 = torch.constant.int 1
%2905 = torch.aten.slice.Tensor %2881, %int0_520, %int0_521, %2904, %int1_522 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2905, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_523 = torch.constant.int 1
%int0_524 = torch.constant.int 0
%int9223372036854775807_525 = torch.constant.int 9223372036854775807
%int1_526 = torch.constant.int 1
%2906 = torch.aten.slice.Tensor %2905, %int1_523, %int0_524, %int9223372036854775807_525, %int1_526 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2906, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_527 = torch.constant.int 0
%2907 = torch.aten.unsqueeze %2906, %int0_527 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2907, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_528 = torch.constant.int 2
%2908 = torch.aten.unsqueeze %2907, %int2_528 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2908, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_529 = torch.constant.int 3
%int0_530 = torch.constant.int 0
%int9223372036854775807_531 = torch.constant.int 9223372036854775807
%int1_532 = torch.constant.int 1
%2909 = torch.aten.slice.Tensor %2908, %int3_529, %int0_530, %int9223372036854775807_531, %int1_532 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2909, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2910 = torch_c.to_builtin_tensor %2676 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_533 = arith.constant 1 : index
%dim_534 = tensor.dim %2910, %c1_533 : tensor<4x?x1x128xf16>
%2911 = flow.tensor.bitcast %2910 : tensor<4x?x1x128xf16>{%dim_534} -> tensor<4x?x1x64xcomplex<f16>>{%dim_534}
%2912 = torch_c.from_builtin_tensor %2911 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %2912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%2913 = torch.aten.mul.Tensor %2912, %2909 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%2914 = torch_c.to_builtin_tensor %2913 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_535 = arith.constant 1 : index
%dim_536 = tensor.dim %2914, %c1_535 : tensor<4x?x1x64xcomplex<f32>>
%2915 = flow.tensor.bitcast %2914 : tensor<4x?x1x64xcomplex<f32>>{%dim_536} -> tensor<4x?x1x128xf32>{%dim_536}
%2916 = torch_c.from_builtin_tensor %2915 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %2916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_537 = torch.constant.int 5
%2917 = torch.prims.convert_element_type %2916, %int5_537 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_538 = torch.constant.int 1
%2918 = torch.aten.size.int %2558, %int1_538 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_539 = torch.constant.int 0
%2919 = torch.aten.add.int %int0_539, %2918 : !torch.int, !torch.int -> !torch.int
%int0_540 = torch.constant.int 0
%int0_541 = torch.constant.int 0
%int1_542 = torch.constant.int 1
%2920 = torch.aten.slice.Tensor %2884, %int0_540, %int0_541, %2919, %int1_542 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2920, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_543 = torch.constant.int 1
%int0_544 = torch.constant.int 0
%int9223372036854775807_545 = torch.constant.int 9223372036854775807
%int1_546 = torch.constant.int 1
%2921 = torch.aten.slice.Tensor %2920, %int1_543, %int0_544, %int9223372036854775807_545, %int1_546 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2921, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_547 = torch.constant.int 0
%2922 = torch.aten.unsqueeze %2921, %int0_547 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2922, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_548 = torch.constant.int 2
%2923 = torch.aten.unsqueeze %2922, %int2_548 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2923, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_549 = torch.constant.int 3
%int0_550 = torch.constant.int 0
%int9223372036854775807_551 = torch.constant.int 9223372036854775807
%int1_552 = torch.constant.int 1
%2924 = torch.aten.slice.Tensor %2923, %int3_549, %int0_550, %int9223372036854775807_551, %int1_552 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2924, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2925 = torch_c.to_builtin_tensor %2678 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_553 = arith.constant 1 : index
%dim_554 = tensor.dim %2925, %c1_553 : tensor<4x?x1x128xf16>
%2926 = flow.tensor.bitcast %2925 : tensor<4x?x1x128xf16>{%dim_554} -> tensor<4x?x1x64xcomplex<f16>>{%dim_554}
%2927 = torch_c.from_builtin_tensor %2926 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %2927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%2928 = torch.aten.mul.Tensor %2927, %2924 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%2929 = torch_c.to_builtin_tensor %2928 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_555 = arith.constant 1 : index
%dim_556 = tensor.dim %2929, %c1_555 : tensor<4x?x1x64xcomplex<f32>>
%2930 = flow.tensor.bitcast %2929 : tensor<4x?x1x64xcomplex<f32>>{%dim_556} -> tensor<4x?x1x128xf32>{%dim_556}
%2931 = torch_c.from_builtin_tensor %2930 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %2931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_557 = torch.constant.int 5
%2932 = torch.prims.convert_element_type %2931, %int5_557 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_558 = torch.constant.int 1
%2933 = torch.aten.size.int %2564, %int1_558 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_559 = torch.constant.int 0
%2934 = torch.aten.add.int %int0_559, %2933 : !torch.int, !torch.int -> !torch.int
%int0_560 = torch.constant.int 0
%int0_561 = torch.constant.int 0
%int1_562 = torch.constant.int 1
%2935 = torch.aten.slice.Tensor %2887, %int0_560, %int0_561, %2934, %int1_562 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2935, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_563 = torch.constant.int 1
%int0_564 = torch.constant.int 0
%int9223372036854775807_565 = torch.constant.int 9223372036854775807
%int1_566 = torch.constant.int 1
%2936 = torch.aten.slice.Tensor %2935, %int1_563, %int0_564, %int9223372036854775807_565, %int1_566 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2936, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_567 = torch.constant.int 0
%2937 = torch.aten.unsqueeze %2936, %int0_567 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2937, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_568 = torch.constant.int 2
%2938 = torch.aten.unsqueeze %2937, %int2_568 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2938, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_569 = torch.constant.int 3
%int0_570 = torch.constant.int 0
%int9223372036854775807_571 = torch.constant.int 9223372036854775807
%int1_572 = torch.constant.int 1
%2939 = torch.aten.slice.Tensor %2938, %int3_569, %int0_570, %int9223372036854775807_571, %int1_572 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2939, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2940 = torch_c.to_builtin_tensor %2680 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_573 = arith.constant 1 : index
%dim_574 = tensor.dim %2940, %c1_573 : tensor<4x?x1x128xf16>
%2941 = flow.tensor.bitcast %2940 : tensor<4x?x1x128xf16>{%dim_574} -> tensor<4x?x1x64xcomplex<f16>>{%dim_574}
%2942 = torch_c.from_builtin_tensor %2941 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %2942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%2943 = torch.aten.mul.Tensor %2942, %2939 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%2944 = torch_c.to_builtin_tensor %2943 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_575 = arith.constant 1 : index
%dim_576 = tensor.dim %2944, %c1_575 : tensor<4x?x1x64xcomplex<f32>>
%2945 = flow.tensor.bitcast %2944 : tensor<4x?x1x64xcomplex<f32>>{%dim_576} -> tensor<4x?x1x128xf32>{%dim_576}
%2946 = torch_c.from_builtin_tensor %2945 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %2946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_577 = torch.constant.int 5
%2947 = torch.prims.convert_element_type %2946, %int5_577 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_578 = torch.constant.int 1
%2948 = torch.aten.size.int %2570, %int1_578 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_579 = torch.constant.int 0
%2949 = torch.aten.add.int %int0_579, %2948 : !torch.int, !torch.int -> !torch.int
%int0_580 = torch.constant.int 0
%int0_581 = torch.constant.int 0
%int1_582 = torch.constant.int 1
%2950 = torch.aten.slice.Tensor %2890, %int0_580, %int0_581, %2949, %int1_582 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2950, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_583 = torch.constant.int 1
%int0_584 = torch.constant.int 0
%int9223372036854775807_585 = torch.constant.int 9223372036854775807
%int1_586 = torch.constant.int 1
%2951 = torch.aten.slice.Tensor %2950, %int1_583, %int0_584, %int9223372036854775807_585, %int1_586 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2951, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_587 = torch.constant.int 0
%2952 = torch.aten.unsqueeze %2951, %int0_587 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2952, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_588 = torch.constant.int 2
%2953 = torch.aten.unsqueeze %2952, %int2_588 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2953, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_589 = torch.constant.int 3
%int0_590 = torch.constant.int 0
%int9223372036854775807_591 = torch.constant.int 9223372036854775807
%int1_592 = torch.constant.int 1
%2954 = torch.aten.slice.Tensor %2953, %int3_589, %int0_590, %int9223372036854775807_591, %int1_592 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2954, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2955 = torch_c.to_builtin_tensor %2682 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_593 = arith.constant 1 : index
%dim_594 = tensor.dim %2955, %c1_593 : tensor<4x?x1x128xf16>
%2956 = flow.tensor.bitcast %2955 : tensor<4x?x1x128xf16>{%dim_594} -> tensor<4x?x1x64xcomplex<f16>>{%dim_594}
%2957 = torch_c.from_builtin_tensor %2956 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %2957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%2958 = torch.aten.mul.Tensor %2957, %2954 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%2959 = torch_c.to_builtin_tensor %2958 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_595 = arith.constant 1 : index
%dim_596 = tensor.dim %2959, %c1_595 : tensor<4x?x1x64xcomplex<f32>>
%2960 = flow.tensor.bitcast %2959 : tensor<4x?x1x64xcomplex<f32>>{%dim_596} -> tensor<4x?x1x128xf32>{%dim_596}
%2961 = torch_c.from_builtin_tensor %2960 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %2961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_597 = torch.constant.int 5
%2962 = torch.prims.convert_element_type %2961, %int5_597 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_598 = torch.constant.int 1
%2963 = torch.aten.size.int %2576, %int1_598 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_599 = torch.constant.int 0
%2964 = torch.aten.add.int %int0_599, %2963 : !torch.int, !torch.int -> !torch.int
%int0_600 = torch.constant.int 0
%int0_601 = torch.constant.int 0
%int1_602 = torch.constant.int 1
%2965 = torch.aten.slice.Tensor %2893, %int0_600, %int0_601, %2964, %int1_602 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2965, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_603 = torch.constant.int 1
%int0_604 = torch.constant.int 0
%int9223372036854775807_605 = torch.constant.int 9223372036854775807
%int1_606 = torch.constant.int 1
%2966 = torch.aten.slice.Tensor %2965, %int1_603, %int0_604, %int9223372036854775807_605, %int1_606 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2966, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_607 = torch.constant.int 0
%2967 = torch.aten.unsqueeze %2966, %int0_607 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2967, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_608 = torch.constant.int 2
%2968 = torch.aten.unsqueeze %2967, %int2_608 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2968, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_609 = torch.constant.int 3
%int0_610 = torch.constant.int 0
%int9223372036854775807_611 = torch.constant.int 9223372036854775807
%int1_612 = torch.constant.int 1
%2969 = torch.aten.slice.Tensor %2968, %int3_609, %int0_610, %int9223372036854775807_611, %int1_612 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2969, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2970 = torch_c.to_builtin_tensor %2684 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_613 = arith.constant 1 : index
%dim_614 = tensor.dim %2970, %c1_613 : tensor<4x?x1x128xf16>
%2971 = flow.tensor.bitcast %2970 : tensor<4x?x1x128xf16>{%dim_614} -> tensor<4x?x1x64xcomplex<f16>>{%dim_614}
%2972 = torch_c.from_builtin_tensor %2971 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %2972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%2973 = torch.aten.mul.Tensor %2972, %2969 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%2974 = torch_c.to_builtin_tensor %2973 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_615 = arith.constant 1 : index
%dim_616 = tensor.dim %2974, %c1_615 : tensor<4x?x1x64xcomplex<f32>>
%2975 = flow.tensor.bitcast %2974 : tensor<4x?x1x64xcomplex<f32>>{%dim_616} -> tensor<4x?x1x128xf32>{%dim_616}
%2976 = torch_c.from_builtin_tensor %2975 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %2976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_617 = torch.constant.int 5
%2977 = torch.prims.convert_element_type %2976, %int5_617 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_618 = torch.constant.int 1
%2978 = torch.aten.size.int %2582, %int1_618 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_619 = torch.constant.int 0
%2979 = torch.aten.add.int %int0_619, %2978 : !torch.int, !torch.int -> !torch.int
%int0_620 = torch.constant.int 0
%int0_621 = torch.constant.int 0
%int1_622 = torch.constant.int 1
%2980 = torch.aten.slice.Tensor %2896, %int0_620, %int0_621, %2979, %int1_622 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2980, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_623 = torch.constant.int 1
%int0_624 = torch.constant.int 0
%int9223372036854775807_625 = torch.constant.int 9223372036854775807
%int1_626 = torch.constant.int 1
%2981 = torch.aten.slice.Tensor %2980, %int1_623, %int0_624, %int9223372036854775807_625, %int1_626 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2981, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_627 = torch.constant.int 0
%2982 = torch.aten.unsqueeze %2981, %int0_627 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2982, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_628 = torch.constant.int 2
%2983 = torch.aten.unsqueeze %2982, %int2_628 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2983, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_629 = torch.constant.int 3
%int0_630 = torch.constant.int 0
%int9223372036854775807_631 = torch.constant.int 9223372036854775807
%int1_632 = torch.constant.int 1
%2984 = torch.aten.slice.Tensor %2983, %int3_629, %int0_630, %int9223372036854775807_631, %int1_632 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2984, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%2985 = torch_c.to_builtin_tensor %2686 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_633 = arith.constant 1 : index
%dim_634 = tensor.dim %2985, %c1_633 : tensor<4x?x1x128xf16>
%2986 = flow.tensor.bitcast %2985 : tensor<4x?x1x128xf16>{%dim_634} -> tensor<4x?x1x64xcomplex<f16>>{%dim_634}
%2987 = torch_c.from_builtin_tensor %2986 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %2987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%2988 = torch.aten.mul.Tensor %2987, %2984 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%2989 = torch_c.to_builtin_tensor %2988 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_635 = arith.constant 1 : index
%dim_636 = tensor.dim %2989, %c1_635 : tensor<4x?x1x64xcomplex<f32>>
%2990 = flow.tensor.bitcast %2989 : tensor<4x?x1x64xcomplex<f32>>{%dim_636} -> tensor<4x?x1x128xf32>{%dim_636}
%2991 = torch_c.from_builtin_tensor %2990 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %2991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_637 = torch.constant.int 5
%2992 = torch.prims.convert_element_type %2991, %int5_637 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %2992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_638 = torch.constant.int 1
%2993 = torch.aten.size.int %2588, %int1_638 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_639 = torch.constant.int 0
%2994 = torch.aten.add.int %int0_639, %2993 : !torch.int, !torch.int -> !torch.int
%int0_640 = torch.constant.int 0
%int0_641 = torch.constant.int 0
%int1_642 = torch.constant.int 1
%2995 = torch.aten.slice.Tensor %2899, %int0_640, %int0_641, %2994, %int1_642 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2995, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_643 = torch.constant.int 1
%int0_644 = torch.constant.int 0
%int9223372036854775807_645 = torch.constant.int 9223372036854775807
%int1_646 = torch.constant.int 1
%2996 = torch.aten.slice.Tensor %2995, %int1_643, %int0_644, %int9223372036854775807_645, %int1_646 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %2996, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_647 = torch.constant.int 0
%2997 = torch.aten.unsqueeze %2996, %int0_647 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %2997, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_648 = torch.constant.int 2
%2998 = torch.aten.unsqueeze %2997, %int2_648 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2998, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_649 = torch.constant.int 3
%int0_650 = torch.constant.int 0
%int9223372036854775807_651 = torch.constant.int 9223372036854775807
%int1_652 = torch.constant.int 1
%2999 = torch.aten.slice.Tensor %2998, %int3_649, %int0_650, %int9223372036854775807_651, %int1_652 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %2999, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%3000 = torch_c.to_builtin_tensor %2688 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_653 = arith.constant 1 : index
%dim_654 = tensor.dim %3000, %c1_653 : tensor<4x?x1x128xf16>
%3001 = flow.tensor.bitcast %3000 : tensor<4x?x1x128xf16>{%dim_654} -> tensor<4x?x1x64xcomplex<f16>>{%dim_654}
%3002 = torch_c.from_builtin_tensor %3001 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %3002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%3003 = torch.aten.mul.Tensor %3002, %2999 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %3003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%3004 = torch_c.to_builtin_tensor %3003 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_655 = arith.constant 1 : index
%dim_656 = tensor.dim %3004, %c1_655 : tensor<4x?x1x64xcomplex<f32>>
%3005 = flow.tensor.bitcast %3004 : tensor<4x?x1x64xcomplex<f32>>{%dim_656} -> tensor<4x?x1x128xf32>{%dim_656}
%3006 = torch_c.from_builtin_tensor %3005 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %3006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_657 = torch.constant.int 5
%3007 = torch.prims.convert_element_type %3006, %int5_657 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %3007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_658 = torch.constant.int 1
%3008 = torch.aten.size.int %2594, %int1_658 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_659 = torch.constant.int 0
%3009 = torch.aten.add.int %int0_659, %3008 : !torch.int, !torch.int -> !torch.int
%int0_660 = torch.constant.int 0
%int0_661 = torch.constant.int 0
%int1_662 = torch.constant.int 1
%3010 = torch.aten.slice.Tensor %2902, %int0_660, %int0_661, %3009, %int1_662 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %3010, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_663 = torch.constant.int 1
%int0_664 = torch.constant.int 0
%int9223372036854775807_665 = torch.constant.int 9223372036854775807
%int1_666 = torch.constant.int 1
%3011 = torch.aten.slice.Tensor %3010, %int1_663, %int0_664, %int9223372036854775807_665, %int1_666 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %3011, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_667 = torch.constant.int 0
%3012 = torch.aten.unsqueeze %3011, %int0_667 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %3012, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_668 = torch.constant.int 2
%3013 = torch.aten.unsqueeze %3012, %int2_668 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %3013, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_669 = torch.constant.int 3
%int0_670 = torch.constant.int 0
%int9223372036854775807_671 = torch.constant.int 9223372036854775807
%int1_672 = torch.constant.int 1
%3014 = torch.aten.slice.Tensor %3013, %int3_669, %int0_670, %int9223372036854775807_671, %int1_672 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %3014, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%3015 = torch_c.to_builtin_tensor %2690 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_673 = arith.constant 1 : index
%dim_674 = tensor.dim %3015, %c1_673 : tensor<4x?x1x128xf16>
%3016 = flow.tensor.bitcast %3015 : tensor<4x?x1x128xf16>{%dim_674} -> tensor<4x?x1x64xcomplex<f16>>{%dim_674}
%3017 = torch_c.from_builtin_tensor %3016 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %3017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%3018 = torch.aten.mul.Tensor %3017, %3014 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %3018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%3019 = torch_c.to_builtin_tensor %3018 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_675 = arith.constant 1 : index
%dim_676 = tensor.dim %3019, %c1_675 : tensor<4x?x1x64xcomplex<f32>>
%3020 = flow.tensor.bitcast %3019 : tensor<4x?x1x64xcomplex<f32>>{%dim_676} -> tensor<4x?x1x128xf32>{%dim_676}
%3021 = torch_c.from_builtin_tensor %3020 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %3021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_677 = torch.constant.int 5
%3022 = torch.prims.convert_element_type %3021, %int5_677 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %3022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int0_678 = torch.constant.int 0
%3023 = torch.aten.size.int %2328, %int0_678 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32 = torch.constant.int 32
%int2_679 = torch.constant.int 2
%int16 = torch.constant.int 16
%int1_680 = torch.constant.int 1
%int128_681 = torch.constant.int 128
%3024 = torch.prim.ListConstruct %3023, %int32, %int2_679, %int16, %int1_680, %int128_681 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3025 = torch.aten.view %2328, %3024 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3025, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_682 = torch.constant.int 0
%3026 = torch.aten.size.int %2329, %int0_682 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_683 = torch.constant.int 32
%int2_684 = torch.constant.int 2
%int16_685 = torch.constant.int 16
%int1_686 = torch.constant.int 1
%int128_687 = torch.constant.int 128
%3027 = torch.prim.ListConstruct %3026, %int32_683, %int2_684, %int16_685, %int1_686, %int128_687 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3028 = torch.aten.view %2329, %3027 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3028, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_688 = torch.constant.int 0
%3029 = torch.aten.size.int %2330, %int0_688 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_689 = torch.constant.int 32
%int2_690 = torch.constant.int 2
%int16_691 = torch.constant.int 16
%int1_692 = torch.constant.int 1
%int128_693 = torch.constant.int 128
%3030 = torch.prim.ListConstruct %3029, %int32_689, %int2_690, %int16_691, %int1_692, %int128_693 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3031 = torch.aten.view %2330, %3030 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3031, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_694 = torch.constant.int 0
%3032 = torch.aten.size.int %2331, %int0_694 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_695 = torch.constant.int 32
%int2_696 = torch.constant.int 2
%int16_697 = torch.constant.int 16
%int1_698 = torch.constant.int 1
%int128_699 = torch.constant.int 128
%3033 = torch.prim.ListConstruct %3032, %int32_695, %int2_696, %int16_697, %int1_698, %int128_699 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3034 = torch.aten.view %2331, %3033 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3034, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_700 = torch.constant.int 0
%3035 = torch.aten.size.int %2332, %int0_700 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_701 = torch.constant.int 32
%int2_702 = torch.constant.int 2
%int16_703 = torch.constant.int 16
%int1_704 = torch.constant.int 1
%int128_705 = torch.constant.int 128
%3036 = torch.prim.ListConstruct %3035, %int32_701, %int2_702, %int16_703, %int1_704, %int128_705 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3037 = torch.aten.view %2332, %3036 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3037, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_706 = torch.constant.int 0
%3038 = torch.aten.size.int %2333, %int0_706 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_707 = torch.constant.int 32
%int2_708 = torch.constant.int 2
%int16_709 = torch.constant.int 16
%int1_710 = torch.constant.int 1
%int128_711 = torch.constant.int 128
%3039 = torch.prim.ListConstruct %3038, %int32_707, %int2_708, %int16_709, %int1_710, %int128_711 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3040 = torch.aten.view %2333, %3039 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3040, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_712 = torch.constant.int 0
%3041 = torch.aten.size.int %2334, %int0_712 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_713 = torch.constant.int 32
%int2_714 = torch.constant.int 2
%int16_715 = torch.constant.int 16
%int1_716 = torch.constant.int 1
%int128_717 = torch.constant.int 128
%3042 = torch.prim.ListConstruct %3041, %int32_713, %int2_714, %int16_715, %int1_716, %int128_717 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3043 = torch.aten.view %2334, %3042 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3043, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int0_718 = torch.constant.int 0
%3044 = torch.aten.size.int %2335, %int0_718 : !torch.vtensor<[?,131072],f16>, !torch.int -> !torch.int
%int32_719 = torch.constant.int 32
%int2_720 = torch.constant.int 2
%int16_721 = torch.constant.int 16
%int1_722 = torch.constant.int 1
%int128_723 = torch.constant.int 128
%3045 = torch.prim.ListConstruct %3044, %int32_719, %int2_720, %int16_721, %int1_722, %int128_723 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3046 = torch.aten.view %2335, %3045 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3046, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_724 = torch.constant.int 32
%3047 = torch.aten.mul.int %3023, %int32_724 : !torch.int, !torch.int -> !torch.int
%int2_725 = torch.constant.int 2
%3048 = torch.aten.mul.int %3047, %int2_725 : !torch.int, !torch.int -> !torch.int
%int16_726 = torch.constant.int 16
%int1_727 = torch.constant.int 1
%int128_728 = torch.constant.int 128
%3049 = torch.prim.ListConstruct %3048, %int16_726, %int1_727, %int128_728 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3050 = torch.aten.view %3025, %3049 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3050, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_729 = torch.constant.int 32
%3051 = torch.aten.mul.int %3026, %int32_729 : !torch.int, !torch.int -> !torch.int
%int2_730 = torch.constant.int 2
%3052 = torch.aten.mul.int %3051, %int2_730 : !torch.int, !torch.int -> !torch.int
%int16_731 = torch.constant.int 16
%int1_732 = torch.constant.int 1
%int128_733 = torch.constant.int 128
%3053 = torch.prim.ListConstruct %3052, %int16_731, %int1_732, %int128_733 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3054 = torch.aten.view %3028, %3053 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3054, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_734 = torch.constant.int 32
%3055 = torch.aten.mul.int %3029, %int32_734 : !torch.int, !torch.int -> !torch.int
%int2_735 = torch.constant.int 2
%3056 = torch.aten.mul.int %3055, %int2_735 : !torch.int, !torch.int -> !torch.int
%int16_736 = torch.constant.int 16
%int1_737 = torch.constant.int 1
%int128_738 = torch.constant.int 128
%3057 = torch.prim.ListConstruct %3056, %int16_736, %int1_737, %int128_738 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3058 = torch.aten.view %3031, %3057 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3058, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_739 = torch.constant.int 32
%3059 = torch.aten.mul.int %3032, %int32_739 : !torch.int, !torch.int -> !torch.int
%int2_740 = torch.constant.int 2
%3060 = torch.aten.mul.int %3059, %int2_740 : !torch.int, !torch.int -> !torch.int
%int16_741 = torch.constant.int 16
%int1_742 = torch.constant.int 1
%int128_743 = torch.constant.int 128
%3061 = torch.prim.ListConstruct %3060, %int16_741, %int1_742, %int128_743 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3062 = torch.aten.view %3034, %3061 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3062, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_744 = torch.constant.int 32
%3063 = torch.aten.mul.int %3035, %int32_744 : !torch.int, !torch.int -> !torch.int
%int2_745 = torch.constant.int 2
%3064 = torch.aten.mul.int %3063, %int2_745 : !torch.int, !torch.int -> !torch.int
%int16_746 = torch.constant.int 16
%int1_747 = torch.constant.int 1
%int128_748 = torch.constant.int 128
%3065 = torch.prim.ListConstruct %3064, %int16_746, %int1_747, %int128_748 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3066 = torch.aten.view %3037, %3065 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3066, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_749 = torch.constant.int 32
%3067 = torch.aten.mul.int %3038, %int32_749 : !torch.int, !torch.int -> !torch.int
%int2_750 = torch.constant.int 2
%3068 = torch.aten.mul.int %3067, %int2_750 : !torch.int, !torch.int -> !torch.int
%int16_751 = torch.constant.int 16
%int1_752 = torch.constant.int 1
%int128_753 = torch.constant.int 128
%3069 = torch.prim.ListConstruct %3068, %int16_751, %int1_752, %int128_753 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3070 = torch.aten.view %3040, %3069 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3070, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_754 = torch.constant.int 32
%3071 = torch.aten.mul.int %3041, %int32_754 : !torch.int, !torch.int -> !torch.int
%int2_755 = torch.constant.int 2
%3072 = torch.aten.mul.int %3071, %int2_755 : !torch.int, !torch.int -> !torch.int
%int16_756 = torch.constant.int 16
%int1_757 = torch.constant.int 1
%int128_758 = torch.constant.int 128
%3073 = torch.prim.ListConstruct %3072, %int16_756, %int1_757, %int128_758 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3074 = torch.aten.view %3043, %3073 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3074, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_759 = torch.constant.int 32
%3075 = torch.aten.mul.int %3044, %int32_759 : !torch.int, !torch.int -> !torch.int
%int2_760 = torch.constant.int 2
%3076 = torch.aten.mul.int %3075, %int2_760 : !torch.int, !torch.int -> !torch.int
%int16_761 = torch.constant.int 16
%int1_762 = torch.constant.int 1
%int128_763 = torch.constant.int 128
%3077 = torch.prim.ListConstruct %3076, %int16_761, %int1_762, %int128_763 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3078 = torch.aten.view %3046, %3077 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3078, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int64_764 = torch.constant.int 64
%3079 = torch.aten.mul.Scalar %2364, %int64_764 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3079, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_765 = torch.constant.int 64
%3080 = torch.aten.mul.Scalar %2367, %int64_765 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3080, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_766 = torch.constant.int 64
%3081 = torch.aten.mul.Scalar %2370, %int64_766 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3081, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_767 = torch.constant.int 64
%3082 = torch.aten.mul.Scalar %2373, %int64_767 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3082, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_768 = torch.constant.int 64
%3083 = torch.aten.mul.Scalar %2376, %int64_768 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3083, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_769 = torch.constant.int 64
%3084 = torch.aten.mul.Scalar %2379, %int64_769 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3084, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_770 = torch.constant.int 64
%3085 = torch.aten.mul.Scalar %2382, %int64_770 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3085, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_771 = torch.constant.int 64
%3086 = torch.aten.mul.Scalar %2385, %int64_771 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3086, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_772 = torch.constant.int 0
%int1_773 = torch.constant.int 1
%3087 = torch.aten.add.Scalar %3079, %int0_772, %int1_773 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3087, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_774 = torch.constant.int 0
%int1_775 = torch.constant.int 1
%3088 = torch.aten.add.Scalar %3080, %int0_774, %int1_775 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3088, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_776 = torch.constant.int 0
%int1_777 = torch.constant.int 1
%3089 = torch.aten.add.Scalar %3081, %int0_776, %int1_777 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3089, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_778 = torch.constant.int 0
%int1_779 = torch.constant.int 1
%3090 = torch.aten.add.Scalar %3082, %int0_778, %int1_779 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3090, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_780 = torch.constant.int 0
%int1_781 = torch.constant.int 1
%3091 = torch.aten.add.Scalar %3083, %int0_780, %int1_781 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3091, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_782 = torch.constant.int 0
%int1_783 = torch.constant.int 1
%3092 = torch.aten.add.Scalar %3084, %int0_782, %int1_783 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3092, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_784 = torch.constant.int 0
%int1_785 = torch.constant.int 1
%3093 = torch.aten.add.Scalar %3085, %int0_784, %int1_785 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3093, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int0_786 = torch.constant.int 0
%int1_787 = torch.constant.int 1
%3094 = torch.aten.add.Scalar %3086, %int0_786, %int1_787 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3094, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_788 = torch.constant.int 1
%3095 = torch.aten.size.int %arg2, %int1_788 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.int
%int4_789 = torch.constant.int 4
%int16_790 = torch.constant.int 16
%int1_791 = torch.constant.int 1
%int128_792 = torch.constant.int 128
%3096 = torch.prim.ListConstruct %int4_789, %3095, %int16_790, %int1_791, %int128_792 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3097 = torch.aten.view %2917, %3096 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3097, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_793 = torch.constant.int 4
%int16_794 = torch.constant.int 16
%int1_795 = torch.constant.int 1
%int128_796 = torch.constant.int 128
%3098 = torch.prim.ListConstruct %int4_793, %3095, %int16_794, %int1_795, %int128_796 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3099 = torch.aten.view %2932, %3098 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3099, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_797 = torch.constant.int 4
%int16_798 = torch.constant.int 16
%int1_799 = torch.constant.int 1
%int128_800 = torch.constant.int 128
%3100 = torch.prim.ListConstruct %int4_797, %3095, %int16_798, %int1_799, %int128_800 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3101 = torch.aten.view %2947, %3100 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3101, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_801 = torch.constant.int 4
%int16_802 = torch.constant.int 16
%int1_803 = torch.constant.int 1
%int128_804 = torch.constant.int 128
%3102 = torch.prim.ListConstruct %int4_801, %3095, %int16_802, %int1_803, %int128_804 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3103 = torch.aten.view %2962, %3102 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3103, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_805 = torch.constant.int 4
%int16_806 = torch.constant.int 16
%int1_807 = torch.constant.int 1
%int128_808 = torch.constant.int 128
%3104 = torch.prim.ListConstruct %int4_805, %3095, %int16_806, %int1_807, %int128_808 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3105 = torch.aten.view %2977, %3104 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3105, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_809 = torch.constant.int 4
%int16_810 = torch.constant.int 16
%int1_811 = torch.constant.int 1
%int128_812 = torch.constant.int 128
%3106 = torch.prim.ListConstruct %int4_809, %3095, %int16_810, %int1_811, %int128_812 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3107 = torch.aten.view %2992, %3106 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3107, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_813 = torch.constant.int 4
%int16_814 = torch.constant.int 16
%int1_815 = torch.constant.int 1
%int128_816 = torch.constant.int 128
%3108 = torch.prim.ListConstruct %int4_813, %3095, %int16_814, %int1_815, %int128_816 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3109 = torch.aten.view %3007, %3108 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3109, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_817 = torch.constant.int 4
%int16_818 = torch.constant.int 16
%int1_819 = torch.constant.int 1
%int128_820 = torch.constant.int 128
%3110 = torch.prim.ListConstruct %int4_817, %3095, %int16_818, %int1_819, %int128_820 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3111 = torch.aten.view %3022, %3110 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3111, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_821 = torch.constant.int 4
%3112 = torch.aten.mul.int %int4_821, %3095 : !torch.int, !torch.int -> !torch.int
%int16_822 = torch.constant.int 16
%int1_823 = torch.constant.int 1
%int128_824 = torch.constant.int 128
%3113 = torch.prim.ListConstruct %3112, %int16_822, %int1_823, %int128_824 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3114 = torch.aten.view %3097, %3113 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3114, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_825 = torch.constant.int 4
%3115 = torch.aten.mul.int %int4_825, %3095 : !torch.int, !torch.int -> !torch.int
%int16_826 = torch.constant.int 16
%int1_827 = torch.constant.int 1
%int128_828 = torch.constant.int 128
%3116 = torch.prim.ListConstruct %3115, %int16_826, %int1_827, %int128_828 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3117 = torch.aten.view %3099, %3116 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3117, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_829 = torch.constant.int 4
%3118 = torch.aten.mul.int %int4_829, %3095 : !torch.int, !torch.int -> !torch.int
%int16_830 = torch.constant.int 16
%int1_831 = torch.constant.int 1
%int128_832 = torch.constant.int 128
%3119 = torch.prim.ListConstruct %3118, %int16_830, %int1_831, %int128_832 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3120 = torch.aten.view %3101, %3119 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3120, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_833 = torch.constant.int 4
%3121 = torch.aten.mul.int %int4_833, %3095 : !torch.int, !torch.int -> !torch.int
%int16_834 = torch.constant.int 16
%int1_835 = torch.constant.int 1
%int128_836 = torch.constant.int 128
%3122 = torch.prim.ListConstruct %3121, %int16_834, %int1_835, %int128_836 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3123 = torch.aten.view %3103, %3122 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3123, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_837 = torch.constant.int 4
%3124 = torch.aten.mul.int %int4_837, %3095 : !torch.int, !torch.int -> !torch.int
%int16_838 = torch.constant.int 16
%int1_839 = torch.constant.int 1
%int128_840 = torch.constant.int 128
%3125 = torch.prim.ListConstruct %3124, %int16_838, %int1_839, %int128_840 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3126 = torch.aten.view %3105, %3125 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3126, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_841 = torch.constant.int 4
%3127 = torch.aten.mul.int %int4_841, %3095 : !torch.int, !torch.int -> !torch.int
%int16_842 = torch.constant.int 16
%int1_843 = torch.constant.int 1
%int128_844 = torch.constant.int 128
%3128 = torch.prim.ListConstruct %3127, %int16_842, %int1_843, %int128_844 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3129 = torch.aten.view %3107, %3128 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3129, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_845 = torch.constant.int 4
%3130 = torch.aten.mul.int %int4_845, %3095 : !torch.int, !torch.int -> !torch.int
%int16_846 = torch.constant.int 16
%int1_847 = torch.constant.int 1
%int128_848 = torch.constant.int 128
%3131 = torch.prim.ListConstruct %3130, %int16_846, %int1_847, %int128_848 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3132 = torch.aten.view %3109, %3131 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3132, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_849 = torch.constant.int 4
%3133 = torch.aten.mul.int %int4_849, %3095 : !torch.int, !torch.int -> !torch.int
%int16_850 = torch.constant.int 16
%int1_851 = torch.constant.int 1
%int128_852 = torch.constant.int 128
%3134 = torch.prim.ListConstruct %3133, %int16_850, %int1_851, %int128_852 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3135 = torch.aten.view %3111, %3134 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3135, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_853 = torch.constant.int 4
%3136 = torch.aten.mul.int %int4_853, %3095 : !torch.int, !torch.int -> !torch.int
%3137 = torch.prim.ListConstruct %3136 : (!torch.int) -> !torch.list<int>
%3138 = torch.aten.view %3087, %3137 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3138, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_854 = torch.constant.int 4
%3139 = torch.aten.mul.int %int4_854, %3095 : !torch.int, !torch.int -> !torch.int
%3140 = torch.prim.ListConstruct %3139 : (!torch.int) -> !torch.list<int>
%3141 = torch.aten.view %3088, %3140 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3141, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_855 = torch.constant.int 4
%3142 = torch.aten.mul.int %int4_855, %3095 : !torch.int, !torch.int -> !torch.int
%3143 = torch.prim.ListConstruct %3142 : (!torch.int) -> !torch.list<int>
%3144 = torch.aten.view %3089, %3143 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3144, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_856 = torch.constant.int 4
%3145 = torch.aten.mul.int %int4_856, %3095 : !torch.int, !torch.int -> !torch.int
%3146 = torch.prim.ListConstruct %3145 : (!torch.int) -> !torch.list<int>
%3147 = torch.aten.view %3090, %3146 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3147, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_857 = torch.constant.int 4
%3148 = torch.aten.mul.int %int4_857, %3095 : !torch.int, !torch.int -> !torch.int
%3149 = torch.prim.ListConstruct %3148 : (!torch.int) -> !torch.list<int>
%3150 = torch.aten.view %3091, %3149 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3150, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_858 = torch.constant.int 4
%3151 = torch.aten.mul.int %int4_858, %3095 : !torch.int, !torch.int -> !torch.int
%3152 = torch.prim.ListConstruct %3151 : (!torch.int) -> !torch.list<int>
%3153 = torch.aten.view %3092, %3152 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3153, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_859 = torch.constant.int 4
%3154 = torch.aten.mul.int %int4_859, %3095 : !torch.int, !torch.int -> !torch.int
%3155 = torch.prim.ListConstruct %3154 : (!torch.int) -> !torch.list<int>
%3156 = torch.aten.view %3093, %3155 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3156, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_860 = torch.constant.int 4
%3157 = torch.aten.mul.int %int4_860, %3095 : !torch.int, !torch.int -> !torch.int
%3158 = torch.prim.ListConstruct %3157 : (!torch.int) -> !torch.list<int>
%3159 = torch.aten.view %3094, %3158 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3159, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_861 = torch.constant.int 4
%int16_862 = torch.constant.int 16
%int1_863 = torch.constant.int 1
%int128_864 = torch.constant.int 128
%3160 = torch.prim.ListConstruct %int4_861, %3095, %int16_862, %int1_863, %int128_864 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3161 = torch.aten.view %2692, %3160 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3161, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_865 = torch.constant.int 4
%int16_866 = torch.constant.int 16
%int1_867 = torch.constant.int 1
%int128_868 = torch.constant.int 128
%3162 = torch.prim.ListConstruct %int4_865, %3095, %int16_866, %int1_867, %int128_868 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3163 = torch.aten.view %2694, %3162 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3163, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_869 = torch.constant.int 4
%int16_870 = torch.constant.int 16
%int1_871 = torch.constant.int 1
%int128_872 = torch.constant.int 128
%3164 = torch.prim.ListConstruct %int4_869, %3095, %int16_870, %int1_871, %int128_872 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3165 = torch.aten.view %2696, %3164 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3165, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_873 = torch.constant.int 4
%int16_874 = torch.constant.int 16
%int1_875 = torch.constant.int 1
%int128_876 = torch.constant.int 128
%3166 = torch.prim.ListConstruct %int4_873, %3095, %int16_874, %int1_875, %int128_876 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3167 = torch.aten.view %2698, %3166 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3167, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_877 = torch.constant.int 4
%int16_878 = torch.constant.int 16
%int1_879 = torch.constant.int 1
%int128_880 = torch.constant.int 128
%3168 = torch.prim.ListConstruct %int4_877, %3095, %int16_878, %int1_879, %int128_880 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3169 = torch.aten.view %2700, %3168 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3169, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_881 = torch.constant.int 4
%int16_882 = torch.constant.int 16
%int1_883 = torch.constant.int 1
%int128_884 = torch.constant.int 128
%3170 = torch.prim.ListConstruct %int4_881, %3095, %int16_882, %int1_883, %int128_884 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3171 = torch.aten.view %2702, %3170 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3171, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_885 = torch.constant.int 4
%int16_886 = torch.constant.int 16
%int1_887 = torch.constant.int 1
%int128_888 = torch.constant.int 128
%3172 = torch.prim.ListConstruct %int4_885, %3095, %int16_886, %int1_887, %int128_888 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3173 = torch.aten.view %2704, %3172 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3173, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_889 = torch.constant.int 4
%int16_890 = torch.constant.int 16
%int1_891 = torch.constant.int 1
%int128_892 = torch.constant.int 128
%3174 = torch.prim.ListConstruct %int4_889, %3095, %int16_890, %int1_891, %int128_892 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3175 = torch.aten.view %2706, %3174 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %3175, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_893 = torch.constant.int 4
%3176 = torch.aten.mul.int %int4_893, %3095 : !torch.int, !torch.int -> !torch.int
%int16_894 = torch.constant.int 16
%int1_895 = torch.constant.int 1
%int128_896 = torch.constant.int 128
%3177 = torch.prim.ListConstruct %3176, %int16_894, %int1_895, %int128_896 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3178 = torch.aten.view %3161, %3177 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3178, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_897 = torch.constant.int 4
%3179 = torch.aten.mul.int %int4_897, %3095 : !torch.int, !torch.int -> !torch.int
%int16_898 = torch.constant.int 16
%int1_899 = torch.constant.int 1
%int128_900 = torch.constant.int 128
%3180 = torch.prim.ListConstruct %3179, %int16_898, %int1_899, %int128_900 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3181 = torch.aten.view %3163, %3180 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3181, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_901 = torch.constant.int 4
%3182 = torch.aten.mul.int %int4_901, %3095 : !torch.int, !torch.int -> !torch.int
%int16_902 = torch.constant.int 16
%int1_903 = torch.constant.int 1
%int128_904 = torch.constant.int 128
%3183 = torch.prim.ListConstruct %3182, %int16_902, %int1_903, %int128_904 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3184 = torch.aten.view %3165, %3183 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3184, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_905 = torch.constant.int 4
%3185 = torch.aten.mul.int %int4_905, %3095 : !torch.int, !torch.int -> !torch.int
%int16_906 = torch.constant.int 16
%int1_907 = torch.constant.int 1
%int128_908 = torch.constant.int 128
%3186 = torch.prim.ListConstruct %3185, %int16_906, %int1_907, %int128_908 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3187 = torch.aten.view %3167, %3186 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3187, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_909 = torch.constant.int 4
%3188 = torch.aten.mul.int %int4_909, %3095 : !torch.int, !torch.int -> !torch.int
%int16_910 = torch.constant.int 16
%int1_911 = torch.constant.int 1
%int128_912 = torch.constant.int 128
%3189 = torch.prim.ListConstruct %3188, %int16_910, %int1_911, %int128_912 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3190 = torch.aten.view %3169, %3189 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3190, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_913 = torch.constant.int 4
%3191 = torch.aten.mul.int %int4_913, %3095 : !torch.int, !torch.int -> !torch.int
%int16_914 = torch.constant.int 16
%int1_915 = torch.constant.int 1
%int128_916 = torch.constant.int 128
%3192 = torch.prim.ListConstruct %3191, %int16_914, %int1_915, %int128_916 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3193 = torch.aten.view %3171, %3192 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3193, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_917 = torch.constant.int 4
%3194 = torch.aten.mul.int %int4_917, %3095 : !torch.int, !torch.int -> !torch.int
%int16_918 = torch.constant.int 16
%int1_919 = torch.constant.int 1
%int128_920 = torch.constant.int 128
%3195 = torch.prim.ListConstruct %3194, %int16_918, %int1_919, %int128_920 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3196 = torch.aten.view %3173, %3195 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3196, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_921 = torch.constant.int 4
%3197 = torch.aten.mul.int %int4_921, %3095 : !torch.int, !torch.int -> !torch.int
%int16_922 = torch.constant.int 16
%int1_923 = torch.constant.int 1
%int128_924 = torch.constant.int 128
%3198 = torch.prim.ListConstruct %3197, %int16_922, %int1_923, %int128_924 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3199 = torch.aten.view %3175, %3198 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3199, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_925 = torch.constant.int 1
%int1_926 = torch.constant.int 1
%3200 = torch.aten.add.Scalar %3087, %int1_925, %int1_926 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3200, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_927 = torch.constant.int 1
%int1_928 = torch.constant.int 1
%3201 = torch.aten.add.Scalar %3088, %int1_927, %int1_928 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3201, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_929 = torch.constant.int 1
%int1_930 = torch.constant.int 1
%3202 = torch.aten.add.Scalar %3089, %int1_929, %int1_930 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3202, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_931 = torch.constant.int 1
%int1_932 = torch.constant.int 1
%3203 = torch.aten.add.Scalar %3090, %int1_931, %int1_932 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3203, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_933 = torch.constant.int 1
%int1_934 = torch.constant.int 1
%3204 = torch.aten.add.Scalar %3091, %int1_933, %int1_934 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3204, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_935 = torch.constant.int 1
%int1_936 = torch.constant.int 1
%3205 = torch.aten.add.Scalar %3092, %int1_935, %int1_936 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3205, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_937 = torch.constant.int 1
%int1_938 = torch.constant.int 1
%3206 = torch.aten.add.Scalar %3093, %int1_937, %int1_938 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3206, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_939 = torch.constant.int 1
%int1_940 = torch.constant.int 1
%3207 = torch.aten.add.Scalar %3094, %int1_939, %int1_940 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %3207, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_941 = torch.constant.int 4
%3208 = torch.aten.mul.int %int4_941, %3095 : !torch.int, !torch.int -> !torch.int
%3209 = torch.prim.ListConstruct %3208 : (!torch.int) -> !torch.list<int>
%3210 = torch.aten.view %3200, %3209 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3210, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_942 = torch.constant.int 4
%3211 = torch.aten.mul.int %int4_942, %3095 : !torch.int, !torch.int -> !torch.int
%3212 = torch.prim.ListConstruct %3211 : (!torch.int) -> !torch.list<int>
%3213 = torch.aten.view %3201, %3212 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3213, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_943 = torch.constant.int 4
%3214 = torch.aten.mul.int %int4_943, %3095 : !torch.int, !torch.int -> !torch.int
%3215 = torch.prim.ListConstruct %3214 : (!torch.int) -> !torch.list<int>
%3216 = torch.aten.view %3202, %3215 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3216, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_944 = torch.constant.int 4
%3217 = torch.aten.mul.int %int4_944, %3095 : !torch.int, !torch.int -> !torch.int
%3218 = torch.prim.ListConstruct %3217 : (!torch.int) -> !torch.list<int>
%3219 = torch.aten.view %3203, %3218 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3219, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_945 = torch.constant.int 4
%3220 = torch.aten.mul.int %int4_945, %3095 : !torch.int, !torch.int -> !torch.int
%3221 = torch.prim.ListConstruct %3220 : (!torch.int) -> !torch.list<int>
%3222 = torch.aten.view %3204, %3221 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3222, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_946 = torch.constant.int 4
%3223 = torch.aten.mul.int %int4_946, %3095 : !torch.int, !torch.int -> !torch.int
%3224 = torch.prim.ListConstruct %3223 : (!torch.int) -> !torch.list<int>
%3225 = torch.aten.view %3205, %3224 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3225, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_947 = torch.constant.int 4
%3226 = torch.aten.mul.int %int4_947, %3095 : !torch.int, !torch.int -> !torch.int
%3227 = torch.prim.ListConstruct %3226 : (!torch.int) -> !torch.list<int>
%3228 = torch.aten.view %3206, %3227 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3228, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_948 = torch.constant.int 4
%3229 = torch.aten.mul.int %int4_948, %3095 : !torch.int, !torch.int -> !torch.int
%3230 = torch.prim.ListConstruct %3229 : (!torch.int) -> !torch.list<int>
%3231 = torch.aten.view %3207, %3230 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3231, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%3232 = torch.prim.ListConstruct %3138, %3210 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_949 = torch.constant.int 0
%3233 = torch.aten.cat %3232, %int0_949 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3233, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3234 = torch.prim.ListConstruct %3141, %3213 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_950 = torch.constant.int 0
%3235 = torch.aten.cat %3234, %int0_950 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3235, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3236 = torch.prim.ListConstruct %3144, %3216 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_951 = torch.constant.int 0
%3237 = torch.aten.cat %3236, %int0_951 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3237, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3238 = torch.prim.ListConstruct %3147, %3219 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_952 = torch.constant.int 0
%3239 = torch.aten.cat %3238, %int0_952 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3239, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3240 = torch.prim.ListConstruct %3150, %3222 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_953 = torch.constant.int 0
%3241 = torch.aten.cat %3240, %int0_953 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3241, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3242 = torch.prim.ListConstruct %3153, %3225 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_954 = torch.constant.int 0
%3243 = torch.aten.cat %3242, %int0_954 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3243, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3244 = torch.prim.ListConstruct %3156, %3228 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_955 = torch.constant.int 0
%3245 = torch.aten.cat %3244, %int0_955 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3245, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3246 = torch.prim.ListConstruct %3159, %3231 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_956 = torch.constant.int 0
%3247 = torch.aten.cat %3246, %int0_956 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %3247, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%3248 = torch.prim.ListConstruct %3114, %3178 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_957 = torch.constant.int 0
%3249 = torch.aten.cat %3248, %int0_957 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3249, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3250 = torch.prim.ListConstruct %3117, %3181 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_958 = torch.constant.int 0
%3251 = torch.aten.cat %3250, %int0_958 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3251, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3252 = torch.prim.ListConstruct %3120, %3184 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_959 = torch.constant.int 0
%3253 = torch.aten.cat %3252, %int0_959 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3253, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3254 = torch.prim.ListConstruct %3123, %3187 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_960 = torch.constant.int 0
%3255 = torch.aten.cat %3254, %int0_960 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3255, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3256 = torch.prim.ListConstruct %3126, %3190 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_961 = torch.constant.int 0
%3257 = torch.aten.cat %3256, %int0_961 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3257, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3258 = torch.prim.ListConstruct %3129, %3193 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_962 = torch.constant.int 0
%3259 = torch.aten.cat %3258, %int0_962 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3259, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3260 = torch.prim.ListConstruct %3132, %3196 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_963 = torch.constant.int 0
%3261 = torch.aten.cat %3260, %int0_963 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3261, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3262 = torch.prim.ListConstruct %3135, %3199 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_964 = torch.constant.int 0
%3263 = torch.aten.cat %3262, %int0_964 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3263, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%3264 = torch.prim.ListConstruct %3233 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_965 = torch.constant.bool false
%3265 = torch.aten.index_put %3050, %3264, %3249, %false_965 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3265, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_966 = torch.constant.int 32
%int2_967 = torch.constant.int 2
%int16_968 = torch.constant.int 16
%int1_969 = torch.constant.int 1
%int128_970 = torch.constant.int 128
%3266 = torch.prim.ListConstruct %3023, %int32_966, %int2_967, %int16_968, %int1_969, %int128_970 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3267 = torch.aten.view %3265, %3266 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3267, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_971 = torch.constant.int 131072
%3268 = torch.prim.ListConstruct %3023, %int131072_971 : (!torch.int, !torch.int) -> !torch.list<int>
%3269 = torch.aten.view %3267, %3268 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3269, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3270 = torch.prim.ListConstruct %3235 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_972 = torch.constant.bool false
%3271 = torch.aten.index_put %3054, %3270, %3251, %false_972 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3271, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_973 = torch.constant.int 32
%int2_974 = torch.constant.int 2
%int16_975 = torch.constant.int 16
%int1_976 = torch.constant.int 1
%int128_977 = torch.constant.int 128
%3272 = torch.prim.ListConstruct %3026, %int32_973, %int2_974, %int16_975, %int1_976, %int128_977 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3273 = torch.aten.view %3271, %3272 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3273, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_978 = torch.constant.int 131072
%3274 = torch.prim.ListConstruct %3026, %int131072_978 : (!torch.int, !torch.int) -> !torch.list<int>
%3275 = torch.aten.view %3273, %3274 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3275, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3276 = torch.prim.ListConstruct %3237 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_979 = torch.constant.bool false
%3277 = torch.aten.index_put %3058, %3276, %3253, %false_979 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3277, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_980 = torch.constant.int 32
%int2_981 = torch.constant.int 2
%int16_982 = torch.constant.int 16
%int1_983 = torch.constant.int 1
%int128_984 = torch.constant.int 128
%3278 = torch.prim.ListConstruct %3029, %int32_980, %int2_981, %int16_982, %int1_983, %int128_984 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3279 = torch.aten.view %3277, %3278 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3279, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_985 = torch.constant.int 131072
%3280 = torch.prim.ListConstruct %3029, %int131072_985 : (!torch.int, !torch.int) -> !torch.list<int>
%3281 = torch.aten.view %3279, %3280 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3281, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3282 = torch.prim.ListConstruct %3239 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_986 = torch.constant.bool false
%3283 = torch.aten.index_put %3062, %3282, %3255, %false_986 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3283, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_987 = torch.constant.int 32
%int2_988 = torch.constant.int 2
%int16_989 = torch.constant.int 16
%int1_990 = torch.constant.int 1
%int128_991 = torch.constant.int 128
%3284 = torch.prim.ListConstruct %3032, %int32_987, %int2_988, %int16_989, %int1_990, %int128_991 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3285 = torch.aten.view %3283, %3284 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3285, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_992 = torch.constant.int 131072
%3286 = torch.prim.ListConstruct %3032, %int131072_992 : (!torch.int, !torch.int) -> !torch.list<int>
%3287 = torch.aten.view %3285, %3286 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3287, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3288 = torch.prim.ListConstruct %3241 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_993 = torch.constant.bool false
%3289 = torch.aten.index_put %3066, %3288, %3257, %false_993 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3289, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_994 = torch.constant.int 32
%int2_995 = torch.constant.int 2
%int16_996 = torch.constant.int 16
%int1_997 = torch.constant.int 1
%int128_998 = torch.constant.int 128
%3290 = torch.prim.ListConstruct %3035, %int32_994, %int2_995, %int16_996, %int1_997, %int128_998 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3291 = torch.aten.view %3289, %3290 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3291, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_999 = torch.constant.int 131072
%3292 = torch.prim.ListConstruct %3035, %int131072_999 : (!torch.int, !torch.int) -> !torch.list<int>
%3293 = torch.aten.view %3291, %3292 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3293, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3294 = torch.prim.ListConstruct %3243 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_1000 = torch.constant.bool false
%3295 = torch.aten.index_put %3070, %3294, %3259, %false_1000 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3295, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_1001 = torch.constant.int 32
%int2_1002 = torch.constant.int 2
%int16_1003 = torch.constant.int 16
%int1_1004 = torch.constant.int 1
%int128_1005 = torch.constant.int 128
%3296 = torch.prim.ListConstruct %3038, %int32_1001, %int2_1002, %int16_1003, %int1_1004, %int128_1005 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3297 = torch.aten.view %3295, %3296 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3297, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_1006 = torch.constant.int 131072
%3298 = torch.prim.ListConstruct %3038, %int131072_1006 : (!torch.int, !torch.int) -> !torch.list<int>
%3299 = torch.aten.view %3297, %3298 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3299, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3300 = torch.prim.ListConstruct %3245 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_1007 = torch.constant.bool false
%3301 = torch.aten.index_put %3074, %3300, %3261, %false_1007 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3301, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_1008 = torch.constant.int 32
%int2_1009 = torch.constant.int 2
%int16_1010 = torch.constant.int 16
%int1_1011 = torch.constant.int 1
%int128_1012 = torch.constant.int 128
%3302 = torch.prim.ListConstruct %3041, %int32_1008, %int2_1009, %int16_1010, %int1_1011, %int128_1012 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3303 = torch.aten.view %3301, %3302 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3303, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_1013 = torch.constant.int 131072
%3304 = torch.prim.ListConstruct %3041, %int131072_1013 : (!torch.int, !torch.int) -> !torch.list<int>
%3305 = torch.aten.view %3303, %3304 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3305, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%3306 = torch.prim.ListConstruct %3247 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_1014 = torch.constant.bool false
%3307 = torch.aten.index_put %3078, %3306, %3263, %false_1014 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %3307, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_1015 = torch.constant.int 32
%int2_1016 = torch.constant.int 2
%int16_1017 = torch.constant.int 16
%int1_1018 = torch.constant.int 1
%int128_1019 = torch.constant.int 128
%3308 = torch.prim.ListConstruct %3044, %int32_1015, %int2_1016, %int16_1017, %int1_1018, %int128_1019 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3309 = torch.aten.view %3307, %3308 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %3309, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_1020 = torch.constant.int 131072
%3310 = torch.prim.ListConstruct %3044, %int131072_1020 : (!torch.int, !torch.int) -> !torch.list<int>
%3311 = torch.aten.view %3309, %3310 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %3311, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2 = torch.constant.int -2
%3312 = torch.aten.unsqueeze %2917, %int-2 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1021 = torch.constant.int -2
%3313 = torch.aten.unsqueeze %2932, %int-2_1021 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1022 = torch.constant.int -2
%3314 = torch.aten.unsqueeze %2947, %int-2_1022 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1023 = torch.constant.int -2
%3315 = torch.aten.unsqueeze %2962, %int-2_1023 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1024 = torch.constant.int -2
%3316 = torch.aten.unsqueeze %2977, %int-2_1024 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1025 = torch.constant.int -2
%3317 = torch.aten.unsqueeze %2992, %int-2_1025 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1026 = torch.constant.int -2
%3318 = torch.aten.unsqueeze %3007, %int-2_1026 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1027 = torch.constant.int -2
%3319 = torch.aten.unsqueeze %3022, %int-2_1027 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_1028 = torch.constant.int 4
%int1_1029 = torch.constant.int 1
%int4_1030 = torch.constant.int 4
%int128_1031 = torch.constant.int 128
%3320 = torch.prim.ListConstruct %int4_1028, %2903, %int1_1029, %int4_1030, %int128_1031 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1032 = torch.constant.bool false
%3321 = torch.aten.expand %3312, %3320, %false_1032 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1033 = torch.constant.int 4
%int1_1034 = torch.constant.int 1
%int4_1035 = torch.constant.int 4
%int128_1036 = torch.constant.int 128
%3322 = torch.prim.ListConstruct %int4_1033, %2903, %int1_1034, %int4_1035, %int128_1036 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1037 = torch.constant.bool false
%3323 = torch.aten.expand %3313, %3322, %false_1037 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1038 = torch.constant.int 4
%int1_1039 = torch.constant.int 1
%int4_1040 = torch.constant.int 4
%int128_1041 = torch.constant.int 128
%3324 = torch.prim.ListConstruct %int4_1038, %2903, %int1_1039, %int4_1040, %int128_1041 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1042 = torch.constant.bool false
%3325 = torch.aten.expand %3314, %3324, %false_1042 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1043 = torch.constant.int 4
%int1_1044 = torch.constant.int 1
%int4_1045 = torch.constant.int 4
%int128_1046 = torch.constant.int 128
%3326 = torch.prim.ListConstruct %int4_1043, %2903, %int1_1044, %int4_1045, %int128_1046 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1047 = torch.constant.bool false
%3327 = torch.aten.expand %3315, %3326, %false_1047 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1048 = torch.constant.int 4
%int1_1049 = torch.constant.int 1
%int4_1050 = torch.constant.int 4
%int128_1051 = torch.constant.int 128
%3328 = torch.prim.ListConstruct %int4_1048, %2903, %int1_1049, %int4_1050, %int128_1051 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1052 = torch.constant.bool false
%3329 = torch.aten.expand %3316, %3328, %false_1052 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1053 = torch.constant.int 4
%int1_1054 = torch.constant.int 1
%int4_1055 = torch.constant.int 4
%int128_1056 = torch.constant.int 128
%3330 = torch.prim.ListConstruct %int4_1053, %2903, %int1_1054, %int4_1055, %int128_1056 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1057 = torch.constant.bool false
%3331 = torch.aten.expand %3317, %3330, %false_1057 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1058 = torch.constant.int 4
%int1_1059 = torch.constant.int 1
%int4_1060 = torch.constant.int 4
%int128_1061 = torch.constant.int 128
%3332 = torch.prim.ListConstruct %int4_1058, %2903, %int1_1059, %int4_1060, %int128_1061 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1062 = torch.constant.bool false
%3333 = torch.aten.expand %3318, %3332, %false_1062 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1063 = torch.constant.int 4
%int1_1064 = torch.constant.int 1
%int4_1065 = torch.constant.int 4
%int128_1066 = torch.constant.int 128
%3334 = torch.prim.ListConstruct %int4_1063, %2903, %int1_1064, %int4_1065, %int128_1066 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1067 = torch.constant.bool false
%3335 = torch.aten.expand %3319, %3334, %false_1067 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1068 = torch.constant.int 4
%int4_1069 = torch.constant.int 4
%int128_1070 = torch.constant.int 128
%3336 = torch.prim.ListConstruct %int4_1068, %2903, %int4_1069, %int128_1070 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3337 = torch.aten.view %3321, %3336 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1071 = torch.constant.int 4
%int4_1072 = torch.constant.int 4
%int128_1073 = torch.constant.int 128
%3338 = torch.prim.ListConstruct %int4_1071, %2903, %int4_1072, %int128_1073 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3339 = torch.aten.view %3323, %3338 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1074 = torch.constant.int 4
%int4_1075 = torch.constant.int 4
%int128_1076 = torch.constant.int 128
%3340 = torch.prim.ListConstruct %int4_1074, %2903, %int4_1075, %int128_1076 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3341 = torch.aten.view %3325, %3340 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1077 = torch.constant.int 4
%int4_1078 = torch.constant.int 4
%int128_1079 = torch.constant.int 128
%3342 = torch.prim.ListConstruct %int4_1077, %2903, %int4_1078, %int128_1079 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3343 = torch.aten.view %3327, %3342 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1080 = torch.constant.int 4
%int4_1081 = torch.constant.int 4
%int128_1082 = torch.constant.int 128
%3344 = torch.prim.ListConstruct %int4_1080, %2903, %int4_1081, %int128_1082 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3345 = torch.aten.view %3329, %3344 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1083 = torch.constant.int 4
%int4_1084 = torch.constant.int 4
%int128_1085 = torch.constant.int 128
%3346 = torch.prim.ListConstruct %int4_1083, %2903, %int4_1084, %int128_1085 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3347 = torch.aten.view %3331, %3346 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1086 = torch.constant.int 4
%int4_1087 = torch.constant.int 4
%int128_1088 = torch.constant.int 128
%3348 = torch.prim.ListConstruct %int4_1086, %2903, %int4_1087, %int128_1088 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3349 = torch.aten.view %3333, %3348 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1089 = torch.constant.int 4
%int4_1090 = torch.constant.int 4
%int128_1091 = torch.constant.int 128
%3350 = torch.prim.ListConstruct %int4_1089, %2903, %int4_1090, %int128_1091 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3351 = torch.aten.view %3335, %3350 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_1092 = torch.constant.int -2
%3352 = torch.aten.unsqueeze %2692, %int-2_1092 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1093 = torch.constant.int -2
%3353 = torch.aten.unsqueeze %2694, %int-2_1093 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1094 = torch.constant.int -2
%3354 = torch.aten.unsqueeze %2696, %int-2_1094 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1095 = torch.constant.int -2
%3355 = torch.aten.unsqueeze %2698, %int-2_1095 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1096 = torch.constant.int -2
%3356 = torch.aten.unsqueeze %2700, %int-2_1096 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1097 = torch.constant.int -2
%3357 = torch.aten.unsqueeze %2702, %int-2_1097 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1098 = torch.constant.int -2
%3358 = torch.aten.unsqueeze %2704, %int-2_1098 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_1099 = torch.constant.int -2
%3359 = torch.aten.unsqueeze %2706, %int-2_1099 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %3359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_1100 = torch.constant.int 1
%3360 = torch.aten.size.int %2616, %int1_1100 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_1101 = torch.constant.int 4
%int1_1102 = torch.constant.int 1
%int4_1103 = torch.constant.int 4
%int128_1104 = torch.constant.int 128
%3361 = torch.prim.ListConstruct %int4_1101, %3360, %int1_1102, %int4_1103, %int128_1104 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1105 = torch.constant.bool false
%3362 = torch.aten.expand %3352, %3361, %false_1105 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1106 = torch.constant.int 4
%int1_1107 = torch.constant.int 1
%int4_1108 = torch.constant.int 4
%int128_1109 = torch.constant.int 128
%3363 = torch.prim.ListConstruct %int4_1106, %3360, %int1_1107, %int4_1108, %int128_1109 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1110 = torch.constant.bool false
%3364 = torch.aten.expand %3353, %3363, %false_1110 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1111 = torch.constant.int 4
%int1_1112 = torch.constant.int 1
%int4_1113 = torch.constant.int 4
%int128_1114 = torch.constant.int 128
%3365 = torch.prim.ListConstruct %int4_1111, %3360, %int1_1112, %int4_1113, %int128_1114 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1115 = torch.constant.bool false
%3366 = torch.aten.expand %3354, %3365, %false_1115 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1116 = torch.constant.int 4
%int1_1117 = torch.constant.int 1
%int4_1118 = torch.constant.int 4
%int128_1119 = torch.constant.int 128
%3367 = torch.prim.ListConstruct %int4_1116, %3360, %int1_1117, %int4_1118, %int128_1119 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1120 = torch.constant.bool false
%3368 = torch.aten.expand %3355, %3367, %false_1120 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1121 = torch.constant.int 4
%int1_1122 = torch.constant.int 1
%int4_1123 = torch.constant.int 4
%int128_1124 = torch.constant.int 128
%3369 = torch.prim.ListConstruct %int4_1121, %3360, %int1_1122, %int4_1123, %int128_1124 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1125 = torch.constant.bool false
%3370 = torch.aten.expand %3356, %3369, %false_1125 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1126 = torch.constant.int 4
%int1_1127 = torch.constant.int 1
%int4_1128 = torch.constant.int 4
%int128_1129 = torch.constant.int 128
%3371 = torch.prim.ListConstruct %int4_1126, %3360, %int1_1127, %int4_1128, %int128_1129 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1130 = torch.constant.bool false
%3372 = torch.aten.expand %3357, %3371, %false_1130 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1131 = torch.constant.int 4
%int1_1132 = torch.constant.int 1
%int4_1133 = torch.constant.int 4
%int128_1134 = torch.constant.int 128
%3373 = torch.prim.ListConstruct %int4_1131, %3360, %int1_1132, %int4_1133, %int128_1134 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1135 = torch.constant.bool false
%3374 = torch.aten.expand %3358, %3373, %false_1135 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1136 = torch.constant.int 4
%int1_1137 = torch.constant.int 1
%int4_1138 = torch.constant.int 4
%int128_1139 = torch.constant.int 128
%3375 = torch.prim.ListConstruct %int4_1136, %3360, %int1_1137, %int4_1138, %int128_1139 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_1140 = torch.constant.bool false
%3376 = torch.aten.expand %3359, %3375, %false_1140 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %3376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_1141 = torch.constant.int 4
%int4_1142 = torch.constant.int 4
%int128_1143 = torch.constant.int 128
%3377 = torch.prim.ListConstruct %int4_1141, %3360, %int4_1142, %int128_1143 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3378 = torch.aten.view %3362, %3377 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1144 = torch.constant.int 4
%int4_1145 = torch.constant.int 4
%int128_1146 = torch.constant.int 128
%3379 = torch.prim.ListConstruct %int4_1144, %3360, %int4_1145, %int128_1146 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3380 = torch.aten.view %3364, %3379 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1147 = torch.constant.int 4
%int4_1148 = torch.constant.int 4
%int128_1149 = torch.constant.int 128
%3381 = torch.prim.ListConstruct %int4_1147, %3360, %int4_1148, %int128_1149 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3382 = torch.aten.view %3366, %3381 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1150 = torch.constant.int 4
%int4_1151 = torch.constant.int 4
%int128_1152 = torch.constant.int 128
%3383 = torch.prim.ListConstruct %int4_1150, %3360, %int4_1151, %int128_1152 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3384 = torch.aten.view %3368, %3383 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1153 = torch.constant.int 4
%int4_1154 = torch.constant.int 4
%int128_1155 = torch.constant.int 128
%3385 = torch.prim.ListConstruct %int4_1153, %3360, %int4_1154, %int128_1155 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3386 = torch.aten.view %3370, %3385 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1156 = torch.constant.int 4
%int4_1157 = torch.constant.int 4
%int128_1158 = torch.constant.int 128
%3387 = torch.prim.ListConstruct %int4_1156, %3360, %int4_1157, %int128_1158 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3388 = torch.aten.view %3372, %3387 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1159 = torch.constant.int 4
%int4_1160 = torch.constant.int 4
%int128_1161 = torch.constant.int 128
%3389 = torch.prim.ListConstruct %int4_1159, %3360, %int4_1160, %int128_1161 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3390 = torch.aten.view %3374, %3389 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1162 = torch.constant.int 4
%int4_1163 = torch.constant.int 4
%int128_1164 = torch.constant.int 128
%3391 = torch.prim.ListConstruct %int4_1162, %3360, %int4_1163, %int128_1164 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3392 = torch.aten.view %3376, %3391 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1165 = torch.constant.int 1
%int2_1166 = torch.constant.int 2
%3393 = torch.aten.transpose.int %2759, %int1_1165, %int2_1166 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3393, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1167 = torch.constant.int 1
%int2_1168 = torch.constant.int 2
%3394 = torch.aten.transpose.int %2774, %int1_1167, %int2_1168 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3394, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1169 = torch.constant.int 1
%int2_1170 = torch.constant.int 2
%3395 = torch.aten.transpose.int %2789, %int1_1169, %int2_1170 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3395, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1171 = torch.constant.int 1
%int2_1172 = torch.constant.int 2
%3396 = torch.aten.transpose.int %2804, %int1_1171, %int2_1172 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3396, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1173 = torch.constant.int 1
%int2_1174 = torch.constant.int 2
%3397 = torch.aten.transpose.int %2819, %int1_1173, %int2_1174 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3397, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1175 = torch.constant.int 1
%int2_1176 = torch.constant.int 2
%3398 = torch.aten.transpose.int %2834, %int1_1175, %int2_1176 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3398, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1177 = torch.constant.int 1
%int2_1178 = torch.constant.int 2
%3399 = torch.aten.transpose.int %2849, %int1_1177, %int2_1178 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3399, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1179 = torch.constant.int 1
%int2_1180 = torch.constant.int 2
%3400 = torch.aten.transpose.int %2864, %int1_1179, %int2_1180 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3400, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1181 = torch.constant.int 1
%int2_1182 = torch.constant.int 2
%3401 = torch.aten.transpose.int %3337, %int1_1181, %int2_1182 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3401, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1183 = torch.constant.int 1
%int2_1184 = torch.constant.int 2
%3402 = torch.aten.transpose.int %3339, %int1_1183, %int2_1184 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3402, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1185 = torch.constant.int 1
%int2_1186 = torch.constant.int 2
%3403 = torch.aten.transpose.int %3341, %int1_1185, %int2_1186 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3403, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1187 = torch.constant.int 1
%int2_1188 = torch.constant.int 2
%3404 = torch.aten.transpose.int %3343, %int1_1187, %int2_1188 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3404, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1189 = torch.constant.int 1
%int2_1190 = torch.constant.int 2
%3405 = torch.aten.transpose.int %3345, %int1_1189, %int2_1190 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3405, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1191 = torch.constant.int 1
%int2_1192 = torch.constant.int 2
%3406 = torch.aten.transpose.int %3347, %int1_1191, %int2_1192 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3406, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1193 = torch.constant.int 1
%int2_1194 = torch.constant.int 2
%3407 = torch.aten.transpose.int %3349, %int1_1193, %int2_1194 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3407, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1195 = torch.constant.int 1
%int2_1196 = torch.constant.int 2
%3408 = torch.aten.transpose.int %3351, %int1_1195, %int2_1196 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3408, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1197 = torch.constant.int 1
%int2_1198 = torch.constant.int 2
%3409 = torch.aten.transpose.int %3378, %int1_1197, %int2_1198 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3409, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1199 = torch.constant.int 1
%int2_1200 = torch.constant.int 2
%3410 = torch.aten.transpose.int %3380, %int1_1199, %int2_1200 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3410, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1201 = torch.constant.int 1
%int2_1202 = torch.constant.int 2
%3411 = torch.aten.transpose.int %3382, %int1_1201, %int2_1202 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3411, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1203 = torch.constant.int 1
%int2_1204 = torch.constant.int 2
%3412 = torch.aten.transpose.int %3384, %int1_1203, %int2_1204 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3412, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1205 = torch.constant.int 1
%int2_1206 = torch.constant.int 2
%3413 = torch.aten.transpose.int %3386, %int1_1205, %int2_1206 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3413, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1207 = torch.constant.int 1
%int2_1208 = torch.constant.int 2
%3414 = torch.aten.transpose.int %3388, %int1_1207, %int2_1208 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3414, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1209 = torch.constant.int 1
%int2_1210 = torch.constant.int 2
%3415 = torch.aten.transpose.int %3390, %int1_1209, %int2_1210 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3415, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1211 = torch.constant.int 1
%int2_1212 = torch.constant.int 2
%3416 = torch.aten.transpose.int %3392, %int1_1211, %int2_1212 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %3416, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00 = torch.constant.float 0.000000e+00
%true_1213 = torch.constant.bool true
%none_1214 = torch.constant.none
%none_1215 = torch.constant.none
%3417:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3393, %3401, %3409, %float0.000000e00, %true_1213, %none_1214, %none_1215) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3417#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1216 = torch.constant.float 0.000000e+00
%true_1217 = torch.constant.bool true
%none_1218 = torch.constant.none
%none_1219 = torch.constant.none
%3418:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3394, %3402, %3410, %float0.000000e00_1216, %true_1217, %none_1218, %none_1219) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3418#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1220 = torch.constant.float 0.000000e+00
%true_1221 = torch.constant.bool true
%none_1222 = torch.constant.none
%none_1223 = torch.constant.none
%3419:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3395, %3403, %3411, %float0.000000e00_1220, %true_1221, %none_1222, %none_1223) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3419#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1224 = torch.constant.float 0.000000e+00
%true_1225 = torch.constant.bool true
%none_1226 = torch.constant.none
%none_1227 = torch.constant.none
%3420:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3396, %3404, %3412, %float0.000000e00_1224, %true_1225, %none_1226, %none_1227) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3420#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1228 = torch.constant.float 0.000000e+00
%true_1229 = torch.constant.bool true
%none_1230 = torch.constant.none
%none_1231 = torch.constant.none
%3421:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3397, %3405, %3413, %float0.000000e00_1228, %true_1229, %none_1230, %none_1231) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3421#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1232 = torch.constant.float 0.000000e+00
%true_1233 = torch.constant.bool true
%none_1234 = torch.constant.none
%none_1235 = torch.constant.none
%3422:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3398, %3406, %3414, %float0.000000e00_1232, %true_1233, %none_1234, %none_1235) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3422#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1236 = torch.constant.float 0.000000e+00
%true_1237 = torch.constant.bool true
%none_1238 = torch.constant.none
%none_1239 = torch.constant.none
%3423:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3399, %3407, %3415, %float0.000000e00_1236, %true_1237, %none_1238, %none_1239) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3423#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_1240 = torch.constant.float 0.000000e+00
%true_1241 = torch.constant.bool true
%none_1242 = torch.constant.none
%none_1243 = torch.constant.none
%3424:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%3400, %3408, %3416, %float0.000000e00_1240, %true_1241, %none_1242, %none_1243) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %3424#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_1244 = torch.constant.int 1
%int2_1245 = torch.constant.int 2
%3425 = torch.aten.transpose.int %3417#0, %int1_1244, %int2_1245 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1246 = torch.constant.int 1
%int2_1247 = torch.constant.int 2
%3426 = torch.aten.transpose.int %3418#0, %int1_1246, %int2_1247 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1248 = torch.constant.int 1
%int2_1249 = torch.constant.int 2
%3427 = torch.aten.transpose.int %3419#0, %int1_1248, %int2_1249 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1250 = torch.constant.int 1
%int2_1251 = torch.constant.int 2
%3428 = torch.aten.transpose.int %3420#0, %int1_1250, %int2_1251 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1252 = torch.constant.int 1
%int2_1253 = torch.constant.int 2
%3429 = torch.aten.transpose.int %3421#0, %int1_1252, %int2_1253 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1254 = torch.constant.int 1
%int2_1255 = torch.constant.int 2
%3430 = torch.aten.transpose.int %3422#0, %int1_1254, %int2_1255 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1256 = torch.constant.int 1
%int2_1257 = torch.constant.int 2
%3431 = torch.aten.transpose.int %3423#0, %int1_1256, %int2_1257 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_1258 = torch.constant.int 1
%int2_1259 = torch.constant.int 2
%3432 = torch.aten.transpose.int %3424#0, %int1_1258, %int2_1259 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %3432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_1260 = torch.constant.int 4
%int512_1261 = torch.constant.int 512
%3433 = torch.prim.ListConstruct %int4_1260, %2745, %int512_1261 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3434 = torch.aten.view %3425, %3433 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1262 = torch.constant.int 4
%int512_1263 = torch.constant.int 512
%3435 = torch.prim.ListConstruct %int4_1262, %2760, %int512_1263 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3436 = torch.aten.view %3426, %3435 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1264 = torch.constant.int 4
%int512_1265 = torch.constant.int 512
%3437 = torch.prim.ListConstruct %int4_1264, %2775, %int512_1265 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3438 = torch.aten.view %3427, %3437 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1266 = torch.constant.int 4
%int512_1267 = torch.constant.int 512
%3439 = torch.prim.ListConstruct %int4_1266, %2790, %int512_1267 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3440 = torch.aten.view %3428, %3439 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1268 = torch.constant.int 4
%int512_1269 = torch.constant.int 512
%3441 = torch.prim.ListConstruct %int4_1268, %2805, %int512_1269 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3442 = torch.aten.view %3429, %3441 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1270 = torch.constant.int 4
%int512_1271 = torch.constant.int 512
%3443 = torch.prim.ListConstruct %int4_1270, %2820, %int512_1271 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3444 = torch.aten.view %3430, %3443 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1272 = torch.constant.int 4
%int512_1273 = torch.constant.int 512
%3445 = torch.prim.ListConstruct %int4_1272, %2835, %int512_1273 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3446 = torch.aten.view %3431, %3445 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1274 = torch.constant.int 4
%int512_1275 = torch.constant.int 512
%3447 = torch.prim.ListConstruct %int4_1274, %2850, %int512_1275 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3448 = torch.aten.view %3432, %3447 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %3448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_1276 = torch.constant.int 1
%int0_1277 = torch.constant.int 0
%3449 = torch.prim.ListConstruct %int1_1276, %int0_1277 : (!torch.int, !torch.int) -> !torch.list<int>
%3450 = torch.aten.permute %40, %3449 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1278 = torch.constant.int 1
%int0_1279 = torch.constant.int 0
%3451 = torch.prim.ListConstruct %int1_1278, %int0_1279 : (!torch.int, !torch.int) -> !torch.list<int>
%3452 = torch.aten.permute %41, %3451 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1280 = torch.constant.int 1
%int0_1281 = torch.constant.int 0
%3453 = torch.prim.ListConstruct %int1_1280, %int0_1281 : (!torch.int, !torch.int) -> !torch.list<int>
%3454 = torch.aten.permute %42, %3453 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1282 = torch.constant.int 1
%int0_1283 = torch.constant.int 0
%3455 = torch.prim.ListConstruct %int1_1282, %int0_1283 : (!torch.int, !torch.int) -> !torch.list<int>
%3456 = torch.aten.permute %43, %3455 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1284 = torch.constant.int 1
%int0_1285 = torch.constant.int 0
%3457 = torch.prim.ListConstruct %int1_1284, %int0_1285 : (!torch.int, !torch.int) -> !torch.list<int>
%3458 = torch.aten.permute %44, %3457 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1286 = torch.constant.int 1
%int0_1287 = torch.constant.int 0
%3459 = torch.prim.ListConstruct %int1_1286, %int0_1287 : (!torch.int, !torch.int) -> !torch.list<int>
%3460 = torch.aten.permute %45, %3459 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1288 = torch.constant.int 1
%int0_1289 = torch.constant.int 0
%3461 = torch.prim.ListConstruct %int1_1288, %int0_1289 : (!torch.int, !torch.int) -> !torch.list<int>
%3462 = torch.aten.permute %46, %3461 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_1290 = torch.constant.int 1
%int0_1291 = torch.constant.int 0
%3463 = torch.prim.ListConstruct %int1_1290, %int0_1291 : (!torch.int, !torch.int) -> !torch.list<int>
%3464 = torch.aten.permute %47, %3463 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_1292 = torch.constant.int 4
%3465 = torch.aten.mul.int %int4_1292, %2745 : !torch.int, !torch.int -> !torch.int
%int512_1293 = torch.constant.int 512
%3466 = torch.prim.ListConstruct %3465, %int512_1293 : (!torch.int, !torch.int) -> !torch.list<int>
%3467 = torch.aten.view %3434, %3466 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3467, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3468 = torch.aten.mm %3467, %3450 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3468, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1294 = torch.constant.int 4
%int4096_1295 = torch.constant.int 4096
%3469 = torch.prim.ListConstruct %int4_1294, %2745, %int4096_1295 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3470 = torch.aten.view %3468, %3469 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1296 = torch.constant.int 4
%3471 = torch.aten.mul.int %int4_1296, %2760 : !torch.int, !torch.int -> !torch.int
%int512_1297 = torch.constant.int 512
%3472 = torch.prim.ListConstruct %3471, %int512_1297 : (!torch.int, !torch.int) -> !torch.list<int>
%3473 = torch.aten.view %3436, %3472 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3473, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3474 = torch.aten.mm %3473, %3452 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3474, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1298 = torch.constant.int 4
%int4096_1299 = torch.constant.int 4096
%3475 = torch.prim.ListConstruct %int4_1298, %2760, %int4096_1299 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3476 = torch.aten.view %3474, %3475 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1300 = torch.constant.int 4
%3477 = torch.aten.mul.int %int4_1300, %2775 : !torch.int, !torch.int -> !torch.int
%int512_1301 = torch.constant.int 512
%3478 = torch.prim.ListConstruct %3477, %int512_1301 : (!torch.int, !torch.int) -> !torch.list<int>
%3479 = torch.aten.view %3438, %3478 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3479, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3480 = torch.aten.mm %3479, %3454 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3480, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1302 = torch.constant.int 4
%int4096_1303 = torch.constant.int 4096
%3481 = torch.prim.ListConstruct %int4_1302, %2775, %int4096_1303 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3482 = torch.aten.view %3480, %3481 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1304 = torch.constant.int 4
%3483 = torch.aten.mul.int %int4_1304, %2790 : !torch.int, !torch.int -> !torch.int
%int512_1305 = torch.constant.int 512
%3484 = torch.prim.ListConstruct %3483, %int512_1305 : (!torch.int, !torch.int) -> !torch.list<int>
%3485 = torch.aten.view %3440, %3484 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3485, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3486 = torch.aten.mm %3485, %3456 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3486, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1306 = torch.constant.int 4
%int4096_1307 = torch.constant.int 4096
%3487 = torch.prim.ListConstruct %int4_1306, %2790, %int4096_1307 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3488 = torch.aten.view %3486, %3487 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1308 = torch.constant.int 4
%3489 = torch.aten.mul.int %int4_1308, %2805 : !torch.int, !torch.int -> !torch.int
%int512_1309 = torch.constant.int 512
%3490 = torch.prim.ListConstruct %3489, %int512_1309 : (!torch.int, !torch.int) -> !torch.list<int>
%3491 = torch.aten.view %3442, %3490 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3491, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3492 = torch.aten.mm %3491, %3458 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3492, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1310 = torch.constant.int 4
%int4096_1311 = torch.constant.int 4096
%3493 = torch.prim.ListConstruct %int4_1310, %2805, %int4096_1311 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3494 = torch.aten.view %3492, %3493 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1312 = torch.constant.int 4
%3495 = torch.aten.mul.int %int4_1312, %2820 : !torch.int, !torch.int -> !torch.int
%int512_1313 = torch.constant.int 512
%3496 = torch.prim.ListConstruct %3495, %int512_1313 : (!torch.int, !torch.int) -> !torch.list<int>
%3497 = torch.aten.view %3444, %3496 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3497, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3498 = torch.aten.mm %3497, %3460 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3498, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1314 = torch.constant.int 4
%int4096_1315 = torch.constant.int 4096
%3499 = torch.prim.ListConstruct %int4_1314, %2820, %int4096_1315 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3500 = torch.aten.view %3498, %3499 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1316 = torch.constant.int 4
%3501 = torch.aten.mul.int %int4_1316, %2835 : !torch.int, !torch.int -> !torch.int
%int512_1317 = torch.constant.int 512
%3502 = torch.prim.ListConstruct %3501, %int512_1317 : (!torch.int, !torch.int) -> !torch.list<int>
%3503 = torch.aten.view %3446, %3502 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3503, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3504 = torch.aten.mm %3503, %3462 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3504, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1318 = torch.constant.int 4
%int4096_1319 = torch.constant.int 4096
%3505 = torch.prim.ListConstruct %int4_1318, %2835, %int4096_1319 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3506 = torch.aten.view %3504, %3505 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_1320 = torch.constant.int 4
%3507 = torch.aten.mul.int %int4_1320, %2850 : !torch.int, !torch.int -> !torch.int
%int512_1321 = torch.constant.int 512
%3508 = torch.prim.ListConstruct %3507, %int512_1321 : (!torch.int, !torch.int) -> !torch.list<int>
%3509 = torch.aten.view %3448, %3508 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %3509, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%3510 = torch.aten.mm %3509, %3464 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3510, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1322 = torch.constant.int 4
%int4096_1323 = torch.constant.int 4096
%3511 = torch.prim.ListConstruct %int4_1322, %2850, %int4096_1323 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3512 = torch.aten.view %3510, %3511 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3513 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1324 = arith.constant 1 : index
%dim_1325 = tensor.dim %3513, %c1_1324 : tensor<4x?x4096xf16>
%3514 = flow.tensor.transfer %3513 : tensor<4x?x4096xf16>{%dim_1325} to #hal.device.promise<@__device_0>
%3515 = torch_c.from_builtin_tensor %3514 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3516 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1326 = arith.constant 1 : index
%dim_1327 = tensor.dim %3516, %c1_1326 : tensor<4x?x4096xf16>
%3517 = flow.tensor.transfer %3516 : tensor<4x?x4096xf16>{%dim_1327} to #hal.device.promise<@__device_0>
%3518 = torch_c.from_builtin_tensor %3517 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3519 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1328 = arith.constant 1 : index
%dim_1329 = tensor.dim %3519, %c1_1328 : tensor<4x?x4096xf16>
%3520 = flow.tensor.transfer %3519 : tensor<4x?x4096xf16>{%dim_1329} to #hal.device.promise<@__device_0>
%3521 = torch_c.from_builtin_tensor %3520 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3522 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1330 = arith.constant 1 : index
%dim_1331 = tensor.dim %3522, %c1_1330 : tensor<4x?x4096xf16>
%3523 = flow.tensor.transfer %3522 : tensor<4x?x4096xf16>{%dim_1331} to #hal.device.promise<@__device_0>
%3524 = torch_c.from_builtin_tensor %3523 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3525 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1332 = arith.constant 1 : index
%dim_1333 = tensor.dim %3525, %c1_1332 : tensor<4x?x4096xf16>
%3526 = flow.tensor.transfer %3525 : tensor<4x?x4096xf16>{%dim_1333} to #hal.device.promise<@__device_0>
%3527 = torch_c.from_builtin_tensor %3526 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3528 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1334 = arith.constant 1 : index
%dim_1335 = tensor.dim %3528, %c1_1334 : tensor<4x?x4096xf16>
%3529 = flow.tensor.transfer %3528 : tensor<4x?x4096xf16>{%dim_1335} to #hal.device.promise<@__device_0>
%3530 = torch_c.from_builtin_tensor %3529 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3531 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1336 = arith.constant 1 : index
%dim_1337 = tensor.dim %3531, %c1_1336 : tensor<4x?x4096xf16>
%3532 = flow.tensor.transfer %3531 : tensor<4x?x4096xf16>{%dim_1337} to #hal.device.promise<@__device_0>
%3533 = torch_c.from_builtin_tensor %3532 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1338 = torch.constant.int 1
%3534 = torch.aten.add.Tensor %3470, %3515, %int1_1338 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1339 = torch.constant.int 1
%3535 = torch.aten.add.Tensor %3534, %3518, %int1_1339 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1340 = torch.constant.int 1
%3536 = torch.aten.add.Tensor %3535, %3521, %int1_1340 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1341 = torch.constant.int 1
%3537 = torch.aten.add.Tensor %3536, %3524, %int1_1341 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1342 = torch.constant.int 1
%3538 = torch.aten.add.Tensor %3537, %3527, %int1_1342 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1343 = torch.constant.int 1
%3539 = torch.aten.add.Tensor %3538, %3530, %int1_1343 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1344 = torch.constant.int 1
%3540 = torch.aten.add.Tensor %3539, %3533, %int1_1344 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3541 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1345 = arith.constant 1 : index
%dim_1346 = tensor.dim %3541, %c1_1345 : tensor<4x?x4096xf16>
%3542 = flow.tensor.transfer %3541 : tensor<4x?x4096xf16>{%dim_1346} to #hal.device.promise<@__device_1>
%3543 = torch_c.from_builtin_tensor %3542 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3544 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1347 = arith.constant 1 : index
%dim_1348 = tensor.dim %3544, %c1_1347 : tensor<4x?x4096xf16>
%3545 = flow.tensor.transfer %3544 : tensor<4x?x4096xf16>{%dim_1348} to #hal.device.promise<@__device_1>
%3546 = torch_c.from_builtin_tensor %3545 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3547 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1349 = arith.constant 1 : index
%dim_1350 = tensor.dim %3547, %c1_1349 : tensor<4x?x4096xf16>
%3548 = flow.tensor.transfer %3547 : tensor<4x?x4096xf16>{%dim_1350} to #hal.device.promise<@__device_1>
%3549 = torch_c.from_builtin_tensor %3548 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3550 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1351 = arith.constant 1 : index
%dim_1352 = tensor.dim %3550, %c1_1351 : tensor<4x?x4096xf16>
%3551 = flow.tensor.transfer %3550 : tensor<4x?x4096xf16>{%dim_1352} to #hal.device.promise<@__device_1>
%3552 = torch_c.from_builtin_tensor %3551 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3553 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1353 = arith.constant 1 : index
%dim_1354 = tensor.dim %3553, %c1_1353 : tensor<4x?x4096xf16>
%3554 = flow.tensor.transfer %3553 : tensor<4x?x4096xf16>{%dim_1354} to #hal.device.promise<@__device_1>
%3555 = torch_c.from_builtin_tensor %3554 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3556 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1355 = arith.constant 1 : index
%dim_1356 = tensor.dim %3556, %c1_1355 : tensor<4x?x4096xf16>
%3557 = flow.tensor.transfer %3556 : tensor<4x?x4096xf16>{%dim_1356} to #hal.device.promise<@__device_1>
%3558 = torch_c.from_builtin_tensor %3557 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3559 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1357 = arith.constant 1 : index
%dim_1358 = tensor.dim %3559, %c1_1357 : tensor<4x?x4096xf16>
%3560 = flow.tensor.transfer %3559 : tensor<4x?x4096xf16>{%dim_1358} to #hal.device.promise<@__device_1>
%3561 = torch_c.from_builtin_tensor %3560 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1359 = torch.constant.int 1
%3562 = torch.aten.add.Tensor %3543, %3476, %int1_1359 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1360 = torch.constant.int 1
%3563 = torch.aten.add.Tensor %3562, %3546, %int1_1360 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1361 = torch.constant.int 1
%3564 = torch.aten.add.Tensor %3563, %3549, %int1_1361 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1362 = torch.constant.int 1
%3565 = torch.aten.add.Tensor %3564, %3552, %int1_1362 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1363 = torch.constant.int 1
%3566 = torch.aten.add.Tensor %3565, %3555, %int1_1363 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1364 = torch.constant.int 1
%3567 = torch.aten.add.Tensor %3566, %3558, %int1_1364 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1365 = torch.constant.int 1
%3568 = torch.aten.add.Tensor %3567, %3561, %int1_1365 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3569 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1366 = arith.constant 1 : index
%dim_1367 = tensor.dim %3569, %c1_1366 : tensor<4x?x4096xf16>
%3570 = flow.tensor.transfer %3569 : tensor<4x?x4096xf16>{%dim_1367} to #hal.device.promise<@__device_2>
%3571 = torch_c.from_builtin_tensor %3570 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3572 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1368 = arith.constant 1 : index
%dim_1369 = tensor.dim %3572, %c1_1368 : tensor<4x?x4096xf16>
%3573 = flow.tensor.transfer %3572 : tensor<4x?x4096xf16>{%dim_1369} to #hal.device.promise<@__device_2>
%3574 = torch_c.from_builtin_tensor %3573 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3575 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1370 = arith.constant 1 : index
%dim_1371 = tensor.dim %3575, %c1_1370 : tensor<4x?x4096xf16>
%3576 = flow.tensor.transfer %3575 : tensor<4x?x4096xf16>{%dim_1371} to #hal.device.promise<@__device_2>
%3577 = torch_c.from_builtin_tensor %3576 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3578 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1372 = arith.constant 1 : index
%dim_1373 = tensor.dim %3578, %c1_1372 : tensor<4x?x4096xf16>
%3579 = flow.tensor.transfer %3578 : tensor<4x?x4096xf16>{%dim_1373} to #hal.device.promise<@__device_2>
%3580 = torch_c.from_builtin_tensor %3579 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3581 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1374 = arith.constant 1 : index
%dim_1375 = tensor.dim %3581, %c1_1374 : tensor<4x?x4096xf16>
%3582 = flow.tensor.transfer %3581 : tensor<4x?x4096xf16>{%dim_1375} to #hal.device.promise<@__device_2>
%3583 = torch_c.from_builtin_tensor %3582 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3584 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1376 = arith.constant 1 : index
%dim_1377 = tensor.dim %3584, %c1_1376 : tensor<4x?x4096xf16>
%3585 = flow.tensor.transfer %3584 : tensor<4x?x4096xf16>{%dim_1377} to #hal.device.promise<@__device_2>
%3586 = torch_c.from_builtin_tensor %3585 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3587 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1378 = arith.constant 1 : index
%dim_1379 = tensor.dim %3587, %c1_1378 : tensor<4x?x4096xf16>
%3588 = flow.tensor.transfer %3587 : tensor<4x?x4096xf16>{%dim_1379} to #hal.device.promise<@__device_2>
%3589 = torch_c.from_builtin_tensor %3588 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1380 = torch.constant.int 1
%3590 = torch.aten.add.Tensor %3571, %3574, %int1_1380 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1381 = torch.constant.int 1
%3591 = torch.aten.add.Tensor %3590, %3482, %int1_1381 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1382 = torch.constant.int 1
%3592 = torch.aten.add.Tensor %3591, %3577, %int1_1382 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1383 = torch.constant.int 1
%3593 = torch.aten.add.Tensor %3592, %3580, %int1_1383 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1384 = torch.constant.int 1
%3594 = torch.aten.add.Tensor %3593, %3583, %int1_1384 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1385 = torch.constant.int 1
%3595 = torch.aten.add.Tensor %3594, %3586, %int1_1385 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1386 = torch.constant.int 1
%3596 = torch.aten.add.Tensor %3595, %3589, %int1_1386 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3597 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1387 = arith.constant 1 : index
%dim_1388 = tensor.dim %3597, %c1_1387 : tensor<4x?x4096xf16>
%3598 = flow.tensor.transfer %3597 : tensor<4x?x4096xf16>{%dim_1388} to #hal.device.promise<@__device_3>
%3599 = torch_c.from_builtin_tensor %3598 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3600 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1389 = arith.constant 1 : index
%dim_1390 = tensor.dim %3600, %c1_1389 : tensor<4x?x4096xf16>
%3601 = flow.tensor.transfer %3600 : tensor<4x?x4096xf16>{%dim_1390} to #hal.device.promise<@__device_3>
%3602 = torch_c.from_builtin_tensor %3601 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3603 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1391 = arith.constant 1 : index
%dim_1392 = tensor.dim %3603, %c1_1391 : tensor<4x?x4096xf16>
%3604 = flow.tensor.transfer %3603 : tensor<4x?x4096xf16>{%dim_1392} to #hal.device.promise<@__device_3>
%3605 = torch_c.from_builtin_tensor %3604 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3606 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1393 = arith.constant 1 : index
%dim_1394 = tensor.dim %3606, %c1_1393 : tensor<4x?x4096xf16>
%3607 = flow.tensor.transfer %3606 : tensor<4x?x4096xf16>{%dim_1394} to #hal.device.promise<@__device_3>
%3608 = torch_c.from_builtin_tensor %3607 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3609 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1395 = arith.constant 1 : index
%dim_1396 = tensor.dim %3609, %c1_1395 : tensor<4x?x4096xf16>
%3610 = flow.tensor.transfer %3609 : tensor<4x?x4096xf16>{%dim_1396} to #hal.device.promise<@__device_3>
%3611 = torch_c.from_builtin_tensor %3610 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3612 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1397 = arith.constant 1 : index
%dim_1398 = tensor.dim %3612, %c1_1397 : tensor<4x?x4096xf16>
%3613 = flow.tensor.transfer %3612 : tensor<4x?x4096xf16>{%dim_1398} to #hal.device.promise<@__device_3>
%3614 = torch_c.from_builtin_tensor %3613 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3615 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1399 = arith.constant 1 : index
%dim_1400 = tensor.dim %3615, %c1_1399 : tensor<4x?x4096xf16>
%3616 = flow.tensor.transfer %3615 : tensor<4x?x4096xf16>{%dim_1400} to #hal.device.promise<@__device_3>
%3617 = torch_c.from_builtin_tensor %3616 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1401 = torch.constant.int 1
%3618 = torch.aten.add.Tensor %3599, %3602, %int1_1401 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1402 = torch.constant.int 1
%3619 = torch.aten.add.Tensor %3618, %3605, %int1_1402 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1403 = torch.constant.int 1
%3620 = torch.aten.add.Tensor %3619, %3488, %int1_1403 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1404 = torch.constant.int 1
%3621 = torch.aten.add.Tensor %3620, %3608, %int1_1404 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1405 = torch.constant.int 1
%3622 = torch.aten.add.Tensor %3621, %3611, %int1_1405 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1406 = torch.constant.int 1
%3623 = torch.aten.add.Tensor %3622, %3614, %int1_1406 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1407 = torch.constant.int 1
%3624 = torch.aten.add.Tensor %3623, %3617, %int1_1407 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3625 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1408 = arith.constant 1 : index
%dim_1409 = tensor.dim %3625, %c1_1408 : tensor<4x?x4096xf16>
%3626 = flow.tensor.transfer %3625 : tensor<4x?x4096xf16>{%dim_1409} to #hal.device.promise<@__device_4>
%3627 = torch_c.from_builtin_tensor %3626 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3628 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1410 = arith.constant 1 : index
%dim_1411 = tensor.dim %3628, %c1_1410 : tensor<4x?x4096xf16>
%3629 = flow.tensor.transfer %3628 : tensor<4x?x4096xf16>{%dim_1411} to #hal.device.promise<@__device_4>
%3630 = torch_c.from_builtin_tensor %3629 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3631 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1412 = arith.constant 1 : index
%dim_1413 = tensor.dim %3631, %c1_1412 : tensor<4x?x4096xf16>
%3632 = flow.tensor.transfer %3631 : tensor<4x?x4096xf16>{%dim_1413} to #hal.device.promise<@__device_4>
%3633 = torch_c.from_builtin_tensor %3632 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3634 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1414 = arith.constant 1 : index
%dim_1415 = tensor.dim %3634, %c1_1414 : tensor<4x?x4096xf16>
%3635 = flow.tensor.transfer %3634 : tensor<4x?x4096xf16>{%dim_1415} to #hal.device.promise<@__device_4>
%3636 = torch_c.from_builtin_tensor %3635 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3637 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1416 = arith.constant 1 : index
%dim_1417 = tensor.dim %3637, %c1_1416 : tensor<4x?x4096xf16>
%3638 = flow.tensor.transfer %3637 : tensor<4x?x4096xf16>{%dim_1417} to #hal.device.promise<@__device_4>
%3639 = torch_c.from_builtin_tensor %3638 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3640 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1418 = arith.constant 1 : index
%dim_1419 = tensor.dim %3640, %c1_1418 : tensor<4x?x4096xf16>
%3641 = flow.tensor.transfer %3640 : tensor<4x?x4096xf16>{%dim_1419} to #hal.device.promise<@__device_4>
%3642 = torch_c.from_builtin_tensor %3641 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3643 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1420 = arith.constant 1 : index
%dim_1421 = tensor.dim %3643, %c1_1420 : tensor<4x?x4096xf16>
%3644 = flow.tensor.transfer %3643 : tensor<4x?x4096xf16>{%dim_1421} to #hal.device.promise<@__device_4>
%3645 = torch_c.from_builtin_tensor %3644 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1422 = torch.constant.int 1
%3646 = torch.aten.add.Tensor %3627, %3630, %int1_1422 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1423 = torch.constant.int 1
%3647 = torch.aten.add.Tensor %3646, %3633, %int1_1423 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1424 = torch.constant.int 1
%3648 = torch.aten.add.Tensor %3647, %3636, %int1_1424 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1425 = torch.constant.int 1
%3649 = torch.aten.add.Tensor %3648, %3494, %int1_1425 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1426 = torch.constant.int 1
%3650 = torch.aten.add.Tensor %3649, %3639, %int1_1426 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1427 = torch.constant.int 1
%3651 = torch.aten.add.Tensor %3650, %3642, %int1_1427 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1428 = torch.constant.int 1
%3652 = torch.aten.add.Tensor %3651, %3645, %int1_1428 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3653 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1429 = arith.constant 1 : index
%dim_1430 = tensor.dim %3653, %c1_1429 : tensor<4x?x4096xf16>
%3654 = flow.tensor.transfer %3653 : tensor<4x?x4096xf16>{%dim_1430} to #hal.device.promise<@__device_5>
%3655 = torch_c.from_builtin_tensor %3654 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3656 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1431 = arith.constant 1 : index
%dim_1432 = tensor.dim %3656, %c1_1431 : tensor<4x?x4096xf16>
%3657 = flow.tensor.transfer %3656 : tensor<4x?x4096xf16>{%dim_1432} to #hal.device.promise<@__device_5>
%3658 = torch_c.from_builtin_tensor %3657 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3659 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1433 = arith.constant 1 : index
%dim_1434 = tensor.dim %3659, %c1_1433 : tensor<4x?x4096xf16>
%3660 = flow.tensor.transfer %3659 : tensor<4x?x4096xf16>{%dim_1434} to #hal.device.promise<@__device_5>
%3661 = torch_c.from_builtin_tensor %3660 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3662 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1435 = arith.constant 1 : index
%dim_1436 = tensor.dim %3662, %c1_1435 : tensor<4x?x4096xf16>
%3663 = flow.tensor.transfer %3662 : tensor<4x?x4096xf16>{%dim_1436} to #hal.device.promise<@__device_5>
%3664 = torch_c.from_builtin_tensor %3663 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3665 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1437 = arith.constant 1 : index
%dim_1438 = tensor.dim %3665, %c1_1437 : tensor<4x?x4096xf16>
%3666 = flow.tensor.transfer %3665 : tensor<4x?x4096xf16>{%dim_1438} to #hal.device.promise<@__device_5>
%3667 = torch_c.from_builtin_tensor %3666 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3668 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1439 = arith.constant 1 : index
%dim_1440 = tensor.dim %3668, %c1_1439 : tensor<4x?x4096xf16>
%3669 = flow.tensor.transfer %3668 : tensor<4x?x4096xf16>{%dim_1440} to #hal.device.promise<@__device_5>
%3670 = torch_c.from_builtin_tensor %3669 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3671 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1441 = arith.constant 1 : index
%dim_1442 = tensor.dim %3671, %c1_1441 : tensor<4x?x4096xf16>
%3672 = flow.tensor.transfer %3671 : tensor<4x?x4096xf16>{%dim_1442} to #hal.device.promise<@__device_5>
%3673 = torch_c.from_builtin_tensor %3672 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1443 = torch.constant.int 1
%3674 = torch.aten.add.Tensor %3655, %3658, %int1_1443 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1444 = torch.constant.int 1
%3675 = torch.aten.add.Tensor %3674, %3661, %int1_1444 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1445 = torch.constant.int 1
%3676 = torch.aten.add.Tensor %3675, %3664, %int1_1445 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1446 = torch.constant.int 1
%3677 = torch.aten.add.Tensor %3676, %3667, %int1_1446 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1447 = torch.constant.int 1
%3678 = torch.aten.add.Tensor %3677, %3500, %int1_1447 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1448 = torch.constant.int 1
%3679 = torch.aten.add.Tensor %3678, %3670, %int1_1448 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1449 = torch.constant.int 1
%3680 = torch.aten.add.Tensor %3679, %3673, %int1_1449 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3681 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1450 = arith.constant 1 : index
%dim_1451 = tensor.dim %3681, %c1_1450 : tensor<4x?x4096xf16>
%3682 = flow.tensor.transfer %3681 : tensor<4x?x4096xf16>{%dim_1451} to #hal.device.promise<@__device_6>
%3683 = torch_c.from_builtin_tensor %3682 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3684 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1452 = arith.constant 1 : index
%dim_1453 = tensor.dim %3684, %c1_1452 : tensor<4x?x4096xf16>
%3685 = flow.tensor.transfer %3684 : tensor<4x?x4096xf16>{%dim_1453} to #hal.device.promise<@__device_6>
%3686 = torch_c.from_builtin_tensor %3685 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3687 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1454 = arith.constant 1 : index
%dim_1455 = tensor.dim %3687, %c1_1454 : tensor<4x?x4096xf16>
%3688 = flow.tensor.transfer %3687 : tensor<4x?x4096xf16>{%dim_1455} to #hal.device.promise<@__device_6>
%3689 = torch_c.from_builtin_tensor %3688 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3690 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1456 = arith.constant 1 : index
%dim_1457 = tensor.dim %3690, %c1_1456 : tensor<4x?x4096xf16>
%3691 = flow.tensor.transfer %3690 : tensor<4x?x4096xf16>{%dim_1457} to #hal.device.promise<@__device_6>
%3692 = torch_c.from_builtin_tensor %3691 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3693 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1458 = arith.constant 1 : index
%dim_1459 = tensor.dim %3693, %c1_1458 : tensor<4x?x4096xf16>
%3694 = flow.tensor.transfer %3693 : tensor<4x?x4096xf16>{%dim_1459} to #hal.device.promise<@__device_6>
%3695 = torch_c.from_builtin_tensor %3694 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3696 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1460 = arith.constant 1 : index
%dim_1461 = tensor.dim %3696, %c1_1460 : tensor<4x?x4096xf16>
%3697 = flow.tensor.transfer %3696 : tensor<4x?x4096xf16>{%dim_1461} to #hal.device.promise<@__device_6>
%3698 = torch_c.from_builtin_tensor %3697 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3699 = torch_c.to_builtin_tensor %3512 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1462 = arith.constant 1 : index
%dim_1463 = tensor.dim %3699, %c1_1462 : tensor<4x?x4096xf16>
%3700 = flow.tensor.transfer %3699 : tensor<4x?x4096xf16>{%dim_1463} to #hal.device.promise<@__device_6>
%3701 = torch_c.from_builtin_tensor %3700 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1464 = torch.constant.int 1
%3702 = torch.aten.add.Tensor %3683, %3686, %int1_1464 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1465 = torch.constant.int 1
%3703 = torch.aten.add.Tensor %3702, %3689, %int1_1465 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1466 = torch.constant.int 1
%3704 = torch.aten.add.Tensor %3703, %3692, %int1_1466 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1467 = torch.constant.int 1
%3705 = torch.aten.add.Tensor %3704, %3695, %int1_1467 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1468 = torch.constant.int 1
%3706 = torch.aten.add.Tensor %3705, %3698, %int1_1468 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1469 = torch.constant.int 1
%3707 = torch.aten.add.Tensor %3706, %3506, %int1_1469 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1470 = torch.constant.int 1
%3708 = torch.aten.add.Tensor %3707, %3701, %int1_1470 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3709 = torch_c.to_builtin_tensor %3470 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1471 = arith.constant 1 : index
%dim_1472 = tensor.dim %3709, %c1_1471 : tensor<4x?x4096xf16>
%3710 = flow.tensor.transfer %3709 : tensor<4x?x4096xf16>{%dim_1472} to #hal.device.promise<@__device_7>
%3711 = torch_c.from_builtin_tensor %3710 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3712 = torch_c.to_builtin_tensor %3476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1473 = arith.constant 1 : index
%dim_1474 = tensor.dim %3712, %c1_1473 : tensor<4x?x4096xf16>
%3713 = flow.tensor.transfer %3712 : tensor<4x?x4096xf16>{%dim_1474} to #hal.device.promise<@__device_7>
%3714 = torch_c.from_builtin_tensor %3713 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3715 = torch_c.to_builtin_tensor %3482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1475 = arith.constant 1 : index
%dim_1476 = tensor.dim %3715, %c1_1475 : tensor<4x?x4096xf16>
%3716 = flow.tensor.transfer %3715 : tensor<4x?x4096xf16>{%dim_1476} to #hal.device.promise<@__device_7>
%3717 = torch_c.from_builtin_tensor %3716 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3718 = torch_c.to_builtin_tensor %3488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1477 = arith.constant 1 : index
%dim_1478 = tensor.dim %3718, %c1_1477 : tensor<4x?x4096xf16>
%3719 = flow.tensor.transfer %3718 : tensor<4x?x4096xf16>{%dim_1478} to #hal.device.promise<@__device_7>
%3720 = torch_c.from_builtin_tensor %3719 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3721 = torch_c.to_builtin_tensor %3494 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1479 = arith.constant 1 : index
%dim_1480 = tensor.dim %3721, %c1_1479 : tensor<4x?x4096xf16>
%3722 = flow.tensor.transfer %3721 : tensor<4x?x4096xf16>{%dim_1480} to #hal.device.promise<@__device_7>
%3723 = torch_c.from_builtin_tensor %3722 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3724 = torch_c.to_builtin_tensor %3500 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1481 = arith.constant 1 : index
%dim_1482 = tensor.dim %3724, %c1_1481 : tensor<4x?x4096xf16>
%3725 = flow.tensor.transfer %3724 : tensor<4x?x4096xf16>{%dim_1482} to #hal.device.promise<@__device_7>
%3726 = torch_c.from_builtin_tensor %3725 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%3727 = torch_c.to_builtin_tensor %3506 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1483 = arith.constant 1 : index
%dim_1484 = tensor.dim %3727, %c1_1483 : tensor<4x?x4096xf16>
%3728 = flow.tensor.transfer %3727 : tensor<4x?x4096xf16>{%dim_1484} to #hal.device.promise<@__device_7>
%3729 = torch_c.from_builtin_tensor %3728 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1485 = torch.constant.int 1
%3730 = torch.aten.add.Tensor %3711, %3714, %int1_1485 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1486 = torch.constant.int 1
%3731 = torch.aten.add.Tensor %3730, %3717, %int1_1486 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1487 = torch.constant.int 1
%3732 = torch.aten.add.Tensor %3731, %3720, %int1_1487 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1488 = torch.constant.int 1
%3733 = torch.aten.add.Tensor %3732, %3723, %int1_1488 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1489 = torch.constant.int 1
%3734 = torch.aten.add.Tensor %3733, %3726, %int1_1489 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1490 = torch.constant.int 1
%3735 = torch.aten.add.Tensor %3734, %3729, %int1_1490 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1491 = torch.constant.int 1
%3736 = torch.aten.add.Tensor %3735, %3512, %int1_1491 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1492 = torch.constant.int 1
%3737 = torch.aten.add.Tensor %2386, %3540, %int1_1492 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1493 = torch.constant.int 1
%3738 = torch.aten.add.Tensor %2387, %3568, %int1_1493 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1494 = torch.constant.int 1
%3739 = torch.aten.add.Tensor %2388, %3596, %int1_1494 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1495 = torch.constant.int 1
%3740 = torch.aten.add.Tensor %2389, %3624, %int1_1495 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1496 = torch.constant.int 1
%3741 = torch.aten.add.Tensor %2390, %3652, %int1_1496 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1497 = torch.constant.int 1
%3742 = torch.aten.add.Tensor %2391, %3680, %int1_1497 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1498 = torch.constant.int 1
%3743 = torch.aten.add.Tensor %2392, %3708, %int1_1498 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1499 = torch.constant.int 1
%3744 = torch.aten.add.Tensor %2393, %3736, %int1_1499 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_1500 = torch.constant.int 6
%3745 = torch.prims.convert_element_type %3737, %int6_1500 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1501 = torch.constant.int 6
%3746 = torch.prims.convert_element_type %3738, %int6_1501 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1502 = torch.constant.int 6
%3747 = torch.prims.convert_element_type %3739, %int6_1502 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1503 = torch.constant.int 6
%3748 = torch.prims.convert_element_type %3740, %int6_1503 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1504 = torch.constant.int 6
%3749 = torch.prims.convert_element_type %3741, %int6_1504 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1505 = torch.constant.int 6
%3750 = torch.prims.convert_element_type %3742, %int6_1505 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1506 = torch.constant.int 6
%3751 = torch.prims.convert_element_type %3743, %int6_1506 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1507 = torch.constant.int 6
%3752 = torch.prims.convert_element_type %3744, %int6_1507 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1508 = torch.constant.int 2
%3753 = torch.aten.pow.Tensor_Scalar %3745, %int2_1508 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1509 = torch.constant.int 2
%3754 = torch.aten.pow.Tensor_Scalar %3746, %int2_1509 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1510 = torch.constant.int 2
%3755 = torch.aten.pow.Tensor_Scalar %3747, %int2_1510 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1511 = torch.constant.int 2
%3756 = torch.aten.pow.Tensor_Scalar %3748, %int2_1511 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1512 = torch.constant.int 2
%3757 = torch.aten.pow.Tensor_Scalar %3749, %int2_1512 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1513 = torch.constant.int 2
%3758 = torch.aten.pow.Tensor_Scalar %3750, %int2_1513 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1514 = torch.constant.int 2
%3759 = torch.aten.pow.Tensor_Scalar %3751, %int2_1514 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1515 = torch.constant.int 2
%3760 = torch.aten.pow.Tensor_Scalar %3752, %int2_1515 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_1516 = torch.constant.int -1
%3761 = torch.prim.ListConstruct %int-1_1516 : (!torch.int) -> !torch.list<int>
%true_1517 = torch.constant.bool true
%none_1518 = torch.constant.none
%3762 = torch.aten.mean.dim %3753, %3761, %true_1517, %none_1518 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1519 = torch.constant.int -1
%3763 = torch.prim.ListConstruct %int-1_1519 : (!torch.int) -> !torch.list<int>
%true_1520 = torch.constant.bool true
%none_1521 = torch.constant.none
%3764 = torch.aten.mean.dim %3754, %3763, %true_1520, %none_1521 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1522 = torch.constant.int -1
%3765 = torch.prim.ListConstruct %int-1_1522 : (!torch.int) -> !torch.list<int>
%true_1523 = torch.constant.bool true
%none_1524 = torch.constant.none
%3766 = torch.aten.mean.dim %3755, %3765, %true_1523, %none_1524 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1525 = torch.constant.int -1
%3767 = torch.prim.ListConstruct %int-1_1525 : (!torch.int) -> !torch.list<int>
%true_1526 = torch.constant.bool true
%none_1527 = torch.constant.none
%3768 = torch.aten.mean.dim %3756, %3767, %true_1526, %none_1527 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1528 = torch.constant.int -1
%3769 = torch.prim.ListConstruct %int-1_1528 : (!torch.int) -> !torch.list<int>
%true_1529 = torch.constant.bool true
%none_1530 = torch.constant.none
%3770 = torch.aten.mean.dim %3757, %3769, %true_1529, %none_1530 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1531 = torch.constant.int -1
%3771 = torch.prim.ListConstruct %int-1_1531 : (!torch.int) -> !torch.list<int>
%true_1532 = torch.constant.bool true
%none_1533 = torch.constant.none
%3772 = torch.aten.mean.dim %3758, %3771, %true_1532, %none_1533 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1534 = torch.constant.int -1
%3773 = torch.prim.ListConstruct %int-1_1534 : (!torch.int) -> !torch.list<int>
%true_1535 = torch.constant.bool true
%none_1536 = torch.constant.none
%3774 = torch.aten.mean.dim %3759, %3773, %true_1535, %none_1536 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1537 = torch.constant.int -1
%3775 = torch.prim.ListConstruct %int-1_1537 : (!torch.int) -> !torch.list<int>
%true_1538 = torch.constant.bool true
%none_1539 = torch.constant.none
%3776 = torch.aten.mean.dim %3760, %3775, %true_1538, %none_1539 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1540 = torch.constant.float 9.9999997473787516E-6
%int1_1541 = torch.constant.int 1
%3777 = torch.aten.add.Scalar %3762, %float9.999990e-06_1540, %int1_1541 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1542 = torch.constant.float 9.9999997473787516E-6
%int1_1543 = torch.constant.int 1
%3778 = torch.aten.add.Scalar %3764, %float9.999990e-06_1542, %int1_1543 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1544 = torch.constant.float 9.9999997473787516E-6
%int1_1545 = torch.constant.int 1
%3779 = torch.aten.add.Scalar %3766, %float9.999990e-06_1544, %int1_1545 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1546 = torch.constant.float 9.9999997473787516E-6
%int1_1547 = torch.constant.int 1
%3780 = torch.aten.add.Scalar %3768, %float9.999990e-06_1546, %int1_1547 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1548 = torch.constant.float 9.9999997473787516E-6
%int1_1549 = torch.constant.int 1
%3781 = torch.aten.add.Scalar %3770, %float9.999990e-06_1548, %int1_1549 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1550 = torch.constant.float 9.9999997473787516E-6
%int1_1551 = torch.constant.int 1
%3782 = torch.aten.add.Scalar %3772, %float9.999990e-06_1550, %int1_1551 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1552 = torch.constant.float 9.9999997473787516E-6
%int1_1553 = torch.constant.int 1
%3783 = torch.aten.add.Scalar %3774, %float9.999990e-06_1552, %int1_1553 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1554 = torch.constant.float 9.9999997473787516E-6
%int1_1555 = torch.constant.int 1
%3784 = torch.aten.add.Scalar %3776, %float9.999990e-06_1554, %int1_1555 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3785 = torch.aten.rsqrt %3777 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3786 = torch.aten.rsqrt %3778 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3787 = torch.aten.rsqrt %3779 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3788 = torch.aten.rsqrt %3780 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3789 = torch.aten.rsqrt %3781 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3790 = torch.aten.rsqrt %3782 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3791 = torch.aten.rsqrt %3783 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3792 = torch.aten.rsqrt %3784 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %3792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%3793 = torch.aten.mul.Tensor %3745, %3785 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3794 = torch.aten.mul.Tensor %3746, %3786 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3795 = torch.aten.mul.Tensor %3747, %3787 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3796 = torch.aten.mul.Tensor %3748, %3788 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3797 = torch.aten.mul.Tensor %3749, %3789 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3798 = torch.aten.mul.Tensor %3750, %3790 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3799 = torch.aten.mul.Tensor %3751, %3791 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3800 = torch.aten.mul.Tensor %3752, %3792 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3801 = torch.aten.mul.Tensor %48, %3793 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3802 = torch.aten.mul.Tensor %49, %3794 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3803 = torch.aten.mul.Tensor %50, %3795 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3804 = torch.aten.mul.Tensor %51, %3796 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3805 = torch.aten.mul.Tensor %52, %3797 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3806 = torch.aten.mul.Tensor %53, %3798 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3807 = torch.aten.mul.Tensor %54, %3799 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%3808 = torch.aten.mul.Tensor %55, %3800 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %3808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_1556 = torch.constant.int 5
%3809 = torch.prims.convert_element_type %3801, %int5_1556 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1557 = torch.constant.int 5
%3810 = torch.prims.convert_element_type %3802, %int5_1557 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1558 = torch.constant.int 5
%3811 = torch.prims.convert_element_type %3803, %int5_1558 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1559 = torch.constant.int 5
%3812 = torch.prims.convert_element_type %3804, %int5_1559 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1560 = torch.constant.int 5
%3813 = torch.prims.convert_element_type %3805, %int5_1560 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1561 = torch.constant.int 5
%3814 = torch.prims.convert_element_type %3806, %int5_1561 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1562 = torch.constant.int 5
%3815 = torch.prims.convert_element_type %3807, %int5_1562 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1563 = torch.constant.int 5
%3816 = torch.prims.convert_element_type %3808, %int5_1563 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1564 = torch.constant.int 1
%int0_1565 = torch.constant.int 0
%3817 = torch.prim.ListConstruct %int1_1564, %int0_1565 : (!torch.int, !torch.int) -> !torch.list<int>
%3818 = torch.aten.permute %56, %3817 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1566 = torch.constant.int 1
%int0_1567 = torch.constant.int 0
%3819 = torch.prim.ListConstruct %int1_1566, %int0_1567 : (!torch.int, !torch.int) -> !torch.list<int>
%3820 = torch.aten.permute %57, %3819 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1568 = torch.constant.int 1
%int0_1569 = torch.constant.int 0
%3821 = torch.prim.ListConstruct %int1_1568, %int0_1569 : (!torch.int, !torch.int) -> !torch.list<int>
%3822 = torch.aten.permute %58, %3821 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1570 = torch.constant.int 1
%int0_1571 = torch.constant.int 0
%3823 = torch.prim.ListConstruct %int1_1570, %int0_1571 : (!torch.int, !torch.int) -> !torch.list<int>
%3824 = torch.aten.permute %59, %3823 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1572 = torch.constant.int 1
%int0_1573 = torch.constant.int 0
%3825 = torch.prim.ListConstruct %int1_1572, %int0_1573 : (!torch.int, !torch.int) -> !torch.list<int>
%3826 = torch.aten.permute %60, %3825 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1574 = torch.constant.int 1
%int0_1575 = torch.constant.int 0
%3827 = torch.prim.ListConstruct %int1_1574, %int0_1575 : (!torch.int, !torch.int) -> !torch.list<int>
%3828 = torch.aten.permute %61, %3827 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1576 = torch.constant.int 1
%int0_1577 = torch.constant.int 0
%3829 = torch.prim.ListConstruct %int1_1576, %int0_1577 : (!torch.int, !torch.int) -> !torch.list<int>
%3830 = torch.aten.permute %62, %3829 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1578 = torch.constant.int 1
%int0_1579 = torch.constant.int 0
%3831 = torch.prim.ListConstruct %int1_1578, %int0_1579 : (!torch.int, !torch.int) -> !torch.list<int>
%3832 = torch.aten.permute %63, %3831 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_1580 = torch.constant.int 4
%3833 = torch.aten.mul.int %int4_1580, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1581 = torch.constant.int 4096
%3834 = torch.prim.ListConstruct %3833, %int4096_1581 : (!torch.int, !torch.int) -> !torch.list<int>
%3835 = torch.aten.view %3809, %3834 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3835, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3836 = torch.aten.mm %3835, %3818 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3836, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1582 = torch.constant.int 4
%int1792 = torch.constant.int 1792
%3837 = torch.prim.ListConstruct %int4_1582, %2482, %int1792 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3838 = torch.aten.view %3836, %3837 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1583 = torch.constant.int 4
%3839 = torch.aten.mul.int %int4_1583, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1584 = torch.constant.int 4096
%3840 = torch.prim.ListConstruct %3839, %int4096_1584 : (!torch.int, !torch.int) -> !torch.list<int>
%3841 = torch.aten.view %3810, %3840 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3841, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3842 = torch.aten.mm %3841, %3820 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3842, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1585 = torch.constant.int 4
%int1792_1586 = torch.constant.int 1792
%3843 = torch.prim.ListConstruct %int4_1585, %2482, %int1792_1586 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3844 = torch.aten.view %3842, %3843 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1587 = torch.constant.int 4
%3845 = torch.aten.mul.int %int4_1587, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1588 = torch.constant.int 4096
%3846 = torch.prim.ListConstruct %3845, %int4096_1588 : (!torch.int, !torch.int) -> !torch.list<int>
%3847 = torch.aten.view %3811, %3846 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3847, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3848 = torch.aten.mm %3847, %3822 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3848, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1589 = torch.constant.int 4
%int1792_1590 = torch.constant.int 1792
%3849 = torch.prim.ListConstruct %int4_1589, %2482, %int1792_1590 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3850 = torch.aten.view %3848, %3849 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1591 = torch.constant.int 4
%3851 = torch.aten.mul.int %int4_1591, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1592 = torch.constant.int 4096
%3852 = torch.prim.ListConstruct %3851, %int4096_1592 : (!torch.int, !torch.int) -> !torch.list<int>
%3853 = torch.aten.view %3812, %3852 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3853, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3854 = torch.aten.mm %3853, %3824 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3854, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1593 = torch.constant.int 4
%int1792_1594 = torch.constant.int 1792
%3855 = torch.prim.ListConstruct %int4_1593, %2482, %int1792_1594 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3856 = torch.aten.view %3854, %3855 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1595 = torch.constant.int 4
%3857 = torch.aten.mul.int %int4_1595, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1596 = torch.constant.int 4096
%3858 = torch.prim.ListConstruct %3857, %int4096_1596 : (!torch.int, !torch.int) -> !torch.list<int>
%3859 = torch.aten.view %3813, %3858 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3859, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3860 = torch.aten.mm %3859, %3826 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3860, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1597 = torch.constant.int 4
%int1792_1598 = torch.constant.int 1792
%3861 = torch.prim.ListConstruct %int4_1597, %2482, %int1792_1598 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3862 = torch.aten.view %3860, %3861 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1599 = torch.constant.int 4
%3863 = torch.aten.mul.int %int4_1599, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1600 = torch.constant.int 4096
%3864 = torch.prim.ListConstruct %3863, %int4096_1600 : (!torch.int, !torch.int) -> !torch.list<int>
%3865 = torch.aten.view %3814, %3864 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3865, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3866 = torch.aten.mm %3865, %3828 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3866, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1601 = torch.constant.int 4
%int1792_1602 = torch.constant.int 1792
%3867 = torch.prim.ListConstruct %int4_1601, %2482, %int1792_1602 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3868 = torch.aten.view %3866, %3867 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1603 = torch.constant.int 4
%3869 = torch.aten.mul.int %int4_1603, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1604 = torch.constant.int 4096
%3870 = torch.prim.ListConstruct %3869, %int4096_1604 : (!torch.int, !torch.int) -> !torch.list<int>
%3871 = torch.aten.view %3815, %3870 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3871, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3872 = torch.aten.mm %3871, %3830 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3872, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1605 = torch.constant.int 4
%int1792_1606 = torch.constant.int 1792
%3873 = torch.prim.ListConstruct %int4_1605, %2482, %int1792_1606 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3874 = torch.aten.view %3872, %3873 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1607 = torch.constant.int 4
%3875 = torch.aten.mul.int %int4_1607, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1608 = torch.constant.int 4096
%3876 = torch.prim.ListConstruct %3875, %int4096_1608 : (!torch.int, !torch.int) -> !torch.list<int>
%3877 = torch.aten.view %3816, %3876 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3877, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3878 = torch.aten.mm %3877, %3832 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3878, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1609 = torch.constant.int 4
%int1792_1610 = torch.constant.int 1792
%3879 = torch.prim.ListConstruct %int4_1609, %2482, %int1792_1610 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3880 = torch.aten.view %3878, %3879 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3881 = torch.aten.silu %3838 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3882 = torch.aten.silu %3844 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3883 = torch.aten.silu %3850 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3884 = torch.aten.silu %3856 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3885 = torch.aten.silu %3862 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3886 = torch.aten.silu %3868 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3887 = torch.aten.silu %3874 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3888 = torch.aten.silu %3880 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_1611 = torch.constant.int 1
%int0_1612 = torch.constant.int 0
%3889 = torch.prim.ListConstruct %int1_1611, %int0_1612 : (!torch.int, !torch.int) -> !torch.list<int>
%3890 = torch.aten.permute %64, %3889 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1613 = torch.constant.int 1
%int0_1614 = torch.constant.int 0
%3891 = torch.prim.ListConstruct %int1_1613, %int0_1614 : (!torch.int, !torch.int) -> !torch.list<int>
%3892 = torch.aten.permute %65, %3891 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1615 = torch.constant.int 1
%int0_1616 = torch.constant.int 0
%3893 = torch.prim.ListConstruct %int1_1615, %int0_1616 : (!torch.int, !torch.int) -> !torch.list<int>
%3894 = torch.aten.permute %66, %3893 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1617 = torch.constant.int 1
%int0_1618 = torch.constant.int 0
%3895 = torch.prim.ListConstruct %int1_1617, %int0_1618 : (!torch.int, !torch.int) -> !torch.list<int>
%3896 = torch.aten.permute %67, %3895 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1619 = torch.constant.int 1
%int0_1620 = torch.constant.int 0
%3897 = torch.prim.ListConstruct %int1_1619, %int0_1620 : (!torch.int, !torch.int) -> !torch.list<int>
%3898 = torch.aten.permute %68, %3897 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1621 = torch.constant.int 1
%int0_1622 = torch.constant.int 0
%3899 = torch.prim.ListConstruct %int1_1621, %int0_1622 : (!torch.int, !torch.int) -> !torch.list<int>
%3900 = torch.aten.permute %69, %3899 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1623 = torch.constant.int 1
%int0_1624 = torch.constant.int 0
%3901 = torch.prim.ListConstruct %int1_1623, %int0_1624 : (!torch.int, !torch.int) -> !torch.list<int>
%3902 = torch.aten.permute %70, %3901 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_1625 = torch.constant.int 1
%int0_1626 = torch.constant.int 0
%3903 = torch.prim.ListConstruct %int1_1625, %int0_1626 : (!torch.int, !torch.int) -> !torch.list<int>
%3904 = torch.aten.permute %71, %3903 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_1627 = torch.constant.int 4
%3905 = torch.aten.mul.int %int4_1627, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1628 = torch.constant.int 4096
%3906 = torch.prim.ListConstruct %3905, %int4096_1628 : (!torch.int, !torch.int) -> !torch.list<int>
%3907 = torch.aten.view %3809, %3906 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3907, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3908 = torch.aten.mm %3907, %3890 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3908, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1629 = torch.constant.int 4
%int1792_1630 = torch.constant.int 1792
%3909 = torch.prim.ListConstruct %int4_1629, %2482, %int1792_1630 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3910 = torch.aten.view %3908, %3909 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1631 = torch.constant.int 4
%3911 = torch.aten.mul.int %int4_1631, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1632 = torch.constant.int 4096
%3912 = torch.prim.ListConstruct %3911, %int4096_1632 : (!torch.int, !torch.int) -> !torch.list<int>
%3913 = torch.aten.view %3810, %3912 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3913, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3914 = torch.aten.mm %3913, %3892 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3914, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1633 = torch.constant.int 4
%int1792_1634 = torch.constant.int 1792
%3915 = torch.prim.ListConstruct %int4_1633, %2482, %int1792_1634 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3916 = torch.aten.view %3914, %3915 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1635 = torch.constant.int 4
%3917 = torch.aten.mul.int %int4_1635, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1636 = torch.constant.int 4096
%3918 = torch.prim.ListConstruct %3917, %int4096_1636 : (!torch.int, !torch.int) -> !torch.list<int>
%3919 = torch.aten.view %3811, %3918 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3919, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3920 = torch.aten.mm %3919, %3894 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3920, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1637 = torch.constant.int 4
%int1792_1638 = torch.constant.int 1792
%3921 = torch.prim.ListConstruct %int4_1637, %2482, %int1792_1638 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3922 = torch.aten.view %3920, %3921 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1639 = torch.constant.int 4
%3923 = torch.aten.mul.int %int4_1639, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1640 = torch.constant.int 4096
%3924 = torch.prim.ListConstruct %3923, %int4096_1640 : (!torch.int, !torch.int) -> !torch.list<int>
%3925 = torch.aten.view %3812, %3924 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3925, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3926 = torch.aten.mm %3925, %3896 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3926, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1641 = torch.constant.int 4
%int1792_1642 = torch.constant.int 1792
%3927 = torch.prim.ListConstruct %int4_1641, %2482, %int1792_1642 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3928 = torch.aten.view %3926, %3927 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1643 = torch.constant.int 4
%3929 = torch.aten.mul.int %int4_1643, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1644 = torch.constant.int 4096
%3930 = torch.prim.ListConstruct %3929, %int4096_1644 : (!torch.int, !torch.int) -> !torch.list<int>
%3931 = torch.aten.view %3813, %3930 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3931, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3932 = torch.aten.mm %3931, %3898 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3932, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1645 = torch.constant.int 4
%int1792_1646 = torch.constant.int 1792
%3933 = torch.prim.ListConstruct %int4_1645, %2482, %int1792_1646 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3934 = torch.aten.view %3932, %3933 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1647 = torch.constant.int 4
%3935 = torch.aten.mul.int %int4_1647, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1648 = torch.constant.int 4096
%3936 = torch.prim.ListConstruct %3935, %int4096_1648 : (!torch.int, !torch.int) -> !torch.list<int>
%3937 = torch.aten.view %3814, %3936 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3937, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3938 = torch.aten.mm %3937, %3900 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3938, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1649 = torch.constant.int 4
%int1792_1650 = torch.constant.int 1792
%3939 = torch.prim.ListConstruct %int4_1649, %2482, %int1792_1650 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3940 = torch.aten.view %3938, %3939 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1651 = torch.constant.int 4
%3941 = torch.aten.mul.int %int4_1651, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1652 = torch.constant.int 4096
%3942 = torch.prim.ListConstruct %3941, %int4096_1652 : (!torch.int, !torch.int) -> !torch.list<int>
%3943 = torch.aten.view %3815, %3942 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3943, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3944 = torch.aten.mm %3943, %3902 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3944, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1653 = torch.constant.int 4
%int1792_1654 = torch.constant.int 1792
%3945 = torch.prim.ListConstruct %int4_1653, %2482, %int1792_1654 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3946 = torch.aten.view %3944, %3945 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_1655 = torch.constant.int 4
%3947 = torch.aten.mul.int %int4_1655, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1656 = torch.constant.int 4096
%3948 = torch.prim.ListConstruct %3947, %int4096_1656 : (!torch.int, !torch.int) -> !torch.list<int>
%3949 = torch.aten.view %3816, %3948 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3949, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%3950 = torch.aten.mm %3949, %3904 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3950, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_1657 = torch.constant.int 4
%int1792_1658 = torch.constant.int 1792
%3951 = torch.prim.ListConstruct %int4_1657, %2482, %int1792_1658 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3952 = torch.aten.view %3950, %3951 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3953 = torch.aten.mul.Tensor %3881, %3910 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3954 = torch.aten.mul.Tensor %3882, %3916 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3955 = torch.aten.mul.Tensor %3883, %3922 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3956 = torch.aten.mul.Tensor %3884, %3928 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3957 = torch.aten.mul.Tensor %3885, %3934 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3958 = torch.aten.mul.Tensor %3886, %3940 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3959 = torch.aten.mul.Tensor %3887, %3946 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%3960 = torch.aten.mul.Tensor %3888, %3952 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %3960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_1659 = torch.constant.int 1
%int0_1660 = torch.constant.int 0
%3961 = torch.prim.ListConstruct %int1_1659, %int0_1660 : (!torch.int, !torch.int) -> !torch.list<int>
%3962 = torch.aten.permute %72, %3961 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1661 = torch.constant.int 1
%int0_1662 = torch.constant.int 0
%3963 = torch.prim.ListConstruct %int1_1661, %int0_1662 : (!torch.int, !torch.int) -> !torch.list<int>
%3964 = torch.aten.permute %73, %3963 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1663 = torch.constant.int 1
%int0_1664 = torch.constant.int 0
%3965 = torch.prim.ListConstruct %int1_1663, %int0_1664 : (!torch.int, !torch.int) -> !torch.list<int>
%3966 = torch.aten.permute %74, %3965 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1665 = torch.constant.int 1
%int0_1666 = torch.constant.int 0
%3967 = torch.prim.ListConstruct %int1_1665, %int0_1666 : (!torch.int, !torch.int) -> !torch.list<int>
%3968 = torch.aten.permute %75, %3967 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1667 = torch.constant.int 1
%int0_1668 = torch.constant.int 0
%3969 = torch.prim.ListConstruct %int1_1667, %int0_1668 : (!torch.int, !torch.int) -> !torch.list<int>
%3970 = torch.aten.permute %76, %3969 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1669 = torch.constant.int 1
%int0_1670 = torch.constant.int 0
%3971 = torch.prim.ListConstruct %int1_1669, %int0_1670 : (!torch.int, !torch.int) -> !torch.list<int>
%3972 = torch.aten.permute %77, %3971 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1671 = torch.constant.int 1
%int0_1672 = torch.constant.int 0
%3973 = torch.prim.ListConstruct %int1_1671, %int0_1672 : (!torch.int, !torch.int) -> !torch.list<int>
%3974 = torch.aten.permute %78, %3973 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1673 = torch.constant.int 1
%int0_1674 = torch.constant.int 0
%3975 = torch.prim.ListConstruct %int1_1673, %int0_1674 : (!torch.int, !torch.int) -> !torch.list<int>
%3976 = torch.aten.permute %79, %3975 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_1675 = torch.constant.int 1
%3977 = torch.aten.size.int %3838, %int1_1675 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1676 = torch.constant.int 4
%3978 = torch.aten.mul.int %int4_1676, %3977 : !torch.int, !torch.int -> !torch.int
%int1792_1677 = torch.constant.int 1792
%3979 = torch.prim.ListConstruct %3978, %int1792_1677 : (!torch.int, !torch.int) -> !torch.list<int>
%3980 = torch.aten.view %3953, %3979 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3980, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%3981 = torch.aten.mm %3980, %3962 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3981, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1678 = torch.constant.int 4
%int4096_1679 = torch.constant.int 4096
%3982 = torch.prim.ListConstruct %int4_1678, %3977, %int4096_1679 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3983 = torch.aten.view %3981, %3982 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1680 = torch.constant.int 1
%3984 = torch.aten.size.int %3844, %int1_1680 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1681 = torch.constant.int 4
%3985 = torch.aten.mul.int %int4_1681, %3984 : !torch.int, !torch.int -> !torch.int
%int1792_1682 = torch.constant.int 1792
%3986 = torch.prim.ListConstruct %3985, %int1792_1682 : (!torch.int, !torch.int) -> !torch.list<int>
%3987 = torch.aten.view %3954, %3986 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3987, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%3988 = torch.aten.mm %3987, %3964 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3988, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1683 = torch.constant.int 4
%int4096_1684 = torch.constant.int 4096
%3989 = torch.prim.ListConstruct %int4_1683, %3984, %int4096_1684 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3990 = torch.aten.view %3988, %3989 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1685 = torch.constant.int 1
%3991 = torch.aten.size.int %3850, %int1_1685 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1686 = torch.constant.int 4
%3992 = torch.aten.mul.int %int4_1686, %3991 : !torch.int, !torch.int -> !torch.int
%int1792_1687 = torch.constant.int 1792
%3993 = torch.prim.ListConstruct %3992, %int1792_1687 : (!torch.int, !torch.int) -> !torch.list<int>
%3994 = torch.aten.view %3955, %3993 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %3994, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%3995 = torch.aten.mm %3994, %3966 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %3995, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1688 = torch.constant.int 4
%int4096_1689 = torch.constant.int 4096
%3996 = torch.prim.ListConstruct %int4_1688, %3991, %int4096_1689 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%3997 = torch.aten.view %3995, %3996 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %3997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1690 = torch.constant.int 1
%3998 = torch.aten.size.int %3856, %int1_1690 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1691 = torch.constant.int 4
%3999 = torch.aten.mul.int %int4_1691, %3998 : !torch.int, !torch.int -> !torch.int
%int1792_1692 = torch.constant.int 1792
%4000 = torch.prim.ListConstruct %3999, %int1792_1692 : (!torch.int, !torch.int) -> !torch.list<int>
%4001 = torch.aten.view %3956, %4000 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %4001, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%4002 = torch.aten.mm %4001, %3968 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4002, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1693 = torch.constant.int 4
%int4096_1694 = torch.constant.int 4096
%4003 = torch.prim.ListConstruct %int4_1693, %3998, %int4096_1694 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4004 = torch.aten.view %4002, %4003 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1695 = torch.constant.int 1
%4005 = torch.aten.size.int %3862, %int1_1695 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1696 = torch.constant.int 4
%4006 = torch.aten.mul.int %int4_1696, %4005 : !torch.int, !torch.int -> !torch.int
%int1792_1697 = torch.constant.int 1792
%4007 = torch.prim.ListConstruct %4006, %int1792_1697 : (!torch.int, !torch.int) -> !torch.list<int>
%4008 = torch.aten.view %3957, %4007 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %4008, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%4009 = torch.aten.mm %4008, %3970 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4009, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1698 = torch.constant.int 4
%int4096_1699 = torch.constant.int 4096
%4010 = torch.prim.ListConstruct %int4_1698, %4005, %int4096_1699 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4011 = torch.aten.view %4009, %4010 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1700 = torch.constant.int 1
%4012 = torch.aten.size.int %3868, %int1_1700 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1701 = torch.constant.int 4
%4013 = torch.aten.mul.int %int4_1701, %4012 : !torch.int, !torch.int -> !torch.int
%int1792_1702 = torch.constant.int 1792
%4014 = torch.prim.ListConstruct %4013, %int1792_1702 : (!torch.int, !torch.int) -> !torch.list<int>
%4015 = torch.aten.view %3958, %4014 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %4015, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%4016 = torch.aten.mm %4015, %3972 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4016, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1703 = torch.constant.int 4
%int4096_1704 = torch.constant.int 4096
%4017 = torch.prim.ListConstruct %int4_1703, %4012, %int4096_1704 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4018 = torch.aten.view %4016, %4017 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1705 = torch.constant.int 1
%4019 = torch.aten.size.int %3874, %int1_1705 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1706 = torch.constant.int 4
%4020 = torch.aten.mul.int %int4_1706, %4019 : !torch.int, !torch.int -> !torch.int
%int1792_1707 = torch.constant.int 1792
%4021 = torch.prim.ListConstruct %4020, %int1792_1707 : (!torch.int, !torch.int) -> !torch.list<int>
%4022 = torch.aten.view %3959, %4021 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %4022, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%4023 = torch.aten.mm %4022, %3974 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4023, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1708 = torch.constant.int 4
%int4096_1709 = torch.constant.int 4096
%4024 = torch.prim.ListConstruct %int4_1708, %4019, %int4096_1709 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4025 = torch.aten.view %4023, %4024 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1710 = torch.constant.int 1
%4026 = torch.aten.size.int %3880, %int1_1710 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_1711 = torch.constant.int 4
%4027 = torch.aten.mul.int %int4_1711, %4026 : !torch.int, !torch.int -> !torch.int
%int1792_1712 = torch.constant.int 1792
%4028 = torch.prim.ListConstruct %4027, %int1792_1712 : (!torch.int, !torch.int) -> !torch.list<int>
%4029 = torch.aten.view %3960, %4028 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %4029, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%4030 = torch.aten.mm %4029, %3976 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4030, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_1713 = torch.constant.int 4
%int4096_1714 = torch.constant.int 4096
%4031 = torch.prim.ListConstruct %int4_1713, %4026, %int4096_1714 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4032 = torch.aten.view %4030, %4031 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4033 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1715 = arith.constant 1 : index
%dim_1716 = tensor.dim %4033, %c1_1715 : tensor<4x?x4096xf16>
%4034 = flow.tensor.transfer %4033 : tensor<4x?x4096xf16>{%dim_1716} to #hal.device.promise<@__device_0>
%4035 = torch_c.from_builtin_tensor %4034 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4036 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1717 = arith.constant 1 : index
%dim_1718 = tensor.dim %4036, %c1_1717 : tensor<4x?x4096xf16>
%4037 = flow.tensor.transfer %4036 : tensor<4x?x4096xf16>{%dim_1718} to #hal.device.promise<@__device_0>
%4038 = torch_c.from_builtin_tensor %4037 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4039 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1719 = arith.constant 1 : index
%dim_1720 = tensor.dim %4039, %c1_1719 : tensor<4x?x4096xf16>
%4040 = flow.tensor.transfer %4039 : tensor<4x?x4096xf16>{%dim_1720} to #hal.device.promise<@__device_0>
%4041 = torch_c.from_builtin_tensor %4040 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4042 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1721 = arith.constant 1 : index
%dim_1722 = tensor.dim %4042, %c1_1721 : tensor<4x?x4096xf16>
%4043 = flow.tensor.transfer %4042 : tensor<4x?x4096xf16>{%dim_1722} to #hal.device.promise<@__device_0>
%4044 = torch_c.from_builtin_tensor %4043 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4045 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1723 = arith.constant 1 : index
%dim_1724 = tensor.dim %4045, %c1_1723 : tensor<4x?x4096xf16>
%4046 = flow.tensor.transfer %4045 : tensor<4x?x4096xf16>{%dim_1724} to #hal.device.promise<@__device_0>
%4047 = torch_c.from_builtin_tensor %4046 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4048 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1725 = arith.constant 1 : index
%dim_1726 = tensor.dim %4048, %c1_1725 : tensor<4x?x4096xf16>
%4049 = flow.tensor.transfer %4048 : tensor<4x?x4096xf16>{%dim_1726} to #hal.device.promise<@__device_0>
%4050 = torch_c.from_builtin_tensor %4049 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4051 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1727 = arith.constant 1 : index
%dim_1728 = tensor.dim %4051, %c1_1727 : tensor<4x?x4096xf16>
%4052 = flow.tensor.transfer %4051 : tensor<4x?x4096xf16>{%dim_1728} to #hal.device.promise<@__device_0>
%4053 = torch_c.from_builtin_tensor %4052 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1729 = torch.constant.int 1
%4054 = torch.aten.add.Tensor %3983, %4035, %int1_1729 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1730 = torch.constant.int 1
%4055 = torch.aten.add.Tensor %4054, %4038, %int1_1730 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1731 = torch.constant.int 1
%4056 = torch.aten.add.Tensor %4055, %4041, %int1_1731 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1732 = torch.constant.int 1
%4057 = torch.aten.add.Tensor %4056, %4044, %int1_1732 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1733 = torch.constant.int 1
%4058 = torch.aten.add.Tensor %4057, %4047, %int1_1733 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1734 = torch.constant.int 1
%4059 = torch.aten.add.Tensor %4058, %4050, %int1_1734 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1735 = torch.constant.int 1
%4060 = torch.aten.add.Tensor %4059, %4053, %int1_1735 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4061 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1736 = arith.constant 1 : index
%dim_1737 = tensor.dim %4061, %c1_1736 : tensor<4x?x4096xf16>
%4062 = flow.tensor.transfer %4061 : tensor<4x?x4096xf16>{%dim_1737} to #hal.device.promise<@__device_1>
%4063 = torch_c.from_builtin_tensor %4062 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4064 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1738 = arith.constant 1 : index
%dim_1739 = tensor.dim %4064, %c1_1738 : tensor<4x?x4096xf16>
%4065 = flow.tensor.transfer %4064 : tensor<4x?x4096xf16>{%dim_1739} to #hal.device.promise<@__device_1>
%4066 = torch_c.from_builtin_tensor %4065 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4067 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1740 = arith.constant 1 : index
%dim_1741 = tensor.dim %4067, %c1_1740 : tensor<4x?x4096xf16>
%4068 = flow.tensor.transfer %4067 : tensor<4x?x4096xf16>{%dim_1741} to #hal.device.promise<@__device_1>
%4069 = torch_c.from_builtin_tensor %4068 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4070 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1742 = arith.constant 1 : index
%dim_1743 = tensor.dim %4070, %c1_1742 : tensor<4x?x4096xf16>
%4071 = flow.tensor.transfer %4070 : tensor<4x?x4096xf16>{%dim_1743} to #hal.device.promise<@__device_1>
%4072 = torch_c.from_builtin_tensor %4071 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4073 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1744 = arith.constant 1 : index
%dim_1745 = tensor.dim %4073, %c1_1744 : tensor<4x?x4096xf16>
%4074 = flow.tensor.transfer %4073 : tensor<4x?x4096xf16>{%dim_1745} to #hal.device.promise<@__device_1>
%4075 = torch_c.from_builtin_tensor %4074 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4076 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1746 = arith.constant 1 : index
%dim_1747 = tensor.dim %4076, %c1_1746 : tensor<4x?x4096xf16>
%4077 = flow.tensor.transfer %4076 : tensor<4x?x4096xf16>{%dim_1747} to #hal.device.promise<@__device_1>
%4078 = torch_c.from_builtin_tensor %4077 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4079 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1748 = arith.constant 1 : index
%dim_1749 = tensor.dim %4079, %c1_1748 : tensor<4x?x4096xf16>
%4080 = flow.tensor.transfer %4079 : tensor<4x?x4096xf16>{%dim_1749} to #hal.device.promise<@__device_1>
%4081 = torch_c.from_builtin_tensor %4080 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1750 = torch.constant.int 1
%4082 = torch.aten.add.Tensor %4063, %3990, %int1_1750 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1751 = torch.constant.int 1
%4083 = torch.aten.add.Tensor %4082, %4066, %int1_1751 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1752 = torch.constant.int 1
%4084 = torch.aten.add.Tensor %4083, %4069, %int1_1752 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1753 = torch.constant.int 1
%4085 = torch.aten.add.Tensor %4084, %4072, %int1_1753 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1754 = torch.constant.int 1
%4086 = torch.aten.add.Tensor %4085, %4075, %int1_1754 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1755 = torch.constant.int 1
%4087 = torch.aten.add.Tensor %4086, %4078, %int1_1755 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1756 = torch.constant.int 1
%4088 = torch.aten.add.Tensor %4087, %4081, %int1_1756 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4089 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1757 = arith.constant 1 : index
%dim_1758 = tensor.dim %4089, %c1_1757 : tensor<4x?x4096xf16>
%4090 = flow.tensor.transfer %4089 : tensor<4x?x4096xf16>{%dim_1758} to #hal.device.promise<@__device_2>
%4091 = torch_c.from_builtin_tensor %4090 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4092 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1759 = arith.constant 1 : index
%dim_1760 = tensor.dim %4092, %c1_1759 : tensor<4x?x4096xf16>
%4093 = flow.tensor.transfer %4092 : tensor<4x?x4096xf16>{%dim_1760} to #hal.device.promise<@__device_2>
%4094 = torch_c.from_builtin_tensor %4093 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4095 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1761 = arith.constant 1 : index
%dim_1762 = tensor.dim %4095, %c1_1761 : tensor<4x?x4096xf16>
%4096 = flow.tensor.transfer %4095 : tensor<4x?x4096xf16>{%dim_1762} to #hal.device.promise<@__device_2>
%4097 = torch_c.from_builtin_tensor %4096 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4098 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1763 = arith.constant 1 : index
%dim_1764 = tensor.dim %4098, %c1_1763 : tensor<4x?x4096xf16>
%4099 = flow.tensor.transfer %4098 : tensor<4x?x4096xf16>{%dim_1764} to #hal.device.promise<@__device_2>
%4100 = torch_c.from_builtin_tensor %4099 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4101 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1765 = arith.constant 1 : index
%dim_1766 = tensor.dim %4101, %c1_1765 : tensor<4x?x4096xf16>
%4102 = flow.tensor.transfer %4101 : tensor<4x?x4096xf16>{%dim_1766} to #hal.device.promise<@__device_2>
%4103 = torch_c.from_builtin_tensor %4102 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4104 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1767 = arith.constant 1 : index
%dim_1768 = tensor.dim %4104, %c1_1767 : tensor<4x?x4096xf16>
%4105 = flow.tensor.transfer %4104 : tensor<4x?x4096xf16>{%dim_1768} to #hal.device.promise<@__device_2>
%4106 = torch_c.from_builtin_tensor %4105 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4107 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1769 = arith.constant 1 : index
%dim_1770 = tensor.dim %4107, %c1_1769 : tensor<4x?x4096xf16>
%4108 = flow.tensor.transfer %4107 : tensor<4x?x4096xf16>{%dim_1770} to #hal.device.promise<@__device_2>
%4109 = torch_c.from_builtin_tensor %4108 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1771 = torch.constant.int 1
%4110 = torch.aten.add.Tensor %4091, %4094, %int1_1771 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1772 = torch.constant.int 1
%4111 = torch.aten.add.Tensor %4110, %3997, %int1_1772 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1773 = torch.constant.int 1
%4112 = torch.aten.add.Tensor %4111, %4097, %int1_1773 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1774 = torch.constant.int 1
%4113 = torch.aten.add.Tensor %4112, %4100, %int1_1774 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1775 = torch.constant.int 1
%4114 = torch.aten.add.Tensor %4113, %4103, %int1_1775 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1776 = torch.constant.int 1
%4115 = torch.aten.add.Tensor %4114, %4106, %int1_1776 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1777 = torch.constant.int 1
%4116 = torch.aten.add.Tensor %4115, %4109, %int1_1777 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4117 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1778 = arith.constant 1 : index
%dim_1779 = tensor.dim %4117, %c1_1778 : tensor<4x?x4096xf16>
%4118 = flow.tensor.transfer %4117 : tensor<4x?x4096xf16>{%dim_1779} to #hal.device.promise<@__device_3>
%4119 = torch_c.from_builtin_tensor %4118 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4120 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1780 = arith.constant 1 : index
%dim_1781 = tensor.dim %4120, %c1_1780 : tensor<4x?x4096xf16>
%4121 = flow.tensor.transfer %4120 : tensor<4x?x4096xf16>{%dim_1781} to #hal.device.promise<@__device_3>
%4122 = torch_c.from_builtin_tensor %4121 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4123 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1782 = arith.constant 1 : index
%dim_1783 = tensor.dim %4123, %c1_1782 : tensor<4x?x4096xf16>
%4124 = flow.tensor.transfer %4123 : tensor<4x?x4096xf16>{%dim_1783} to #hal.device.promise<@__device_3>
%4125 = torch_c.from_builtin_tensor %4124 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4126 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1784 = arith.constant 1 : index
%dim_1785 = tensor.dim %4126, %c1_1784 : tensor<4x?x4096xf16>
%4127 = flow.tensor.transfer %4126 : tensor<4x?x4096xf16>{%dim_1785} to #hal.device.promise<@__device_3>
%4128 = torch_c.from_builtin_tensor %4127 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4129 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1786 = arith.constant 1 : index
%dim_1787 = tensor.dim %4129, %c1_1786 : tensor<4x?x4096xf16>
%4130 = flow.tensor.transfer %4129 : tensor<4x?x4096xf16>{%dim_1787} to #hal.device.promise<@__device_3>
%4131 = torch_c.from_builtin_tensor %4130 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4132 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1788 = arith.constant 1 : index
%dim_1789 = tensor.dim %4132, %c1_1788 : tensor<4x?x4096xf16>
%4133 = flow.tensor.transfer %4132 : tensor<4x?x4096xf16>{%dim_1789} to #hal.device.promise<@__device_3>
%4134 = torch_c.from_builtin_tensor %4133 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4135 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1790 = arith.constant 1 : index
%dim_1791 = tensor.dim %4135, %c1_1790 : tensor<4x?x4096xf16>
%4136 = flow.tensor.transfer %4135 : tensor<4x?x4096xf16>{%dim_1791} to #hal.device.promise<@__device_3>
%4137 = torch_c.from_builtin_tensor %4136 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1792 = torch.constant.int 1
%4138 = torch.aten.add.Tensor %4119, %4122, %int1_1792 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1793 = torch.constant.int 1
%4139 = torch.aten.add.Tensor %4138, %4125, %int1_1793 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1794 = torch.constant.int 1
%4140 = torch.aten.add.Tensor %4139, %4004, %int1_1794 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1795 = torch.constant.int 1
%4141 = torch.aten.add.Tensor %4140, %4128, %int1_1795 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1796 = torch.constant.int 1
%4142 = torch.aten.add.Tensor %4141, %4131, %int1_1796 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1797 = torch.constant.int 1
%4143 = torch.aten.add.Tensor %4142, %4134, %int1_1797 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1798 = torch.constant.int 1
%4144 = torch.aten.add.Tensor %4143, %4137, %int1_1798 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4145 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1799 = arith.constant 1 : index
%dim_1800 = tensor.dim %4145, %c1_1799 : tensor<4x?x4096xf16>
%4146 = flow.tensor.transfer %4145 : tensor<4x?x4096xf16>{%dim_1800} to #hal.device.promise<@__device_4>
%4147 = torch_c.from_builtin_tensor %4146 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4148 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1801 = arith.constant 1 : index
%dim_1802 = tensor.dim %4148, %c1_1801 : tensor<4x?x4096xf16>
%4149 = flow.tensor.transfer %4148 : tensor<4x?x4096xf16>{%dim_1802} to #hal.device.promise<@__device_4>
%4150 = torch_c.from_builtin_tensor %4149 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4151 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1803 = arith.constant 1 : index
%dim_1804 = tensor.dim %4151, %c1_1803 : tensor<4x?x4096xf16>
%4152 = flow.tensor.transfer %4151 : tensor<4x?x4096xf16>{%dim_1804} to #hal.device.promise<@__device_4>
%4153 = torch_c.from_builtin_tensor %4152 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4154 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1805 = arith.constant 1 : index
%dim_1806 = tensor.dim %4154, %c1_1805 : tensor<4x?x4096xf16>
%4155 = flow.tensor.transfer %4154 : tensor<4x?x4096xf16>{%dim_1806} to #hal.device.promise<@__device_4>
%4156 = torch_c.from_builtin_tensor %4155 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4157 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1807 = arith.constant 1 : index
%dim_1808 = tensor.dim %4157, %c1_1807 : tensor<4x?x4096xf16>
%4158 = flow.tensor.transfer %4157 : tensor<4x?x4096xf16>{%dim_1808} to #hal.device.promise<@__device_4>
%4159 = torch_c.from_builtin_tensor %4158 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4160 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1809 = arith.constant 1 : index
%dim_1810 = tensor.dim %4160, %c1_1809 : tensor<4x?x4096xf16>
%4161 = flow.tensor.transfer %4160 : tensor<4x?x4096xf16>{%dim_1810} to #hal.device.promise<@__device_4>
%4162 = torch_c.from_builtin_tensor %4161 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4163 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1811 = arith.constant 1 : index
%dim_1812 = tensor.dim %4163, %c1_1811 : tensor<4x?x4096xf16>
%4164 = flow.tensor.transfer %4163 : tensor<4x?x4096xf16>{%dim_1812} to #hal.device.promise<@__device_4>
%4165 = torch_c.from_builtin_tensor %4164 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1813 = torch.constant.int 1
%4166 = torch.aten.add.Tensor %4147, %4150, %int1_1813 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1814 = torch.constant.int 1
%4167 = torch.aten.add.Tensor %4166, %4153, %int1_1814 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1815 = torch.constant.int 1
%4168 = torch.aten.add.Tensor %4167, %4156, %int1_1815 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1816 = torch.constant.int 1
%4169 = torch.aten.add.Tensor %4168, %4011, %int1_1816 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1817 = torch.constant.int 1
%4170 = torch.aten.add.Tensor %4169, %4159, %int1_1817 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1818 = torch.constant.int 1
%4171 = torch.aten.add.Tensor %4170, %4162, %int1_1818 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1819 = torch.constant.int 1
%4172 = torch.aten.add.Tensor %4171, %4165, %int1_1819 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4173 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1820 = arith.constant 1 : index
%dim_1821 = tensor.dim %4173, %c1_1820 : tensor<4x?x4096xf16>
%4174 = flow.tensor.transfer %4173 : tensor<4x?x4096xf16>{%dim_1821} to #hal.device.promise<@__device_5>
%4175 = torch_c.from_builtin_tensor %4174 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4176 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1822 = arith.constant 1 : index
%dim_1823 = tensor.dim %4176, %c1_1822 : tensor<4x?x4096xf16>
%4177 = flow.tensor.transfer %4176 : tensor<4x?x4096xf16>{%dim_1823} to #hal.device.promise<@__device_5>
%4178 = torch_c.from_builtin_tensor %4177 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4179 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1824 = arith.constant 1 : index
%dim_1825 = tensor.dim %4179, %c1_1824 : tensor<4x?x4096xf16>
%4180 = flow.tensor.transfer %4179 : tensor<4x?x4096xf16>{%dim_1825} to #hal.device.promise<@__device_5>
%4181 = torch_c.from_builtin_tensor %4180 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4182 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1826 = arith.constant 1 : index
%dim_1827 = tensor.dim %4182, %c1_1826 : tensor<4x?x4096xf16>
%4183 = flow.tensor.transfer %4182 : tensor<4x?x4096xf16>{%dim_1827} to #hal.device.promise<@__device_5>
%4184 = torch_c.from_builtin_tensor %4183 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4185 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1828 = arith.constant 1 : index
%dim_1829 = tensor.dim %4185, %c1_1828 : tensor<4x?x4096xf16>
%4186 = flow.tensor.transfer %4185 : tensor<4x?x4096xf16>{%dim_1829} to #hal.device.promise<@__device_5>
%4187 = torch_c.from_builtin_tensor %4186 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4188 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1830 = arith.constant 1 : index
%dim_1831 = tensor.dim %4188, %c1_1830 : tensor<4x?x4096xf16>
%4189 = flow.tensor.transfer %4188 : tensor<4x?x4096xf16>{%dim_1831} to #hal.device.promise<@__device_5>
%4190 = torch_c.from_builtin_tensor %4189 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4191 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1832 = arith.constant 1 : index
%dim_1833 = tensor.dim %4191, %c1_1832 : tensor<4x?x4096xf16>
%4192 = flow.tensor.transfer %4191 : tensor<4x?x4096xf16>{%dim_1833} to #hal.device.promise<@__device_5>
%4193 = torch_c.from_builtin_tensor %4192 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1834 = torch.constant.int 1
%4194 = torch.aten.add.Tensor %4175, %4178, %int1_1834 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1835 = torch.constant.int 1
%4195 = torch.aten.add.Tensor %4194, %4181, %int1_1835 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1836 = torch.constant.int 1
%4196 = torch.aten.add.Tensor %4195, %4184, %int1_1836 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1837 = torch.constant.int 1
%4197 = torch.aten.add.Tensor %4196, %4187, %int1_1837 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1838 = torch.constant.int 1
%4198 = torch.aten.add.Tensor %4197, %4018, %int1_1838 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1839 = torch.constant.int 1
%4199 = torch.aten.add.Tensor %4198, %4190, %int1_1839 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1840 = torch.constant.int 1
%4200 = torch.aten.add.Tensor %4199, %4193, %int1_1840 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4201 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1841 = arith.constant 1 : index
%dim_1842 = tensor.dim %4201, %c1_1841 : tensor<4x?x4096xf16>
%4202 = flow.tensor.transfer %4201 : tensor<4x?x4096xf16>{%dim_1842} to #hal.device.promise<@__device_6>
%4203 = torch_c.from_builtin_tensor %4202 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4204 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1843 = arith.constant 1 : index
%dim_1844 = tensor.dim %4204, %c1_1843 : tensor<4x?x4096xf16>
%4205 = flow.tensor.transfer %4204 : tensor<4x?x4096xf16>{%dim_1844} to #hal.device.promise<@__device_6>
%4206 = torch_c.from_builtin_tensor %4205 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4207 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1845 = arith.constant 1 : index
%dim_1846 = tensor.dim %4207, %c1_1845 : tensor<4x?x4096xf16>
%4208 = flow.tensor.transfer %4207 : tensor<4x?x4096xf16>{%dim_1846} to #hal.device.promise<@__device_6>
%4209 = torch_c.from_builtin_tensor %4208 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4210 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1847 = arith.constant 1 : index
%dim_1848 = tensor.dim %4210, %c1_1847 : tensor<4x?x4096xf16>
%4211 = flow.tensor.transfer %4210 : tensor<4x?x4096xf16>{%dim_1848} to #hal.device.promise<@__device_6>
%4212 = torch_c.from_builtin_tensor %4211 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4213 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1849 = arith.constant 1 : index
%dim_1850 = tensor.dim %4213, %c1_1849 : tensor<4x?x4096xf16>
%4214 = flow.tensor.transfer %4213 : tensor<4x?x4096xf16>{%dim_1850} to #hal.device.promise<@__device_6>
%4215 = torch_c.from_builtin_tensor %4214 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4216 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1851 = arith.constant 1 : index
%dim_1852 = tensor.dim %4216, %c1_1851 : tensor<4x?x4096xf16>
%4217 = flow.tensor.transfer %4216 : tensor<4x?x4096xf16>{%dim_1852} to #hal.device.promise<@__device_6>
%4218 = torch_c.from_builtin_tensor %4217 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4219 = torch_c.to_builtin_tensor %4032 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1853 = arith.constant 1 : index
%dim_1854 = tensor.dim %4219, %c1_1853 : tensor<4x?x4096xf16>
%4220 = flow.tensor.transfer %4219 : tensor<4x?x4096xf16>{%dim_1854} to #hal.device.promise<@__device_6>
%4221 = torch_c.from_builtin_tensor %4220 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1855 = torch.constant.int 1
%4222 = torch.aten.add.Tensor %4203, %4206, %int1_1855 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1856 = torch.constant.int 1
%4223 = torch.aten.add.Tensor %4222, %4209, %int1_1856 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1857 = torch.constant.int 1
%4224 = torch.aten.add.Tensor %4223, %4212, %int1_1857 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1858 = torch.constant.int 1
%4225 = torch.aten.add.Tensor %4224, %4215, %int1_1858 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1859 = torch.constant.int 1
%4226 = torch.aten.add.Tensor %4225, %4218, %int1_1859 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1860 = torch.constant.int 1
%4227 = torch.aten.add.Tensor %4226, %4025, %int1_1860 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1861 = torch.constant.int 1
%4228 = torch.aten.add.Tensor %4227, %4221, %int1_1861 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4229 = torch_c.to_builtin_tensor %3983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1862 = arith.constant 1 : index
%dim_1863 = tensor.dim %4229, %c1_1862 : tensor<4x?x4096xf16>
%4230 = flow.tensor.transfer %4229 : tensor<4x?x4096xf16>{%dim_1863} to #hal.device.promise<@__device_7>
%4231 = torch_c.from_builtin_tensor %4230 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4232 = torch_c.to_builtin_tensor %3990 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1864 = arith.constant 1 : index
%dim_1865 = tensor.dim %4232, %c1_1864 : tensor<4x?x4096xf16>
%4233 = flow.tensor.transfer %4232 : tensor<4x?x4096xf16>{%dim_1865} to #hal.device.promise<@__device_7>
%4234 = torch_c.from_builtin_tensor %4233 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4235 = torch_c.to_builtin_tensor %3997 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1866 = arith.constant 1 : index
%dim_1867 = tensor.dim %4235, %c1_1866 : tensor<4x?x4096xf16>
%4236 = flow.tensor.transfer %4235 : tensor<4x?x4096xf16>{%dim_1867} to #hal.device.promise<@__device_7>
%4237 = torch_c.from_builtin_tensor %4236 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4238 = torch_c.to_builtin_tensor %4004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1868 = arith.constant 1 : index
%dim_1869 = tensor.dim %4238, %c1_1868 : tensor<4x?x4096xf16>
%4239 = flow.tensor.transfer %4238 : tensor<4x?x4096xf16>{%dim_1869} to #hal.device.promise<@__device_7>
%4240 = torch_c.from_builtin_tensor %4239 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4241 = torch_c.to_builtin_tensor %4011 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1870 = arith.constant 1 : index
%dim_1871 = tensor.dim %4241, %c1_1870 : tensor<4x?x4096xf16>
%4242 = flow.tensor.transfer %4241 : tensor<4x?x4096xf16>{%dim_1871} to #hal.device.promise<@__device_7>
%4243 = torch_c.from_builtin_tensor %4242 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4244 = torch_c.to_builtin_tensor %4018 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1872 = arith.constant 1 : index
%dim_1873 = tensor.dim %4244, %c1_1872 : tensor<4x?x4096xf16>
%4245 = flow.tensor.transfer %4244 : tensor<4x?x4096xf16>{%dim_1873} to #hal.device.promise<@__device_7>
%4246 = torch_c.from_builtin_tensor %4245 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%4247 = torch_c.to_builtin_tensor %4025 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_1874 = arith.constant 1 : index
%dim_1875 = tensor.dim %4247, %c1_1874 : tensor<4x?x4096xf16>
%4248 = flow.tensor.transfer %4247 : tensor<4x?x4096xf16>{%dim_1875} to #hal.device.promise<@__device_7>
%4249 = torch_c.from_builtin_tensor %4248 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1876 = torch.constant.int 1
%4250 = torch.aten.add.Tensor %4231, %4234, %int1_1876 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1877 = torch.constant.int 1
%4251 = torch.aten.add.Tensor %4250, %4237, %int1_1877 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1878 = torch.constant.int 1
%4252 = torch.aten.add.Tensor %4251, %4240, %int1_1878 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1879 = torch.constant.int 1
%4253 = torch.aten.add.Tensor %4252, %4243, %int1_1879 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1880 = torch.constant.int 1
%4254 = torch.aten.add.Tensor %4253, %4246, %int1_1880 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1881 = torch.constant.int 1
%4255 = torch.aten.add.Tensor %4254, %4249, %int1_1881 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1882 = torch.constant.int 1
%4256 = torch.aten.add.Tensor %4255, %4032, %int1_1882 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1883 = torch.constant.int 1
%4257 = torch.aten.add.Tensor %3737, %4060, %int1_1883 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1884 = torch.constant.int 1
%4258 = torch.aten.add.Tensor %3738, %4088, %int1_1884 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1885 = torch.constant.int 1
%4259 = torch.aten.add.Tensor %3739, %4116, %int1_1885 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1886 = torch.constant.int 1
%4260 = torch.aten.add.Tensor %3740, %4144, %int1_1886 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1887 = torch.constant.int 1
%4261 = torch.aten.add.Tensor %3741, %4172, %int1_1887 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1888 = torch.constant.int 1
%4262 = torch.aten.add.Tensor %3742, %4200, %int1_1888 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1889 = torch.constant.int 1
%4263 = torch.aten.add.Tensor %3743, %4228, %int1_1889 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4263, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1890 = torch.constant.int 1
%4264 = torch.aten.add.Tensor %3744, %4256, %int1_1890 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_1891 = torch.constant.int 6
%4265 = torch.prims.convert_element_type %4257, %int6_1891 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1892 = torch.constant.int 6
%4266 = torch.prims.convert_element_type %4258, %int6_1892 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1893 = torch.constant.int 6
%4267 = torch.prims.convert_element_type %4259, %int6_1893 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1894 = torch.constant.int 6
%4268 = torch.prims.convert_element_type %4260, %int6_1894 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1895 = torch.constant.int 6
%4269 = torch.prims.convert_element_type %4261, %int6_1895 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1896 = torch.constant.int 6
%4270 = torch.prims.convert_element_type %4262, %int6_1896 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1897 = torch.constant.int 6
%4271 = torch.prims.convert_element_type %4263, %int6_1897 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_1898 = torch.constant.int 6
%4272 = torch.prims.convert_element_type %4264, %int6_1898 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1899 = torch.constant.int 2
%4273 = torch.aten.pow.Tensor_Scalar %4265, %int2_1899 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1900 = torch.constant.int 2
%4274 = torch.aten.pow.Tensor_Scalar %4266, %int2_1900 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1901 = torch.constant.int 2
%4275 = torch.aten.pow.Tensor_Scalar %4267, %int2_1901 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1902 = torch.constant.int 2
%4276 = torch.aten.pow.Tensor_Scalar %4268, %int2_1902 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1903 = torch.constant.int 2
%4277 = torch.aten.pow.Tensor_Scalar %4269, %int2_1903 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1904 = torch.constant.int 2
%4278 = torch.aten.pow.Tensor_Scalar %4270, %int2_1904 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1905 = torch.constant.int 2
%4279 = torch.aten.pow.Tensor_Scalar %4271, %int2_1905 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_1906 = torch.constant.int 2
%4280 = torch.aten.pow.Tensor_Scalar %4272, %int2_1906 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_1907 = torch.constant.int -1
%4281 = torch.prim.ListConstruct %int-1_1907 : (!torch.int) -> !torch.list<int>
%true_1908 = torch.constant.bool true
%none_1909 = torch.constant.none
%4282 = torch.aten.mean.dim %4273, %4281, %true_1908, %none_1909 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1910 = torch.constant.int -1
%4283 = torch.prim.ListConstruct %int-1_1910 : (!torch.int) -> !torch.list<int>
%true_1911 = torch.constant.bool true
%none_1912 = torch.constant.none
%4284 = torch.aten.mean.dim %4274, %4283, %true_1911, %none_1912 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1913 = torch.constant.int -1
%4285 = torch.prim.ListConstruct %int-1_1913 : (!torch.int) -> !torch.list<int>
%true_1914 = torch.constant.bool true
%none_1915 = torch.constant.none
%4286 = torch.aten.mean.dim %4275, %4285, %true_1914, %none_1915 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1916 = torch.constant.int -1
%4287 = torch.prim.ListConstruct %int-1_1916 : (!torch.int) -> !torch.list<int>
%true_1917 = torch.constant.bool true
%none_1918 = torch.constant.none
%4288 = torch.aten.mean.dim %4276, %4287, %true_1917, %none_1918 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1919 = torch.constant.int -1
%4289 = torch.prim.ListConstruct %int-1_1919 : (!torch.int) -> !torch.list<int>
%true_1920 = torch.constant.bool true
%none_1921 = torch.constant.none
%4290 = torch.aten.mean.dim %4277, %4289, %true_1920, %none_1921 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1922 = torch.constant.int -1
%4291 = torch.prim.ListConstruct %int-1_1922 : (!torch.int) -> !torch.list<int>
%true_1923 = torch.constant.bool true
%none_1924 = torch.constant.none
%4292 = torch.aten.mean.dim %4278, %4291, %true_1923, %none_1924 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1925 = torch.constant.int -1
%4293 = torch.prim.ListConstruct %int-1_1925 : (!torch.int) -> !torch.list<int>
%true_1926 = torch.constant.bool true
%none_1927 = torch.constant.none
%4294 = torch.aten.mean.dim %4279, %4293, %true_1926, %none_1927 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_1928 = torch.constant.int -1
%4295 = torch.prim.ListConstruct %int-1_1928 : (!torch.int) -> !torch.list<int>
%true_1929 = torch.constant.bool true
%none_1930 = torch.constant.none
%4296 = torch.aten.mean.dim %4280, %4295, %true_1929, %none_1930 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1931 = torch.constant.float 9.9999997473787516E-6
%int1_1932 = torch.constant.int 1
%4297 = torch.aten.add.Scalar %4282, %float9.999990e-06_1931, %int1_1932 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1933 = torch.constant.float 9.9999997473787516E-6
%int1_1934 = torch.constant.int 1
%4298 = torch.aten.add.Scalar %4284, %float9.999990e-06_1933, %int1_1934 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1935 = torch.constant.float 9.9999997473787516E-6
%int1_1936 = torch.constant.int 1
%4299 = torch.aten.add.Scalar %4286, %float9.999990e-06_1935, %int1_1936 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1937 = torch.constant.float 9.9999997473787516E-6
%int1_1938 = torch.constant.int 1
%4300 = torch.aten.add.Scalar %4288, %float9.999990e-06_1937, %int1_1938 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1939 = torch.constant.float 9.9999997473787516E-6
%int1_1940 = torch.constant.int 1
%4301 = torch.aten.add.Scalar %4290, %float9.999990e-06_1939, %int1_1940 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1941 = torch.constant.float 9.9999997473787516E-6
%int1_1942 = torch.constant.int 1
%4302 = torch.aten.add.Scalar %4292, %float9.999990e-06_1941, %int1_1942 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1943 = torch.constant.float 9.9999997473787516E-6
%int1_1944 = torch.constant.int 1
%4303 = torch.aten.add.Scalar %4294, %float9.999990e-06_1943, %int1_1944 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_1945 = torch.constant.float 9.9999997473787516E-6
%int1_1946 = torch.constant.int 1
%4304 = torch.aten.add.Scalar %4296, %float9.999990e-06_1945, %int1_1946 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4305 = torch.aten.rsqrt %4297 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4306 = torch.aten.rsqrt %4298 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4307 = torch.aten.rsqrt %4299 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4308 = torch.aten.rsqrt %4300 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4309 = torch.aten.rsqrt %4301 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4310 = torch.aten.rsqrt %4302 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4311 = torch.aten.rsqrt %4303 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4312 = torch.aten.rsqrt %4304 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %4312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%4313 = torch.aten.mul.Tensor %4265, %4305 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4314 = torch.aten.mul.Tensor %4266, %4306 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4315 = torch.aten.mul.Tensor %4267, %4307 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4316 = torch.aten.mul.Tensor %4268, %4308 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4317 = torch.aten.mul.Tensor %4269, %4309 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4318 = torch.aten.mul.Tensor %4270, %4310 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4319 = torch.aten.mul.Tensor %4271, %4311 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4320 = torch.aten.mul.Tensor %4272, %4312 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4321 = torch.aten.mul.Tensor %80, %4313 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4322 = torch.aten.mul.Tensor %81, %4314 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4323 = torch.aten.mul.Tensor %82, %4315 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4324 = torch.aten.mul.Tensor %83, %4316 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4325 = torch.aten.mul.Tensor %84, %4317 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4326 = torch.aten.mul.Tensor %85, %4318 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4327 = torch.aten.mul.Tensor %86, %4319 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%4328 = torch.aten.mul.Tensor %87, %4320 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %4328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_1947 = torch.constant.int 5
%4329 = torch.prims.convert_element_type %4321, %int5_1947 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1948 = torch.constant.int 5
%4330 = torch.prims.convert_element_type %4322, %int5_1948 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1949 = torch.constant.int 5
%4331 = torch.prims.convert_element_type %4323, %int5_1949 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1950 = torch.constant.int 5
%4332 = torch.prims.convert_element_type %4324, %int5_1950 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1951 = torch.constant.int 5
%4333 = torch.prims.convert_element_type %4325, %int5_1951 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1952 = torch.constant.int 5
%4334 = torch.prims.convert_element_type %4326, %int5_1952 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1953 = torch.constant.int 5
%4335 = torch.prims.convert_element_type %4327, %int5_1953 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_1954 = torch.constant.int 5
%4336 = torch.prims.convert_element_type %4328, %int5_1954 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %4336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_1955 = torch.constant.int 1
%int0_1956 = torch.constant.int 0
%4337 = torch.prim.ListConstruct %int1_1955, %int0_1956 : (!torch.int, !torch.int) -> !torch.list<int>
%4338 = torch.aten.permute %88, %4337 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1957 = torch.constant.int 1
%int0_1958 = torch.constant.int 0
%4339 = torch.prim.ListConstruct %int1_1957, %int0_1958 : (!torch.int, !torch.int) -> !torch.list<int>
%4340 = torch.aten.permute %89, %4339 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1959 = torch.constant.int 1
%int0_1960 = torch.constant.int 0
%4341 = torch.prim.ListConstruct %int1_1959, %int0_1960 : (!torch.int, !torch.int) -> !torch.list<int>
%4342 = torch.aten.permute %90, %4341 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1961 = torch.constant.int 1
%int0_1962 = torch.constant.int 0
%4343 = torch.prim.ListConstruct %int1_1961, %int0_1962 : (!torch.int, !torch.int) -> !torch.list<int>
%4344 = torch.aten.permute %91, %4343 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1963 = torch.constant.int 1
%int0_1964 = torch.constant.int 0
%4345 = torch.prim.ListConstruct %int1_1963, %int0_1964 : (!torch.int, !torch.int) -> !torch.list<int>
%4346 = torch.aten.permute %92, %4345 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1965 = torch.constant.int 1
%int0_1966 = torch.constant.int 0
%4347 = torch.prim.ListConstruct %int1_1965, %int0_1966 : (!torch.int, !torch.int) -> !torch.list<int>
%4348 = torch.aten.permute %93, %4347 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1967 = torch.constant.int 1
%int0_1968 = torch.constant.int 0
%4349 = torch.prim.ListConstruct %int1_1967, %int0_1968 : (!torch.int, !torch.int) -> !torch.list<int>
%4350 = torch.aten.permute %94, %4349 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_1969 = torch.constant.int 1
%int0_1970 = torch.constant.int 0
%4351 = torch.prim.ListConstruct %int1_1969, %int0_1970 : (!torch.int, !torch.int) -> !torch.list<int>
%4352 = torch.aten.permute %95, %4351 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_1971 = torch.constant.int 4
%4353 = torch.aten.mul.int %int4_1971, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1972 = torch.constant.int 4096
%4354 = torch.prim.ListConstruct %4353, %int4096_1972 : (!torch.int, !torch.int) -> !torch.list<int>
%4355 = torch.aten.view %4329, %4354 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4355, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4356 = torch.aten.mm %4355, %4338 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4356, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1973 = torch.constant.int 4
%int512_1974 = torch.constant.int 512
%4357 = torch.prim.ListConstruct %int4_1973, %2482, %int512_1974 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4358 = torch.aten.view %4356, %4357 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1975 = torch.constant.int 4
%4359 = torch.aten.mul.int %int4_1975, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1976 = torch.constant.int 4096
%4360 = torch.prim.ListConstruct %4359, %int4096_1976 : (!torch.int, !torch.int) -> !torch.list<int>
%4361 = torch.aten.view %4330, %4360 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4361, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4362 = torch.aten.mm %4361, %4340 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4362, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1977 = torch.constant.int 4
%int512_1978 = torch.constant.int 512
%4363 = torch.prim.ListConstruct %int4_1977, %2482, %int512_1978 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4364 = torch.aten.view %4362, %4363 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1979 = torch.constant.int 4
%4365 = torch.aten.mul.int %int4_1979, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1980 = torch.constant.int 4096
%4366 = torch.prim.ListConstruct %4365, %int4096_1980 : (!torch.int, !torch.int) -> !torch.list<int>
%4367 = torch.aten.view %4331, %4366 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4367, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4368 = torch.aten.mm %4367, %4342 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4368, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1981 = torch.constant.int 4
%int512_1982 = torch.constant.int 512
%4369 = torch.prim.ListConstruct %int4_1981, %2482, %int512_1982 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4370 = torch.aten.view %4368, %4369 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1983 = torch.constant.int 4
%4371 = torch.aten.mul.int %int4_1983, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1984 = torch.constant.int 4096
%4372 = torch.prim.ListConstruct %4371, %int4096_1984 : (!torch.int, !torch.int) -> !torch.list<int>
%4373 = torch.aten.view %4332, %4372 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4373, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4374 = torch.aten.mm %4373, %4344 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4374, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1985 = torch.constant.int 4
%int512_1986 = torch.constant.int 512
%4375 = torch.prim.ListConstruct %int4_1985, %2482, %int512_1986 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4376 = torch.aten.view %4374, %4375 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1987 = torch.constant.int 4
%4377 = torch.aten.mul.int %int4_1987, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1988 = torch.constant.int 4096
%4378 = torch.prim.ListConstruct %4377, %int4096_1988 : (!torch.int, !torch.int) -> !torch.list<int>
%4379 = torch.aten.view %4333, %4378 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4379, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4380 = torch.aten.mm %4379, %4346 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4380, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1989 = torch.constant.int 4
%int512_1990 = torch.constant.int 512
%4381 = torch.prim.ListConstruct %int4_1989, %2482, %int512_1990 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4382 = torch.aten.view %4380, %4381 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1991 = torch.constant.int 4
%4383 = torch.aten.mul.int %int4_1991, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1992 = torch.constant.int 4096
%4384 = torch.prim.ListConstruct %4383, %int4096_1992 : (!torch.int, !torch.int) -> !torch.list<int>
%4385 = torch.aten.view %4334, %4384 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4385, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4386 = torch.aten.mm %4385, %4348 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4386, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1993 = torch.constant.int 4
%int512_1994 = torch.constant.int 512
%4387 = torch.prim.ListConstruct %int4_1993, %2482, %int512_1994 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4388 = torch.aten.view %4386, %4387 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1995 = torch.constant.int 4
%4389 = torch.aten.mul.int %int4_1995, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_1996 = torch.constant.int 4096
%4390 = torch.prim.ListConstruct %4389, %int4096_1996 : (!torch.int, !torch.int) -> !torch.list<int>
%4391 = torch.aten.view %4335, %4390 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4391, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4392 = torch.aten.mm %4391, %4350 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4392, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_1997 = torch.constant.int 4
%int512_1998 = torch.constant.int 512
%4393 = torch.prim.ListConstruct %int4_1997, %2482, %int512_1998 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4394 = torch.aten.view %4392, %4393 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_1999 = torch.constant.int 4
%4395 = torch.aten.mul.int %int4_1999, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2000 = torch.constant.int 4096
%4396 = torch.prim.ListConstruct %4395, %int4096_2000 : (!torch.int, !torch.int) -> !torch.list<int>
%4397 = torch.aten.view %4336, %4396 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4397, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4398 = torch.aten.mm %4397, %4352 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %4398, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_2001 = torch.constant.int 4
%int512_2002 = torch.constant.int 512
%4399 = torch.prim.ListConstruct %int4_2001, %2482, %int512_2002 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4400 = torch.aten.view %4398, %4399 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %4400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_2003 = torch.constant.int 1
%int0_2004 = torch.constant.int 0
%4401 = torch.prim.ListConstruct %int1_2003, %int0_2004 : (!torch.int, !torch.int) -> !torch.list<int>
%4402 = torch.aten.permute %96, %4401 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2005 = torch.constant.int 1
%int0_2006 = torch.constant.int 0
%4403 = torch.prim.ListConstruct %int1_2005, %int0_2006 : (!torch.int, !torch.int) -> !torch.list<int>
%4404 = torch.aten.permute %97, %4403 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2007 = torch.constant.int 1
%int0_2008 = torch.constant.int 0
%4405 = torch.prim.ListConstruct %int1_2007, %int0_2008 : (!torch.int, !torch.int) -> !torch.list<int>
%4406 = torch.aten.permute %98, %4405 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2009 = torch.constant.int 1
%int0_2010 = torch.constant.int 0
%4407 = torch.prim.ListConstruct %int1_2009, %int0_2010 : (!torch.int, !torch.int) -> !torch.list<int>
%4408 = torch.aten.permute %99, %4407 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2011 = torch.constant.int 1
%int0_2012 = torch.constant.int 0
%4409 = torch.prim.ListConstruct %int1_2011, %int0_2012 : (!torch.int, !torch.int) -> !torch.list<int>
%4410 = torch.aten.permute %100, %4409 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2013 = torch.constant.int 1
%int0_2014 = torch.constant.int 0
%4411 = torch.prim.ListConstruct %int1_2013, %int0_2014 : (!torch.int, !torch.int) -> !torch.list<int>
%4412 = torch.aten.permute %101, %4411 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2015 = torch.constant.int 1
%int0_2016 = torch.constant.int 0
%4413 = torch.prim.ListConstruct %int1_2015, %int0_2016 : (!torch.int, !torch.int) -> !torch.list<int>
%4414 = torch.aten.permute %102, %4413 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2017 = torch.constant.int 1
%int0_2018 = torch.constant.int 0
%4415 = torch.prim.ListConstruct %int1_2017, %int0_2018 : (!torch.int, !torch.int) -> !torch.list<int>
%4416 = torch.aten.permute %103, %4415 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_2019 = torch.constant.int 4
%4417 = torch.aten.mul.int %int4_2019, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2020 = torch.constant.int 4096
%4418 = torch.prim.ListConstruct %4417, %int4096_2020 : (!torch.int, !torch.int) -> !torch.list<int>
%4419 = torch.aten.view %4329, %4418 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4419, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4420 = torch.aten.mm %4419, %4402 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4420, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2021 = torch.constant.int 4
%int128_2022 = torch.constant.int 128
%4421 = torch.prim.ListConstruct %int4_2021, %2482, %int128_2022 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4422 = torch.aten.view %4420, %4421 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2023 = torch.constant.int 4
%4423 = torch.aten.mul.int %int4_2023, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2024 = torch.constant.int 4096
%4424 = torch.prim.ListConstruct %4423, %int4096_2024 : (!torch.int, !torch.int) -> !torch.list<int>
%4425 = torch.aten.view %4330, %4424 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4425, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4426 = torch.aten.mm %4425, %4404 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4426, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2025 = torch.constant.int 4
%int128_2026 = torch.constant.int 128
%4427 = torch.prim.ListConstruct %int4_2025, %2482, %int128_2026 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4428 = torch.aten.view %4426, %4427 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2027 = torch.constant.int 4
%4429 = torch.aten.mul.int %int4_2027, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2028 = torch.constant.int 4096
%4430 = torch.prim.ListConstruct %4429, %int4096_2028 : (!torch.int, !torch.int) -> !torch.list<int>
%4431 = torch.aten.view %4331, %4430 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4431, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4432 = torch.aten.mm %4431, %4406 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4432, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2029 = torch.constant.int 4
%int128_2030 = torch.constant.int 128
%4433 = torch.prim.ListConstruct %int4_2029, %2482, %int128_2030 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4434 = torch.aten.view %4432, %4433 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2031 = torch.constant.int 4
%4435 = torch.aten.mul.int %int4_2031, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2032 = torch.constant.int 4096
%4436 = torch.prim.ListConstruct %4435, %int4096_2032 : (!torch.int, !torch.int) -> !torch.list<int>
%4437 = torch.aten.view %4332, %4436 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4437, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4438 = torch.aten.mm %4437, %4408 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4438, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2033 = torch.constant.int 4
%int128_2034 = torch.constant.int 128
%4439 = torch.prim.ListConstruct %int4_2033, %2482, %int128_2034 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4440 = torch.aten.view %4438, %4439 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2035 = torch.constant.int 4
%4441 = torch.aten.mul.int %int4_2035, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2036 = torch.constant.int 4096
%4442 = torch.prim.ListConstruct %4441, %int4096_2036 : (!torch.int, !torch.int) -> !torch.list<int>
%4443 = torch.aten.view %4333, %4442 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4443, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4444 = torch.aten.mm %4443, %4410 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4444, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2037 = torch.constant.int 4
%int128_2038 = torch.constant.int 128
%4445 = torch.prim.ListConstruct %int4_2037, %2482, %int128_2038 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4446 = torch.aten.view %4444, %4445 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2039 = torch.constant.int 4
%4447 = torch.aten.mul.int %int4_2039, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2040 = torch.constant.int 4096
%4448 = torch.prim.ListConstruct %4447, %int4096_2040 : (!torch.int, !torch.int) -> !torch.list<int>
%4449 = torch.aten.view %4334, %4448 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4449, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4450 = torch.aten.mm %4449, %4412 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4450, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2041 = torch.constant.int 4
%int128_2042 = torch.constant.int 128
%4451 = torch.prim.ListConstruct %int4_2041, %2482, %int128_2042 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4452 = torch.aten.view %4450, %4451 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2043 = torch.constant.int 4
%4453 = torch.aten.mul.int %int4_2043, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2044 = torch.constant.int 4096
%4454 = torch.prim.ListConstruct %4453, %int4096_2044 : (!torch.int, !torch.int) -> !torch.list<int>
%4455 = torch.aten.view %4335, %4454 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4455, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4456 = torch.aten.mm %4455, %4414 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4456, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2045 = torch.constant.int 4
%int128_2046 = torch.constant.int 128
%4457 = torch.prim.ListConstruct %int4_2045, %2482, %int128_2046 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4458 = torch.aten.view %4456, %4457 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2047 = torch.constant.int 4
%4459 = torch.aten.mul.int %int4_2047, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2048 = torch.constant.int 4096
%4460 = torch.prim.ListConstruct %4459, %int4096_2048 : (!torch.int, !torch.int) -> !torch.list<int>
%4461 = torch.aten.view %4336, %4460 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4461, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4462 = torch.aten.mm %4461, %4416 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4462, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2049 = torch.constant.int 4
%int128_2050 = torch.constant.int 128
%4463 = torch.prim.ListConstruct %int4_2049, %2482, %int128_2050 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4464 = torch.aten.view %4462, %4463 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_2051 = torch.constant.int 1
%int0_2052 = torch.constant.int 0
%4465 = torch.prim.ListConstruct %int1_2051, %int0_2052 : (!torch.int, !torch.int) -> !torch.list<int>
%4466 = torch.aten.permute %104, %4465 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2053 = torch.constant.int 1
%int0_2054 = torch.constant.int 0
%4467 = torch.prim.ListConstruct %int1_2053, %int0_2054 : (!torch.int, !torch.int) -> !torch.list<int>
%4468 = torch.aten.permute %105, %4467 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2055 = torch.constant.int 1
%int0_2056 = torch.constant.int 0
%4469 = torch.prim.ListConstruct %int1_2055, %int0_2056 : (!torch.int, !torch.int) -> !torch.list<int>
%4470 = torch.aten.permute %106, %4469 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2057 = torch.constant.int 1
%int0_2058 = torch.constant.int 0
%4471 = torch.prim.ListConstruct %int1_2057, %int0_2058 : (!torch.int, !torch.int) -> !torch.list<int>
%4472 = torch.aten.permute %107, %4471 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2059 = torch.constant.int 1
%int0_2060 = torch.constant.int 0
%4473 = torch.prim.ListConstruct %int1_2059, %int0_2060 : (!torch.int, !torch.int) -> !torch.list<int>
%4474 = torch.aten.permute %108, %4473 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2061 = torch.constant.int 1
%int0_2062 = torch.constant.int 0
%4475 = torch.prim.ListConstruct %int1_2061, %int0_2062 : (!torch.int, !torch.int) -> !torch.list<int>
%4476 = torch.aten.permute %109, %4475 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2063 = torch.constant.int 1
%int0_2064 = torch.constant.int 0
%4477 = torch.prim.ListConstruct %int1_2063, %int0_2064 : (!torch.int, !torch.int) -> !torch.list<int>
%4478 = torch.aten.permute %110, %4477 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_2065 = torch.constant.int 1
%int0_2066 = torch.constant.int 0
%4479 = torch.prim.ListConstruct %int1_2065, %int0_2066 : (!torch.int, !torch.int) -> !torch.list<int>
%4480 = torch.aten.permute %111, %4479 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_2067 = torch.constant.int 4
%4481 = torch.aten.mul.int %int4_2067, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2068 = torch.constant.int 4096
%4482 = torch.prim.ListConstruct %4481, %int4096_2068 : (!torch.int, !torch.int) -> !torch.list<int>
%4483 = torch.aten.view %4329, %4482 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4483, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4484 = torch.aten.mm %4483, %4466 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4484, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2069 = torch.constant.int 4
%int128_2070 = torch.constant.int 128
%4485 = torch.prim.ListConstruct %int4_2069, %2482, %int128_2070 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4486 = torch.aten.view %4484, %4485 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2071 = torch.constant.int 4
%4487 = torch.aten.mul.int %int4_2071, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2072 = torch.constant.int 4096
%4488 = torch.prim.ListConstruct %4487, %int4096_2072 : (!torch.int, !torch.int) -> !torch.list<int>
%4489 = torch.aten.view %4330, %4488 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4489, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4490 = torch.aten.mm %4489, %4468 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4490, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2073 = torch.constant.int 4
%int128_2074 = torch.constant.int 128
%4491 = torch.prim.ListConstruct %int4_2073, %2482, %int128_2074 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4492 = torch.aten.view %4490, %4491 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2075 = torch.constant.int 4
%4493 = torch.aten.mul.int %int4_2075, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2076 = torch.constant.int 4096
%4494 = torch.prim.ListConstruct %4493, %int4096_2076 : (!torch.int, !torch.int) -> !torch.list<int>
%4495 = torch.aten.view %4331, %4494 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4495, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4496 = torch.aten.mm %4495, %4470 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4496, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2077 = torch.constant.int 4
%int128_2078 = torch.constant.int 128
%4497 = torch.prim.ListConstruct %int4_2077, %2482, %int128_2078 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4498 = torch.aten.view %4496, %4497 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2079 = torch.constant.int 4
%4499 = torch.aten.mul.int %int4_2079, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2080 = torch.constant.int 4096
%4500 = torch.prim.ListConstruct %4499, %int4096_2080 : (!torch.int, !torch.int) -> !torch.list<int>
%4501 = torch.aten.view %4332, %4500 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4501, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4502 = torch.aten.mm %4501, %4472 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4502, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2081 = torch.constant.int 4
%int128_2082 = torch.constant.int 128
%4503 = torch.prim.ListConstruct %int4_2081, %2482, %int128_2082 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4504 = torch.aten.view %4502, %4503 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2083 = torch.constant.int 4
%4505 = torch.aten.mul.int %int4_2083, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2084 = torch.constant.int 4096
%4506 = torch.prim.ListConstruct %4505, %int4096_2084 : (!torch.int, !torch.int) -> !torch.list<int>
%4507 = torch.aten.view %4333, %4506 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4507, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4508 = torch.aten.mm %4507, %4474 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4508, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2085 = torch.constant.int 4
%int128_2086 = torch.constant.int 128
%4509 = torch.prim.ListConstruct %int4_2085, %2482, %int128_2086 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4510 = torch.aten.view %4508, %4509 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2087 = torch.constant.int 4
%4511 = torch.aten.mul.int %int4_2087, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2088 = torch.constant.int 4096
%4512 = torch.prim.ListConstruct %4511, %int4096_2088 : (!torch.int, !torch.int) -> !torch.list<int>
%4513 = torch.aten.view %4334, %4512 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4513, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4514 = torch.aten.mm %4513, %4476 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4514, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2089 = torch.constant.int 4
%int128_2090 = torch.constant.int 128
%4515 = torch.prim.ListConstruct %int4_2089, %2482, %int128_2090 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4516 = torch.aten.view %4514, %4515 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2091 = torch.constant.int 4
%4517 = torch.aten.mul.int %int4_2091, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2092 = torch.constant.int 4096
%4518 = torch.prim.ListConstruct %4517, %int4096_2092 : (!torch.int, !torch.int) -> !torch.list<int>
%4519 = torch.aten.view %4335, %4518 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4519, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4520 = torch.aten.mm %4519, %4478 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4520, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2093 = torch.constant.int 4
%int128_2094 = torch.constant.int 128
%4521 = torch.prim.ListConstruct %int4_2093, %2482, %int128_2094 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4522 = torch.aten.view %4520, %4521 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2095 = torch.constant.int 4
%4523 = torch.aten.mul.int %int4_2095, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_2096 = torch.constant.int 4096
%4524 = torch.prim.ListConstruct %4523, %int4096_2096 : (!torch.int, !torch.int) -> !torch.list<int>
%4525 = torch.aten.view %4336, %4524 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %4525, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%4526 = torch.aten.mm %4525, %4480 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %4526, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_2097 = torch.constant.int 4
%int128_2098 = torch.constant.int 128
%4527 = torch.prim.ListConstruct %int4_2097, %2482, %int128_2098 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4528 = torch.aten.view %4526, %4527 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %4528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_2099 = torch.constant.int 4
%int4_2100 = torch.constant.int 4
%int128_2101 = torch.constant.int 128
%4529 = torch.prim.ListConstruct %int4_2099, %2482, %int4_2100, %int128_2101 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4530 = torch.aten.view %4358, %4529 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2102 = torch.constant.int 4
%int4_2103 = torch.constant.int 4
%int128_2104 = torch.constant.int 128
%4531 = torch.prim.ListConstruct %int4_2102, %2482, %int4_2103, %int128_2104 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4532 = torch.aten.view %4364, %4531 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2105 = torch.constant.int 4
%int4_2106 = torch.constant.int 4
%int128_2107 = torch.constant.int 128
%4533 = torch.prim.ListConstruct %int4_2105, %2482, %int4_2106, %int128_2107 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4534 = torch.aten.view %4370, %4533 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2108 = torch.constant.int 4
%int4_2109 = torch.constant.int 4
%int128_2110 = torch.constant.int 128
%4535 = torch.prim.ListConstruct %int4_2108, %2482, %int4_2109, %int128_2110 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4536 = torch.aten.view %4376, %4535 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2111 = torch.constant.int 4
%int4_2112 = torch.constant.int 4
%int128_2113 = torch.constant.int 128
%4537 = torch.prim.ListConstruct %int4_2111, %2482, %int4_2112, %int128_2113 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4538 = torch.aten.view %4382, %4537 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2114 = torch.constant.int 4
%int4_2115 = torch.constant.int 4
%int128_2116 = torch.constant.int 128
%4539 = torch.prim.ListConstruct %int4_2114, %2482, %int4_2115, %int128_2116 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4540 = torch.aten.view %4388, %4539 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2117 = torch.constant.int 4
%int4_2118 = torch.constant.int 4
%int128_2119 = torch.constant.int 128
%4541 = torch.prim.ListConstruct %int4_2117, %2482, %int4_2118, %int128_2119 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4542 = torch.aten.view %4394, %4541 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2120 = torch.constant.int 4
%int4_2121 = torch.constant.int 4
%int128_2122 = torch.constant.int 128
%4543 = torch.prim.ListConstruct %int4_2120, %2482, %int4_2121, %int128_2122 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4544 = torch.aten.view %4400, %4543 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2123 = torch.constant.int 4
%int1_2124 = torch.constant.int 1
%int128_2125 = torch.constant.int 128
%4545 = torch.prim.ListConstruct %int4_2123, %2482, %int1_2124, %int128_2125 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4546 = torch.aten.view %4422, %4545 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2126 = torch.constant.int 4
%int1_2127 = torch.constant.int 1
%int128_2128 = torch.constant.int 128
%4547 = torch.prim.ListConstruct %int4_2126, %2482, %int1_2127, %int128_2128 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4548 = torch.aten.view %4428, %4547 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2129 = torch.constant.int 4
%int1_2130 = torch.constant.int 1
%int128_2131 = torch.constant.int 128
%4549 = torch.prim.ListConstruct %int4_2129, %2482, %int1_2130, %int128_2131 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4550 = torch.aten.view %4434, %4549 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2132 = torch.constant.int 4
%int1_2133 = torch.constant.int 1
%int128_2134 = torch.constant.int 128
%4551 = torch.prim.ListConstruct %int4_2132, %2482, %int1_2133, %int128_2134 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4552 = torch.aten.view %4440, %4551 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2135 = torch.constant.int 4
%int1_2136 = torch.constant.int 1
%int128_2137 = torch.constant.int 128
%4553 = torch.prim.ListConstruct %int4_2135, %2482, %int1_2136, %int128_2137 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4554 = torch.aten.view %4446, %4553 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2138 = torch.constant.int 4
%int1_2139 = torch.constant.int 1
%int128_2140 = torch.constant.int 128
%4555 = torch.prim.ListConstruct %int4_2138, %2482, %int1_2139, %int128_2140 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4556 = torch.aten.view %4452, %4555 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2141 = torch.constant.int 4
%int1_2142 = torch.constant.int 1
%int128_2143 = torch.constant.int 128
%4557 = torch.prim.ListConstruct %int4_2141, %2482, %int1_2142, %int128_2143 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4558 = torch.aten.view %4458, %4557 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2144 = torch.constant.int 4
%int1_2145 = torch.constant.int 1
%int128_2146 = torch.constant.int 128
%4559 = torch.prim.ListConstruct %int4_2144, %2482, %int1_2145, %int128_2146 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4560 = torch.aten.view %4464, %4559 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2147 = torch.constant.int 4
%int1_2148 = torch.constant.int 1
%int128_2149 = torch.constant.int 128
%4561 = torch.prim.ListConstruct %int4_2147, %2482, %int1_2148, %int128_2149 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4562 = torch.aten.view %4486, %4561 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2150 = torch.constant.int 4
%int1_2151 = torch.constant.int 1
%int128_2152 = torch.constant.int 128
%4563 = torch.prim.ListConstruct %int4_2150, %2482, %int1_2151, %int128_2152 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4564 = torch.aten.view %4492, %4563 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2153 = torch.constant.int 4
%int1_2154 = torch.constant.int 1
%int128_2155 = torch.constant.int 128
%4565 = torch.prim.ListConstruct %int4_2153, %2482, %int1_2154, %int128_2155 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4566 = torch.aten.view %4498, %4565 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2156 = torch.constant.int 4
%int1_2157 = torch.constant.int 1
%int128_2158 = torch.constant.int 128
%4567 = torch.prim.ListConstruct %int4_2156, %2482, %int1_2157, %int128_2158 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4568 = torch.aten.view %4504, %4567 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2159 = torch.constant.int 4
%int1_2160 = torch.constant.int 1
%int128_2161 = torch.constant.int 128
%4569 = torch.prim.ListConstruct %int4_2159, %2482, %int1_2160, %int128_2161 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4570 = torch.aten.view %4510, %4569 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2162 = torch.constant.int 4
%int1_2163 = torch.constant.int 1
%int128_2164 = torch.constant.int 128
%4571 = torch.prim.ListConstruct %int4_2162, %2482, %int1_2163, %int128_2164 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4572 = torch.aten.view %4516, %4571 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2165 = torch.constant.int 4
%int1_2166 = torch.constant.int 1
%int128_2167 = torch.constant.int 128
%4573 = torch.prim.ListConstruct %int4_2165, %2482, %int1_2166, %int128_2167 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4574 = torch.aten.view %4522, %4573 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_2168 = torch.constant.int 4
%int1_2169 = torch.constant.int 1
%int128_2170 = torch.constant.int 128
%4575 = torch.prim.ListConstruct %int4_2168, %2482, %int1_2169, %int128_2170 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4576 = torch.aten.view %4528, %4575 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_2171 = torch.constant.int 131072
%none_2172 = torch.constant.none
%none_2173 = torch.constant.none
%cpu_2174 = torch.constant.device "cpu"
%false_2175 = torch.constant.bool false
%4577 = torch.aten.arange %int131072_2171, %none_2172, %none_2173, %cpu_2174, %false_2175 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_2176 = torch.constant.int 0
%int128_2177 = torch.constant.int 128
%int2_2178 = torch.constant.int 2
%none_2179 = torch.constant.none
%none_2180 = torch.constant.none
%cpu_2181 = torch.constant.device "cpu"
%false_2182 = torch.constant.bool false
%4578 = torch.aten.arange.start_step %int0_2176, %int128_2177, %int2_2178, %none_2179, %none_2180, %cpu_2181, %false_2182 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_2183 = torch.constant.int 0
%int0_2184 = torch.constant.int 0
%int64_2185 = torch.constant.int 64
%int1_2186 = torch.constant.int 1
%4579 = torch.aten.slice.Tensor %4578, %int0_2183, %int0_2184, %int64_2185, %int1_2186 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_2187 = torch.constant.int 6
%4580 = torch.prims.convert_element_type %4579, %int6_2187 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_2188 = torch.constant.int 128
%4581 = torch.aten.div.Scalar %4580, %int128_2188 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_2189 = torch.constant.float 5.000000e+05
%4582 = torch.aten.pow.Scalar %float5.000000e05_2189, %4581 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%4583 = torch.aten.reciprocal %4582 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_2190 = torch.constant.float 1.000000e+00
%4584 = torch.aten.mul.Scalar %4583, %float1.000000e00_2190 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_2191 = torch.constant.int 131072
%int1_2192 = torch.constant.int 1
%4585 = torch.prim.ListConstruct %int131072_2191, %int1_2192 : (!torch.int, !torch.int) -> !torch.list<int>
%4586 = torch.aten.view %4577, %4585 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%4587 = torch.aten.mul.Tensor %4586, %4584 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%4588 = torch.aten.cos %4587 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%4589 = torch.aten.sin %4587 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%4590 = torch.aten.complex %4588, %4589 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%4591 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4592 = flow.tensor.transfer %4591 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%4593 = torch_c.from_builtin_tensor %4592 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4594 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4595 = flow.tensor.transfer %4594 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%4596 = torch_c.from_builtin_tensor %4595 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4597 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4598 = flow.tensor.transfer %4597 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%4599 = torch_c.from_builtin_tensor %4598 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4600 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4601 = flow.tensor.transfer %4600 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%4602 = torch_c.from_builtin_tensor %4601 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4603 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4604 = flow.tensor.transfer %4603 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%4605 = torch_c.from_builtin_tensor %4604 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4606 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4607 = flow.tensor.transfer %4606 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%4608 = torch_c.from_builtin_tensor %4607 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4609 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4610 = flow.tensor.transfer %4609 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%4611 = torch_c.from_builtin_tensor %4610 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4612 = torch_c.to_builtin_tensor %4590 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4613 = flow.tensor.transfer %4612 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%4614 = torch_c.from_builtin_tensor %4613 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_2193 = torch.constant.int 1
%4615 = torch.aten.size.int %4358, %int1_2193 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2194 = torch.constant.int 0
%4616 = torch.aten.add.int %int0_2194, %4615 : !torch.int, !torch.int -> !torch.int
%int0_2195 = torch.constant.int 0
%int0_2196 = torch.constant.int 0
%int1_2197 = torch.constant.int 1
%4617 = torch.aten.slice.Tensor %4593, %int0_2195, %int0_2196, %4616, %int1_2197 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4617, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2198 = torch.constant.int 1
%int0_2199 = torch.constant.int 0
%int9223372036854775807_2200 = torch.constant.int 9223372036854775807
%int1_2201 = torch.constant.int 1
%4618 = torch.aten.slice.Tensor %4617, %int1_2198, %int0_2199, %int9223372036854775807_2200, %int1_2201 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4618, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2202 = torch.constant.int 0
%4619 = torch.aten.unsqueeze %4618, %int0_2202 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4619, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2203 = torch.constant.int 2
%4620 = torch.aten.unsqueeze %4619, %int2_2203 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4620, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2204 = torch.constant.int 3
%int0_2205 = torch.constant.int 0
%int9223372036854775807_2206 = torch.constant.int 9223372036854775807
%int1_2207 = torch.constant.int 1
%4621 = torch.aten.slice.Tensor %4620, %int3_2204, %int0_2205, %int9223372036854775807_2206, %int1_2207 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4621, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4622 = torch_c.to_builtin_tensor %4530 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2208 = arith.constant 1 : index
%dim_2209 = tensor.dim %4622, %c1_2208 : tensor<4x?x4x128xf16>
%4623 = flow.tensor.bitcast %4622 : tensor<4x?x4x128xf16>{%dim_2209} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2209}
%4624 = torch_c.from_builtin_tensor %4623 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4625 = torch.aten.mul.Tensor %4624, %4621 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4626 = torch_c.to_builtin_tensor %4625 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2210 = arith.constant 1 : index
%dim_2211 = tensor.dim %4626, %c1_2210 : tensor<4x?x4x64xcomplex<f32>>
%4627 = flow.tensor.bitcast %4626 : tensor<4x?x4x64xcomplex<f32>>{%dim_2211} -> tensor<4x?x4x128xf32>{%dim_2211}
%4628 = torch_c.from_builtin_tensor %4627 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2212 = torch.constant.int 5
%4629 = torch.prims.convert_element_type %4628, %int5_2212 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2213 = torch.constant.int 1
%4630 = torch.aten.size.int %4364, %int1_2213 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2214 = torch.constant.int 0
%4631 = torch.aten.add.int %int0_2214, %4630 : !torch.int, !torch.int -> !torch.int
%int0_2215 = torch.constant.int 0
%int0_2216 = torch.constant.int 0
%int1_2217 = torch.constant.int 1
%4632 = torch.aten.slice.Tensor %4596, %int0_2215, %int0_2216, %4631, %int1_2217 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4632, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2218 = torch.constant.int 1
%int0_2219 = torch.constant.int 0
%int9223372036854775807_2220 = torch.constant.int 9223372036854775807
%int1_2221 = torch.constant.int 1
%4633 = torch.aten.slice.Tensor %4632, %int1_2218, %int0_2219, %int9223372036854775807_2220, %int1_2221 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4633, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2222 = torch.constant.int 0
%4634 = torch.aten.unsqueeze %4633, %int0_2222 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4634, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2223 = torch.constant.int 2
%4635 = torch.aten.unsqueeze %4634, %int2_2223 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4635, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2224 = torch.constant.int 3
%int0_2225 = torch.constant.int 0
%int9223372036854775807_2226 = torch.constant.int 9223372036854775807
%int1_2227 = torch.constant.int 1
%4636 = torch.aten.slice.Tensor %4635, %int3_2224, %int0_2225, %int9223372036854775807_2226, %int1_2227 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4636, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4637 = torch_c.to_builtin_tensor %4532 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2228 = arith.constant 1 : index
%dim_2229 = tensor.dim %4637, %c1_2228 : tensor<4x?x4x128xf16>
%4638 = flow.tensor.bitcast %4637 : tensor<4x?x4x128xf16>{%dim_2229} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2229}
%4639 = torch_c.from_builtin_tensor %4638 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4640 = torch.aten.mul.Tensor %4639, %4636 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4641 = torch_c.to_builtin_tensor %4640 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2230 = arith.constant 1 : index
%dim_2231 = tensor.dim %4641, %c1_2230 : tensor<4x?x4x64xcomplex<f32>>
%4642 = flow.tensor.bitcast %4641 : tensor<4x?x4x64xcomplex<f32>>{%dim_2231} -> tensor<4x?x4x128xf32>{%dim_2231}
%4643 = torch_c.from_builtin_tensor %4642 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2232 = torch.constant.int 5
%4644 = torch.prims.convert_element_type %4643, %int5_2232 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2233 = torch.constant.int 1
%4645 = torch.aten.size.int %4370, %int1_2233 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2234 = torch.constant.int 0
%4646 = torch.aten.add.int %int0_2234, %4645 : !torch.int, !torch.int -> !torch.int
%int0_2235 = torch.constant.int 0
%int0_2236 = torch.constant.int 0
%int1_2237 = torch.constant.int 1
%4647 = torch.aten.slice.Tensor %4599, %int0_2235, %int0_2236, %4646, %int1_2237 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4647, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2238 = torch.constant.int 1
%int0_2239 = torch.constant.int 0
%int9223372036854775807_2240 = torch.constant.int 9223372036854775807
%int1_2241 = torch.constant.int 1
%4648 = torch.aten.slice.Tensor %4647, %int1_2238, %int0_2239, %int9223372036854775807_2240, %int1_2241 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4648, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2242 = torch.constant.int 0
%4649 = torch.aten.unsqueeze %4648, %int0_2242 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4649, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2243 = torch.constant.int 2
%4650 = torch.aten.unsqueeze %4649, %int2_2243 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4650, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2244 = torch.constant.int 3
%int0_2245 = torch.constant.int 0
%int9223372036854775807_2246 = torch.constant.int 9223372036854775807
%int1_2247 = torch.constant.int 1
%4651 = torch.aten.slice.Tensor %4650, %int3_2244, %int0_2245, %int9223372036854775807_2246, %int1_2247 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4651, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4652 = torch_c.to_builtin_tensor %4534 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2248 = arith.constant 1 : index
%dim_2249 = tensor.dim %4652, %c1_2248 : tensor<4x?x4x128xf16>
%4653 = flow.tensor.bitcast %4652 : tensor<4x?x4x128xf16>{%dim_2249} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2249}
%4654 = torch_c.from_builtin_tensor %4653 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4655 = torch.aten.mul.Tensor %4654, %4651 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4656 = torch_c.to_builtin_tensor %4655 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2250 = arith.constant 1 : index
%dim_2251 = tensor.dim %4656, %c1_2250 : tensor<4x?x4x64xcomplex<f32>>
%4657 = flow.tensor.bitcast %4656 : tensor<4x?x4x64xcomplex<f32>>{%dim_2251} -> tensor<4x?x4x128xf32>{%dim_2251}
%4658 = torch_c.from_builtin_tensor %4657 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2252 = torch.constant.int 5
%4659 = torch.prims.convert_element_type %4658, %int5_2252 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2253 = torch.constant.int 1
%4660 = torch.aten.size.int %4376, %int1_2253 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2254 = torch.constant.int 0
%4661 = torch.aten.add.int %int0_2254, %4660 : !torch.int, !torch.int -> !torch.int
%int0_2255 = torch.constant.int 0
%int0_2256 = torch.constant.int 0
%int1_2257 = torch.constant.int 1
%4662 = torch.aten.slice.Tensor %4602, %int0_2255, %int0_2256, %4661, %int1_2257 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4662, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2258 = torch.constant.int 1
%int0_2259 = torch.constant.int 0
%int9223372036854775807_2260 = torch.constant.int 9223372036854775807
%int1_2261 = torch.constant.int 1
%4663 = torch.aten.slice.Tensor %4662, %int1_2258, %int0_2259, %int9223372036854775807_2260, %int1_2261 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4663, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2262 = torch.constant.int 0
%4664 = torch.aten.unsqueeze %4663, %int0_2262 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4664, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2263 = torch.constant.int 2
%4665 = torch.aten.unsqueeze %4664, %int2_2263 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4665, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2264 = torch.constant.int 3
%int0_2265 = torch.constant.int 0
%int9223372036854775807_2266 = torch.constant.int 9223372036854775807
%int1_2267 = torch.constant.int 1
%4666 = torch.aten.slice.Tensor %4665, %int3_2264, %int0_2265, %int9223372036854775807_2266, %int1_2267 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4666, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4667 = torch_c.to_builtin_tensor %4536 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2268 = arith.constant 1 : index
%dim_2269 = tensor.dim %4667, %c1_2268 : tensor<4x?x4x128xf16>
%4668 = flow.tensor.bitcast %4667 : tensor<4x?x4x128xf16>{%dim_2269} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2269}
%4669 = torch_c.from_builtin_tensor %4668 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4670 = torch.aten.mul.Tensor %4669, %4666 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4671 = torch_c.to_builtin_tensor %4670 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2270 = arith.constant 1 : index
%dim_2271 = tensor.dim %4671, %c1_2270 : tensor<4x?x4x64xcomplex<f32>>
%4672 = flow.tensor.bitcast %4671 : tensor<4x?x4x64xcomplex<f32>>{%dim_2271} -> tensor<4x?x4x128xf32>{%dim_2271}
%4673 = torch_c.from_builtin_tensor %4672 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2272 = torch.constant.int 5
%4674 = torch.prims.convert_element_type %4673, %int5_2272 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2273 = torch.constant.int 1
%4675 = torch.aten.size.int %4382, %int1_2273 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2274 = torch.constant.int 0
%4676 = torch.aten.add.int %int0_2274, %4675 : !torch.int, !torch.int -> !torch.int
%int0_2275 = torch.constant.int 0
%int0_2276 = torch.constant.int 0
%int1_2277 = torch.constant.int 1
%4677 = torch.aten.slice.Tensor %4605, %int0_2275, %int0_2276, %4676, %int1_2277 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4677, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2278 = torch.constant.int 1
%int0_2279 = torch.constant.int 0
%int9223372036854775807_2280 = torch.constant.int 9223372036854775807
%int1_2281 = torch.constant.int 1
%4678 = torch.aten.slice.Tensor %4677, %int1_2278, %int0_2279, %int9223372036854775807_2280, %int1_2281 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4678, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2282 = torch.constant.int 0
%4679 = torch.aten.unsqueeze %4678, %int0_2282 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4679, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2283 = torch.constant.int 2
%4680 = torch.aten.unsqueeze %4679, %int2_2283 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4680, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2284 = torch.constant.int 3
%int0_2285 = torch.constant.int 0
%int9223372036854775807_2286 = torch.constant.int 9223372036854775807
%int1_2287 = torch.constant.int 1
%4681 = torch.aten.slice.Tensor %4680, %int3_2284, %int0_2285, %int9223372036854775807_2286, %int1_2287 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4681, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4682 = torch_c.to_builtin_tensor %4538 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2288 = arith.constant 1 : index
%dim_2289 = tensor.dim %4682, %c1_2288 : tensor<4x?x4x128xf16>
%4683 = flow.tensor.bitcast %4682 : tensor<4x?x4x128xf16>{%dim_2289} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2289}
%4684 = torch_c.from_builtin_tensor %4683 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4685 = torch.aten.mul.Tensor %4684, %4681 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4686 = torch_c.to_builtin_tensor %4685 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2290 = arith.constant 1 : index
%dim_2291 = tensor.dim %4686, %c1_2290 : tensor<4x?x4x64xcomplex<f32>>
%4687 = flow.tensor.bitcast %4686 : tensor<4x?x4x64xcomplex<f32>>{%dim_2291} -> tensor<4x?x4x128xf32>{%dim_2291}
%4688 = torch_c.from_builtin_tensor %4687 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2292 = torch.constant.int 5
%4689 = torch.prims.convert_element_type %4688, %int5_2292 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2293 = torch.constant.int 1
%4690 = torch.aten.size.int %4388, %int1_2293 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2294 = torch.constant.int 0
%4691 = torch.aten.add.int %int0_2294, %4690 : !torch.int, !torch.int -> !torch.int
%int0_2295 = torch.constant.int 0
%int0_2296 = torch.constant.int 0
%int1_2297 = torch.constant.int 1
%4692 = torch.aten.slice.Tensor %4608, %int0_2295, %int0_2296, %4691, %int1_2297 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4692, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2298 = torch.constant.int 1
%int0_2299 = torch.constant.int 0
%int9223372036854775807_2300 = torch.constant.int 9223372036854775807
%int1_2301 = torch.constant.int 1
%4693 = torch.aten.slice.Tensor %4692, %int1_2298, %int0_2299, %int9223372036854775807_2300, %int1_2301 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4693, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2302 = torch.constant.int 0
%4694 = torch.aten.unsqueeze %4693, %int0_2302 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4694, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2303 = torch.constant.int 2
%4695 = torch.aten.unsqueeze %4694, %int2_2303 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4695, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2304 = torch.constant.int 3
%int0_2305 = torch.constant.int 0
%int9223372036854775807_2306 = torch.constant.int 9223372036854775807
%int1_2307 = torch.constant.int 1
%4696 = torch.aten.slice.Tensor %4695, %int3_2304, %int0_2305, %int9223372036854775807_2306, %int1_2307 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4696, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4697 = torch_c.to_builtin_tensor %4540 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2308 = arith.constant 1 : index
%dim_2309 = tensor.dim %4697, %c1_2308 : tensor<4x?x4x128xf16>
%4698 = flow.tensor.bitcast %4697 : tensor<4x?x4x128xf16>{%dim_2309} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2309}
%4699 = torch_c.from_builtin_tensor %4698 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4700 = torch.aten.mul.Tensor %4699, %4696 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4701 = torch_c.to_builtin_tensor %4700 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2310 = arith.constant 1 : index
%dim_2311 = tensor.dim %4701, %c1_2310 : tensor<4x?x4x64xcomplex<f32>>
%4702 = flow.tensor.bitcast %4701 : tensor<4x?x4x64xcomplex<f32>>{%dim_2311} -> tensor<4x?x4x128xf32>{%dim_2311}
%4703 = torch_c.from_builtin_tensor %4702 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2312 = torch.constant.int 5
%4704 = torch.prims.convert_element_type %4703, %int5_2312 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2313 = torch.constant.int 1
%4705 = torch.aten.size.int %4394, %int1_2313 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2314 = torch.constant.int 0
%4706 = torch.aten.add.int %int0_2314, %4705 : !torch.int, !torch.int -> !torch.int
%int0_2315 = torch.constant.int 0
%int0_2316 = torch.constant.int 0
%int1_2317 = torch.constant.int 1
%4707 = torch.aten.slice.Tensor %4611, %int0_2315, %int0_2316, %4706, %int1_2317 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4707, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2318 = torch.constant.int 1
%int0_2319 = torch.constant.int 0
%int9223372036854775807_2320 = torch.constant.int 9223372036854775807
%int1_2321 = torch.constant.int 1
%4708 = torch.aten.slice.Tensor %4707, %int1_2318, %int0_2319, %int9223372036854775807_2320, %int1_2321 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4708, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2322 = torch.constant.int 0
%4709 = torch.aten.unsqueeze %4708, %int0_2322 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4709, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2323 = torch.constant.int 2
%4710 = torch.aten.unsqueeze %4709, %int2_2323 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4710, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2324 = torch.constant.int 3
%int0_2325 = torch.constant.int 0
%int9223372036854775807_2326 = torch.constant.int 9223372036854775807
%int1_2327 = torch.constant.int 1
%4711 = torch.aten.slice.Tensor %4710, %int3_2324, %int0_2325, %int9223372036854775807_2326, %int1_2327 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4711, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4712 = torch_c.to_builtin_tensor %4542 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2328 = arith.constant 1 : index
%dim_2329 = tensor.dim %4712, %c1_2328 : tensor<4x?x4x128xf16>
%4713 = flow.tensor.bitcast %4712 : tensor<4x?x4x128xf16>{%dim_2329} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2329}
%4714 = torch_c.from_builtin_tensor %4713 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4715 = torch.aten.mul.Tensor %4714, %4711 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4716 = torch_c.to_builtin_tensor %4715 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2330 = arith.constant 1 : index
%dim_2331 = tensor.dim %4716, %c1_2330 : tensor<4x?x4x64xcomplex<f32>>
%4717 = flow.tensor.bitcast %4716 : tensor<4x?x4x64xcomplex<f32>>{%dim_2331} -> tensor<4x?x4x128xf32>{%dim_2331}
%4718 = torch_c.from_builtin_tensor %4717 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2332 = torch.constant.int 5
%4719 = torch.prims.convert_element_type %4718, %int5_2332 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_2333 = torch.constant.int 1
%4720 = torch.aten.size.int %4400, %int1_2333 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_2334 = torch.constant.int 0
%4721 = torch.aten.add.int %int0_2334, %4720 : !torch.int, !torch.int -> !torch.int
%int0_2335 = torch.constant.int 0
%int0_2336 = torch.constant.int 0
%int1_2337 = torch.constant.int 1
%4722 = torch.aten.slice.Tensor %4614, %int0_2335, %int0_2336, %4721, %int1_2337 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4722, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2338 = torch.constant.int 1
%int0_2339 = torch.constant.int 0
%int9223372036854775807_2340 = torch.constant.int 9223372036854775807
%int1_2341 = torch.constant.int 1
%4723 = torch.aten.slice.Tensor %4722, %int1_2338, %int0_2339, %int9223372036854775807_2340, %int1_2341 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4723, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2342 = torch.constant.int 0
%4724 = torch.aten.unsqueeze %4723, %int0_2342 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4724, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2343 = torch.constant.int 2
%4725 = torch.aten.unsqueeze %4724, %int2_2343 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4725, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2344 = torch.constant.int 3
%int0_2345 = torch.constant.int 0
%int9223372036854775807_2346 = torch.constant.int 9223372036854775807
%int1_2347 = torch.constant.int 1
%4726 = torch.aten.slice.Tensor %4725, %int3_2344, %int0_2345, %int9223372036854775807_2346, %int1_2347 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4726, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4727 = torch_c.to_builtin_tensor %4544 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_2348 = arith.constant 1 : index
%dim_2349 = tensor.dim %4727, %c1_2348 : tensor<4x?x4x128xf16>
%4728 = flow.tensor.bitcast %4727 : tensor<4x?x4x128xf16>{%dim_2349} -> tensor<4x?x4x64xcomplex<f16>>{%dim_2349}
%4729 = torch_c.from_builtin_tensor %4728 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %4729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%4730 = torch.aten.mul.Tensor %4729, %4726 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %4730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%4731 = torch_c.to_builtin_tensor %4730 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_2350 = arith.constant 1 : index
%dim_2351 = tensor.dim %4731, %c1_2350 : tensor<4x?x4x64xcomplex<f32>>
%4732 = flow.tensor.bitcast %4731 : tensor<4x?x4x64xcomplex<f32>>{%dim_2351} -> tensor<4x?x4x128xf32>{%dim_2351}
%4733 = torch_c.from_builtin_tensor %4732 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %4733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_2352 = torch.constant.int 5
%4734 = torch.prims.convert_element_type %4733, %int5_2352 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %4734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_2353 = torch.constant.int 131072
%none_2354 = torch.constant.none
%none_2355 = torch.constant.none
%cpu_2356 = torch.constant.device "cpu"
%false_2357 = torch.constant.bool false
%4735 = torch.aten.arange %int131072_2353, %none_2354, %none_2355, %cpu_2356, %false_2357 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_2358 = torch.constant.int 0
%int128_2359 = torch.constant.int 128
%int2_2360 = torch.constant.int 2
%none_2361 = torch.constant.none
%none_2362 = torch.constant.none
%cpu_2363 = torch.constant.device "cpu"
%false_2364 = torch.constant.bool false
%4736 = torch.aten.arange.start_step %int0_2358, %int128_2359, %int2_2360, %none_2361, %none_2362, %cpu_2363, %false_2364 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_2365 = torch.constant.int 0
%int0_2366 = torch.constant.int 0
%int64_2367 = torch.constant.int 64
%int1_2368 = torch.constant.int 1
%4737 = torch.aten.slice.Tensor %4736, %int0_2365, %int0_2366, %int64_2367, %int1_2368 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_2369 = torch.constant.int 6
%4738 = torch.prims.convert_element_type %4737, %int6_2369 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_2370 = torch.constant.int 128
%4739 = torch.aten.div.Scalar %4738, %int128_2370 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_2371 = torch.constant.float 5.000000e+05
%4740 = torch.aten.pow.Scalar %float5.000000e05_2371, %4739 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%4741 = torch.aten.reciprocal %4740 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_2372 = torch.constant.float 1.000000e+00
%4742 = torch.aten.mul.Scalar %4741, %float1.000000e00_2372 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_2373 = torch.constant.int 131072
%int1_2374 = torch.constant.int 1
%4743 = torch.prim.ListConstruct %int131072_2373, %int1_2374 : (!torch.int, !torch.int) -> !torch.list<int>
%4744 = torch.aten.view %4735, %4743 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%4745 = torch.aten.mul.Tensor %4744, %4742 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%4746 = torch.aten.cos %4745 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%4747 = torch.aten.sin %4745 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%4748 = torch.aten.complex %4746, %4747 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%4749 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4750 = flow.tensor.transfer %4749 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%4751 = torch_c.from_builtin_tensor %4750 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4752 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4753 = flow.tensor.transfer %4752 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%4754 = torch_c.from_builtin_tensor %4753 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4755 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4756 = flow.tensor.transfer %4755 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%4757 = torch_c.from_builtin_tensor %4756 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4758 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4759 = flow.tensor.transfer %4758 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%4760 = torch_c.from_builtin_tensor %4759 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4761 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4762 = flow.tensor.transfer %4761 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%4763 = torch_c.from_builtin_tensor %4762 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4764 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4765 = flow.tensor.transfer %4764 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%4766 = torch_c.from_builtin_tensor %4765 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4767 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4768 = flow.tensor.transfer %4767 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%4769 = torch_c.from_builtin_tensor %4768 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%4770 = torch_c.to_builtin_tensor %4748 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%4771 = flow.tensor.transfer %4770 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%4772 = torch_c.from_builtin_tensor %4771 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_2375 = torch.constant.int 1
%4773 = torch.aten.size.int %4422, %int1_2375 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2376 = torch.constant.int 0
%4774 = torch.aten.add.int %int0_2376, %4773 : !torch.int, !torch.int -> !torch.int
%int0_2377 = torch.constant.int 0
%int0_2378 = torch.constant.int 0
%int1_2379 = torch.constant.int 1
%4775 = torch.aten.slice.Tensor %4751, %int0_2377, %int0_2378, %4774, %int1_2379 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4775, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2380 = torch.constant.int 1
%int0_2381 = torch.constant.int 0
%int9223372036854775807_2382 = torch.constant.int 9223372036854775807
%int1_2383 = torch.constant.int 1
%4776 = torch.aten.slice.Tensor %4775, %int1_2380, %int0_2381, %int9223372036854775807_2382, %int1_2383 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4776, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2384 = torch.constant.int 0
%4777 = torch.aten.unsqueeze %4776, %int0_2384 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4777, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2385 = torch.constant.int 2
%4778 = torch.aten.unsqueeze %4777, %int2_2385 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4778, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2386 = torch.constant.int 3
%int0_2387 = torch.constant.int 0
%int9223372036854775807_2388 = torch.constant.int 9223372036854775807
%int1_2389 = torch.constant.int 1
%4779 = torch.aten.slice.Tensor %4778, %int3_2386, %int0_2387, %int9223372036854775807_2388, %int1_2389 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4779, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4780 = torch_c.to_builtin_tensor %4546 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2390 = arith.constant 1 : index
%dim_2391 = tensor.dim %4780, %c1_2390 : tensor<4x?x1x128xf16>
%4781 = flow.tensor.bitcast %4780 : tensor<4x?x1x128xf16>{%dim_2391} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2391}
%4782 = torch_c.from_builtin_tensor %4781 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4783 = torch.aten.mul.Tensor %4782, %4779 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4784 = torch_c.to_builtin_tensor %4783 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2392 = arith.constant 1 : index
%dim_2393 = tensor.dim %4784, %c1_2392 : tensor<4x?x1x64xcomplex<f32>>
%4785 = flow.tensor.bitcast %4784 : tensor<4x?x1x64xcomplex<f32>>{%dim_2393} -> tensor<4x?x1x128xf32>{%dim_2393}
%4786 = torch_c.from_builtin_tensor %4785 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2394 = torch.constant.int 5
%4787 = torch.prims.convert_element_type %4786, %int5_2394 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2395 = torch.constant.int 1
%4788 = torch.aten.size.int %4428, %int1_2395 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2396 = torch.constant.int 0
%4789 = torch.aten.add.int %int0_2396, %4788 : !torch.int, !torch.int -> !torch.int
%int0_2397 = torch.constant.int 0
%int0_2398 = torch.constant.int 0
%int1_2399 = torch.constant.int 1
%4790 = torch.aten.slice.Tensor %4754, %int0_2397, %int0_2398, %4789, %int1_2399 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4790, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2400 = torch.constant.int 1
%int0_2401 = torch.constant.int 0
%int9223372036854775807_2402 = torch.constant.int 9223372036854775807
%int1_2403 = torch.constant.int 1
%4791 = torch.aten.slice.Tensor %4790, %int1_2400, %int0_2401, %int9223372036854775807_2402, %int1_2403 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4791, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2404 = torch.constant.int 0
%4792 = torch.aten.unsqueeze %4791, %int0_2404 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4792, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2405 = torch.constant.int 2
%4793 = torch.aten.unsqueeze %4792, %int2_2405 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4793, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2406 = torch.constant.int 3
%int0_2407 = torch.constant.int 0
%int9223372036854775807_2408 = torch.constant.int 9223372036854775807
%int1_2409 = torch.constant.int 1
%4794 = torch.aten.slice.Tensor %4793, %int3_2406, %int0_2407, %int9223372036854775807_2408, %int1_2409 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4794, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4795 = torch_c.to_builtin_tensor %4548 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2410 = arith.constant 1 : index
%dim_2411 = tensor.dim %4795, %c1_2410 : tensor<4x?x1x128xf16>
%4796 = flow.tensor.bitcast %4795 : tensor<4x?x1x128xf16>{%dim_2411} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2411}
%4797 = torch_c.from_builtin_tensor %4796 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4798 = torch.aten.mul.Tensor %4797, %4794 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4799 = torch_c.to_builtin_tensor %4798 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2412 = arith.constant 1 : index
%dim_2413 = tensor.dim %4799, %c1_2412 : tensor<4x?x1x64xcomplex<f32>>
%4800 = flow.tensor.bitcast %4799 : tensor<4x?x1x64xcomplex<f32>>{%dim_2413} -> tensor<4x?x1x128xf32>{%dim_2413}
%4801 = torch_c.from_builtin_tensor %4800 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2414 = torch.constant.int 5
%4802 = torch.prims.convert_element_type %4801, %int5_2414 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2415 = torch.constant.int 1
%4803 = torch.aten.size.int %4434, %int1_2415 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2416 = torch.constant.int 0
%4804 = torch.aten.add.int %int0_2416, %4803 : !torch.int, !torch.int -> !torch.int
%int0_2417 = torch.constant.int 0
%int0_2418 = torch.constant.int 0
%int1_2419 = torch.constant.int 1
%4805 = torch.aten.slice.Tensor %4757, %int0_2417, %int0_2418, %4804, %int1_2419 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4805, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2420 = torch.constant.int 1
%int0_2421 = torch.constant.int 0
%int9223372036854775807_2422 = torch.constant.int 9223372036854775807
%int1_2423 = torch.constant.int 1
%4806 = torch.aten.slice.Tensor %4805, %int1_2420, %int0_2421, %int9223372036854775807_2422, %int1_2423 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4806, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2424 = torch.constant.int 0
%4807 = torch.aten.unsqueeze %4806, %int0_2424 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4807, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2425 = torch.constant.int 2
%4808 = torch.aten.unsqueeze %4807, %int2_2425 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4808, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2426 = torch.constant.int 3
%int0_2427 = torch.constant.int 0
%int9223372036854775807_2428 = torch.constant.int 9223372036854775807
%int1_2429 = torch.constant.int 1
%4809 = torch.aten.slice.Tensor %4808, %int3_2426, %int0_2427, %int9223372036854775807_2428, %int1_2429 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4809, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4810 = torch_c.to_builtin_tensor %4550 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2430 = arith.constant 1 : index
%dim_2431 = tensor.dim %4810, %c1_2430 : tensor<4x?x1x128xf16>
%4811 = flow.tensor.bitcast %4810 : tensor<4x?x1x128xf16>{%dim_2431} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2431}
%4812 = torch_c.from_builtin_tensor %4811 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4813 = torch.aten.mul.Tensor %4812, %4809 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4814 = torch_c.to_builtin_tensor %4813 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2432 = arith.constant 1 : index
%dim_2433 = tensor.dim %4814, %c1_2432 : tensor<4x?x1x64xcomplex<f32>>
%4815 = flow.tensor.bitcast %4814 : tensor<4x?x1x64xcomplex<f32>>{%dim_2433} -> tensor<4x?x1x128xf32>{%dim_2433}
%4816 = torch_c.from_builtin_tensor %4815 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2434 = torch.constant.int 5
%4817 = torch.prims.convert_element_type %4816, %int5_2434 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2435 = torch.constant.int 1
%4818 = torch.aten.size.int %4440, %int1_2435 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2436 = torch.constant.int 0
%4819 = torch.aten.add.int %int0_2436, %4818 : !torch.int, !torch.int -> !torch.int
%int0_2437 = torch.constant.int 0
%int0_2438 = torch.constant.int 0
%int1_2439 = torch.constant.int 1
%4820 = torch.aten.slice.Tensor %4760, %int0_2437, %int0_2438, %4819, %int1_2439 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4820, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2440 = torch.constant.int 1
%int0_2441 = torch.constant.int 0
%int9223372036854775807_2442 = torch.constant.int 9223372036854775807
%int1_2443 = torch.constant.int 1
%4821 = torch.aten.slice.Tensor %4820, %int1_2440, %int0_2441, %int9223372036854775807_2442, %int1_2443 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4821, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2444 = torch.constant.int 0
%4822 = torch.aten.unsqueeze %4821, %int0_2444 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4822, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2445 = torch.constant.int 2
%4823 = torch.aten.unsqueeze %4822, %int2_2445 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4823, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2446 = torch.constant.int 3
%int0_2447 = torch.constant.int 0
%int9223372036854775807_2448 = torch.constant.int 9223372036854775807
%int1_2449 = torch.constant.int 1
%4824 = torch.aten.slice.Tensor %4823, %int3_2446, %int0_2447, %int9223372036854775807_2448, %int1_2449 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4824, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4825 = torch_c.to_builtin_tensor %4552 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2450 = arith.constant 1 : index
%dim_2451 = tensor.dim %4825, %c1_2450 : tensor<4x?x1x128xf16>
%4826 = flow.tensor.bitcast %4825 : tensor<4x?x1x128xf16>{%dim_2451} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2451}
%4827 = torch_c.from_builtin_tensor %4826 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4828 = torch.aten.mul.Tensor %4827, %4824 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4829 = torch_c.to_builtin_tensor %4828 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2452 = arith.constant 1 : index
%dim_2453 = tensor.dim %4829, %c1_2452 : tensor<4x?x1x64xcomplex<f32>>
%4830 = flow.tensor.bitcast %4829 : tensor<4x?x1x64xcomplex<f32>>{%dim_2453} -> tensor<4x?x1x128xf32>{%dim_2453}
%4831 = torch_c.from_builtin_tensor %4830 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2454 = torch.constant.int 5
%4832 = torch.prims.convert_element_type %4831, %int5_2454 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2455 = torch.constant.int 1
%4833 = torch.aten.size.int %4446, %int1_2455 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2456 = torch.constant.int 0
%4834 = torch.aten.add.int %int0_2456, %4833 : !torch.int, !torch.int -> !torch.int
%int0_2457 = torch.constant.int 0
%int0_2458 = torch.constant.int 0
%int1_2459 = torch.constant.int 1
%4835 = torch.aten.slice.Tensor %4763, %int0_2457, %int0_2458, %4834, %int1_2459 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4835, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2460 = torch.constant.int 1
%int0_2461 = torch.constant.int 0
%int9223372036854775807_2462 = torch.constant.int 9223372036854775807
%int1_2463 = torch.constant.int 1
%4836 = torch.aten.slice.Tensor %4835, %int1_2460, %int0_2461, %int9223372036854775807_2462, %int1_2463 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4836, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2464 = torch.constant.int 0
%4837 = torch.aten.unsqueeze %4836, %int0_2464 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4837, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2465 = torch.constant.int 2
%4838 = torch.aten.unsqueeze %4837, %int2_2465 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4838, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2466 = torch.constant.int 3
%int0_2467 = torch.constant.int 0
%int9223372036854775807_2468 = torch.constant.int 9223372036854775807
%int1_2469 = torch.constant.int 1
%4839 = torch.aten.slice.Tensor %4838, %int3_2466, %int0_2467, %int9223372036854775807_2468, %int1_2469 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4839, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4840 = torch_c.to_builtin_tensor %4554 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2470 = arith.constant 1 : index
%dim_2471 = tensor.dim %4840, %c1_2470 : tensor<4x?x1x128xf16>
%4841 = flow.tensor.bitcast %4840 : tensor<4x?x1x128xf16>{%dim_2471} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2471}
%4842 = torch_c.from_builtin_tensor %4841 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4843 = torch.aten.mul.Tensor %4842, %4839 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4844 = torch_c.to_builtin_tensor %4843 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2472 = arith.constant 1 : index
%dim_2473 = tensor.dim %4844, %c1_2472 : tensor<4x?x1x64xcomplex<f32>>
%4845 = flow.tensor.bitcast %4844 : tensor<4x?x1x64xcomplex<f32>>{%dim_2473} -> tensor<4x?x1x128xf32>{%dim_2473}
%4846 = torch_c.from_builtin_tensor %4845 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2474 = torch.constant.int 5
%4847 = torch.prims.convert_element_type %4846, %int5_2474 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2475 = torch.constant.int 1
%4848 = torch.aten.size.int %4452, %int1_2475 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2476 = torch.constant.int 0
%4849 = torch.aten.add.int %int0_2476, %4848 : !torch.int, !torch.int -> !torch.int
%int0_2477 = torch.constant.int 0
%int0_2478 = torch.constant.int 0
%int1_2479 = torch.constant.int 1
%4850 = torch.aten.slice.Tensor %4766, %int0_2477, %int0_2478, %4849, %int1_2479 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4850, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2480 = torch.constant.int 1
%int0_2481 = torch.constant.int 0
%int9223372036854775807_2482 = torch.constant.int 9223372036854775807
%int1_2483 = torch.constant.int 1
%4851 = torch.aten.slice.Tensor %4850, %int1_2480, %int0_2481, %int9223372036854775807_2482, %int1_2483 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4851, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2484 = torch.constant.int 0
%4852 = torch.aten.unsqueeze %4851, %int0_2484 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4852, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2485 = torch.constant.int 2
%4853 = torch.aten.unsqueeze %4852, %int2_2485 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4853, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2486 = torch.constant.int 3
%int0_2487 = torch.constant.int 0
%int9223372036854775807_2488 = torch.constant.int 9223372036854775807
%int1_2489 = torch.constant.int 1
%4854 = torch.aten.slice.Tensor %4853, %int3_2486, %int0_2487, %int9223372036854775807_2488, %int1_2489 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4854, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4855 = torch_c.to_builtin_tensor %4556 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2490 = arith.constant 1 : index
%dim_2491 = tensor.dim %4855, %c1_2490 : tensor<4x?x1x128xf16>
%4856 = flow.tensor.bitcast %4855 : tensor<4x?x1x128xf16>{%dim_2491} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2491}
%4857 = torch_c.from_builtin_tensor %4856 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4858 = torch.aten.mul.Tensor %4857, %4854 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4859 = torch_c.to_builtin_tensor %4858 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2492 = arith.constant 1 : index
%dim_2493 = tensor.dim %4859, %c1_2492 : tensor<4x?x1x64xcomplex<f32>>
%4860 = flow.tensor.bitcast %4859 : tensor<4x?x1x64xcomplex<f32>>{%dim_2493} -> tensor<4x?x1x128xf32>{%dim_2493}
%4861 = torch_c.from_builtin_tensor %4860 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2494 = torch.constant.int 5
%4862 = torch.prims.convert_element_type %4861, %int5_2494 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2495 = torch.constant.int 1
%4863 = torch.aten.size.int %4458, %int1_2495 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2496 = torch.constant.int 0
%4864 = torch.aten.add.int %int0_2496, %4863 : !torch.int, !torch.int -> !torch.int
%int0_2497 = torch.constant.int 0
%int0_2498 = torch.constant.int 0
%int1_2499 = torch.constant.int 1
%4865 = torch.aten.slice.Tensor %4769, %int0_2497, %int0_2498, %4864, %int1_2499 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4865, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2500 = torch.constant.int 1
%int0_2501 = torch.constant.int 0
%int9223372036854775807_2502 = torch.constant.int 9223372036854775807
%int1_2503 = torch.constant.int 1
%4866 = torch.aten.slice.Tensor %4865, %int1_2500, %int0_2501, %int9223372036854775807_2502, %int1_2503 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4866, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2504 = torch.constant.int 0
%4867 = torch.aten.unsqueeze %4866, %int0_2504 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4867, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2505 = torch.constant.int 2
%4868 = torch.aten.unsqueeze %4867, %int2_2505 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4868, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2506 = torch.constant.int 3
%int0_2507 = torch.constant.int 0
%int9223372036854775807_2508 = torch.constant.int 9223372036854775807
%int1_2509 = torch.constant.int 1
%4869 = torch.aten.slice.Tensor %4868, %int3_2506, %int0_2507, %int9223372036854775807_2508, %int1_2509 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4869, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4870 = torch_c.to_builtin_tensor %4558 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2510 = arith.constant 1 : index
%dim_2511 = tensor.dim %4870, %c1_2510 : tensor<4x?x1x128xf16>
%4871 = flow.tensor.bitcast %4870 : tensor<4x?x1x128xf16>{%dim_2511} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2511}
%4872 = torch_c.from_builtin_tensor %4871 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4873 = torch.aten.mul.Tensor %4872, %4869 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4874 = torch_c.to_builtin_tensor %4873 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2512 = arith.constant 1 : index
%dim_2513 = tensor.dim %4874, %c1_2512 : tensor<4x?x1x64xcomplex<f32>>
%4875 = flow.tensor.bitcast %4874 : tensor<4x?x1x64xcomplex<f32>>{%dim_2513} -> tensor<4x?x1x128xf32>{%dim_2513}
%4876 = torch_c.from_builtin_tensor %4875 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2514 = torch.constant.int 5
%4877 = torch.prims.convert_element_type %4876, %int5_2514 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_2515 = torch.constant.int 1
%4878 = torch.aten.size.int %4464, %int1_2515 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_2516 = torch.constant.int 0
%4879 = torch.aten.add.int %int0_2516, %4878 : !torch.int, !torch.int -> !torch.int
%int0_2517 = torch.constant.int 0
%int0_2518 = torch.constant.int 0
%int1_2519 = torch.constant.int 1
%4880 = torch.aten.slice.Tensor %4772, %int0_2517, %int0_2518, %4879, %int1_2519 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4880, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_2520 = torch.constant.int 1
%int0_2521 = torch.constant.int 0
%int9223372036854775807_2522 = torch.constant.int 9223372036854775807
%int1_2523 = torch.constant.int 1
%4881 = torch.aten.slice.Tensor %4880, %int1_2520, %int0_2521, %int9223372036854775807_2522, %int1_2523 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %4881, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_2524 = torch.constant.int 0
%4882 = torch.aten.unsqueeze %4881, %int0_2524 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %4882, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_2525 = torch.constant.int 2
%4883 = torch.aten.unsqueeze %4882, %int2_2525 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4883, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_2526 = torch.constant.int 3
%int0_2527 = torch.constant.int 0
%int9223372036854775807_2528 = torch.constant.int 9223372036854775807
%int1_2529 = torch.constant.int 1
%4884 = torch.aten.slice.Tensor %4883, %int3_2526, %int0_2527, %int9223372036854775807_2528, %int1_2529 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4884, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%4885 = torch_c.to_builtin_tensor %4560 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_2530 = arith.constant 1 : index
%dim_2531 = tensor.dim %4885, %c1_2530 : tensor<4x?x1x128xf16>
%4886 = flow.tensor.bitcast %4885 : tensor<4x?x1x128xf16>{%dim_2531} -> tensor<4x?x1x64xcomplex<f16>>{%dim_2531}
%4887 = torch_c.from_builtin_tensor %4886 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %4887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%4888 = torch.aten.mul.Tensor %4887, %4884 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %4888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%4889 = torch_c.to_builtin_tensor %4888 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_2532 = arith.constant 1 : index
%dim_2533 = tensor.dim %4889, %c1_2532 : tensor<4x?x1x64xcomplex<f32>>
%4890 = flow.tensor.bitcast %4889 : tensor<4x?x1x64xcomplex<f32>>{%dim_2533} -> tensor<4x?x1x128xf32>{%dim_2533}
%4891 = torch_c.from_builtin_tensor %4890 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %4891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_2534 = torch.constant.int 5
%4892 = torch.prims.convert_element_type %4891, %int5_2534 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %4892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_2535 = torch.constant.int 64
%4893 = torch.aten.mul.Scalar %2364, %int64_2535 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4893, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2536 = torch.constant.int 64
%4894 = torch.aten.mul.Scalar %2367, %int64_2536 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4894, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2537 = torch.constant.int 64
%4895 = torch.aten.mul.Scalar %2370, %int64_2537 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4895, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2538 = torch.constant.int 64
%4896 = torch.aten.mul.Scalar %2373, %int64_2538 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4896, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2539 = torch.constant.int 64
%4897 = torch.aten.mul.Scalar %2376, %int64_2539 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4897, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2540 = torch.constant.int 64
%4898 = torch.aten.mul.Scalar %2379, %int64_2540 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4898, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2541 = torch.constant.int 64
%4899 = torch.aten.mul.Scalar %2382, %int64_2541 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4899, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_2542 = torch.constant.int 64
%4900 = torch.aten.mul.Scalar %2385, %int64_2542 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4900, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2543 = torch.constant.int 2
%int1_2544 = torch.constant.int 1
%4901 = torch.aten.add.Scalar %4893, %int2_2543, %int1_2544 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4901, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2545 = torch.constant.int 2
%int1_2546 = torch.constant.int 1
%4902 = torch.aten.add.Scalar %4894, %int2_2545, %int1_2546 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4902, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2547 = torch.constant.int 2
%int1_2548 = torch.constant.int 1
%4903 = torch.aten.add.Scalar %4895, %int2_2547, %int1_2548 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4903, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2549 = torch.constant.int 2
%int1_2550 = torch.constant.int 1
%4904 = torch.aten.add.Scalar %4896, %int2_2549, %int1_2550 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4904, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2551 = torch.constant.int 2
%int1_2552 = torch.constant.int 1
%4905 = torch.aten.add.Scalar %4897, %int2_2551, %int1_2552 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4905, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2553 = torch.constant.int 2
%int1_2554 = torch.constant.int 1
%4906 = torch.aten.add.Scalar %4898, %int2_2553, %int1_2554 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4906, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2555 = torch.constant.int 2
%int1_2556 = torch.constant.int 1
%4907 = torch.aten.add.Scalar %4899, %int2_2555, %int1_2556 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4907, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int2_2557 = torch.constant.int 2
%int1_2558 = torch.constant.int 1
%4908 = torch.aten.add.Scalar %4900, %int2_2557, %int1_2558 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %4908, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_2559 = torch.constant.int 4
%int16_2560 = torch.constant.int 16
%int1_2561 = torch.constant.int 1
%int128_2562 = torch.constant.int 128
%4909 = torch.prim.ListConstruct %int4_2559, %3095, %int16_2560, %int1_2561, %int128_2562 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4910 = torch.aten.view %4787, %4909 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4910, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2563 = torch.constant.int 4
%int16_2564 = torch.constant.int 16
%int1_2565 = torch.constant.int 1
%int128_2566 = torch.constant.int 128
%4911 = torch.prim.ListConstruct %int4_2563, %3095, %int16_2564, %int1_2565, %int128_2566 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4912 = torch.aten.view %4802, %4911 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4912, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2567 = torch.constant.int 4
%int16_2568 = torch.constant.int 16
%int1_2569 = torch.constant.int 1
%int128_2570 = torch.constant.int 128
%4913 = torch.prim.ListConstruct %int4_2567, %3095, %int16_2568, %int1_2569, %int128_2570 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4914 = torch.aten.view %4817, %4913 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4914, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2571 = torch.constant.int 4
%int16_2572 = torch.constant.int 16
%int1_2573 = torch.constant.int 1
%int128_2574 = torch.constant.int 128
%4915 = torch.prim.ListConstruct %int4_2571, %3095, %int16_2572, %int1_2573, %int128_2574 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4916 = torch.aten.view %4832, %4915 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4916, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2575 = torch.constant.int 4
%int16_2576 = torch.constant.int 16
%int1_2577 = torch.constant.int 1
%int128_2578 = torch.constant.int 128
%4917 = torch.prim.ListConstruct %int4_2575, %3095, %int16_2576, %int1_2577, %int128_2578 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4918 = torch.aten.view %4847, %4917 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4918, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2579 = torch.constant.int 4
%int16_2580 = torch.constant.int 16
%int1_2581 = torch.constant.int 1
%int128_2582 = torch.constant.int 128
%4919 = torch.prim.ListConstruct %int4_2579, %3095, %int16_2580, %int1_2581, %int128_2582 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4920 = torch.aten.view %4862, %4919 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4920, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2583 = torch.constant.int 4
%int16_2584 = torch.constant.int 16
%int1_2585 = torch.constant.int 1
%int128_2586 = torch.constant.int 128
%4921 = torch.prim.ListConstruct %int4_2583, %3095, %int16_2584, %int1_2585, %int128_2586 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4922 = torch.aten.view %4877, %4921 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4922, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2587 = torch.constant.int 4
%int16_2588 = torch.constant.int 16
%int1_2589 = torch.constant.int 1
%int128_2590 = torch.constant.int 128
%4923 = torch.prim.ListConstruct %int4_2587, %3095, %int16_2588, %int1_2589, %int128_2590 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4924 = torch.aten.view %4892, %4923 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4924, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2591 = torch.constant.int 4
%4925 = torch.aten.mul.int %int4_2591, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2592 = torch.constant.int 16
%int1_2593 = torch.constant.int 1
%int128_2594 = torch.constant.int 128
%4926 = torch.prim.ListConstruct %4925, %int16_2592, %int1_2593, %int128_2594 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4927 = torch.aten.view %4910, %4926 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4927, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2595 = torch.constant.int 4
%4928 = torch.aten.mul.int %int4_2595, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2596 = torch.constant.int 16
%int1_2597 = torch.constant.int 1
%int128_2598 = torch.constant.int 128
%4929 = torch.prim.ListConstruct %4928, %int16_2596, %int1_2597, %int128_2598 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4930 = torch.aten.view %4912, %4929 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4930, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2599 = torch.constant.int 4
%4931 = torch.aten.mul.int %int4_2599, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2600 = torch.constant.int 16
%int1_2601 = torch.constant.int 1
%int128_2602 = torch.constant.int 128
%4932 = torch.prim.ListConstruct %4931, %int16_2600, %int1_2601, %int128_2602 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4933 = torch.aten.view %4914, %4932 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4933, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2603 = torch.constant.int 4
%4934 = torch.aten.mul.int %int4_2603, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2604 = torch.constant.int 16
%int1_2605 = torch.constant.int 1
%int128_2606 = torch.constant.int 128
%4935 = torch.prim.ListConstruct %4934, %int16_2604, %int1_2605, %int128_2606 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4936 = torch.aten.view %4916, %4935 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4936, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2607 = torch.constant.int 4
%4937 = torch.aten.mul.int %int4_2607, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2608 = torch.constant.int 16
%int1_2609 = torch.constant.int 1
%int128_2610 = torch.constant.int 128
%4938 = torch.prim.ListConstruct %4937, %int16_2608, %int1_2609, %int128_2610 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4939 = torch.aten.view %4918, %4938 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4939, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2611 = torch.constant.int 4
%4940 = torch.aten.mul.int %int4_2611, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2612 = torch.constant.int 16
%int1_2613 = torch.constant.int 1
%int128_2614 = torch.constant.int 128
%4941 = torch.prim.ListConstruct %4940, %int16_2612, %int1_2613, %int128_2614 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4942 = torch.aten.view %4920, %4941 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4942, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2615 = torch.constant.int 4
%4943 = torch.aten.mul.int %int4_2615, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2616 = torch.constant.int 16
%int1_2617 = torch.constant.int 1
%int128_2618 = torch.constant.int 128
%4944 = torch.prim.ListConstruct %4943, %int16_2616, %int1_2617, %int128_2618 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4945 = torch.aten.view %4922, %4944 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4945, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2619 = torch.constant.int 4
%4946 = torch.aten.mul.int %int4_2619, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2620 = torch.constant.int 16
%int1_2621 = torch.constant.int 1
%int128_2622 = torch.constant.int 128
%4947 = torch.prim.ListConstruct %4946, %int16_2620, %int1_2621, %int128_2622 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4948 = torch.aten.view %4924, %4947 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4948, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2623 = torch.constant.int 4
%4949 = torch.aten.mul.int %int4_2623, %3095 : !torch.int, !torch.int -> !torch.int
%4950 = torch.prim.ListConstruct %4949 : (!torch.int) -> !torch.list<int>
%4951 = torch.aten.view %4901, %4950 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4951, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2624 = torch.constant.int 4
%4952 = torch.aten.mul.int %int4_2624, %3095 : !torch.int, !torch.int -> !torch.int
%4953 = torch.prim.ListConstruct %4952 : (!torch.int) -> !torch.list<int>
%4954 = torch.aten.view %4902, %4953 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4954, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2625 = torch.constant.int 4
%4955 = torch.aten.mul.int %int4_2625, %3095 : !torch.int, !torch.int -> !torch.int
%4956 = torch.prim.ListConstruct %4955 : (!torch.int) -> !torch.list<int>
%4957 = torch.aten.view %4903, %4956 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4957, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2626 = torch.constant.int 4
%4958 = torch.aten.mul.int %int4_2626, %3095 : !torch.int, !torch.int -> !torch.int
%4959 = torch.prim.ListConstruct %4958 : (!torch.int) -> !torch.list<int>
%4960 = torch.aten.view %4904, %4959 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4960, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2627 = torch.constant.int 4
%4961 = torch.aten.mul.int %int4_2627, %3095 : !torch.int, !torch.int -> !torch.int
%4962 = torch.prim.ListConstruct %4961 : (!torch.int) -> !torch.list<int>
%4963 = torch.aten.view %4905, %4962 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4963, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2628 = torch.constant.int 4
%4964 = torch.aten.mul.int %int4_2628, %3095 : !torch.int, !torch.int -> !torch.int
%4965 = torch.prim.ListConstruct %4964 : (!torch.int) -> !torch.list<int>
%4966 = torch.aten.view %4906, %4965 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4966, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2629 = torch.constant.int 4
%4967 = torch.aten.mul.int %int4_2629, %3095 : !torch.int, !torch.int -> !torch.int
%4968 = torch.prim.ListConstruct %4967 : (!torch.int) -> !torch.list<int>
%4969 = torch.aten.view %4907, %4968 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4969, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2630 = torch.constant.int 4
%4970 = torch.aten.mul.int %int4_2630, %3095 : !torch.int, !torch.int -> !torch.int
%4971 = torch.prim.ListConstruct %4970 : (!torch.int) -> !torch.list<int>
%4972 = torch.aten.view %4908, %4971 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %4972, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2631 = torch.constant.int 4
%int16_2632 = torch.constant.int 16
%int1_2633 = torch.constant.int 1
%int128_2634 = torch.constant.int 128
%4973 = torch.prim.ListConstruct %int4_2631, %3095, %int16_2632, %int1_2633, %int128_2634 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4974 = torch.aten.view %4562, %4973 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4974, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2635 = torch.constant.int 4
%int16_2636 = torch.constant.int 16
%int1_2637 = torch.constant.int 1
%int128_2638 = torch.constant.int 128
%4975 = torch.prim.ListConstruct %int4_2635, %3095, %int16_2636, %int1_2637, %int128_2638 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4976 = torch.aten.view %4564, %4975 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4976, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2639 = torch.constant.int 4
%int16_2640 = torch.constant.int 16
%int1_2641 = torch.constant.int 1
%int128_2642 = torch.constant.int 128
%4977 = torch.prim.ListConstruct %int4_2639, %3095, %int16_2640, %int1_2641, %int128_2642 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4978 = torch.aten.view %4566, %4977 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4978, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2643 = torch.constant.int 4
%int16_2644 = torch.constant.int 16
%int1_2645 = torch.constant.int 1
%int128_2646 = torch.constant.int 128
%4979 = torch.prim.ListConstruct %int4_2643, %3095, %int16_2644, %int1_2645, %int128_2646 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4980 = torch.aten.view %4568, %4979 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4980, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2647 = torch.constant.int 4
%int16_2648 = torch.constant.int 16
%int1_2649 = torch.constant.int 1
%int128_2650 = torch.constant.int 128
%4981 = torch.prim.ListConstruct %int4_2647, %3095, %int16_2648, %int1_2649, %int128_2650 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4982 = torch.aten.view %4570, %4981 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4982, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2651 = torch.constant.int 4
%int16_2652 = torch.constant.int 16
%int1_2653 = torch.constant.int 1
%int128_2654 = torch.constant.int 128
%4983 = torch.prim.ListConstruct %int4_2651, %3095, %int16_2652, %int1_2653, %int128_2654 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4984 = torch.aten.view %4572, %4983 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4984, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2655 = torch.constant.int 4
%int16_2656 = torch.constant.int 16
%int1_2657 = torch.constant.int 1
%int128_2658 = torch.constant.int 128
%4985 = torch.prim.ListConstruct %int4_2655, %3095, %int16_2656, %int1_2657, %int128_2658 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4986 = torch.aten.view %4574, %4985 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4986, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2659 = torch.constant.int 4
%int16_2660 = torch.constant.int 16
%int1_2661 = torch.constant.int 1
%int128_2662 = torch.constant.int 128
%4987 = torch.prim.ListConstruct %int4_2659, %3095, %int16_2660, %int1_2661, %int128_2662 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4988 = torch.aten.view %4576, %4987 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %4988, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_2663 = torch.constant.int 4
%4989 = torch.aten.mul.int %int4_2663, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2664 = torch.constant.int 16
%int1_2665 = torch.constant.int 1
%int128_2666 = torch.constant.int 128
%4990 = torch.prim.ListConstruct %4989, %int16_2664, %int1_2665, %int128_2666 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4991 = torch.aten.view %4974, %4990 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4991, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2667 = torch.constant.int 4
%4992 = torch.aten.mul.int %int4_2667, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2668 = torch.constant.int 16
%int1_2669 = torch.constant.int 1
%int128_2670 = torch.constant.int 128
%4993 = torch.prim.ListConstruct %4992, %int16_2668, %int1_2669, %int128_2670 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4994 = torch.aten.view %4976, %4993 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4994, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2671 = torch.constant.int 4
%4995 = torch.aten.mul.int %int4_2671, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2672 = torch.constant.int 16
%int1_2673 = torch.constant.int 1
%int128_2674 = torch.constant.int 128
%4996 = torch.prim.ListConstruct %4995, %int16_2672, %int1_2673, %int128_2674 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%4997 = torch.aten.view %4978, %4996 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %4997, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2675 = torch.constant.int 4
%4998 = torch.aten.mul.int %int4_2675, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2676 = torch.constant.int 16
%int1_2677 = torch.constant.int 1
%int128_2678 = torch.constant.int 128
%4999 = torch.prim.ListConstruct %4998, %int16_2676, %int1_2677, %int128_2678 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5000 = torch.aten.view %4980, %4999 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5000, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2679 = torch.constant.int 4
%5001 = torch.aten.mul.int %int4_2679, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2680 = torch.constant.int 16
%int1_2681 = torch.constant.int 1
%int128_2682 = torch.constant.int 128
%5002 = torch.prim.ListConstruct %5001, %int16_2680, %int1_2681, %int128_2682 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5003 = torch.aten.view %4982, %5002 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5003, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2683 = torch.constant.int 4
%5004 = torch.aten.mul.int %int4_2683, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2684 = torch.constant.int 16
%int1_2685 = torch.constant.int 1
%int128_2686 = torch.constant.int 128
%5005 = torch.prim.ListConstruct %5004, %int16_2684, %int1_2685, %int128_2686 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5006 = torch.aten.view %4984, %5005 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5006, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2687 = torch.constant.int 4
%5007 = torch.aten.mul.int %int4_2687, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2688 = torch.constant.int 16
%int1_2689 = torch.constant.int 1
%int128_2690 = torch.constant.int 128
%5008 = torch.prim.ListConstruct %5007, %int16_2688, %int1_2689, %int128_2690 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5009 = torch.aten.view %4986, %5008 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5009, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_2691 = torch.constant.int 4
%5010 = torch.aten.mul.int %int4_2691, %3095 : !torch.int, !torch.int -> !torch.int
%int16_2692 = torch.constant.int 16
%int1_2693 = torch.constant.int 1
%int128_2694 = torch.constant.int 128
%5011 = torch.prim.ListConstruct %5010, %int16_2692, %int1_2693, %int128_2694 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5012 = torch.aten.view %4988, %5011 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5012, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_2695 = torch.constant.int 1
%int1_2696 = torch.constant.int 1
%5013 = torch.aten.add.Scalar %4901, %int1_2695, %int1_2696 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5013, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2697 = torch.constant.int 1
%int1_2698 = torch.constant.int 1
%5014 = torch.aten.add.Scalar %4902, %int1_2697, %int1_2698 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5014, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2699 = torch.constant.int 1
%int1_2700 = torch.constant.int 1
%5015 = torch.aten.add.Scalar %4903, %int1_2699, %int1_2700 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5015, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2701 = torch.constant.int 1
%int1_2702 = torch.constant.int 1
%5016 = torch.aten.add.Scalar %4904, %int1_2701, %int1_2702 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5016, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2703 = torch.constant.int 1
%int1_2704 = torch.constant.int 1
%5017 = torch.aten.add.Scalar %4905, %int1_2703, %int1_2704 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5017, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2705 = torch.constant.int 1
%int1_2706 = torch.constant.int 1
%5018 = torch.aten.add.Scalar %4906, %int1_2705, %int1_2706 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5018, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2707 = torch.constant.int 1
%int1_2708 = torch.constant.int 1
%5019 = torch.aten.add.Scalar %4907, %int1_2707, %int1_2708 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5019, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_2709 = torch.constant.int 1
%int1_2710 = torch.constant.int 1
%5020 = torch.aten.add.Scalar %4908, %int1_2709, %int1_2710 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %5020, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_2711 = torch.constant.int 4
%5021 = torch.aten.mul.int %int4_2711, %3095 : !torch.int, !torch.int -> !torch.int
%5022 = torch.prim.ListConstruct %5021 : (!torch.int) -> !torch.list<int>
%5023 = torch.aten.view %5013, %5022 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5023, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2712 = torch.constant.int 4
%5024 = torch.aten.mul.int %int4_2712, %3095 : !torch.int, !torch.int -> !torch.int
%5025 = torch.prim.ListConstruct %5024 : (!torch.int) -> !torch.list<int>
%5026 = torch.aten.view %5014, %5025 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5026, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2713 = torch.constant.int 4
%5027 = torch.aten.mul.int %int4_2713, %3095 : !torch.int, !torch.int -> !torch.int
%5028 = torch.prim.ListConstruct %5027 : (!torch.int) -> !torch.list<int>
%5029 = torch.aten.view %5015, %5028 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5029, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2714 = torch.constant.int 4
%5030 = torch.aten.mul.int %int4_2714, %3095 : !torch.int, !torch.int -> !torch.int
%5031 = torch.prim.ListConstruct %5030 : (!torch.int) -> !torch.list<int>
%5032 = torch.aten.view %5016, %5031 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5032, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2715 = torch.constant.int 4
%5033 = torch.aten.mul.int %int4_2715, %3095 : !torch.int, !torch.int -> !torch.int
%5034 = torch.prim.ListConstruct %5033 : (!torch.int) -> !torch.list<int>
%5035 = torch.aten.view %5017, %5034 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5035, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2716 = torch.constant.int 4
%5036 = torch.aten.mul.int %int4_2716, %3095 : !torch.int, !torch.int -> !torch.int
%5037 = torch.prim.ListConstruct %5036 : (!torch.int) -> !torch.list<int>
%5038 = torch.aten.view %5018, %5037 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5038, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2717 = torch.constant.int 4
%5039 = torch.aten.mul.int %int4_2717, %3095 : !torch.int, !torch.int -> !torch.int
%5040 = torch.prim.ListConstruct %5039 : (!torch.int) -> !torch.list<int>
%5041 = torch.aten.view %5019, %5040 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5041, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_2718 = torch.constant.int 4
%5042 = torch.aten.mul.int %int4_2718, %3095 : !torch.int, !torch.int -> !torch.int
%5043 = torch.prim.ListConstruct %5042 : (!torch.int) -> !torch.list<int>
%5044 = torch.aten.view %5020, %5043 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5044, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%5045 = torch.prim.ListConstruct %4951, %5023 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2719 = torch.constant.int 0
%5046 = torch.aten.cat %5045, %int0_2719 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5046, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5047 = torch.prim.ListConstruct %4954, %5026 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2720 = torch.constant.int 0
%5048 = torch.aten.cat %5047, %int0_2720 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5048, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5049 = torch.prim.ListConstruct %4957, %5029 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2721 = torch.constant.int 0
%5050 = torch.aten.cat %5049, %int0_2721 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5050, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5051 = torch.prim.ListConstruct %4960, %5032 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2722 = torch.constant.int 0
%5052 = torch.aten.cat %5051, %int0_2722 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5052, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5053 = torch.prim.ListConstruct %4963, %5035 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2723 = torch.constant.int 0
%5054 = torch.aten.cat %5053, %int0_2723 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5054, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5055 = torch.prim.ListConstruct %4966, %5038 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2724 = torch.constant.int 0
%5056 = torch.aten.cat %5055, %int0_2724 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5056, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5057 = torch.prim.ListConstruct %4969, %5041 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2725 = torch.constant.int 0
%5058 = torch.aten.cat %5057, %int0_2725 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5058, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5059 = torch.prim.ListConstruct %4972, %5044 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_2726 = torch.constant.int 0
%5060 = torch.aten.cat %5059, %int0_2726 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %5060, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%5061 = torch.prim.ListConstruct %4927, %4991 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2727 = torch.constant.int 0
%5062 = torch.aten.cat %5061, %int0_2727 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5062, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5063 = torch.prim.ListConstruct %4930, %4994 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2728 = torch.constant.int 0
%5064 = torch.aten.cat %5063, %int0_2728 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5064, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5065 = torch.prim.ListConstruct %4933, %4997 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2729 = torch.constant.int 0
%5066 = torch.aten.cat %5065, %int0_2729 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5066, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5067 = torch.prim.ListConstruct %4936, %5000 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2730 = torch.constant.int 0
%5068 = torch.aten.cat %5067, %int0_2730 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5068, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5069 = torch.prim.ListConstruct %4939, %5003 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2731 = torch.constant.int 0
%5070 = torch.aten.cat %5069, %int0_2731 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5070, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5071 = torch.prim.ListConstruct %4942, %5006 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2732 = torch.constant.int 0
%5072 = torch.aten.cat %5071, %int0_2732 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5072, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5073 = torch.prim.ListConstruct %4945, %5009 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2733 = torch.constant.int 0
%5074 = torch.aten.cat %5073, %int0_2733 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5074, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5075 = torch.prim.ListConstruct %4948, %5012 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_2734 = torch.constant.int 0
%5076 = torch.aten.cat %5075, %int0_2734 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5076, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2735 = torch.constant.int 32
%int2_2736 = torch.constant.int 2
%int16_2737 = torch.constant.int 16
%int1_2738 = torch.constant.int 1
%int128_2739 = torch.constant.int 128
%5077 = torch.prim.ListConstruct %3023, %int32_2735, %int2_2736, %int16_2737, %int1_2738, %int128_2739 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5078 = torch.aten.view %3269, %5077 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5078, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2740 = torch.constant.int 32
%5079 = torch.aten.mul.int %3023, %int32_2740 : !torch.int, !torch.int -> !torch.int
%int2_2741 = torch.constant.int 2
%5080 = torch.aten.mul.int %5079, %int2_2741 : !torch.int, !torch.int -> !torch.int
%int16_2742 = torch.constant.int 16
%int1_2743 = torch.constant.int 1
%int128_2744 = torch.constant.int 128
%5081 = torch.prim.ListConstruct %5080, %int16_2742, %int1_2743, %int128_2744 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5082 = torch.aten.view %5078, %5081 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5082, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5083 = torch.prim.ListConstruct %5046 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2745 = torch.constant.bool false
%5084 = torch.aten.index_put %5082, %5083, %5062, %false_2745 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5084, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2746 = torch.constant.int 32
%int2_2747 = torch.constant.int 2
%int16_2748 = torch.constant.int 16
%int1_2749 = torch.constant.int 1
%int128_2750 = torch.constant.int 128
%5085 = torch.prim.ListConstruct %3023, %int32_2746, %int2_2747, %int16_2748, %int1_2749, %int128_2750 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5086 = torch.aten.view %5084, %5085 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5086, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2751 = torch.constant.int 131072
%5087 = torch.prim.ListConstruct %3023, %int131072_2751 : (!torch.int, !torch.int) -> !torch.list<int>
%5088 = torch.aten.view %5086, %5087 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5088, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2752 = torch.constant.int 32
%int2_2753 = torch.constant.int 2
%int16_2754 = torch.constant.int 16
%int1_2755 = torch.constant.int 1
%int128_2756 = torch.constant.int 128
%5089 = torch.prim.ListConstruct %3026, %int32_2752, %int2_2753, %int16_2754, %int1_2755, %int128_2756 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5090 = torch.aten.view %3275, %5089 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5090, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2757 = torch.constant.int 32
%5091 = torch.aten.mul.int %3026, %int32_2757 : !torch.int, !torch.int -> !torch.int
%int2_2758 = torch.constant.int 2
%5092 = torch.aten.mul.int %5091, %int2_2758 : !torch.int, !torch.int -> !torch.int
%int16_2759 = torch.constant.int 16
%int1_2760 = torch.constant.int 1
%int128_2761 = torch.constant.int 128
%5093 = torch.prim.ListConstruct %5092, %int16_2759, %int1_2760, %int128_2761 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5094 = torch.aten.view %5090, %5093 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5094, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5095 = torch.prim.ListConstruct %5048 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2762 = torch.constant.bool false
%5096 = torch.aten.index_put %5094, %5095, %5064, %false_2762 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5096, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2763 = torch.constant.int 32
%int2_2764 = torch.constant.int 2
%int16_2765 = torch.constant.int 16
%int1_2766 = torch.constant.int 1
%int128_2767 = torch.constant.int 128
%5097 = torch.prim.ListConstruct %3026, %int32_2763, %int2_2764, %int16_2765, %int1_2766, %int128_2767 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5098 = torch.aten.view %5096, %5097 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5098, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2768 = torch.constant.int 131072
%5099 = torch.prim.ListConstruct %3026, %int131072_2768 : (!torch.int, !torch.int) -> !torch.list<int>
%5100 = torch.aten.view %5098, %5099 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5100, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2769 = torch.constant.int 32
%int2_2770 = torch.constant.int 2
%int16_2771 = torch.constant.int 16
%int1_2772 = torch.constant.int 1
%int128_2773 = torch.constant.int 128
%5101 = torch.prim.ListConstruct %3029, %int32_2769, %int2_2770, %int16_2771, %int1_2772, %int128_2773 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5102 = torch.aten.view %3281, %5101 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5102, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2774 = torch.constant.int 32
%5103 = torch.aten.mul.int %3029, %int32_2774 : !torch.int, !torch.int -> !torch.int
%int2_2775 = torch.constant.int 2
%5104 = torch.aten.mul.int %5103, %int2_2775 : !torch.int, !torch.int -> !torch.int
%int16_2776 = torch.constant.int 16
%int1_2777 = torch.constant.int 1
%int128_2778 = torch.constant.int 128
%5105 = torch.prim.ListConstruct %5104, %int16_2776, %int1_2777, %int128_2778 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5106 = torch.aten.view %5102, %5105 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5106, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5107 = torch.prim.ListConstruct %5050 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2779 = torch.constant.bool false
%5108 = torch.aten.index_put %5106, %5107, %5066, %false_2779 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5108, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2780 = torch.constant.int 32
%int2_2781 = torch.constant.int 2
%int16_2782 = torch.constant.int 16
%int1_2783 = torch.constant.int 1
%int128_2784 = torch.constant.int 128
%5109 = torch.prim.ListConstruct %3029, %int32_2780, %int2_2781, %int16_2782, %int1_2783, %int128_2784 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5110 = torch.aten.view %5108, %5109 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5110, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2785 = torch.constant.int 131072
%5111 = torch.prim.ListConstruct %3029, %int131072_2785 : (!torch.int, !torch.int) -> !torch.list<int>
%5112 = torch.aten.view %5110, %5111 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5112, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2786 = torch.constant.int 32
%int2_2787 = torch.constant.int 2
%int16_2788 = torch.constant.int 16
%int1_2789 = torch.constant.int 1
%int128_2790 = torch.constant.int 128
%5113 = torch.prim.ListConstruct %3032, %int32_2786, %int2_2787, %int16_2788, %int1_2789, %int128_2790 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5114 = torch.aten.view %3287, %5113 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5114, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2791 = torch.constant.int 32
%5115 = torch.aten.mul.int %3032, %int32_2791 : !torch.int, !torch.int -> !torch.int
%int2_2792 = torch.constant.int 2
%5116 = torch.aten.mul.int %5115, %int2_2792 : !torch.int, !torch.int -> !torch.int
%int16_2793 = torch.constant.int 16
%int1_2794 = torch.constant.int 1
%int128_2795 = torch.constant.int 128
%5117 = torch.prim.ListConstruct %5116, %int16_2793, %int1_2794, %int128_2795 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5118 = torch.aten.view %5114, %5117 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5118, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5119 = torch.prim.ListConstruct %5052 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2796 = torch.constant.bool false
%5120 = torch.aten.index_put %5118, %5119, %5068, %false_2796 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5120, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2797 = torch.constant.int 32
%int2_2798 = torch.constant.int 2
%int16_2799 = torch.constant.int 16
%int1_2800 = torch.constant.int 1
%int128_2801 = torch.constant.int 128
%5121 = torch.prim.ListConstruct %3032, %int32_2797, %int2_2798, %int16_2799, %int1_2800, %int128_2801 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5122 = torch.aten.view %5120, %5121 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5122, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2802 = torch.constant.int 131072
%5123 = torch.prim.ListConstruct %3032, %int131072_2802 : (!torch.int, !torch.int) -> !torch.list<int>
%5124 = torch.aten.view %5122, %5123 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5124, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2803 = torch.constant.int 32
%int2_2804 = torch.constant.int 2
%int16_2805 = torch.constant.int 16
%int1_2806 = torch.constant.int 1
%int128_2807 = torch.constant.int 128
%5125 = torch.prim.ListConstruct %3035, %int32_2803, %int2_2804, %int16_2805, %int1_2806, %int128_2807 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5126 = torch.aten.view %3293, %5125 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5126, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2808 = torch.constant.int 32
%5127 = torch.aten.mul.int %3035, %int32_2808 : !torch.int, !torch.int -> !torch.int
%int2_2809 = torch.constant.int 2
%5128 = torch.aten.mul.int %5127, %int2_2809 : !torch.int, !torch.int -> !torch.int
%int16_2810 = torch.constant.int 16
%int1_2811 = torch.constant.int 1
%int128_2812 = torch.constant.int 128
%5129 = torch.prim.ListConstruct %5128, %int16_2810, %int1_2811, %int128_2812 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5130 = torch.aten.view %5126, %5129 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5130, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5131 = torch.prim.ListConstruct %5054 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2813 = torch.constant.bool false
%5132 = torch.aten.index_put %5130, %5131, %5070, %false_2813 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5132, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2814 = torch.constant.int 32
%int2_2815 = torch.constant.int 2
%int16_2816 = torch.constant.int 16
%int1_2817 = torch.constant.int 1
%int128_2818 = torch.constant.int 128
%5133 = torch.prim.ListConstruct %3035, %int32_2814, %int2_2815, %int16_2816, %int1_2817, %int128_2818 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5134 = torch.aten.view %5132, %5133 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5134, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2819 = torch.constant.int 131072
%5135 = torch.prim.ListConstruct %3035, %int131072_2819 : (!torch.int, !torch.int) -> !torch.list<int>
%5136 = torch.aten.view %5134, %5135 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5136, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2820 = torch.constant.int 32
%int2_2821 = torch.constant.int 2
%int16_2822 = torch.constant.int 16
%int1_2823 = torch.constant.int 1
%int128_2824 = torch.constant.int 128
%5137 = torch.prim.ListConstruct %3038, %int32_2820, %int2_2821, %int16_2822, %int1_2823, %int128_2824 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5138 = torch.aten.view %3299, %5137 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5138, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2825 = torch.constant.int 32
%5139 = torch.aten.mul.int %3038, %int32_2825 : !torch.int, !torch.int -> !torch.int
%int2_2826 = torch.constant.int 2
%5140 = torch.aten.mul.int %5139, %int2_2826 : !torch.int, !torch.int -> !torch.int
%int16_2827 = torch.constant.int 16
%int1_2828 = torch.constant.int 1
%int128_2829 = torch.constant.int 128
%5141 = torch.prim.ListConstruct %5140, %int16_2827, %int1_2828, %int128_2829 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5142 = torch.aten.view %5138, %5141 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5142, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5143 = torch.prim.ListConstruct %5056 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2830 = torch.constant.bool false
%5144 = torch.aten.index_put %5142, %5143, %5072, %false_2830 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5144, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2831 = torch.constant.int 32
%int2_2832 = torch.constant.int 2
%int16_2833 = torch.constant.int 16
%int1_2834 = torch.constant.int 1
%int128_2835 = torch.constant.int 128
%5145 = torch.prim.ListConstruct %3038, %int32_2831, %int2_2832, %int16_2833, %int1_2834, %int128_2835 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5146 = torch.aten.view %5144, %5145 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5146, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2836 = torch.constant.int 131072
%5147 = torch.prim.ListConstruct %3038, %int131072_2836 : (!torch.int, !torch.int) -> !torch.list<int>
%5148 = torch.aten.view %5146, %5147 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5148, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2837 = torch.constant.int 32
%int2_2838 = torch.constant.int 2
%int16_2839 = torch.constant.int 16
%int1_2840 = torch.constant.int 1
%int128_2841 = torch.constant.int 128
%5149 = torch.prim.ListConstruct %3041, %int32_2837, %int2_2838, %int16_2839, %int1_2840, %int128_2841 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5150 = torch.aten.view %3305, %5149 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5150, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2842 = torch.constant.int 32
%5151 = torch.aten.mul.int %3041, %int32_2842 : !torch.int, !torch.int -> !torch.int
%int2_2843 = torch.constant.int 2
%5152 = torch.aten.mul.int %5151, %int2_2843 : !torch.int, !torch.int -> !torch.int
%int16_2844 = torch.constant.int 16
%int1_2845 = torch.constant.int 1
%int128_2846 = torch.constant.int 128
%5153 = torch.prim.ListConstruct %5152, %int16_2844, %int1_2845, %int128_2846 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5154 = torch.aten.view %5150, %5153 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5154, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5155 = torch.prim.ListConstruct %5058 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2847 = torch.constant.bool false
%5156 = torch.aten.index_put %5154, %5155, %5074, %false_2847 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5156, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2848 = torch.constant.int 32
%int2_2849 = torch.constant.int 2
%int16_2850 = torch.constant.int 16
%int1_2851 = torch.constant.int 1
%int128_2852 = torch.constant.int 128
%5157 = torch.prim.ListConstruct %3041, %int32_2848, %int2_2849, %int16_2850, %int1_2851, %int128_2852 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5158 = torch.aten.view %5156, %5157 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5158, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2853 = torch.constant.int 131072
%5159 = torch.prim.ListConstruct %3041, %int131072_2853 : (!torch.int, !torch.int) -> !torch.list<int>
%5160 = torch.aten.view %5158, %5159 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5160, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_2854 = torch.constant.int 32
%int2_2855 = torch.constant.int 2
%int16_2856 = torch.constant.int 16
%int1_2857 = torch.constant.int 1
%int128_2858 = torch.constant.int 128
%5161 = torch.prim.ListConstruct %3044, %int32_2854, %int2_2855, %int16_2856, %int1_2857, %int128_2858 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5162 = torch.aten.view %3311, %5161 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5162, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_2859 = torch.constant.int 32
%5163 = torch.aten.mul.int %3044, %int32_2859 : !torch.int, !torch.int -> !torch.int
%int2_2860 = torch.constant.int 2
%5164 = torch.aten.mul.int %5163, %int2_2860 : !torch.int, !torch.int -> !torch.int
%int16_2861 = torch.constant.int 16
%int1_2862 = torch.constant.int 1
%int128_2863 = torch.constant.int 128
%5165 = torch.prim.ListConstruct %5164, %int16_2861, %int1_2862, %int128_2863 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5166 = torch.aten.view %5162, %5165 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5166, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%5167 = torch.prim.ListConstruct %5060 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_2864 = torch.constant.bool false
%5168 = torch.aten.index_put %5166, %5167, %5076, %false_2864 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %5168, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_2865 = torch.constant.int 32
%int2_2866 = torch.constant.int 2
%int16_2867 = torch.constant.int 16
%int1_2868 = torch.constant.int 1
%int128_2869 = torch.constant.int 128
%5169 = torch.prim.ListConstruct %3044, %int32_2865, %int2_2866, %int16_2867, %int1_2868, %int128_2869 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5170 = torch.aten.view %5168, %5169 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %5170, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_2870 = torch.constant.int 131072
%5171 = torch.prim.ListConstruct %3044, %int131072_2870 : (!torch.int, !torch.int) -> !torch.list<int>
%5172 = torch.aten.view %5170, %5171 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %5172, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_2871 = torch.constant.int -2
%5173 = torch.aten.unsqueeze %4787, %int-2_2871 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2872 = torch.constant.int -2
%5174 = torch.aten.unsqueeze %4802, %int-2_2872 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2873 = torch.constant.int -2
%5175 = torch.aten.unsqueeze %4817, %int-2_2873 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2874 = torch.constant.int -2
%5176 = torch.aten.unsqueeze %4832, %int-2_2874 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2875 = torch.constant.int -2
%5177 = torch.aten.unsqueeze %4847, %int-2_2875 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2876 = torch.constant.int -2
%5178 = torch.aten.unsqueeze %4862, %int-2_2876 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2877 = torch.constant.int -2
%5179 = torch.aten.unsqueeze %4877, %int-2_2877 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2878 = torch.constant.int -2
%5180 = torch.aten.unsqueeze %4892, %int-2_2878 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_2879 = torch.constant.int 4
%int1_2880 = torch.constant.int 1
%int4_2881 = torch.constant.int 4
%int128_2882 = torch.constant.int 128
%5181 = torch.prim.ListConstruct %int4_2879, %4773, %int1_2880, %int4_2881, %int128_2882 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2883 = torch.constant.bool false
%5182 = torch.aten.expand %5173, %5181, %false_2883 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2884 = torch.constant.int 4
%int1_2885 = torch.constant.int 1
%int4_2886 = torch.constant.int 4
%int128_2887 = torch.constant.int 128
%5183 = torch.prim.ListConstruct %int4_2884, %4773, %int1_2885, %int4_2886, %int128_2887 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2888 = torch.constant.bool false
%5184 = torch.aten.expand %5174, %5183, %false_2888 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2889 = torch.constant.int 4
%int1_2890 = torch.constant.int 1
%int4_2891 = torch.constant.int 4
%int128_2892 = torch.constant.int 128
%5185 = torch.prim.ListConstruct %int4_2889, %4773, %int1_2890, %int4_2891, %int128_2892 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2893 = torch.constant.bool false
%5186 = torch.aten.expand %5175, %5185, %false_2893 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2894 = torch.constant.int 4
%int1_2895 = torch.constant.int 1
%int4_2896 = torch.constant.int 4
%int128_2897 = torch.constant.int 128
%5187 = torch.prim.ListConstruct %int4_2894, %4773, %int1_2895, %int4_2896, %int128_2897 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2898 = torch.constant.bool false
%5188 = torch.aten.expand %5176, %5187, %false_2898 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2899 = torch.constant.int 4
%int1_2900 = torch.constant.int 1
%int4_2901 = torch.constant.int 4
%int128_2902 = torch.constant.int 128
%5189 = torch.prim.ListConstruct %int4_2899, %4773, %int1_2900, %int4_2901, %int128_2902 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2903 = torch.constant.bool false
%5190 = torch.aten.expand %5177, %5189, %false_2903 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2904 = torch.constant.int 4
%int1_2905 = torch.constant.int 1
%int4_2906 = torch.constant.int 4
%int128_2907 = torch.constant.int 128
%5191 = torch.prim.ListConstruct %int4_2904, %4773, %int1_2905, %int4_2906, %int128_2907 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2908 = torch.constant.bool false
%5192 = torch.aten.expand %5178, %5191, %false_2908 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2909 = torch.constant.int 4
%int1_2910 = torch.constant.int 1
%int4_2911 = torch.constant.int 4
%int128_2912 = torch.constant.int 128
%5193 = torch.prim.ListConstruct %int4_2909, %4773, %int1_2910, %int4_2911, %int128_2912 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2913 = torch.constant.bool false
%5194 = torch.aten.expand %5179, %5193, %false_2913 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2914 = torch.constant.int 4
%int1_2915 = torch.constant.int 1
%int4_2916 = torch.constant.int 4
%int128_2917 = torch.constant.int 128
%5195 = torch.prim.ListConstruct %int4_2914, %4773, %int1_2915, %int4_2916, %int128_2917 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2918 = torch.constant.bool false
%5196 = torch.aten.expand %5180, %5195, %false_2918 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2919 = torch.constant.int 4
%int4_2920 = torch.constant.int 4
%int128_2921 = torch.constant.int 128
%5197 = torch.prim.ListConstruct %int4_2919, %4773, %int4_2920, %int128_2921 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5198 = torch.aten.view %5182, %5197 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2922 = torch.constant.int 4
%int4_2923 = torch.constant.int 4
%int128_2924 = torch.constant.int 128
%5199 = torch.prim.ListConstruct %int4_2922, %4773, %int4_2923, %int128_2924 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5200 = torch.aten.view %5184, %5199 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2925 = torch.constant.int 4
%int4_2926 = torch.constant.int 4
%int128_2927 = torch.constant.int 128
%5201 = torch.prim.ListConstruct %int4_2925, %4773, %int4_2926, %int128_2927 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5202 = torch.aten.view %5186, %5201 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2928 = torch.constant.int 4
%int4_2929 = torch.constant.int 4
%int128_2930 = torch.constant.int 128
%5203 = torch.prim.ListConstruct %int4_2928, %4773, %int4_2929, %int128_2930 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5204 = torch.aten.view %5188, %5203 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2931 = torch.constant.int 4
%int4_2932 = torch.constant.int 4
%int128_2933 = torch.constant.int 128
%5205 = torch.prim.ListConstruct %int4_2931, %4773, %int4_2932, %int128_2933 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5206 = torch.aten.view %5190, %5205 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2934 = torch.constant.int 4
%int4_2935 = torch.constant.int 4
%int128_2936 = torch.constant.int 128
%5207 = torch.prim.ListConstruct %int4_2934, %4773, %int4_2935, %int128_2936 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5208 = torch.aten.view %5192, %5207 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2937 = torch.constant.int 4
%int4_2938 = torch.constant.int 4
%int128_2939 = torch.constant.int 128
%5209 = torch.prim.ListConstruct %int4_2937, %4773, %int4_2938, %int128_2939 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5210 = torch.aten.view %5194, %5209 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2940 = torch.constant.int 4
%int4_2941 = torch.constant.int 4
%int128_2942 = torch.constant.int 128
%5211 = torch.prim.ListConstruct %int4_2940, %4773, %int4_2941, %int128_2942 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5212 = torch.aten.view %5196, %5211 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_2943 = torch.constant.int -2
%5213 = torch.aten.unsqueeze %4562, %int-2_2943 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2944 = torch.constant.int -2
%5214 = torch.aten.unsqueeze %4564, %int-2_2944 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2945 = torch.constant.int -2
%5215 = torch.aten.unsqueeze %4566, %int-2_2945 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2946 = torch.constant.int -2
%5216 = torch.aten.unsqueeze %4568, %int-2_2946 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2947 = torch.constant.int -2
%5217 = torch.aten.unsqueeze %4570, %int-2_2947 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2948 = torch.constant.int -2
%5218 = torch.aten.unsqueeze %4572, %int-2_2948 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2949 = torch.constant.int -2
%5219 = torch.aten.unsqueeze %4574, %int-2_2949 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_2950 = torch.constant.int -2
%5220 = torch.aten.unsqueeze %4576, %int-2_2950 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %5220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_2951 = torch.constant.int 1
%5221 = torch.aten.size.int %4486, %int1_2951 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_2952 = torch.constant.int 4
%int1_2953 = torch.constant.int 1
%int4_2954 = torch.constant.int 4
%int128_2955 = torch.constant.int 128
%5222 = torch.prim.ListConstruct %int4_2952, %5221, %int1_2953, %int4_2954, %int128_2955 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2956 = torch.constant.bool false
%5223 = torch.aten.expand %5213, %5222, %false_2956 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2957 = torch.constant.int 4
%int1_2958 = torch.constant.int 1
%int4_2959 = torch.constant.int 4
%int128_2960 = torch.constant.int 128
%5224 = torch.prim.ListConstruct %int4_2957, %5221, %int1_2958, %int4_2959, %int128_2960 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2961 = torch.constant.bool false
%5225 = torch.aten.expand %5214, %5224, %false_2961 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2962 = torch.constant.int 4
%int1_2963 = torch.constant.int 1
%int4_2964 = torch.constant.int 4
%int128_2965 = torch.constant.int 128
%5226 = torch.prim.ListConstruct %int4_2962, %5221, %int1_2963, %int4_2964, %int128_2965 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2966 = torch.constant.bool false
%5227 = torch.aten.expand %5215, %5226, %false_2966 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2967 = torch.constant.int 4
%int1_2968 = torch.constant.int 1
%int4_2969 = torch.constant.int 4
%int128_2970 = torch.constant.int 128
%5228 = torch.prim.ListConstruct %int4_2967, %5221, %int1_2968, %int4_2969, %int128_2970 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2971 = torch.constant.bool false
%5229 = torch.aten.expand %5216, %5228, %false_2971 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2972 = torch.constant.int 4
%int1_2973 = torch.constant.int 1
%int4_2974 = torch.constant.int 4
%int128_2975 = torch.constant.int 128
%5230 = torch.prim.ListConstruct %int4_2972, %5221, %int1_2973, %int4_2974, %int128_2975 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2976 = torch.constant.bool false
%5231 = torch.aten.expand %5217, %5230, %false_2976 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2977 = torch.constant.int 4
%int1_2978 = torch.constant.int 1
%int4_2979 = torch.constant.int 4
%int128_2980 = torch.constant.int 128
%5232 = torch.prim.ListConstruct %int4_2977, %5221, %int1_2978, %int4_2979, %int128_2980 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2981 = torch.constant.bool false
%5233 = torch.aten.expand %5218, %5232, %false_2981 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2982 = torch.constant.int 4
%int1_2983 = torch.constant.int 1
%int4_2984 = torch.constant.int 4
%int128_2985 = torch.constant.int 128
%5234 = torch.prim.ListConstruct %int4_2982, %5221, %int1_2983, %int4_2984, %int128_2985 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2986 = torch.constant.bool false
%5235 = torch.aten.expand %5219, %5234, %false_2986 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2987 = torch.constant.int 4
%int1_2988 = torch.constant.int 1
%int4_2989 = torch.constant.int 4
%int128_2990 = torch.constant.int 128
%5236 = torch.prim.ListConstruct %int4_2987, %5221, %int1_2988, %int4_2989, %int128_2990 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_2991 = torch.constant.bool false
%5237 = torch.aten.expand %5220, %5236, %false_2991 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %5237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_2992 = torch.constant.int 4
%int4_2993 = torch.constant.int 4
%int128_2994 = torch.constant.int 128
%5238 = torch.prim.ListConstruct %int4_2992, %5221, %int4_2993, %int128_2994 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5239 = torch.aten.view %5223, %5238 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2995 = torch.constant.int 4
%int4_2996 = torch.constant.int 4
%int128_2997 = torch.constant.int 128
%5240 = torch.prim.ListConstruct %int4_2995, %5221, %int4_2996, %int128_2997 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5241 = torch.aten.view %5225, %5240 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_2998 = torch.constant.int 4
%int4_2999 = torch.constant.int 4
%int128_3000 = torch.constant.int 128
%5242 = torch.prim.ListConstruct %int4_2998, %5221, %int4_2999, %int128_3000 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5243 = torch.aten.view %5227, %5242 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3001 = torch.constant.int 4
%int4_3002 = torch.constant.int 4
%int128_3003 = torch.constant.int 128
%5244 = torch.prim.ListConstruct %int4_3001, %5221, %int4_3002, %int128_3003 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5245 = torch.aten.view %5229, %5244 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3004 = torch.constant.int 4
%int4_3005 = torch.constant.int 4
%int128_3006 = torch.constant.int 128
%5246 = torch.prim.ListConstruct %int4_3004, %5221, %int4_3005, %int128_3006 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5247 = torch.aten.view %5231, %5246 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3007 = torch.constant.int 4
%int4_3008 = torch.constant.int 4
%int128_3009 = torch.constant.int 128
%5248 = torch.prim.ListConstruct %int4_3007, %5221, %int4_3008, %int128_3009 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5249 = torch.aten.view %5233, %5248 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3010 = torch.constant.int 4
%int4_3011 = torch.constant.int 4
%int128_3012 = torch.constant.int 128
%5250 = torch.prim.ListConstruct %int4_3010, %5221, %int4_3011, %int128_3012 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5251 = torch.aten.view %5235, %5250 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3013 = torch.constant.int 4
%int4_3014 = torch.constant.int 4
%int128_3015 = torch.constant.int 128
%5252 = torch.prim.ListConstruct %int4_3013, %5221, %int4_3014, %int128_3015 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5253 = torch.aten.view %5237, %5252 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3016 = torch.constant.int 1
%int2_3017 = torch.constant.int 2
%5254 = torch.aten.transpose.int %4629, %int1_3016, %int2_3017 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5254, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3018 = torch.constant.int 1
%int2_3019 = torch.constant.int 2
%5255 = torch.aten.transpose.int %4644, %int1_3018, %int2_3019 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5255, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3020 = torch.constant.int 1
%int2_3021 = torch.constant.int 2
%5256 = torch.aten.transpose.int %4659, %int1_3020, %int2_3021 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5256, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3022 = torch.constant.int 1
%int2_3023 = torch.constant.int 2
%5257 = torch.aten.transpose.int %4674, %int1_3022, %int2_3023 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5257, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3024 = torch.constant.int 1
%int2_3025 = torch.constant.int 2
%5258 = torch.aten.transpose.int %4689, %int1_3024, %int2_3025 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5258, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3026 = torch.constant.int 1
%int2_3027 = torch.constant.int 2
%5259 = torch.aten.transpose.int %4704, %int1_3026, %int2_3027 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5259, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3028 = torch.constant.int 1
%int2_3029 = torch.constant.int 2
%5260 = torch.aten.transpose.int %4719, %int1_3028, %int2_3029 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5260, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3030 = torch.constant.int 1
%int2_3031 = torch.constant.int 2
%5261 = torch.aten.transpose.int %4734, %int1_3030, %int2_3031 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5261, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3032 = torch.constant.int 1
%int2_3033 = torch.constant.int 2
%5262 = torch.aten.transpose.int %5198, %int1_3032, %int2_3033 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5262, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3034 = torch.constant.int 1
%int2_3035 = torch.constant.int 2
%5263 = torch.aten.transpose.int %5200, %int1_3034, %int2_3035 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5263, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3036 = torch.constant.int 1
%int2_3037 = torch.constant.int 2
%5264 = torch.aten.transpose.int %5202, %int1_3036, %int2_3037 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5264, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3038 = torch.constant.int 1
%int2_3039 = torch.constant.int 2
%5265 = torch.aten.transpose.int %5204, %int1_3038, %int2_3039 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5265, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3040 = torch.constant.int 1
%int2_3041 = torch.constant.int 2
%5266 = torch.aten.transpose.int %5206, %int1_3040, %int2_3041 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5266, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3042 = torch.constant.int 1
%int2_3043 = torch.constant.int 2
%5267 = torch.aten.transpose.int %5208, %int1_3042, %int2_3043 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5267, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3044 = torch.constant.int 1
%int2_3045 = torch.constant.int 2
%5268 = torch.aten.transpose.int %5210, %int1_3044, %int2_3045 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5268, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3046 = torch.constant.int 1
%int2_3047 = torch.constant.int 2
%5269 = torch.aten.transpose.int %5212, %int1_3046, %int2_3047 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5269, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3048 = torch.constant.int 1
%int2_3049 = torch.constant.int 2
%5270 = torch.aten.transpose.int %5239, %int1_3048, %int2_3049 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5270, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3050 = torch.constant.int 1
%int2_3051 = torch.constant.int 2
%5271 = torch.aten.transpose.int %5241, %int1_3050, %int2_3051 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5271, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3052 = torch.constant.int 1
%int2_3053 = torch.constant.int 2
%5272 = torch.aten.transpose.int %5243, %int1_3052, %int2_3053 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5272, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3054 = torch.constant.int 1
%int2_3055 = torch.constant.int 2
%5273 = torch.aten.transpose.int %5245, %int1_3054, %int2_3055 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5273, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3056 = torch.constant.int 1
%int2_3057 = torch.constant.int 2
%5274 = torch.aten.transpose.int %5247, %int1_3056, %int2_3057 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5274, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3058 = torch.constant.int 1
%int2_3059 = torch.constant.int 2
%5275 = torch.aten.transpose.int %5249, %int1_3058, %int2_3059 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5275, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3060 = torch.constant.int 1
%int2_3061 = torch.constant.int 2
%5276 = torch.aten.transpose.int %5251, %int1_3060, %int2_3061 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5276, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3062 = torch.constant.int 1
%int2_3063 = torch.constant.int 2
%5277 = torch.aten.transpose.int %5253, %int1_3062, %int2_3063 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %5277, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3064 = torch.constant.float 0.000000e+00
%true_3065 = torch.constant.bool true
%none_3066 = torch.constant.none
%none_3067 = torch.constant.none
%5278:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5254, %5262, %5270, %float0.000000e00_3064, %true_3065, %none_3066, %none_3067) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5278#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3068 = torch.constant.float 0.000000e+00
%true_3069 = torch.constant.bool true
%none_3070 = torch.constant.none
%none_3071 = torch.constant.none
%5279:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5255, %5263, %5271, %float0.000000e00_3068, %true_3069, %none_3070, %none_3071) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5279#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3072 = torch.constant.float 0.000000e+00
%true_3073 = torch.constant.bool true
%none_3074 = torch.constant.none
%none_3075 = torch.constant.none
%5280:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5256, %5264, %5272, %float0.000000e00_3072, %true_3073, %none_3074, %none_3075) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5280#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3076 = torch.constant.float 0.000000e+00
%true_3077 = torch.constant.bool true
%none_3078 = torch.constant.none
%none_3079 = torch.constant.none
%5281:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5257, %5265, %5273, %float0.000000e00_3076, %true_3077, %none_3078, %none_3079) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5281#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3080 = torch.constant.float 0.000000e+00
%true_3081 = torch.constant.bool true
%none_3082 = torch.constant.none
%none_3083 = torch.constant.none
%5282:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5258, %5266, %5274, %float0.000000e00_3080, %true_3081, %none_3082, %none_3083) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5282#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3084 = torch.constant.float 0.000000e+00
%true_3085 = torch.constant.bool true
%none_3086 = torch.constant.none
%none_3087 = torch.constant.none
%5283:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5259, %5267, %5275, %float0.000000e00_3084, %true_3085, %none_3086, %none_3087) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5283#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3088 = torch.constant.float 0.000000e+00
%true_3089 = torch.constant.bool true
%none_3090 = torch.constant.none
%none_3091 = torch.constant.none
%5284:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5260, %5268, %5276, %float0.000000e00_3088, %true_3089, %none_3090, %none_3091) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5284#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_3092 = torch.constant.float 0.000000e+00
%true_3093 = torch.constant.bool true
%none_3094 = torch.constant.none
%none_3095 = torch.constant.none
%5285:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%5261, %5269, %5277, %float0.000000e00_3092, %true_3093, %none_3094, %none_3095) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %5285#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_3096 = torch.constant.int 1
%int2_3097 = torch.constant.int 2
%5286 = torch.aten.transpose.int %5278#0, %int1_3096, %int2_3097 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3098 = torch.constant.int 1
%int2_3099 = torch.constant.int 2
%5287 = torch.aten.transpose.int %5279#0, %int1_3098, %int2_3099 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3100 = torch.constant.int 1
%int2_3101 = torch.constant.int 2
%5288 = torch.aten.transpose.int %5280#0, %int1_3100, %int2_3101 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3102 = torch.constant.int 1
%int2_3103 = torch.constant.int 2
%5289 = torch.aten.transpose.int %5281#0, %int1_3102, %int2_3103 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3104 = torch.constant.int 1
%int2_3105 = torch.constant.int 2
%5290 = torch.aten.transpose.int %5282#0, %int1_3104, %int2_3105 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3106 = torch.constant.int 1
%int2_3107 = torch.constant.int 2
%5291 = torch.aten.transpose.int %5283#0, %int1_3106, %int2_3107 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3108 = torch.constant.int 1
%int2_3109 = torch.constant.int 2
%5292 = torch.aten.transpose.int %5284#0, %int1_3108, %int2_3109 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_3110 = torch.constant.int 1
%int2_3111 = torch.constant.int 2
%5293 = torch.aten.transpose.int %5285#0, %int1_3110, %int2_3111 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %5293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3112 = torch.constant.int 4
%int512_3113 = torch.constant.int 512
%5294 = torch.prim.ListConstruct %int4_3112, %4615, %int512_3113 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5295 = torch.aten.view %5286, %5294 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3114 = torch.constant.int 4
%int512_3115 = torch.constant.int 512
%5296 = torch.prim.ListConstruct %int4_3114, %4630, %int512_3115 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5297 = torch.aten.view %5287, %5296 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3116 = torch.constant.int 4
%int512_3117 = torch.constant.int 512
%5298 = torch.prim.ListConstruct %int4_3116, %4645, %int512_3117 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5299 = torch.aten.view %5288, %5298 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3118 = torch.constant.int 4
%int512_3119 = torch.constant.int 512
%5300 = torch.prim.ListConstruct %int4_3118, %4660, %int512_3119 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5301 = torch.aten.view %5289, %5300 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3120 = torch.constant.int 4
%int512_3121 = torch.constant.int 512
%5302 = torch.prim.ListConstruct %int4_3120, %4675, %int512_3121 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5303 = torch.aten.view %5290, %5302 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3122 = torch.constant.int 4
%int512_3123 = torch.constant.int 512
%5304 = torch.prim.ListConstruct %int4_3122, %4690, %int512_3123 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5305 = torch.aten.view %5291, %5304 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3124 = torch.constant.int 4
%int512_3125 = torch.constant.int 512
%5306 = torch.prim.ListConstruct %int4_3124, %4705, %int512_3125 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5307 = torch.aten.view %5292, %5306 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3126 = torch.constant.int 4
%int512_3127 = torch.constant.int 512
%5308 = torch.prim.ListConstruct %int4_3126, %4720, %int512_3127 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5309 = torch.aten.view %5293, %5308 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %5309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_3128 = torch.constant.int 1
%int0_3129 = torch.constant.int 0
%5310 = torch.prim.ListConstruct %int1_3128, %int0_3129 : (!torch.int, !torch.int) -> !torch.list<int>
%5311 = torch.aten.permute %112, %5310 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3130 = torch.constant.int 1
%int0_3131 = torch.constant.int 0
%5312 = torch.prim.ListConstruct %int1_3130, %int0_3131 : (!torch.int, !torch.int) -> !torch.list<int>
%5313 = torch.aten.permute %113, %5312 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3132 = torch.constant.int 1
%int0_3133 = torch.constant.int 0
%5314 = torch.prim.ListConstruct %int1_3132, %int0_3133 : (!torch.int, !torch.int) -> !torch.list<int>
%5315 = torch.aten.permute %114, %5314 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3134 = torch.constant.int 1
%int0_3135 = torch.constant.int 0
%5316 = torch.prim.ListConstruct %int1_3134, %int0_3135 : (!torch.int, !torch.int) -> !torch.list<int>
%5317 = torch.aten.permute %115, %5316 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3136 = torch.constant.int 1
%int0_3137 = torch.constant.int 0
%5318 = torch.prim.ListConstruct %int1_3136, %int0_3137 : (!torch.int, !torch.int) -> !torch.list<int>
%5319 = torch.aten.permute %116, %5318 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3138 = torch.constant.int 1
%int0_3139 = torch.constant.int 0
%5320 = torch.prim.ListConstruct %int1_3138, %int0_3139 : (!torch.int, !torch.int) -> !torch.list<int>
%5321 = torch.aten.permute %117, %5320 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3140 = torch.constant.int 1
%int0_3141 = torch.constant.int 0
%5322 = torch.prim.ListConstruct %int1_3140, %int0_3141 : (!torch.int, !torch.int) -> !torch.list<int>
%5323 = torch.aten.permute %118, %5322 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_3142 = torch.constant.int 1
%int0_3143 = torch.constant.int 0
%5324 = torch.prim.ListConstruct %int1_3142, %int0_3143 : (!torch.int, !torch.int) -> !torch.list<int>
%5325 = torch.aten.permute %119, %5324 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_3144 = torch.constant.int 4
%5326 = torch.aten.mul.int %int4_3144, %4615 : !torch.int, !torch.int -> !torch.int
%int512_3145 = torch.constant.int 512
%5327 = torch.prim.ListConstruct %5326, %int512_3145 : (!torch.int, !torch.int) -> !torch.list<int>
%5328 = torch.aten.view %5295, %5327 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5328, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5329 = torch.aten.mm %5328, %5311 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5329, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3146 = torch.constant.int 4
%int4096_3147 = torch.constant.int 4096
%5330 = torch.prim.ListConstruct %int4_3146, %4615, %int4096_3147 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5331 = torch.aten.view %5329, %5330 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3148 = torch.constant.int 4
%5332 = torch.aten.mul.int %int4_3148, %4630 : !torch.int, !torch.int -> !torch.int
%int512_3149 = torch.constant.int 512
%5333 = torch.prim.ListConstruct %5332, %int512_3149 : (!torch.int, !torch.int) -> !torch.list<int>
%5334 = torch.aten.view %5297, %5333 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5334, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5335 = torch.aten.mm %5334, %5313 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5335, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3150 = torch.constant.int 4
%int4096_3151 = torch.constant.int 4096
%5336 = torch.prim.ListConstruct %int4_3150, %4630, %int4096_3151 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5337 = torch.aten.view %5335, %5336 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3152 = torch.constant.int 4
%5338 = torch.aten.mul.int %int4_3152, %4645 : !torch.int, !torch.int -> !torch.int
%int512_3153 = torch.constant.int 512
%5339 = torch.prim.ListConstruct %5338, %int512_3153 : (!torch.int, !torch.int) -> !torch.list<int>
%5340 = torch.aten.view %5299, %5339 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5340, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5341 = torch.aten.mm %5340, %5315 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5341, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3154 = torch.constant.int 4
%int4096_3155 = torch.constant.int 4096
%5342 = torch.prim.ListConstruct %int4_3154, %4645, %int4096_3155 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5343 = torch.aten.view %5341, %5342 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3156 = torch.constant.int 4
%5344 = torch.aten.mul.int %int4_3156, %4660 : !torch.int, !torch.int -> !torch.int
%int512_3157 = torch.constant.int 512
%5345 = torch.prim.ListConstruct %5344, %int512_3157 : (!torch.int, !torch.int) -> !torch.list<int>
%5346 = torch.aten.view %5301, %5345 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5346, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5347 = torch.aten.mm %5346, %5317 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5347, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3158 = torch.constant.int 4
%int4096_3159 = torch.constant.int 4096
%5348 = torch.prim.ListConstruct %int4_3158, %4660, %int4096_3159 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5349 = torch.aten.view %5347, %5348 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3160 = torch.constant.int 4
%5350 = torch.aten.mul.int %int4_3160, %4675 : !torch.int, !torch.int -> !torch.int
%int512_3161 = torch.constant.int 512
%5351 = torch.prim.ListConstruct %5350, %int512_3161 : (!torch.int, !torch.int) -> !torch.list<int>
%5352 = torch.aten.view %5303, %5351 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5352, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5353 = torch.aten.mm %5352, %5319 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5353, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3162 = torch.constant.int 4
%int4096_3163 = torch.constant.int 4096
%5354 = torch.prim.ListConstruct %int4_3162, %4675, %int4096_3163 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5355 = torch.aten.view %5353, %5354 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3164 = torch.constant.int 4
%5356 = torch.aten.mul.int %int4_3164, %4690 : !torch.int, !torch.int -> !torch.int
%int512_3165 = torch.constant.int 512
%5357 = torch.prim.ListConstruct %5356, %int512_3165 : (!torch.int, !torch.int) -> !torch.list<int>
%5358 = torch.aten.view %5305, %5357 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5358, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5359 = torch.aten.mm %5358, %5321 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5359, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3166 = torch.constant.int 4
%int4096_3167 = torch.constant.int 4096
%5360 = torch.prim.ListConstruct %int4_3166, %4690, %int4096_3167 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5361 = torch.aten.view %5359, %5360 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3168 = torch.constant.int 4
%5362 = torch.aten.mul.int %int4_3168, %4705 : !torch.int, !torch.int -> !torch.int
%int512_3169 = torch.constant.int 512
%5363 = torch.prim.ListConstruct %5362, %int512_3169 : (!torch.int, !torch.int) -> !torch.list<int>
%5364 = torch.aten.view %5307, %5363 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5364, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5365 = torch.aten.mm %5364, %5323 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5365, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3170 = torch.constant.int 4
%int4096_3171 = torch.constant.int 4096
%5366 = torch.prim.ListConstruct %int4_3170, %4705, %int4096_3171 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5367 = torch.aten.view %5365, %5366 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_3172 = torch.constant.int 4
%5368 = torch.aten.mul.int %int4_3172, %4720 : !torch.int, !torch.int -> !torch.int
%int512_3173 = torch.constant.int 512
%5369 = torch.prim.ListConstruct %5368, %int512_3173 : (!torch.int, !torch.int) -> !torch.list<int>
%5370 = torch.aten.view %5309, %5369 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %5370, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%5371 = torch.aten.mm %5370, %5325 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5371, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3174 = torch.constant.int 4
%int4096_3175 = torch.constant.int 4096
%5372 = torch.prim.ListConstruct %int4_3174, %4720, %int4096_3175 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5373 = torch.aten.view %5371, %5372 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5374 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3176 = arith.constant 1 : index
%dim_3177 = tensor.dim %5374, %c1_3176 : tensor<4x?x4096xf16>
%5375 = flow.tensor.transfer %5374 : tensor<4x?x4096xf16>{%dim_3177} to #hal.device.promise<@__device_0>
%5376 = torch_c.from_builtin_tensor %5375 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5377 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3178 = arith.constant 1 : index
%dim_3179 = tensor.dim %5377, %c1_3178 : tensor<4x?x4096xf16>
%5378 = flow.tensor.transfer %5377 : tensor<4x?x4096xf16>{%dim_3179} to #hal.device.promise<@__device_0>
%5379 = torch_c.from_builtin_tensor %5378 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5380 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3180 = arith.constant 1 : index
%dim_3181 = tensor.dim %5380, %c1_3180 : tensor<4x?x4096xf16>
%5381 = flow.tensor.transfer %5380 : tensor<4x?x4096xf16>{%dim_3181} to #hal.device.promise<@__device_0>
%5382 = torch_c.from_builtin_tensor %5381 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5383 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3182 = arith.constant 1 : index
%dim_3183 = tensor.dim %5383, %c1_3182 : tensor<4x?x4096xf16>
%5384 = flow.tensor.transfer %5383 : tensor<4x?x4096xf16>{%dim_3183} to #hal.device.promise<@__device_0>
%5385 = torch_c.from_builtin_tensor %5384 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5386 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3184 = arith.constant 1 : index
%dim_3185 = tensor.dim %5386, %c1_3184 : tensor<4x?x4096xf16>
%5387 = flow.tensor.transfer %5386 : tensor<4x?x4096xf16>{%dim_3185} to #hal.device.promise<@__device_0>
%5388 = torch_c.from_builtin_tensor %5387 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5389 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3186 = arith.constant 1 : index
%dim_3187 = tensor.dim %5389, %c1_3186 : tensor<4x?x4096xf16>
%5390 = flow.tensor.transfer %5389 : tensor<4x?x4096xf16>{%dim_3187} to #hal.device.promise<@__device_0>
%5391 = torch_c.from_builtin_tensor %5390 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5392 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3188 = arith.constant 1 : index
%dim_3189 = tensor.dim %5392, %c1_3188 : tensor<4x?x4096xf16>
%5393 = flow.tensor.transfer %5392 : tensor<4x?x4096xf16>{%dim_3189} to #hal.device.promise<@__device_0>
%5394 = torch_c.from_builtin_tensor %5393 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3190 = torch.constant.int 1
%5395 = torch.aten.add.Tensor %5331, %5376, %int1_3190 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3191 = torch.constant.int 1
%5396 = torch.aten.add.Tensor %5395, %5379, %int1_3191 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3192 = torch.constant.int 1
%5397 = torch.aten.add.Tensor %5396, %5382, %int1_3192 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3193 = torch.constant.int 1
%5398 = torch.aten.add.Tensor %5397, %5385, %int1_3193 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3194 = torch.constant.int 1
%5399 = torch.aten.add.Tensor %5398, %5388, %int1_3194 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3195 = torch.constant.int 1
%5400 = torch.aten.add.Tensor %5399, %5391, %int1_3195 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3196 = torch.constant.int 1
%5401 = torch.aten.add.Tensor %5400, %5394, %int1_3196 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5402 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3197 = arith.constant 1 : index
%dim_3198 = tensor.dim %5402, %c1_3197 : tensor<4x?x4096xf16>
%5403 = flow.tensor.transfer %5402 : tensor<4x?x4096xf16>{%dim_3198} to #hal.device.promise<@__device_1>
%5404 = torch_c.from_builtin_tensor %5403 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5405 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3199 = arith.constant 1 : index
%dim_3200 = tensor.dim %5405, %c1_3199 : tensor<4x?x4096xf16>
%5406 = flow.tensor.transfer %5405 : tensor<4x?x4096xf16>{%dim_3200} to #hal.device.promise<@__device_1>
%5407 = torch_c.from_builtin_tensor %5406 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5408 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3201 = arith.constant 1 : index
%dim_3202 = tensor.dim %5408, %c1_3201 : tensor<4x?x4096xf16>
%5409 = flow.tensor.transfer %5408 : tensor<4x?x4096xf16>{%dim_3202} to #hal.device.promise<@__device_1>
%5410 = torch_c.from_builtin_tensor %5409 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5411 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3203 = arith.constant 1 : index
%dim_3204 = tensor.dim %5411, %c1_3203 : tensor<4x?x4096xf16>
%5412 = flow.tensor.transfer %5411 : tensor<4x?x4096xf16>{%dim_3204} to #hal.device.promise<@__device_1>
%5413 = torch_c.from_builtin_tensor %5412 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5414 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3205 = arith.constant 1 : index
%dim_3206 = tensor.dim %5414, %c1_3205 : tensor<4x?x4096xf16>
%5415 = flow.tensor.transfer %5414 : tensor<4x?x4096xf16>{%dim_3206} to #hal.device.promise<@__device_1>
%5416 = torch_c.from_builtin_tensor %5415 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5417 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3207 = arith.constant 1 : index
%dim_3208 = tensor.dim %5417, %c1_3207 : tensor<4x?x4096xf16>
%5418 = flow.tensor.transfer %5417 : tensor<4x?x4096xf16>{%dim_3208} to #hal.device.promise<@__device_1>
%5419 = torch_c.from_builtin_tensor %5418 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5420 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3209 = arith.constant 1 : index
%dim_3210 = tensor.dim %5420, %c1_3209 : tensor<4x?x4096xf16>
%5421 = flow.tensor.transfer %5420 : tensor<4x?x4096xf16>{%dim_3210} to #hal.device.promise<@__device_1>
%5422 = torch_c.from_builtin_tensor %5421 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3211 = torch.constant.int 1
%5423 = torch.aten.add.Tensor %5404, %5337, %int1_3211 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3212 = torch.constant.int 1
%5424 = torch.aten.add.Tensor %5423, %5407, %int1_3212 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3213 = torch.constant.int 1
%5425 = torch.aten.add.Tensor %5424, %5410, %int1_3213 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3214 = torch.constant.int 1
%5426 = torch.aten.add.Tensor %5425, %5413, %int1_3214 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3215 = torch.constant.int 1
%5427 = torch.aten.add.Tensor %5426, %5416, %int1_3215 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3216 = torch.constant.int 1
%5428 = torch.aten.add.Tensor %5427, %5419, %int1_3216 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3217 = torch.constant.int 1
%5429 = torch.aten.add.Tensor %5428, %5422, %int1_3217 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5430 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3218 = arith.constant 1 : index
%dim_3219 = tensor.dim %5430, %c1_3218 : tensor<4x?x4096xf16>
%5431 = flow.tensor.transfer %5430 : tensor<4x?x4096xf16>{%dim_3219} to #hal.device.promise<@__device_2>
%5432 = torch_c.from_builtin_tensor %5431 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5433 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3220 = arith.constant 1 : index
%dim_3221 = tensor.dim %5433, %c1_3220 : tensor<4x?x4096xf16>
%5434 = flow.tensor.transfer %5433 : tensor<4x?x4096xf16>{%dim_3221} to #hal.device.promise<@__device_2>
%5435 = torch_c.from_builtin_tensor %5434 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5436 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3222 = arith.constant 1 : index
%dim_3223 = tensor.dim %5436, %c1_3222 : tensor<4x?x4096xf16>
%5437 = flow.tensor.transfer %5436 : tensor<4x?x4096xf16>{%dim_3223} to #hal.device.promise<@__device_2>
%5438 = torch_c.from_builtin_tensor %5437 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5439 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3224 = arith.constant 1 : index
%dim_3225 = tensor.dim %5439, %c1_3224 : tensor<4x?x4096xf16>
%5440 = flow.tensor.transfer %5439 : tensor<4x?x4096xf16>{%dim_3225} to #hal.device.promise<@__device_2>
%5441 = torch_c.from_builtin_tensor %5440 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5442 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3226 = arith.constant 1 : index
%dim_3227 = tensor.dim %5442, %c1_3226 : tensor<4x?x4096xf16>
%5443 = flow.tensor.transfer %5442 : tensor<4x?x4096xf16>{%dim_3227} to #hal.device.promise<@__device_2>
%5444 = torch_c.from_builtin_tensor %5443 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5445 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3228 = arith.constant 1 : index
%dim_3229 = tensor.dim %5445, %c1_3228 : tensor<4x?x4096xf16>
%5446 = flow.tensor.transfer %5445 : tensor<4x?x4096xf16>{%dim_3229} to #hal.device.promise<@__device_2>
%5447 = torch_c.from_builtin_tensor %5446 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5448 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3230 = arith.constant 1 : index
%dim_3231 = tensor.dim %5448, %c1_3230 : tensor<4x?x4096xf16>
%5449 = flow.tensor.transfer %5448 : tensor<4x?x4096xf16>{%dim_3231} to #hal.device.promise<@__device_2>
%5450 = torch_c.from_builtin_tensor %5449 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3232 = torch.constant.int 1
%5451 = torch.aten.add.Tensor %5432, %5435, %int1_3232 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3233 = torch.constant.int 1
%5452 = torch.aten.add.Tensor %5451, %5343, %int1_3233 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3234 = torch.constant.int 1
%5453 = torch.aten.add.Tensor %5452, %5438, %int1_3234 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3235 = torch.constant.int 1
%5454 = torch.aten.add.Tensor %5453, %5441, %int1_3235 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3236 = torch.constant.int 1
%5455 = torch.aten.add.Tensor %5454, %5444, %int1_3236 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3237 = torch.constant.int 1
%5456 = torch.aten.add.Tensor %5455, %5447, %int1_3237 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3238 = torch.constant.int 1
%5457 = torch.aten.add.Tensor %5456, %5450, %int1_3238 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5458 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3239 = arith.constant 1 : index
%dim_3240 = tensor.dim %5458, %c1_3239 : tensor<4x?x4096xf16>
%5459 = flow.tensor.transfer %5458 : tensor<4x?x4096xf16>{%dim_3240} to #hal.device.promise<@__device_3>
%5460 = torch_c.from_builtin_tensor %5459 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5461 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3241 = arith.constant 1 : index
%dim_3242 = tensor.dim %5461, %c1_3241 : tensor<4x?x4096xf16>
%5462 = flow.tensor.transfer %5461 : tensor<4x?x4096xf16>{%dim_3242} to #hal.device.promise<@__device_3>
%5463 = torch_c.from_builtin_tensor %5462 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5464 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3243 = arith.constant 1 : index
%dim_3244 = tensor.dim %5464, %c1_3243 : tensor<4x?x4096xf16>
%5465 = flow.tensor.transfer %5464 : tensor<4x?x4096xf16>{%dim_3244} to #hal.device.promise<@__device_3>
%5466 = torch_c.from_builtin_tensor %5465 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5467 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3245 = arith.constant 1 : index
%dim_3246 = tensor.dim %5467, %c1_3245 : tensor<4x?x4096xf16>
%5468 = flow.tensor.transfer %5467 : tensor<4x?x4096xf16>{%dim_3246} to #hal.device.promise<@__device_3>
%5469 = torch_c.from_builtin_tensor %5468 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5470 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3247 = arith.constant 1 : index
%dim_3248 = tensor.dim %5470, %c1_3247 : tensor<4x?x4096xf16>
%5471 = flow.tensor.transfer %5470 : tensor<4x?x4096xf16>{%dim_3248} to #hal.device.promise<@__device_3>
%5472 = torch_c.from_builtin_tensor %5471 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5473 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3249 = arith.constant 1 : index
%dim_3250 = tensor.dim %5473, %c1_3249 : tensor<4x?x4096xf16>
%5474 = flow.tensor.transfer %5473 : tensor<4x?x4096xf16>{%dim_3250} to #hal.device.promise<@__device_3>
%5475 = torch_c.from_builtin_tensor %5474 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5476 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3251 = arith.constant 1 : index
%dim_3252 = tensor.dim %5476, %c1_3251 : tensor<4x?x4096xf16>
%5477 = flow.tensor.transfer %5476 : tensor<4x?x4096xf16>{%dim_3252} to #hal.device.promise<@__device_3>
%5478 = torch_c.from_builtin_tensor %5477 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3253 = torch.constant.int 1
%5479 = torch.aten.add.Tensor %5460, %5463, %int1_3253 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3254 = torch.constant.int 1
%5480 = torch.aten.add.Tensor %5479, %5466, %int1_3254 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3255 = torch.constant.int 1
%5481 = torch.aten.add.Tensor %5480, %5349, %int1_3255 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3256 = torch.constant.int 1
%5482 = torch.aten.add.Tensor %5481, %5469, %int1_3256 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3257 = torch.constant.int 1
%5483 = torch.aten.add.Tensor %5482, %5472, %int1_3257 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3258 = torch.constant.int 1
%5484 = torch.aten.add.Tensor %5483, %5475, %int1_3258 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3259 = torch.constant.int 1
%5485 = torch.aten.add.Tensor %5484, %5478, %int1_3259 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5486 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3260 = arith.constant 1 : index
%dim_3261 = tensor.dim %5486, %c1_3260 : tensor<4x?x4096xf16>
%5487 = flow.tensor.transfer %5486 : tensor<4x?x4096xf16>{%dim_3261} to #hal.device.promise<@__device_4>
%5488 = torch_c.from_builtin_tensor %5487 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5489 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3262 = arith.constant 1 : index
%dim_3263 = tensor.dim %5489, %c1_3262 : tensor<4x?x4096xf16>
%5490 = flow.tensor.transfer %5489 : tensor<4x?x4096xf16>{%dim_3263} to #hal.device.promise<@__device_4>
%5491 = torch_c.from_builtin_tensor %5490 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5492 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3264 = arith.constant 1 : index
%dim_3265 = tensor.dim %5492, %c1_3264 : tensor<4x?x4096xf16>
%5493 = flow.tensor.transfer %5492 : tensor<4x?x4096xf16>{%dim_3265} to #hal.device.promise<@__device_4>
%5494 = torch_c.from_builtin_tensor %5493 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5495 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3266 = arith.constant 1 : index
%dim_3267 = tensor.dim %5495, %c1_3266 : tensor<4x?x4096xf16>
%5496 = flow.tensor.transfer %5495 : tensor<4x?x4096xf16>{%dim_3267} to #hal.device.promise<@__device_4>
%5497 = torch_c.from_builtin_tensor %5496 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5498 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3268 = arith.constant 1 : index
%dim_3269 = tensor.dim %5498, %c1_3268 : tensor<4x?x4096xf16>
%5499 = flow.tensor.transfer %5498 : tensor<4x?x4096xf16>{%dim_3269} to #hal.device.promise<@__device_4>
%5500 = torch_c.from_builtin_tensor %5499 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5501 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3270 = arith.constant 1 : index
%dim_3271 = tensor.dim %5501, %c1_3270 : tensor<4x?x4096xf16>
%5502 = flow.tensor.transfer %5501 : tensor<4x?x4096xf16>{%dim_3271} to #hal.device.promise<@__device_4>
%5503 = torch_c.from_builtin_tensor %5502 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5504 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3272 = arith.constant 1 : index
%dim_3273 = tensor.dim %5504, %c1_3272 : tensor<4x?x4096xf16>
%5505 = flow.tensor.transfer %5504 : tensor<4x?x4096xf16>{%dim_3273} to #hal.device.promise<@__device_4>
%5506 = torch_c.from_builtin_tensor %5505 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3274 = torch.constant.int 1
%5507 = torch.aten.add.Tensor %5488, %5491, %int1_3274 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3275 = torch.constant.int 1
%5508 = torch.aten.add.Tensor %5507, %5494, %int1_3275 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3276 = torch.constant.int 1
%5509 = torch.aten.add.Tensor %5508, %5497, %int1_3276 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3277 = torch.constant.int 1
%5510 = torch.aten.add.Tensor %5509, %5355, %int1_3277 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3278 = torch.constant.int 1
%5511 = torch.aten.add.Tensor %5510, %5500, %int1_3278 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3279 = torch.constant.int 1
%5512 = torch.aten.add.Tensor %5511, %5503, %int1_3279 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3280 = torch.constant.int 1
%5513 = torch.aten.add.Tensor %5512, %5506, %int1_3280 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5514 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3281 = arith.constant 1 : index
%dim_3282 = tensor.dim %5514, %c1_3281 : tensor<4x?x4096xf16>
%5515 = flow.tensor.transfer %5514 : tensor<4x?x4096xf16>{%dim_3282} to #hal.device.promise<@__device_5>
%5516 = torch_c.from_builtin_tensor %5515 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5517 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3283 = arith.constant 1 : index
%dim_3284 = tensor.dim %5517, %c1_3283 : tensor<4x?x4096xf16>
%5518 = flow.tensor.transfer %5517 : tensor<4x?x4096xf16>{%dim_3284} to #hal.device.promise<@__device_5>
%5519 = torch_c.from_builtin_tensor %5518 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5520 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3285 = arith.constant 1 : index
%dim_3286 = tensor.dim %5520, %c1_3285 : tensor<4x?x4096xf16>
%5521 = flow.tensor.transfer %5520 : tensor<4x?x4096xf16>{%dim_3286} to #hal.device.promise<@__device_5>
%5522 = torch_c.from_builtin_tensor %5521 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5523 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3287 = arith.constant 1 : index
%dim_3288 = tensor.dim %5523, %c1_3287 : tensor<4x?x4096xf16>
%5524 = flow.tensor.transfer %5523 : tensor<4x?x4096xf16>{%dim_3288} to #hal.device.promise<@__device_5>
%5525 = torch_c.from_builtin_tensor %5524 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5526 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3289 = arith.constant 1 : index
%dim_3290 = tensor.dim %5526, %c1_3289 : tensor<4x?x4096xf16>
%5527 = flow.tensor.transfer %5526 : tensor<4x?x4096xf16>{%dim_3290} to #hal.device.promise<@__device_5>
%5528 = torch_c.from_builtin_tensor %5527 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5529 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3291 = arith.constant 1 : index
%dim_3292 = tensor.dim %5529, %c1_3291 : tensor<4x?x4096xf16>
%5530 = flow.tensor.transfer %5529 : tensor<4x?x4096xf16>{%dim_3292} to #hal.device.promise<@__device_5>
%5531 = torch_c.from_builtin_tensor %5530 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5532 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3293 = arith.constant 1 : index
%dim_3294 = tensor.dim %5532, %c1_3293 : tensor<4x?x4096xf16>
%5533 = flow.tensor.transfer %5532 : tensor<4x?x4096xf16>{%dim_3294} to #hal.device.promise<@__device_5>
%5534 = torch_c.from_builtin_tensor %5533 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3295 = torch.constant.int 1
%5535 = torch.aten.add.Tensor %5516, %5519, %int1_3295 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3296 = torch.constant.int 1
%5536 = torch.aten.add.Tensor %5535, %5522, %int1_3296 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3297 = torch.constant.int 1
%5537 = torch.aten.add.Tensor %5536, %5525, %int1_3297 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3298 = torch.constant.int 1
%5538 = torch.aten.add.Tensor %5537, %5528, %int1_3298 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3299 = torch.constant.int 1
%5539 = torch.aten.add.Tensor %5538, %5361, %int1_3299 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3300 = torch.constant.int 1
%5540 = torch.aten.add.Tensor %5539, %5531, %int1_3300 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3301 = torch.constant.int 1
%5541 = torch.aten.add.Tensor %5540, %5534, %int1_3301 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5541, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5542 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3302 = arith.constant 1 : index
%dim_3303 = tensor.dim %5542, %c1_3302 : tensor<4x?x4096xf16>
%5543 = flow.tensor.transfer %5542 : tensor<4x?x4096xf16>{%dim_3303} to #hal.device.promise<@__device_6>
%5544 = torch_c.from_builtin_tensor %5543 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5545 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3304 = arith.constant 1 : index
%dim_3305 = tensor.dim %5545, %c1_3304 : tensor<4x?x4096xf16>
%5546 = flow.tensor.transfer %5545 : tensor<4x?x4096xf16>{%dim_3305} to #hal.device.promise<@__device_6>
%5547 = torch_c.from_builtin_tensor %5546 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5548 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3306 = arith.constant 1 : index
%dim_3307 = tensor.dim %5548, %c1_3306 : tensor<4x?x4096xf16>
%5549 = flow.tensor.transfer %5548 : tensor<4x?x4096xf16>{%dim_3307} to #hal.device.promise<@__device_6>
%5550 = torch_c.from_builtin_tensor %5549 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5551 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3308 = arith.constant 1 : index
%dim_3309 = tensor.dim %5551, %c1_3308 : tensor<4x?x4096xf16>
%5552 = flow.tensor.transfer %5551 : tensor<4x?x4096xf16>{%dim_3309} to #hal.device.promise<@__device_6>
%5553 = torch_c.from_builtin_tensor %5552 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5554 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3310 = arith.constant 1 : index
%dim_3311 = tensor.dim %5554, %c1_3310 : tensor<4x?x4096xf16>
%5555 = flow.tensor.transfer %5554 : tensor<4x?x4096xf16>{%dim_3311} to #hal.device.promise<@__device_6>
%5556 = torch_c.from_builtin_tensor %5555 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5557 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3312 = arith.constant 1 : index
%dim_3313 = tensor.dim %5557, %c1_3312 : tensor<4x?x4096xf16>
%5558 = flow.tensor.transfer %5557 : tensor<4x?x4096xf16>{%dim_3313} to #hal.device.promise<@__device_6>
%5559 = torch_c.from_builtin_tensor %5558 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5560 = torch_c.to_builtin_tensor %5373 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3314 = arith.constant 1 : index
%dim_3315 = tensor.dim %5560, %c1_3314 : tensor<4x?x4096xf16>
%5561 = flow.tensor.transfer %5560 : tensor<4x?x4096xf16>{%dim_3315} to #hal.device.promise<@__device_6>
%5562 = torch_c.from_builtin_tensor %5561 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3316 = torch.constant.int 1
%5563 = torch.aten.add.Tensor %5544, %5547, %int1_3316 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3317 = torch.constant.int 1
%5564 = torch.aten.add.Tensor %5563, %5550, %int1_3317 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3318 = torch.constant.int 1
%5565 = torch.aten.add.Tensor %5564, %5553, %int1_3318 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3319 = torch.constant.int 1
%5566 = torch.aten.add.Tensor %5565, %5556, %int1_3319 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3320 = torch.constant.int 1
%5567 = torch.aten.add.Tensor %5566, %5559, %int1_3320 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3321 = torch.constant.int 1
%5568 = torch.aten.add.Tensor %5567, %5367, %int1_3321 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3322 = torch.constant.int 1
%5569 = torch.aten.add.Tensor %5568, %5562, %int1_3322 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5570 = torch_c.to_builtin_tensor %5331 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3323 = arith.constant 1 : index
%dim_3324 = tensor.dim %5570, %c1_3323 : tensor<4x?x4096xf16>
%5571 = flow.tensor.transfer %5570 : tensor<4x?x4096xf16>{%dim_3324} to #hal.device.promise<@__device_7>
%5572 = torch_c.from_builtin_tensor %5571 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5573 = torch_c.to_builtin_tensor %5337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3325 = arith.constant 1 : index
%dim_3326 = tensor.dim %5573, %c1_3325 : tensor<4x?x4096xf16>
%5574 = flow.tensor.transfer %5573 : tensor<4x?x4096xf16>{%dim_3326} to #hal.device.promise<@__device_7>
%5575 = torch_c.from_builtin_tensor %5574 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5576 = torch_c.to_builtin_tensor %5343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3327 = arith.constant 1 : index
%dim_3328 = tensor.dim %5576, %c1_3327 : tensor<4x?x4096xf16>
%5577 = flow.tensor.transfer %5576 : tensor<4x?x4096xf16>{%dim_3328} to #hal.device.promise<@__device_7>
%5578 = torch_c.from_builtin_tensor %5577 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5579 = torch_c.to_builtin_tensor %5349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3329 = arith.constant 1 : index
%dim_3330 = tensor.dim %5579, %c1_3329 : tensor<4x?x4096xf16>
%5580 = flow.tensor.transfer %5579 : tensor<4x?x4096xf16>{%dim_3330} to #hal.device.promise<@__device_7>
%5581 = torch_c.from_builtin_tensor %5580 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5582 = torch_c.to_builtin_tensor %5355 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3331 = arith.constant 1 : index
%dim_3332 = tensor.dim %5582, %c1_3331 : tensor<4x?x4096xf16>
%5583 = flow.tensor.transfer %5582 : tensor<4x?x4096xf16>{%dim_3332} to #hal.device.promise<@__device_7>
%5584 = torch_c.from_builtin_tensor %5583 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5585 = torch_c.to_builtin_tensor %5361 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3333 = arith.constant 1 : index
%dim_3334 = tensor.dim %5585, %c1_3333 : tensor<4x?x4096xf16>
%5586 = flow.tensor.transfer %5585 : tensor<4x?x4096xf16>{%dim_3334} to #hal.device.promise<@__device_7>
%5587 = torch_c.from_builtin_tensor %5586 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5588 = torch_c.to_builtin_tensor %5367 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3335 = arith.constant 1 : index
%dim_3336 = tensor.dim %5588, %c1_3335 : tensor<4x?x4096xf16>
%5589 = flow.tensor.transfer %5588 : tensor<4x?x4096xf16>{%dim_3336} to #hal.device.promise<@__device_7>
%5590 = torch_c.from_builtin_tensor %5589 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3337 = torch.constant.int 1
%5591 = torch.aten.add.Tensor %5572, %5575, %int1_3337 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3338 = torch.constant.int 1
%5592 = torch.aten.add.Tensor %5591, %5578, %int1_3338 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3339 = torch.constant.int 1
%5593 = torch.aten.add.Tensor %5592, %5581, %int1_3339 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3340 = torch.constant.int 1
%5594 = torch.aten.add.Tensor %5593, %5584, %int1_3340 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3341 = torch.constant.int 1
%5595 = torch.aten.add.Tensor %5594, %5587, %int1_3341 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3342 = torch.constant.int 1
%5596 = torch.aten.add.Tensor %5595, %5590, %int1_3342 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3343 = torch.constant.int 1
%5597 = torch.aten.add.Tensor %5596, %5373, %int1_3343 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3344 = torch.constant.int 1
%5598 = torch.aten.add.Tensor %4257, %5401, %int1_3344 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3345 = torch.constant.int 1
%5599 = torch.aten.add.Tensor %4258, %5429, %int1_3345 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3346 = torch.constant.int 1
%5600 = torch.aten.add.Tensor %4259, %5457, %int1_3346 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3347 = torch.constant.int 1
%5601 = torch.aten.add.Tensor %4260, %5485, %int1_3347 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3348 = torch.constant.int 1
%5602 = torch.aten.add.Tensor %4261, %5513, %int1_3348 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3349 = torch.constant.int 1
%5603 = torch.aten.add.Tensor %4262, %5541, %int1_3349 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3350 = torch.constant.int 1
%5604 = torch.aten.add.Tensor %4263, %5569, %int1_3350 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3351 = torch.constant.int 1
%5605 = torch.aten.add.Tensor %4264, %5597, %int1_3351 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_3352 = torch.constant.int 6
%5606 = torch.prims.convert_element_type %5598, %int6_3352 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3353 = torch.constant.int 6
%5607 = torch.prims.convert_element_type %5599, %int6_3353 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3354 = torch.constant.int 6
%5608 = torch.prims.convert_element_type %5600, %int6_3354 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3355 = torch.constant.int 6
%5609 = torch.prims.convert_element_type %5601, %int6_3355 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3356 = torch.constant.int 6
%5610 = torch.prims.convert_element_type %5602, %int6_3356 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3357 = torch.constant.int 6
%5611 = torch.prims.convert_element_type %5603, %int6_3357 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3358 = torch.constant.int 6
%5612 = torch.prims.convert_element_type %5604, %int6_3358 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3359 = torch.constant.int 6
%5613 = torch.prims.convert_element_type %5605, %int6_3359 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3360 = torch.constant.int 2
%5614 = torch.aten.pow.Tensor_Scalar %5606, %int2_3360 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3361 = torch.constant.int 2
%5615 = torch.aten.pow.Tensor_Scalar %5607, %int2_3361 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3362 = torch.constant.int 2
%5616 = torch.aten.pow.Tensor_Scalar %5608, %int2_3362 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3363 = torch.constant.int 2
%5617 = torch.aten.pow.Tensor_Scalar %5609, %int2_3363 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3364 = torch.constant.int 2
%5618 = torch.aten.pow.Tensor_Scalar %5610, %int2_3364 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3365 = torch.constant.int 2
%5619 = torch.aten.pow.Tensor_Scalar %5611, %int2_3365 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3366 = torch.constant.int 2
%5620 = torch.aten.pow.Tensor_Scalar %5612, %int2_3366 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3367 = torch.constant.int 2
%5621 = torch.aten.pow.Tensor_Scalar %5613, %int2_3367 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_3368 = torch.constant.int -1
%5622 = torch.prim.ListConstruct %int-1_3368 : (!torch.int) -> !torch.list<int>
%true_3369 = torch.constant.bool true
%none_3370 = torch.constant.none
%5623 = torch.aten.mean.dim %5614, %5622, %true_3369, %none_3370 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3371 = torch.constant.int -1
%5624 = torch.prim.ListConstruct %int-1_3371 : (!torch.int) -> !torch.list<int>
%true_3372 = torch.constant.bool true
%none_3373 = torch.constant.none
%5625 = torch.aten.mean.dim %5615, %5624, %true_3372, %none_3373 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3374 = torch.constant.int -1
%5626 = torch.prim.ListConstruct %int-1_3374 : (!torch.int) -> !torch.list<int>
%true_3375 = torch.constant.bool true
%none_3376 = torch.constant.none
%5627 = torch.aten.mean.dim %5616, %5626, %true_3375, %none_3376 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3377 = torch.constant.int -1
%5628 = torch.prim.ListConstruct %int-1_3377 : (!torch.int) -> !torch.list<int>
%true_3378 = torch.constant.bool true
%none_3379 = torch.constant.none
%5629 = torch.aten.mean.dim %5617, %5628, %true_3378, %none_3379 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3380 = torch.constant.int -1
%5630 = torch.prim.ListConstruct %int-1_3380 : (!torch.int) -> !torch.list<int>
%true_3381 = torch.constant.bool true
%none_3382 = torch.constant.none
%5631 = torch.aten.mean.dim %5618, %5630, %true_3381, %none_3382 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3383 = torch.constant.int -1
%5632 = torch.prim.ListConstruct %int-1_3383 : (!torch.int) -> !torch.list<int>
%true_3384 = torch.constant.bool true
%none_3385 = torch.constant.none
%5633 = torch.aten.mean.dim %5619, %5632, %true_3384, %none_3385 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3386 = torch.constant.int -1
%5634 = torch.prim.ListConstruct %int-1_3386 : (!torch.int) -> !torch.list<int>
%true_3387 = torch.constant.bool true
%none_3388 = torch.constant.none
%5635 = torch.aten.mean.dim %5620, %5634, %true_3387, %none_3388 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3389 = torch.constant.int -1
%5636 = torch.prim.ListConstruct %int-1_3389 : (!torch.int) -> !torch.list<int>
%true_3390 = torch.constant.bool true
%none_3391 = torch.constant.none
%5637 = torch.aten.mean.dim %5621, %5636, %true_3390, %none_3391 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3392 = torch.constant.float 9.9999997473787516E-6
%int1_3393 = torch.constant.int 1
%5638 = torch.aten.add.Scalar %5623, %float9.999990e-06_3392, %int1_3393 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3394 = torch.constant.float 9.9999997473787516E-6
%int1_3395 = torch.constant.int 1
%5639 = torch.aten.add.Scalar %5625, %float9.999990e-06_3394, %int1_3395 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3396 = torch.constant.float 9.9999997473787516E-6
%int1_3397 = torch.constant.int 1
%5640 = torch.aten.add.Scalar %5627, %float9.999990e-06_3396, %int1_3397 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3398 = torch.constant.float 9.9999997473787516E-6
%int1_3399 = torch.constant.int 1
%5641 = torch.aten.add.Scalar %5629, %float9.999990e-06_3398, %int1_3399 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3400 = torch.constant.float 9.9999997473787516E-6
%int1_3401 = torch.constant.int 1
%5642 = torch.aten.add.Scalar %5631, %float9.999990e-06_3400, %int1_3401 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3402 = torch.constant.float 9.9999997473787516E-6
%int1_3403 = torch.constant.int 1
%5643 = torch.aten.add.Scalar %5633, %float9.999990e-06_3402, %int1_3403 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3404 = torch.constant.float 9.9999997473787516E-6
%int1_3405 = torch.constant.int 1
%5644 = torch.aten.add.Scalar %5635, %float9.999990e-06_3404, %int1_3405 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3406 = torch.constant.float 9.9999997473787516E-6
%int1_3407 = torch.constant.int 1
%5645 = torch.aten.add.Scalar %5637, %float9.999990e-06_3406, %int1_3407 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5646 = torch.aten.rsqrt %5638 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5647 = torch.aten.rsqrt %5639 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5648 = torch.aten.rsqrt %5640 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5649 = torch.aten.rsqrt %5641 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5650 = torch.aten.rsqrt %5642 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5651 = torch.aten.rsqrt %5643 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5652 = torch.aten.rsqrt %5644 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5653 = torch.aten.rsqrt %5645 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %5653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%5654 = torch.aten.mul.Tensor %5606, %5646 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5655 = torch.aten.mul.Tensor %5607, %5647 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5656 = torch.aten.mul.Tensor %5608, %5648 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5657 = torch.aten.mul.Tensor %5609, %5649 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5658 = torch.aten.mul.Tensor %5610, %5650 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5659 = torch.aten.mul.Tensor %5611, %5651 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5660 = torch.aten.mul.Tensor %5612, %5652 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5661 = torch.aten.mul.Tensor %5613, %5653 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5662 = torch.aten.mul.Tensor %120, %5654 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5663 = torch.aten.mul.Tensor %121, %5655 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5664 = torch.aten.mul.Tensor %122, %5656 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5665 = torch.aten.mul.Tensor %123, %5657 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5666 = torch.aten.mul.Tensor %124, %5658 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5667 = torch.aten.mul.Tensor %125, %5659 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5668 = torch.aten.mul.Tensor %126, %5660 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%5669 = torch.aten.mul.Tensor %127, %5661 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %5669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_3408 = torch.constant.int 5
%5670 = torch.prims.convert_element_type %5662, %int5_3408 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3409 = torch.constant.int 5
%5671 = torch.prims.convert_element_type %5663, %int5_3409 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3410 = torch.constant.int 5
%5672 = torch.prims.convert_element_type %5664, %int5_3410 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3411 = torch.constant.int 5
%5673 = torch.prims.convert_element_type %5665, %int5_3411 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3412 = torch.constant.int 5
%5674 = torch.prims.convert_element_type %5666, %int5_3412 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3413 = torch.constant.int 5
%5675 = torch.prims.convert_element_type %5667, %int5_3413 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3414 = torch.constant.int 5
%5676 = torch.prims.convert_element_type %5668, %int5_3414 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3415 = torch.constant.int 5
%5677 = torch.prims.convert_element_type %5669, %int5_3415 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3416 = torch.constant.int 1
%int0_3417 = torch.constant.int 0
%5678 = torch.prim.ListConstruct %int1_3416, %int0_3417 : (!torch.int, !torch.int) -> !torch.list<int>
%5679 = torch.aten.permute %128, %5678 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3418 = torch.constant.int 1
%int0_3419 = torch.constant.int 0
%5680 = torch.prim.ListConstruct %int1_3418, %int0_3419 : (!torch.int, !torch.int) -> !torch.list<int>
%5681 = torch.aten.permute %129, %5680 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3420 = torch.constant.int 1
%int0_3421 = torch.constant.int 0
%5682 = torch.prim.ListConstruct %int1_3420, %int0_3421 : (!torch.int, !torch.int) -> !torch.list<int>
%5683 = torch.aten.permute %130, %5682 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3422 = torch.constant.int 1
%int0_3423 = torch.constant.int 0
%5684 = torch.prim.ListConstruct %int1_3422, %int0_3423 : (!torch.int, !torch.int) -> !torch.list<int>
%5685 = torch.aten.permute %131, %5684 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3424 = torch.constant.int 1
%int0_3425 = torch.constant.int 0
%5686 = torch.prim.ListConstruct %int1_3424, %int0_3425 : (!torch.int, !torch.int) -> !torch.list<int>
%5687 = torch.aten.permute %132, %5686 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3426 = torch.constant.int 1
%int0_3427 = torch.constant.int 0
%5688 = torch.prim.ListConstruct %int1_3426, %int0_3427 : (!torch.int, !torch.int) -> !torch.list<int>
%5689 = torch.aten.permute %133, %5688 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3428 = torch.constant.int 1
%int0_3429 = torch.constant.int 0
%5690 = torch.prim.ListConstruct %int1_3428, %int0_3429 : (!torch.int, !torch.int) -> !torch.list<int>
%5691 = torch.aten.permute %134, %5690 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3430 = torch.constant.int 1
%int0_3431 = torch.constant.int 0
%5692 = torch.prim.ListConstruct %int1_3430, %int0_3431 : (!torch.int, !torch.int) -> !torch.list<int>
%5693 = torch.aten.permute %135, %5692 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_3432 = torch.constant.int 4
%5694 = torch.aten.mul.int %int4_3432, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3433 = torch.constant.int 4096
%5695 = torch.prim.ListConstruct %5694, %int4096_3433 : (!torch.int, !torch.int) -> !torch.list<int>
%5696 = torch.aten.view %5670, %5695 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5696, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5697 = torch.aten.mm %5696, %5679 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5697, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3434 = torch.constant.int 4
%int1792_3435 = torch.constant.int 1792
%5698 = torch.prim.ListConstruct %int4_3434, %2482, %int1792_3435 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5699 = torch.aten.view %5697, %5698 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3436 = torch.constant.int 4
%5700 = torch.aten.mul.int %int4_3436, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3437 = torch.constant.int 4096
%5701 = torch.prim.ListConstruct %5700, %int4096_3437 : (!torch.int, !torch.int) -> !torch.list<int>
%5702 = torch.aten.view %5671, %5701 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5702, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5703 = torch.aten.mm %5702, %5681 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5703, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3438 = torch.constant.int 4
%int1792_3439 = torch.constant.int 1792
%5704 = torch.prim.ListConstruct %int4_3438, %2482, %int1792_3439 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5705 = torch.aten.view %5703, %5704 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3440 = torch.constant.int 4
%5706 = torch.aten.mul.int %int4_3440, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3441 = torch.constant.int 4096
%5707 = torch.prim.ListConstruct %5706, %int4096_3441 : (!torch.int, !torch.int) -> !torch.list<int>
%5708 = torch.aten.view %5672, %5707 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5708, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5709 = torch.aten.mm %5708, %5683 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5709, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3442 = torch.constant.int 4
%int1792_3443 = torch.constant.int 1792
%5710 = torch.prim.ListConstruct %int4_3442, %2482, %int1792_3443 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5711 = torch.aten.view %5709, %5710 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3444 = torch.constant.int 4
%5712 = torch.aten.mul.int %int4_3444, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3445 = torch.constant.int 4096
%5713 = torch.prim.ListConstruct %5712, %int4096_3445 : (!torch.int, !torch.int) -> !torch.list<int>
%5714 = torch.aten.view %5673, %5713 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5714, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5715 = torch.aten.mm %5714, %5685 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5715, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3446 = torch.constant.int 4
%int1792_3447 = torch.constant.int 1792
%5716 = torch.prim.ListConstruct %int4_3446, %2482, %int1792_3447 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5717 = torch.aten.view %5715, %5716 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3448 = torch.constant.int 4
%5718 = torch.aten.mul.int %int4_3448, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3449 = torch.constant.int 4096
%5719 = torch.prim.ListConstruct %5718, %int4096_3449 : (!torch.int, !torch.int) -> !torch.list<int>
%5720 = torch.aten.view %5674, %5719 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5720, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5721 = torch.aten.mm %5720, %5687 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5721, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3450 = torch.constant.int 4
%int1792_3451 = torch.constant.int 1792
%5722 = torch.prim.ListConstruct %int4_3450, %2482, %int1792_3451 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5723 = torch.aten.view %5721, %5722 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3452 = torch.constant.int 4
%5724 = torch.aten.mul.int %int4_3452, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3453 = torch.constant.int 4096
%5725 = torch.prim.ListConstruct %5724, %int4096_3453 : (!torch.int, !torch.int) -> !torch.list<int>
%5726 = torch.aten.view %5675, %5725 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5726, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5727 = torch.aten.mm %5726, %5689 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5727, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3454 = torch.constant.int 4
%int1792_3455 = torch.constant.int 1792
%5728 = torch.prim.ListConstruct %int4_3454, %2482, %int1792_3455 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5729 = torch.aten.view %5727, %5728 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3456 = torch.constant.int 4
%5730 = torch.aten.mul.int %int4_3456, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3457 = torch.constant.int 4096
%5731 = torch.prim.ListConstruct %5730, %int4096_3457 : (!torch.int, !torch.int) -> !torch.list<int>
%5732 = torch.aten.view %5676, %5731 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5732, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5733 = torch.aten.mm %5732, %5691 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5733, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3458 = torch.constant.int 4
%int1792_3459 = torch.constant.int 1792
%5734 = torch.prim.ListConstruct %int4_3458, %2482, %int1792_3459 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5735 = torch.aten.view %5733, %5734 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3460 = torch.constant.int 4
%5736 = torch.aten.mul.int %int4_3460, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3461 = torch.constant.int 4096
%5737 = torch.prim.ListConstruct %5736, %int4096_3461 : (!torch.int, !torch.int) -> !torch.list<int>
%5738 = torch.aten.view %5677, %5737 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5738, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5739 = torch.aten.mm %5738, %5693 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5739, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3462 = torch.constant.int 4
%int1792_3463 = torch.constant.int 1792
%5740 = torch.prim.ListConstruct %int4_3462, %2482, %int1792_3463 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5741 = torch.aten.view %5739, %5740 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5742 = torch.aten.silu %5699 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5743 = torch.aten.silu %5705 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5744 = torch.aten.silu %5711 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5745 = torch.aten.silu %5717 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5746 = torch.aten.silu %5723 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5747 = torch.aten.silu %5729 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5748 = torch.aten.silu %5735 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5749 = torch.aten.silu %5741 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_3464 = torch.constant.int 1
%int0_3465 = torch.constant.int 0
%5750 = torch.prim.ListConstruct %int1_3464, %int0_3465 : (!torch.int, !torch.int) -> !torch.list<int>
%5751 = torch.aten.permute %136, %5750 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3466 = torch.constant.int 1
%int0_3467 = torch.constant.int 0
%5752 = torch.prim.ListConstruct %int1_3466, %int0_3467 : (!torch.int, !torch.int) -> !torch.list<int>
%5753 = torch.aten.permute %137, %5752 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3468 = torch.constant.int 1
%int0_3469 = torch.constant.int 0
%5754 = torch.prim.ListConstruct %int1_3468, %int0_3469 : (!torch.int, !torch.int) -> !torch.list<int>
%5755 = torch.aten.permute %138, %5754 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3470 = torch.constant.int 1
%int0_3471 = torch.constant.int 0
%5756 = torch.prim.ListConstruct %int1_3470, %int0_3471 : (!torch.int, !torch.int) -> !torch.list<int>
%5757 = torch.aten.permute %139, %5756 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3472 = torch.constant.int 1
%int0_3473 = torch.constant.int 0
%5758 = torch.prim.ListConstruct %int1_3472, %int0_3473 : (!torch.int, !torch.int) -> !torch.list<int>
%5759 = torch.aten.permute %140, %5758 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3474 = torch.constant.int 1
%int0_3475 = torch.constant.int 0
%5760 = torch.prim.ListConstruct %int1_3474, %int0_3475 : (!torch.int, !torch.int) -> !torch.list<int>
%5761 = torch.aten.permute %141, %5760 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3476 = torch.constant.int 1
%int0_3477 = torch.constant.int 0
%5762 = torch.prim.ListConstruct %int1_3476, %int0_3477 : (!torch.int, !torch.int) -> !torch.list<int>
%5763 = torch.aten.permute %142, %5762 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_3478 = torch.constant.int 1
%int0_3479 = torch.constant.int 0
%5764 = torch.prim.ListConstruct %int1_3478, %int0_3479 : (!torch.int, !torch.int) -> !torch.list<int>
%5765 = torch.aten.permute %143, %5764 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_3480 = torch.constant.int 4
%5766 = torch.aten.mul.int %int4_3480, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3481 = torch.constant.int 4096
%5767 = torch.prim.ListConstruct %5766, %int4096_3481 : (!torch.int, !torch.int) -> !torch.list<int>
%5768 = torch.aten.view %5670, %5767 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5768, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5769 = torch.aten.mm %5768, %5751 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5769, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3482 = torch.constant.int 4
%int1792_3483 = torch.constant.int 1792
%5770 = torch.prim.ListConstruct %int4_3482, %2482, %int1792_3483 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5771 = torch.aten.view %5769, %5770 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3484 = torch.constant.int 4
%5772 = torch.aten.mul.int %int4_3484, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3485 = torch.constant.int 4096
%5773 = torch.prim.ListConstruct %5772, %int4096_3485 : (!torch.int, !torch.int) -> !torch.list<int>
%5774 = torch.aten.view %5671, %5773 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5774, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5775 = torch.aten.mm %5774, %5753 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5775, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3486 = torch.constant.int 4
%int1792_3487 = torch.constant.int 1792
%5776 = torch.prim.ListConstruct %int4_3486, %2482, %int1792_3487 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5777 = torch.aten.view %5775, %5776 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3488 = torch.constant.int 4
%5778 = torch.aten.mul.int %int4_3488, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3489 = torch.constant.int 4096
%5779 = torch.prim.ListConstruct %5778, %int4096_3489 : (!torch.int, !torch.int) -> !torch.list<int>
%5780 = torch.aten.view %5672, %5779 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5780, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5781 = torch.aten.mm %5780, %5755 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5781, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3490 = torch.constant.int 4
%int1792_3491 = torch.constant.int 1792
%5782 = torch.prim.ListConstruct %int4_3490, %2482, %int1792_3491 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5783 = torch.aten.view %5781, %5782 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3492 = torch.constant.int 4
%5784 = torch.aten.mul.int %int4_3492, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3493 = torch.constant.int 4096
%5785 = torch.prim.ListConstruct %5784, %int4096_3493 : (!torch.int, !torch.int) -> !torch.list<int>
%5786 = torch.aten.view %5673, %5785 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5786, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5787 = torch.aten.mm %5786, %5757 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5787, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3494 = torch.constant.int 4
%int1792_3495 = torch.constant.int 1792
%5788 = torch.prim.ListConstruct %int4_3494, %2482, %int1792_3495 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5789 = torch.aten.view %5787, %5788 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3496 = torch.constant.int 4
%5790 = torch.aten.mul.int %int4_3496, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3497 = torch.constant.int 4096
%5791 = torch.prim.ListConstruct %5790, %int4096_3497 : (!torch.int, !torch.int) -> !torch.list<int>
%5792 = torch.aten.view %5674, %5791 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5792, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5793 = torch.aten.mm %5792, %5759 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5793, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3498 = torch.constant.int 4
%int1792_3499 = torch.constant.int 1792
%5794 = torch.prim.ListConstruct %int4_3498, %2482, %int1792_3499 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5795 = torch.aten.view %5793, %5794 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3500 = torch.constant.int 4
%5796 = torch.aten.mul.int %int4_3500, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3501 = torch.constant.int 4096
%5797 = torch.prim.ListConstruct %5796, %int4096_3501 : (!torch.int, !torch.int) -> !torch.list<int>
%5798 = torch.aten.view %5675, %5797 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5798, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5799 = torch.aten.mm %5798, %5761 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5799, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3502 = torch.constant.int 4
%int1792_3503 = torch.constant.int 1792
%5800 = torch.prim.ListConstruct %int4_3502, %2482, %int1792_3503 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5801 = torch.aten.view %5799, %5800 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3504 = torch.constant.int 4
%5802 = torch.aten.mul.int %int4_3504, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3505 = torch.constant.int 4096
%5803 = torch.prim.ListConstruct %5802, %int4096_3505 : (!torch.int, !torch.int) -> !torch.list<int>
%5804 = torch.aten.view %5676, %5803 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5804, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5805 = torch.aten.mm %5804, %5763 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5805, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3506 = torch.constant.int 4
%int1792_3507 = torch.constant.int 1792
%5806 = torch.prim.ListConstruct %int4_3506, %2482, %int1792_3507 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5807 = torch.aten.view %5805, %5806 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_3508 = torch.constant.int 4
%5808 = torch.aten.mul.int %int4_3508, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3509 = torch.constant.int 4096
%5809 = torch.prim.ListConstruct %5808, %int4096_3509 : (!torch.int, !torch.int) -> !torch.list<int>
%5810 = torch.aten.view %5677, %5809 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5810, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%5811 = torch.aten.mm %5810, %5765 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5811, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_3510 = torch.constant.int 4
%int1792_3511 = torch.constant.int 1792
%5812 = torch.prim.ListConstruct %int4_3510, %2482, %int1792_3511 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5813 = torch.aten.view %5811, %5812 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5814 = torch.aten.mul.Tensor %5742, %5771 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5815 = torch.aten.mul.Tensor %5743, %5777 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5816 = torch.aten.mul.Tensor %5744, %5783 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5817 = torch.aten.mul.Tensor %5745, %5789 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5818 = torch.aten.mul.Tensor %5746, %5795 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5819 = torch.aten.mul.Tensor %5747, %5801 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5820 = torch.aten.mul.Tensor %5748, %5807 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%5821 = torch.aten.mul.Tensor %5749, %5813 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %5821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_3512 = torch.constant.int 1
%int0_3513 = torch.constant.int 0
%5822 = torch.prim.ListConstruct %int1_3512, %int0_3513 : (!torch.int, !torch.int) -> !torch.list<int>
%5823 = torch.aten.permute %144, %5822 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3514 = torch.constant.int 1
%int0_3515 = torch.constant.int 0
%5824 = torch.prim.ListConstruct %int1_3514, %int0_3515 : (!torch.int, !torch.int) -> !torch.list<int>
%5825 = torch.aten.permute %145, %5824 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3516 = torch.constant.int 1
%int0_3517 = torch.constant.int 0
%5826 = torch.prim.ListConstruct %int1_3516, %int0_3517 : (!torch.int, !torch.int) -> !torch.list<int>
%5827 = torch.aten.permute %146, %5826 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3518 = torch.constant.int 1
%int0_3519 = torch.constant.int 0
%5828 = torch.prim.ListConstruct %int1_3518, %int0_3519 : (!torch.int, !torch.int) -> !torch.list<int>
%5829 = torch.aten.permute %147, %5828 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3520 = torch.constant.int 1
%int0_3521 = torch.constant.int 0
%5830 = torch.prim.ListConstruct %int1_3520, %int0_3521 : (!torch.int, !torch.int) -> !torch.list<int>
%5831 = torch.aten.permute %148, %5830 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3522 = torch.constant.int 1
%int0_3523 = torch.constant.int 0
%5832 = torch.prim.ListConstruct %int1_3522, %int0_3523 : (!torch.int, !torch.int) -> !torch.list<int>
%5833 = torch.aten.permute %149, %5832 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3524 = torch.constant.int 1
%int0_3525 = torch.constant.int 0
%5834 = torch.prim.ListConstruct %int1_3524, %int0_3525 : (!torch.int, !torch.int) -> !torch.list<int>
%5835 = torch.aten.permute %150, %5834 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3526 = torch.constant.int 1
%int0_3527 = torch.constant.int 0
%5836 = torch.prim.ListConstruct %int1_3526, %int0_3527 : (!torch.int, !torch.int) -> !torch.list<int>
%5837 = torch.aten.permute %151, %5836 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_3528 = torch.constant.int 1
%5838 = torch.aten.size.int %5699, %int1_3528 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3529 = torch.constant.int 4
%5839 = torch.aten.mul.int %int4_3529, %5838 : !torch.int, !torch.int -> !torch.int
%int1792_3530 = torch.constant.int 1792
%5840 = torch.prim.ListConstruct %5839, %int1792_3530 : (!torch.int, !torch.int) -> !torch.list<int>
%5841 = torch.aten.view %5814, %5840 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5841, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5842 = torch.aten.mm %5841, %5823 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5842, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3531 = torch.constant.int 4
%int4096_3532 = torch.constant.int 4096
%5843 = torch.prim.ListConstruct %int4_3531, %5838, %int4096_3532 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5844 = torch.aten.view %5842, %5843 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3533 = torch.constant.int 1
%5845 = torch.aten.size.int %5705, %int1_3533 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3534 = torch.constant.int 4
%5846 = torch.aten.mul.int %int4_3534, %5845 : !torch.int, !torch.int -> !torch.int
%int1792_3535 = torch.constant.int 1792
%5847 = torch.prim.ListConstruct %5846, %int1792_3535 : (!torch.int, !torch.int) -> !torch.list<int>
%5848 = torch.aten.view %5815, %5847 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5848, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5849 = torch.aten.mm %5848, %5825 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5849, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3536 = torch.constant.int 4
%int4096_3537 = torch.constant.int 4096
%5850 = torch.prim.ListConstruct %int4_3536, %5845, %int4096_3537 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5851 = torch.aten.view %5849, %5850 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3538 = torch.constant.int 1
%5852 = torch.aten.size.int %5711, %int1_3538 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3539 = torch.constant.int 4
%5853 = torch.aten.mul.int %int4_3539, %5852 : !torch.int, !torch.int -> !torch.int
%int1792_3540 = torch.constant.int 1792
%5854 = torch.prim.ListConstruct %5853, %int1792_3540 : (!torch.int, !torch.int) -> !torch.list<int>
%5855 = torch.aten.view %5816, %5854 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5855, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5856 = torch.aten.mm %5855, %5827 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5856, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3541 = torch.constant.int 4
%int4096_3542 = torch.constant.int 4096
%5857 = torch.prim.ListConstruct %int4_3541, %5852, %int4096_3542 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5858 = torch.aten.view %5856, %5857 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3543 = torch.constant.int 1
%5859 = torch.aten.size.int %5717, %int1_3543 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3544 = torch.constant.int 4
%5860 = torch.aten.mul.int %int4_3544, %5859 : !torch.int, !torch.int -> !torch.int
%int1792_3545 = torch.constant.int 1792
%5861 = torch.prim.ListConstruct %5860, %int1792_3545 : (!torch.int, !torch.int) -> !torch.list<int>
%5862 = torch.aten.view %5817, %5861 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5862, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5863 = torch.aten.mm %5862, %5829 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5863, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3546 = torch.constant.int 4
%int4096_3547 = torch.constant.int 4096
%5864 = torch.prim.ListConstruct %int4_3546, %5859, %int4096_3547 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5865 = torch.aten.view %5863, %5864 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3548 = torch.constant.int 1
%5866 = torch.aten.size.int %5723, %int1_3548 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3549 = torch.constant.int 4
%5867 = torch.aten.mul.int %int4_3549, %5866 : !torch.int, !torch.int -> !torch.int
%int1792_3550 = torch.constant.int 1792
%5868 = torch.prim.ListConstruct %5867, %int1792_3550 : (!torch.int, !torch.int) -> !torch.list<int>
%5869 = torch.aten.view %5818, %5868 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5869, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5870 = torch.aten.mm %5869, %5831 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5870, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3551 = torch.constant.int 4
%int4096_3552 = torch.constant.int 4096
%5871 = torch.prim.ListConstruct %int4_3551, %5866, %int4096_3552 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5872 = torch.aten.view %5870, %5871 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3553 = torch.constant.int 1
%5873 = torch.aten.size.int %5729, %int1_3553 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3554 = torch.constant.int 4
%5874 = torch.aten.mul.int %int4_3554, %5873 : !torch.int, !torch.int -> !torch.int
%int1792_3555 = torch.constant.int 1792
%5875 = torch.prim.ListConstruct %5874, %int1792_3555 : (!torch.int, !torch.int) -> !torch.list<int>
%5876 = torch.aten.view %5819, %5875 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5876, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5877 = torch.aten.mm %5876, %5833 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5877, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3556 = torch.constant.int 4
%int4096_3557 = torch.constant.int 4096
%5878 = torch.prim.ListConstruct %int4_3556, %5873, %int4096_3557 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5879 = torch.aten.view %5877, %5878 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3558 = torch.constant.int 1
%5880 = torch.aten.size.int %5735, %int1_3558 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3559 = torch.constant.int 4
%5881 = torch.aten.mul.int %int4_3559, %5880 : !torch.int, !torch.int -> !torch.int
%int1792_3560 = torch.constant.int 1792
%5882 = torch.prim.ListConstruct %5881, %int1792_3560 : (!torch.int, !torch.int) -> !torch.list<int>
%5883 = torch.aten.view %5820, %5882 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5883, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5884 = torch.aten.mm %5883, %5835 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5884, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3561 = torch.constant.int 4
%int4096_3562 = torch.constant.int 4096
%5885 = torch.prim.ListConstruct %int4_3561, %5880, %int4096_3562 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5886 = torch.aten.view %5884, %5885 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3563 = torch.constant.int 1
%5887 = torch.aten.size.int %5741, %int1_3563 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_3564 = torch.constant.int 4
%5888 = torch.aten.mul.int %int4_3564, %5887 : !torch.int, !torch.int -> !torch.int
%int1792_3565 = torch.constant.int 1792
%5889 = torch.prim.ListConstruct %5888, %int1792_3565 : (!torch.int, !torch.int) -> !torch.list<int>
%5890 = torch.aten.view %5821, %5889 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %5890, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%5891 = torch.aten.mm %5890, %5837 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %5891, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_3566 = torch.constant.int 4
%int4096_3567 = torch.constant.int 4096
%5892 = torch.prim.ListConstruct %int4_3566, %5887, %int4096_3567 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%5893 = torch.aten.view %5891, %5892 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5894 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3568 = arith.constant 1 : index
%dim_3569 = tensor.dim %5894, %c1_3568 : tensor<4x?x4096xf16>
%5895 = flow.tensor.transfer %5894 : tensor<4x?x4096xf16>{%dim_3569} to #hal.device.promise<@__device_0>
%5896 = torch_c.from_builtin_tensor %5895 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5897 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3570 = arith.constant 1 : index
%dim_3571 = tensor.dim %5897, %c1_3570 : tensor<4x?x4096xf16>
%5898 = flow.tensor.transfer %5897 : tensor<4x?x4096xf16>{%dim_3571} to #hal.device.promise<@__device_0>
%5899 = torch_c.from_builtin_tensor %5898 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5900 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3572 = arith.constant 1 : index
%dim_3573 = tensor.dim %5900, %c1_3572 : tensor<4x?x4096xf16>
%5901 = flow.tensor.transfer %5900 : tensor<4x?x4096xf16>{%dim_3573} to #hal.device.promise<@__device_0>
%5902 = torch_c.from_builtin_tensor %5901 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5903 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3574 = arith.constant 1 : index
%dim_3575 = tensor.dim %5903, %c1_3574 : tensor<4x?x4096xf16>
%5904 = flow.tensor.transfer %5903 : tensor<4x?x4096xf16>{%dim_3575} to #hal.device.promise<@__device_0>
%5905 = torch_c.from_builtin_tensor %5904 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5906 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3576 = arith.constant 1 : index
%dim_3577 = tensor.dim %5906, %c1_3576 : tensor<4x?x4096xf16>
%5907 = flow.tensor.transfer %5906 : tensor<4x?x4096xf16>{%dim_3577} to #hal.device.promise<@__device_0>
%5908 = torch_c.from_builtin_tensor %5907 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5909 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3578 = arith.constant 1 : index
%dim_3579 = tensor.dim %5909, %c1_3578 : tensor<4x?x4096xf16>
%5910 = flow.tensor.transfer %5909 : tensor<4x?x4096xf16>{%dim_3579} to #hal.device.promise<@__device_0>
%5911 = torch_c.from_builtin_tensor %5910 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5912 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3580 = arith.constant 1 : index
%dim_3581 = tensor.dim %5912, %c1_3580 : tensor<4x?x4096xf16>
%5913 = flow.tensor.transfer %5912 : tensor<4x?x4096xf16>{%dim_3581} to #hal.device.promise<@__device_0>
%5914 = torch_c.from_builtin_tensor %5913 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3582 = torch.constant.int 1
%5915 = torch.aten.add.Tensor %5844, %5896, %int1_3582 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3583 = torch.constant.int 1
%5916 = torch.aten.add.Tensor %5915, %5899, %int1_3583 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3584 = torch.constant.int 1
%5917 = torch.aten.add.Tensor %5916, %5902, %int1_3584 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3585 = torch.constant.int 1
%5918 = torch.aten.add.Tensor %5917, %5905, %int1_3585 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3586 = torch.constant.int 1
%5919 = torch.aten.add.Tensor %5918, %5908, %int1_3586 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3587 = torch.constant.int 1
%5920 = torch.aten.add.Tensor %5919, %5911, %int1_3587 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3588 = torch.constant.int 1
%5921 = torch.aten.add.Tensor %5920, %5914, %int1_3588 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5922 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3589 = arith.constant 1 : index
%dim_3590 = tensor.dim %5922, %c1_3589 : tensor<4x?x4096xf16>
%5923 = flow.tensor.transfer %5922 : tensor<4x?x4096xf16>{%dim_3590} to #hal.device.promise<@__device_1>
%5924 = torch_c.from_builtin_tensor %5923 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5925 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3591 = arith.constant 1 : index
%dim_3592 = tensor.dim %5925, %c1_3591 : tensor<4x?x4096xf16>
%5926 = flow.tensor.transfer %5925 : tensor<4x?x4096xf16>{%dim_3592} to #hal.device.promise<@__device_1>
%5927 = torch_c.from_builtin_tensor %5926 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5928 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3593 = arith.constant 1 : index
%dim_3594 = tensor.dim %5928, %c1_3593 : tensor<4x?x4096xf16>
%5929 = flow.tensor.transfer %5928 : tensor<4x?x4096xf16>{%dim_3594} to #hal.device.promise<@__device_1>
%5930 = torch_c.from_builtin_tensor %5929 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5931 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3595 = arith.constant 1 : index
%dim_3596 = tensor.dim %5931, %c1_3595 : tensor<4x?x4096xf16>
%5932 = flow.tensor.transfer %5931 : tensor<4x?x4096xf16>{%dim_3596} to #hal.device.promise<@__device_1>
%5933 = torch_c.from_builtin_tensor %5932 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5934 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3597 = arith.constant 1 : index
%dim_3598 = tensor.dim %5934, %c1_3597 : tensor<4x?x4096xf16>
%5935 = flow.tensor.transfer %5934 : tensor<4x?x4096xf16>{%dim_3598} to #hal.device.promise<@__device_1>
%5936 = torch_c.from_builtin_tensor %5935 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5937 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3599 = arith.constant 1 : index
%dim_3600 = tensor.dim %5937, %c1_3599 : tensor<4x?x4096xf16>
%5938 = flow.tensor.transfer %5937 : tensor<4x?x4096xf16>{%dim_3600} to #hal.device.promise<@__device_1>
%5939 = torch_c.from_builtin_tensor %5938 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5940 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3601 = arith.constant 1 : index
%dim_3602 = tensor.dim %5940, %c1_3601 : tensor<4x?x4096xf16>
%5941 = flow.tensor.transfer %5940 : tensor<4x?x4096xf16>{%dim_3602} to #hal.device.promise<@__device_1>
%5942 = torch_c.from_builtin_tensor %5941 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3603 = torch.constant.int 1
%5943 = torch.aten.add.Tensor %5924, %5851, %int1_3603 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3604 = torch.constant.int 1
%5944 = torch.aten.add.Tensor %5943, %5927, %int1_3604 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3605 = torch.constant.int 1
%5945 = torch.aten.add.Tensor %5944, %5930, %int1_3605 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3606 = torch.constant.int 1
%5946 = torch.aten.add.Tensor %5945, %5933, %int1_3606 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3607 = torch.constant.int 1
%5947 = torch.aten.add.Tensor %5946, %5936, %int1_3607 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3608 = torch.constant.int 1
%5948 = torch.aten.add.Tensor %5947, %5939, %int1_3608 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3609 = torch.constant.int 1
%5949 = torch.aten.add.Tensor %5948, %5942, %int1_3609 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5950 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3610 = arith.constant 1 : index
%dim_3611 = tensor.dim %5950, %c1_3610 : tensor<4x?x4096xf16>
%5951 = flow.tensor.transfer %5950 : tensor<4x?x4096xf16>{%dim_3611} to #hal.device.promise<@__device_2>
%5952 = torch_c.from_builtin_tensor %5951 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5953 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3612 = arith.constant 1 : index
%dim_3613 = tensor.dim %5953, %c1_3612 : tensor<4x?x4096xf16>
%5954 = flow.tensor.transfer %5953 : tensor<4x?x4096xf16>{%dim_3613} to #hal.device.promise<@__device_2>
%5955 = torch_c.from_builtin_tensor %5954 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5956 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3614 = arith.constant 1 : index
%dim_3615 = tensor.dim %5956, %c1_3614 : tensor<4x?x4096xf16>
%5957 = flow.tensor.transfer %5956 : tensor<4x?x4096xf16>{%dim_3615} to #hal.device.promise<@__device_2>
%5958 = torch_c.from_builtin_tensor %5957 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5959 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3616 = arith.constant 1 : index
%dim_3617 = tensor.dim %5959, %c1_3616 : tensor<4x?x4096xf16>
%5960 = flow.tensor.transfer %5959 : tensor<4x?x4096xf16>{%dim_3617} to #hal.device.promise<@__device_2>
%5961 = torch_c.from_builtin_tensor %5960 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5962 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3618 = arith.constant 1 : index
%dim_3619 = tensor.dim %5962, %c1_3618 : tensor<4x?x4096xf16>
%5963 = flow.tensor.transfer %5962 : tensor<4x?x4096xf16>{%dim_3619} to #hal.device.promise<@__device_2>
%5964 = torch_c.from_builtin_tensor %5963 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5965 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3620 = arith.constant 1 : index
%dim_3621 = tensor.dim %5965, %c1_3620 : tensor<4x?x4096xf16>
%5966 = flow.tensor.transfer %5965 : tensor<4x?x4096xf16>{%dim_3621} to #hal.device.promise<@__device_2>
%5967 = torch_c.from_builtin_tensor %5966 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5968 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3622 = arith.constant 1 : index
%dim_3623 = tensor.dim %5968, %c1_3622 : tensor<4x?x4096xf16>
%5969 = flow.tensor.transfer %5968 : tensor<4x?x4096xf16>{%dim_3623} to #hal.device.promise<@__device_2>
%5970 = torch_c.from_builtin_tensor %5969 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3624 = torch.constant.int 1
%5971 = torch.aten.add.Tensor %5952, %5955, %int1_3624 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3625 = torch.constant.int 1
%5972 = torch.aten.add.Tensor %5971, %5858, %int1_3625 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3626 = torch.constant.int 1
%5973 = torch.aten.add.Tensor %5972, %5958, %int1_3626 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3627 = torch.constant.int 1
%5974 = torch.aten.add.Tensor %5973, %5961, %int1_3627 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3628 = torch.constant.int 1
%5975 = torch.aten.add.Tensor %5974, %5964, %int1_3628 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3629 = torch.constant.int 1
%5976 = torch.aten.add.Tensor %5975, %5967, %int1_3629 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3630 = torch.constant.int 1
%5977 = torch.aten.add.Tensor %5976, %5970, %int1_3630 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5978 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3631 = arith.constant 1 : index
%dim_3632 = tensor.dim %5978, %c1_3631 : tensor<4x?x4096xf16>
%5979 = flow.tensor.transfer %5978 : tensor<4x?x4096xf16>{%dim_3632} to #hal.device.promise<@__device_3>
%5980 = torch_c.from_builtin_tensor %5979 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5981 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3633 = arith.constant 1 : index
%dim_3634 = tensor.dim %5981, %c1_3633 : tensor<4x?x4096xf16>
%5982 = flow.tensor.transfer %5981 : tensor<4x?x4096xf16>{%dim_3634} to #hal.device.promise<@__device_3>
%5983 = torch_c.from_builtin_tensor %5982 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5984 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3635 = arith.constant 1 : index
%dim_3636 = tensor.dim %5984, %c1_3635 : tensor<4x?x4096xf16>
%5985 = flow.tensor.transfer %5984 : tensor<4x?x4096xf16>{%dim_3636} to #hal.device.promise<@__device_3>
%5986 = torch_c.from_builtin_tensor %5985 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5987 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3637 = arith.constant 1 : index
%dim_3638 = tensor.dim %5987, %c1_3637 : tensor<4x?x4096xf16>
%5988 = flow.tensor.transfer %5987 : tensor<4x?x4096xf16>{%dim_3638} to #hal.device.promise<@__device_3>
%5989 = torch_c.from_builtin_tensor %5988 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5990 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3639 = arith.constant 1 : index
%dim_3640 = tensor.dim %5990, %c1_3639 : tensor<4x?x4096xf16>
%5991 = flow.tensor.transfer %5990 : tensor<4x?x4096xf16>{%dim_3640} to #hal.device.promise<@__device_3>
%5992 = torch_c.from_builtin_tensor %5991 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5993 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3641 = arith.constant 1 : index
%dim_3642 = tensor.dim %5993, %c1_3641 : tensor<4x?x4096xf16>
%5994 = flow.tensor.transfer %5993 : tensor<4x?x4096xf16>{%dim_3642} to #hal.device.promise<@__device_3>
%5995 = torch_c.from_builtin_tensor %5994 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%5996 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3643 = arith.constant 1 : index
%dim_3644 = tensor.dim %5996, %c1_3643 : tensor<4x?x4096xf16>
%5997 = flow.tensor.transfer %5996 : tensor<4x?x4096xf16>{%dim_3644} to #hal.device.promise<@__device_3>
%5998 = torch_c.from_builtin_tensor %5997 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3645 = torch.constant.int 1
%5999 = torch.aten.add.Tensor %5980, %5983, %int1_3645 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %5999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3646 = torch.constant.int 1
%6000 = torch.aten.add.Tensor %5999, %5986, %int1_3646 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3647 = torch.constant.int 1
%6001 = torch.aten.add.Tensor %6000, %5865, %int1_3647 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3648 = torch.constant.int 1
%6002 = torch.aten.add.Tensor %6001, %5989, %int1_3648 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3649 = torch.constant.int 1
%6003 = torch.aten.add.Tensor %6002, %5992, %int1_3649 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3650 = torch.constant.int 1
%6004 = torch.aten.add.Tensor %6003, %5995, %int1_3650 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3651 = torch.constant.int 1
%6005 = torch.aten.add.Tensor %6004, %5998, %int1_3651 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6006 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3652 = arith.constant 1 : index
%dim_3653 = tensor.dim %6006, %c1_3652 : tensor<4x?x4096xf16>
%6007 = flow.tensor.transfer %6006 : tensor<4x?x4096xf16>{%dim_3653} to #hal.device.promise<@__device_4>
%6008 = torch_c.from_builtin_tensor %6007 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6009 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3654 = arith.constant 1 : index
%dim_3655 = tensor.dim %6009, %c1_3654 : tensor<4x?x4096xf16>
%6010 = flow.tensor.transfer %6009 : tensor<4x?x4096xf16>{%dim_3655} to #hal.device.promise<@__device_4>
%6011 = torch_c.from_builtin_tensor %6010 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6012 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3656 = arith.constant 1 : index
%dim_3657 = tensor.dim %6012, %c1_3656 : tensor<4x?x4096xf16>
%6013 = flow.tensor.transfer %6012 : tensor<4x?x4096xf16>{%dim_3657} to #hal.device.promise<@__device_4>
%6014 = torch_c.from_builtin_tensor %6013 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6015 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3658 = arith.constant 1 : index
%dim_3659 = tensor.dim %6015, %c1_3658 : tensor<4x?x4096xf16>
%6016 = flow.tensor.transfer %6015 : tensor<4x?x4096xf16>{%dim_3659} to #hal.device.promise<@__device_4>
%6017 = torch_c.from_builtin_tensor %6016 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6018 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3660 = arith.constant 1 : index
%dim_3661 = tensor.dim %6018, %c1_3660 : tensor<4x?x4096xf16>
%6019 = flow.tensor.transfer %6018 : tensor<4x?x4096xf16>{%dim_3661} to #hal.device.promise<@__device_4>
%6020 = torch_c.from_builtin_tensor %6019 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6021 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3662 = arith.constant 1 : index
%dim_3663 = tensor.dim %6021, %c1_3662 : tensor<4x?x4096xf16>
%6022 = flow.tensor.transfer %6021 : tensor<4x?x4096xf16>{%dim_3663} to #hal.device.promise<@__device_4>
%6023 = torch_c.from_builtin_tensor %6022 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6024 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3664 = arith.constant 1 : index
%dim_3665 = tensor.dim %6024, %c1_3664 : tensor<4x?x4096xf16>
%6025 = flow.tensor.transfer %6024 : tensor<4x?x4096xf16>{%dim_3665} to #hal.device.promise<@__device_4>
%6026 = torch_c.from_builtin_tensor %6025 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3666 = torch.constant.int 1
%6027 = torch.aten.add.Tensor %6008, %6011, %int1_3666 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3667 = torch.constant.int 1
%6028 = torch.aten.add.Tensor %6027, %6014, %int1_3667 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3668 = torch.constant.int 1
%6029 = torch.aten.add.Tensor %6028, %6017, %int1_3668 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3669 = torch.constant.int 1
%6030 = torch.aten.add.Tensor %6029, %5872, %int1_3669 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3670 = torch.constant.int 1
%6031 = torch.aten.add.Tensor %6030, %6020, %int1_3670 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3671 = torch.constant.int 1
%6032 = torch.aten.add.Tensor %6031, %6023, %int1_3671 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3672 = torch.constant.int 1
%6033 = torch.aten.add.Tensor %6032, %6026, %int1_3672 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6034 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3673 = arith.constant 1 : index
%dim_3674 = tensor.dim %6034, %c1_3673 : tensor<4x?x4096xf16>
%6035 = flow.tensor.transfer %6034 : tensor<4x?x4096xf16>{%dim_3674} to #hal.device.promise<@__device_5>
%6036 = torch_c.from_builtin_tensor %6035 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6037 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3675 = arith.constant 1 : index
%dim_3676 = tensor.dim %6037, %c1_3675 : tensor<4x?x4096xf16>
%6038 = flow.tensor.transfer %6037 : tensor<4x?x4096xf16>{%dim_3676} to #hal.device.promise<@__device_5>
%6039 = torch_c.from_builtin_tensor %6038 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6040 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3677 = arith.constant 1 : index
%dim_3678 = tensor.dim %6040, %c1_3677 : tensor<4x?x4096xf16>
%6041 = flow.tensor.transfer %6040 : tensor<4x?x4096xf16>{%dim_3678} to #hal.device.promise<@__device_5>
%6042 = torch_c.from_builtin_tensor %6041 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6043 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3679 = arith.constant 1 : index
%dim_3680 = tensor.dim %6043, %c1_3679 : tensor<4x?x4096xf16>
%6044 = flow.tensor.transfer %6043 : tensor<4x?x4096xf16>{%dim_3680} to #hal.device.promise<@__device_5>
%6045 = torch_c.from_builtin_tensor %6044 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6046 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3681 = arith.constant 1 : index
%dim_3682 = tensor.dim %6046, %c1_3681 : tensor<4x?x4096xf16>
%6047 = flow.tensor.transfer %6046 : tensor<4x?x4096xf16>{%dim_3682} to #hal.device.promise<@__device_5>
%6048 = torch_c.from_builtin_tensor %6047 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6049 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3683 = arith.constant 1 : index
%dim_3684 = tensor.dim %6049, %c1_3683 : tensor<4x?x4096xf16>
%6050 = flow.tensor.transfer %6049 : tensor<4x?x4096xf16>{%dim_3684} to #hal.device.promise<@__device_5>
%6051 = torch_c.from_builtin_tensor %6050 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6052 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3685 = arith.constant 1 : index
%dim_3686 = tensor.dim %6052, %c1_3685 : tensor<4x?x4096xf16>
%6053 = flow.tensor.transfer %6052 : tensor<4x?x4096xf16>{%dim_3686} to #hal.device.promise<@__device_5>
%6054 = torch_c.from_builtin_tensor %6053 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3687 = torch.constant.int 1
%6055 = torch.aten.add.Tensor %6036, %6039, %int1_3687 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3688 = torch.constant.int 1
%6056 = torch.aten.add.Tensor %6055, %6042, %int1_3688 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3689 = torch.constant.int 1
%6057 = torch.aten.add.Tensor %6056, %6045, %int1_3689 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3690 = torch.constant.int 1
%6058 = torch.aten.add.Tensor %6057, %6048, %int1_3690 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3691 = torch.constant.int 1
%6059 = torch.aten.add.Tensor %6058, %5879, %int1_3691 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3692 = torch.constant.int 1
%6060 = torch.aten.add.Tensor %6059, %6051, %int1_3692 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3693 = torch.constant.int 1
%6061 = torch.aten.add.Tensor %6060, %6054, %int1_3693 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6062 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3694 = arith.constant 1 : index
%dim_3695 = tensor.dim %6062, %c1_3694 : tensor<4x?x4096xf16>
%6063 = flow.tensor.transfer %6062 : tensor<4x?x4096xf16>{%dim_3695} to #hal.device.promise<@__device_6>
%6064 = torch_c.from_builtin_tensor %6063 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6065 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3696 = arith.constant 1 : index
%dim_3697 = tensor.dim %6065, %c1_3696 : tensor<4x?x4096xf16>
%6066 = flow.tensor.transfer %6065 : tensor<4x?x4096xf16>{%dim_3697} to #hal.device.promise<@__device_6>
%6067 = torch_c.from_builtin_tensor %6066 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6068 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3698 = arith.constant 1 : index
%dim_3699 = tensor.dim %6068, %c1_3698 : tensor<4x?x4096xf16>
%6069 = flow.tensor.transfer %6068 : tensor<4x?x4096xf16>{%dim_3699} to #hal.device.promise<@__device_6>
%6070 = torch_c.from_builtin_tensor %6069 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6071 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3700 = arith.constant 1 : index
%dim_3701 = tensor.dim %6071, %c1_3700 : tensor<4x?x4096xf16>
%6072 = flow.tensor.transfer %6071 : tensor<4x?x4096xf16>{%dim_3701} to #hal.device.promise<@__device_6>
%6073 = torch_c.from_builtin_tensor %6072 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6074 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3702 = arith.constant 1 : index
%dim_3703 = tensor.dim %6074, %c1_3702 : tensor<4x?x4096xf16>
%6075 = flow.tensor.transfer %6074 : tensor<4x?x4096xf16>{%dim_3703} to #hal.device.promise<@__device_6>
%6076 = torch_c.from_builtin_tensor %6075 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6077 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3704 = arith.constant 1 : index
%dim_3705 = tensor.dim %6077, %c1_3704 : tensor<4x?x4096xf16>
%6078 = flow.tensor.transfer %6077 : tensor<4x?x4096xf16>{%dim_3705} to #hal.device.promise<@__device_6>
%6079 = torch_c.from_builtin_tensor %6078 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6080 = torch_c.to_builtin_tensor %5893 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3706 = arith.constant 1 : index
%dim_3707 = tensor.dim %6080, %c1_3706 : tensor<4x?x4096xf16>
%6081 = flow.tensor.transfer %6080 : tensor<4x?x4096xf16>{%dim_3707} to #hal.device.promise<@__device_6>
%6082 = torch_c.from_builtin_tensor %6081 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3708 = torch.constant.int 1
%6083 = torch.aten.add.Tensor %6064, %6067, %int1_3708 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3709 = torch.constant.int 1
%6084 = torch.aten.add.Tensor %6083, %6070, %int1_3709 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3710 = torch.constant.int 1
%6085 = torch.aten.add.Tensor %6084, %6073, %int1_3710 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3711 = torch.constant.int 1
%6086 = torch.aten.add.Tensor %6085, %6076, %int1_3711 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3712 = torch.constant.int 1
%6087 = torch.aten.add.Tensor %6086, %6079, %int1_3712 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3713 = torch.constant.int 1
%6088 = torch.aten.add.Tensor %6087, %5886, %int1_3713 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3714 = torch.constant.int 1
%6089 = torch.aten.add.Tensor %6088, %6082, %int1_3714 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6090 = torch_c.to_builtin_tensor %5844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3715 = arith.constant 1 : index
%dim_3716 = tensor.dim %6090, %c1_3715 : tensor<4x?x4096xf16>
%6091 = flow.tensor.transfer %6090 : tensor<4x?x4096xf16>{%dim_3716} to #hal.device.promise<@__device_7>
%6092 = torch_c.from_builtin_tensor %6091 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6093 = torch_c.to_builtin_tensor %5851 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3717 = arith.constant 1 : index
%dim_3718 = tensor.dim %6093, %c1_3717 : tensor<4x?x4096xf16>
%6094 = flow.tensor.transfer %6093 : tensor<4x?x4096xf16>{%dim_3718} to #hal.device.promise<@__device_7>
%6095 = torch_c.from_builtin_tensor %6094 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6096 = torch_c.to_builtin_tensor %5858 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3719 = arith.constant 1 : index
%dim_3720 = tensor.dim %6096, %c1_3719 : tensor<4x?x4096xf16>
%6097 = flow.tensor.transfer %6096 : tensor<4x?x4096xf16>{%dim_3720} to #hal.device.promise<@__device_7>
%6098 = torch_c.from_builtin_tensor %6097 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6099 = torch_c.to_builtin_tensor %5865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3721 = arith.constant 1 : index
%dim_3722 = tensor.dim %6099, %c1_3721 : tensor<4x?x4096xf16>
%6100 = flow.tensor.transfer %6099 : tensor<4x?x4096xf16>{%dim_3722} to #hal.device.promise<@__device_7>
%6101 = torch_c.from_builtin_tensor %6100 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6102 = torch_c.to_builtin_tensor %5872 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3723 = arith.constant 1 : index
%dim_3724 = tensor.dim %6102, %c1_3723 : tensor<4x?x4096xf16>
%6103 = flow.tensor.transfer %6102 : tensor<4x?x4096xf16>{%dim_3724} to #hal.device.promise<@__device_7>
%6104 = torch_c.from_builtin_tensor %6103 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6105 = torch_c.to_builtin_tensor %5879 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3725 = arith.constant 1 : index
%dim_3726 = tensor.dim %6105, %c1_3725 : tensor<4x?x4096xf16>
%6106 = flow.tensor.transfer %6105 : tensor<4x?x4096xf16>{%dim_3726} to #hal.device.promise<@__device_7>
%6107 = torch_c.from_builtin_tensor %6106 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%6108 = torch_c.to_builtin_tensor %5886 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_3727 = arith.constant 1 : index
%dim_3728 = tensor.dim %6108, %c1_3727 : tensor<4x?x4096xf16>
%6109 = flow.tensor.transfer %6108 : tensor<4x?x4096xf16>{%dim_3728} to #hal.device.promise<@__device_7>
%6110 = torch_c.from_builtin_tensor %6109 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3729 = torch.constant.int 1
%6111 = torch.aten.add.Tensor %6092, %6095, %int1_3729 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3730 = torch.constant.int 1
%6112 = torch.aten.add.Tensor %6111, %6098, %int1_3730 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3731 = torch.constant.int 1
%6113 = torch.aten.add.Tensor %6112, %6101, %int1_3731 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3732 = torch.constant.int 1
%6114 = torch.aten.add.Tensor %6113, %6104, %int1_3732 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3733 = torch.constant.int 1
%6115 = torch.aten.add.Tensor %6114, %6107, %int1_3733 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3734 = torch.constant.int 1
%6116 = torch.aten.add.Tensor %6115, %6110, %int1_3734 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3735 = torch.constant.int 1
%6117 = torch.aten.add.Tensor %6116, %5893, %int1_3735 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3736 = torch.constant.int 1
%6118 = torch.aten.add.Tensor %5598, %5921, %int1_3736 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3737 = torch.constant.int 1
%6119 = torch.aten.add.Tensor %5599, %5949, %int1_3737 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3738 = torch.constant.int 1
%6120 = torch.aten.add.Tensor %5600, %5977, %int1_3738 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3739 = torch.constant.int 1
%6121 = torch.aten.add.Tensor %5601, %6005, %int1_3739 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3740 = torch.constant.int 1
%6122 = torch.aten.add.Tensor %5602, %6033, %int1_3740 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3741 = torch.constant.int 1
%6123 = torch.aten.add.Tensor %5603, %6061, %int1_3741 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3742 = torch.constant.int 1
%6124 = torch.aten.add.Tensor %5604, %6089, %int1_3742 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6124, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3743 = torch.constant.int 1
%6125 = torch.aten.add.Tensor %5605, %6117, %int1_3743 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_3744 = torch.constant.int 6
%6126 = torch.prims.convert_element_type %6118, %int6_3744 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3745 = torch.constant.int 6
%6127 = torch.prims.convert_element_type %6119, %int6_3745 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3746 = torch.constant.int 6
%6128 = torch.prims.convert_element_type %6120, %int6_3746 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3747 = torch.constant.int 6
%6129 = torch.prims.convert_element_type %6121, %int6_3747 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3748 = torch.constant.int 6
%6130 = torch.prims.convert_element_type %6122, %int6_3748 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3749 = torch.constant.int 6
%6131 = torch.prims.convert_element_type %6123, %int6_3749 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3750 = torch.constant.int 6
%6132 = torch.prims.convert_element_type %6124, %int6_3750 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_3751 = torch.constant.int 6
%6133 = torch.prims.convert_element_type %6125, %int6_3751 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3752 = torch.constant.int 2
%6134 = torch.aten.pow.Tensor_Scalar %6126, %int2_3752 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3753 = torch.constant.int 2
%6135 = torch.aten.pow.Tensor_Scalar %6127, %int2_3753 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3754 = torch.constant.int 2
%6136 = torch.aten.pow.Tensor_Scalar %6128, %int2_3754 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3755 = torch.constant.int 2
%6137 = torch.aten.pow.Tensor_Scalar %6129, %int2_3755 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3756 = torch.constant.int 2
%6138 = torch.aten.pow.Tensor_Scalar %6130, %int2_3756 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3757 = torch.constant.int 2
%6139 = torch.aten.pow.Tensor_Scalar %6131, %int2_3757 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3758 = torch.constant.int 2
%6140 = torch.aten.pow.Tensor_Scalar %6132, %int2_3758 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_3759 = torch.constant.int 2
%6141 = torch.aten.pow.Tensor_Scalar %6133, %int2_3759 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_3760 = torch.constant.int -1
%6142 = torch.prim.ListConstruct %int-1_3760 : (!torch.int) -> !torch.list<int>
%true_3761 = torch.constant.bool true
%none_3762 = torch.constant.none
%6143 = torch.aten.mean.dim %6134, %6142, %true_3761, %none_3762 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3763 = torch.constant.int -1
%6144 = torch.prim.ListConstruct %int-1_3763 : (!torch.int) -> !torch.list<int>
%true_3764 = torch.constant.bool true
%none_3765 = torch.constant.none
%6145 = torch.aten.mean.dim %6135, %6144, %true_3764, %none_3765 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3766 = torch.constant.int -1
%6146 = torch.prim.ListConstruct %int-1_3766 : (!torch.int) -> !torch.list<int>
%true_3767 = torch.constant.bool true
%none_3768 = torch.constant.none
%6147 = torch.aten.mean.dim %6136, %6146, %true_3767, %none_3768 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3769 = torch.constant.int -1
%6148 = torch.prim.ListConstruct %int-1_3769 : (!torch.int) -> !torch.list<int>
%true_3770 = torch.constant.bool true
%none_3771 = torch.constant.none
%6149 = torch.aten.mean.dim %6137, %6148, %true_3770, %none_3771 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3772 = torch.constant.int -1
%6150 = torch.prim.ListConstruct %int-1_3772 : (!torch.int) -> !torch.list<int>
%true_3773 = torch.constant.bool true
%none_3774 = torch.constant.none
%6151 = torch.aten.mean.dim %6138, %6150, %true_3773, %none_3774 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3775 = torch.constant.int -1
%6152 = torch.prim.ListConstruct %int-1_3775 : (!torch.int) -> !torch.list<int>
%true_3776 = torch.constant.bool true
%none_3777 = torch.constant.none
%6153 = torch.aten.mean.dim %6139, %6152, %true_3776, %none_3777 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3778 = torch.constant.int -1
%6154 = torch.prim.ListConstruct %int-1_3778 : (!torch.int) -> !torch.list<int>
%true_3779 = torch.constant.bool true
%none_3780 = torch.constant.none
%6155 = torch.aten.mean.dim %6140, %6154, %true_3779, %none_3780 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_3781 = torch.constant.int -1
%6156 = torch.prim.ListConstruct %int-1_3781 : (!torch.int) -> !torch.list<int>
%true_3782 = torch.constant.bool true
%none_3783 = torch.constant.none
%6157 = torch.aten.mean.dim %6141, %6156, %true_3782, %none_3783 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3784 = torch.constant.float 9.9999997473787516E-6
%int1_3785 = torch.constant.int 1
%6158 = torch.aten.add.Scalar %6143, %float9.999990e-06_3784, %int1_3785 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3786 = torch.constant.float 9.9999997473787516E-6
%int1_3787 = torch.constant.int 1
%6159 = torch.aten.add.Scalar %6145, %float9.999990e-06_3786, %int1_3787 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3788 = torch.constant.float 9.9999997473787516E-6
%int1_3789 = torch.constant.int 1
%6160 = torch.aten.add.Scalar %6147, %float9.999990e-06_3788, %int1_3789 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3790 = torch.constant.float 9.9999997473787516E-6
%int1_3791 = torch.constant.int 1
%6161 = torch.aten.add.Scalar %6149, %float9.999990e-06_3790, %int1_3791 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3792 = torch.constant.float 9.9999997473787516E-6
%int1_3793 = torch.constant.int 1
%6162 = torch.aten.add.Scalar %6151, %float9.999990e-06_3792, %int1_3793 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3794 = torch.constant.float 9.9999997473787516E-6
%int1_3795 = torch.constant.int 1
%6163 = torch.aten.add.Scalar %6153, %float9.999990e-06_3794, %int1_3795 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3796 = torch.constant.float 9.9999997473787516E-6
%int1_3797 = torch.constant.int 1
%6164 = torch.aten.add.Scalar %6155, %float9.999990e-06_3796, %int1_3797 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_3798 = torch.constant.float 9.9999997473787516E-6
%int1_3799 = torch.constant.int 1
%6165 = torch.aten.add.Scalar %6157, %float9.999990e-06_3798, %int1_3799 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6166 = torch.aten.rsqrt %6158 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6167 = torch.aten.rsqrt %6159 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6168 = torch.aten.rsqrt %6160 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6169 = torch.aten.rsqrt %6161 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6170 = torch.aten.rsqrt %6162 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6171 = torch.aten.rsqrt %6163 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6172 = torch.aten.rsqrt %6164 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6173 = torch.aten.rsqrt %6165 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %6173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%6174 = torch.aten.mul.Tensor %6126, %6166 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6175 = torch.aten.mul.Tensor %6127, %6167 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6176 = torch.aten.mul.Tensor %6128, %6168 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6177 = torch.aten.mul.Tensor %6129, %6169 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6178 = torch.aten.mul.Tensor %6130, %6170 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6179 = torch.aten.mul.Tensor %6131, %6171 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6180 = torch.aten.mul.Tensor %6132, %6172 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6181 = torch.aten.mul.Tensor %6133, %6173 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6182 = torch.aten.mul.Tensor %152, %6174 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6183 = torch.aten.mul.Tensor %153, %6175 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6184 = torch.aten.mul.Tensor %154, %6176 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6185 = torch.aten.mul.Tensor %155, %6177 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6186 = torch.aten.mul.Tensor %156, %6178 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6187 = torch.aten.mul.Tensor %157, %6179 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6188 = torch.aten.mul.Tensor %158, %6180 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%6189 = torch.aten.mul.Tensor %159, %6181 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %6189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_3800 = torch.constant.int 5
%6190 = torch.prims.convert_element_type %6182, %int5_3800 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3801 = torch.constant.int 5
%6191 = torch.prims.convert_element_type %6183, %int5_3801 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3802 = torch.constant.int 5
%6192 = torch.prims.convert_element_type %6184, %int5_3802 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3803 = torch.constant.int 5
%6193 = torch.prims.convert_element_type %6185, %int5_3803 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3804 = torch.constant.int 5
%6194 = torch.prims.convert_element_type %6186, %int5_3804 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3805 = torch.constant.int 5
%6195 = torch.prims.convert_element_type %6187, %int5_3805 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3806 = torch.constant.int 5
%6196 = torch.prims.convert_element_type %6188, %int5_3806 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_3807 = torch.constant.int 5
%6197 = torch.prims.convert_element_type %6189, %int5_3807 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %6197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_3808 = torch.constant.int 1
%int0_3809 = torch.constant.int 0
%6198 = torch.prim.ListConstruct %int1_3808, %int0_3809 : (!torch.int, !torch.int) -> !torch.list<int>
%6199 = torch.aten.permute %160, %6198 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3810 = torch.constant.int 1
%int0_3811 = torch.constant.int 0
%6200 = torch.prim.ListConstruct %int1_3810, %int0_3811 : (!torch.int, !torch.int) -> !torch.list<int>
%6201 = torch.aten.permute %161, %6200 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3812 = torch.constant.int 1
%int0_3813 = torch.constant.int 0
%6202 = torch.prim.ListConstruct %int1_3812, %int0_3813 : (!torch.int, !torch.int) -> !torch.list<int>
%6203 = torch.aten.permute %162, %6202 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3814 = torch.constant.int 1
%int0_3815 = torch.constant.int 0
%6204 = torch.prim.ListConstruct %int1_3814, %int0_3815 : (!torch.int, !torch.int) -> !torch.list<int>
%6205 = torch.aten.permute %163, %6204 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3816 = torch.constant.int 1
%int0_3817 = torch.constant.int 0
%6206 = torch.prim.ListConstruct %int1_3816, %int0_3817 : (!torch.int, !torch.int) -> !torch.list<int>
%6207 = torch.aten.permute %164, %6206 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3818 = torch.constant.int 1
%int0_3819 = torch.constant.int 0
%6208 = torch.prim.ListConstruct %int1_3818, %int0_3819 : (!torch.int, !torch.int) -> !torch.list<int>
%6209 = torch.aten.permute %165, %6208 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3820 = torch.constant.int 1
%int0_3821 = torch.constant.int 0
%6210 = torch.prim.ListConstruct %int1_3820, %int0_3821 : (!torch.int, !torch.int) -> !torch.list<int>
%6211 = torch.aten.permute %166, %6210 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_3822 = torch.constant.int 1
%int0_3823 = torch.constant.int 0
%6212 = torch.prim.ListConstruct %int1_3822, %int0_3823 : (!torch.int, !torch.int) -> !torch.list<int>
%6213 = torch.aten.permute %167, %6212 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_3824 = torch.constant.int 4
%6214 = torch.aten.mul.int %int4_3824, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3825 = torch.constant.int 4096
%6215 = torch.prim.ListConstruct %6214, %int4096_3825 : (!torch.int, !torch.int) -> !torch.list<int>
%6216 = torch.aten.view %6190, %6215 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6216, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6217 = torch.aten.mm %6216, %6199 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6217, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3826 = torch.constant.int 4
%int512_3827 = torch.constant.int 512
%6218 = torch.prim.ListConstruct %int4_3826, %2482, %int512_3827 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6219 = torch.aten.view %6217, %6218 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3828 = torch.constant.int 4
%6220 = torch.aten.mul.int %int4_3828, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3829 = torch.constant.int 4096
%6221 = torch.prim.ListConstruct %6220, %int4096_3829 : (!torch.int, !torch.int) -> !torch.list<int>
%6222 = torch.aten.view %6191, %6221 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6222, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6223 = torch.aten.mm %6222, %6201 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6223, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3830 = torch.constant.int 4
%int512_3831 = torch.constant.int 512
%6224 = torch.prim.ListConstruct %int4_3830, %2482, %int512_3831 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6225 = torch.aten.view %6223, %6224 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3832 = torch.constant.int 4
%6226 = torch.aten.mul.int %int4_3832, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3833 = torch.constant.int 4096
%6227 = torch.prim.ListConstruct %6226, %int4096_3833 : (!torch.int, !torch.int) -> !torch.list<int>
%6228 = torch.aten.view %6192, %6227 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6228, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6229 = torch.aten.mm %6228, %6203 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6229, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3834 = torch.constant.int 4
%int512_3835 = torch.constant.int 512
%6230 = torch.prim.ListConstruct %int4_3834, %2482, %int512_3835 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6231 = torch.aten.view %6229, %6230 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3836 = torch.constant.int 4
%6232 = torch.aten.mul.int %int4_3836, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3837 = torch.constant.int 4096
%6233 = torch.prim.ListConstruct %6232, %int4096_3837 : (!torch.int, !torch.int) -> !torch.list<int>
%6234 = torch.aten.view %6193, %6233 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6234, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6235 = torch.aten.mm %6234, %6205 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6235, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3838 = torch.constant.int 4
%int512_3839 = torch.constant.int 512
%6236 = torch.prim.ListConstruct %int4_3838, %2482, %int512_3839 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6237 = torch.aten.view %6235, %6236 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3840 = torch.constant.int 4
%6238 = torch.aten.mul.int %int4_3840, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3841 = torch.constant.int 4096
%6239 = torch.prim.ListConstruct %6238, %int4096_3841 : (!torch.int, !torch.int) -> !torch.list<int>
%6240 = torch.aten.view %6194, %6239 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6240, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6241 = torch.aten.mm %6240, %6207 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6241, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3842 = torch.constant.int 4
%int512_3843 = torch.constant.int 512
%6242 = torch.prim.ListConstruct %int4_3842, %2482, %int512_3843 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6243 = torch.aten.view %6241, %6242 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3844 = torch.constant.int 4
%6244 = torch.aten.mul.int %int4_3844, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3845 = torch.constant.int 4096
%6245 = torch.prim.ListConstruct %6244, %int4096_3845 : (!torch.int, !torch.int) -> !torch.list<int>
%6246 = torch.aten.view %6195, %6245 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6246, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6247 = torch.aten.mm %6246, %6209 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6247, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3846 = torch.constant.int 4
%int512_3847 = torch.constant.int 512
%6248 = torch.prim.ListConstruct %int4_3846, %2482, %int512_3847 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6249 = torch.aten.view %6247, %6248 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3848 = torch.constant.int 4
%6250 = torch.aten.mul.int %int4_3848, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3849 = torch.constant.int 4096
%6251 = torch.prim.ListConstruct %6250, %int4096_3849 : (!torch.int, !torch.int) -> !torch.list<int>
%6252 = torch.aten.view %6196, %6251 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6252, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6253 = torch.aten.mm %6252, %6211 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6253, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3850 = torch.constant.int 4
%int512_3851 = torch.constant.int 512
%6254 = torch.prim.ListConstruct %int4_3850, %2482, %int512_3851 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6255 = torch.aten.view %6253, %6254 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_3852 = torch.constant.int 4
%6256 = torch.aten.mul.int %int4_3852, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3853 = torch.constant.int 4096
%6257 = torch.prim.ListConstruct %6256, %int4096_3853 : (!torch.int, !torch.int) -> !torch.list<int>
%6258 = torch.aten.view %6197, %6257 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6258, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6259 = torch.aten.mm %6258, %6213 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %6259, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_3854 = torch.constant.int 4
%int512_3855 = torch.constant.int 512
%6260 = torch.prim.ListConstruct %int4_3854, %2482, %int512_3855 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6261 = torch.aten.view %6259, %6260 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %6261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_3856 = torch.constant.int 1
%int0_3857 = torch.constant.int 0
%6262 = torch.prim.ListConstruct %int1_3856, %int0_3857 : (!torch.int, !torch.int) -> !torch.list<int>
%6263 = torch.aten.permute %168, %6262 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3858 = torch.constant.int 1
%int0_3859 = torch.constant.int 0
%6264 = torch.prim.ListConstruct %int1_3858, %int0_3859 : (!torch.int, !torch.int) -> !torch.list<int>
%6265 = torch.aten.permute %169, %6264 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3860 = torch.constant.int 1
%int0_3861 = torch.constant.int 0
%6266 = torch.prim.ListConstruct %int1_3860, %int0_3861 : (!torch.int, !torch.int) -> !torch.list<int>
%6267 = torch.aten.permute %170, %6266 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3862 = torch.constant.int 1
%int0_3863 = torch.constant.int 0
%6268 = torch.prim.ListConstruct %int1_3862, %int0_3863 : (!torch.int, !torch.int) -> !torch.list<int>
%6269 = torch.aten.permute %171, %6268 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3864 = torch.constant.int 1
%int0_3865 = torch.constant.int 0
%6270 = torch.prim.ListConstruct %int1_3864, %int0_3865 : (!torch.int, !torch.int) -> !torch.list<int>
%6271 = torch.aten.permute %172, %6270 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3866 = torch.constant.int 1
%int0_3867 = torch.constant.int 0
%6272 = torch.prim.ListConstruct %int1_3866, %int0_3867 : (!torch.int, !torch.int) -> !torch.list<int>
%6273 = torch.aten.permute %173, %6272 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3868 = torch.constant.int 1
%int0_3869 = torch.constant.int 0
%6274 = torch.prim.ListConstruct %int1_3868, %int0_3869 : (!torch.int, !torch.int) -> !torch.list<int>
%6275 = torch.aten.permute %174, %6274 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3870 = torch.constant.int 1
%int0_3871 = torch.constant.int 0
%6276 = torch.prim.ListConstruct %int1_3870, %int0_3871 : (!torch.int, !torch.int) -> !torch.list<int>
%6277 = torch.aten.permute %175, %6276 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_3872 = torch.constant.int 4
%6278 = torch.aten.mul.int %int4_3872, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3873 = torch.constant.int 4096
%6279 = torch.prim.ListConstruct %6278, %int4096_3873 : (!torch.int, !torch.int) -> !torch.list<int>
%6280 = torch.aten.view %6190, %6279 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6280, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6281 = torch.aten.mm %6280, %6263 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6281, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3874 = torch.constant.int 4
%int128_3875 = torch.constant.int 128
%6282 = torch.prim.ListConstruct %int4_3874, %2482, %int128_3875 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6283 = torch.aten.view %6281, %6282 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3876 = torch.constant.int 4
%6284 = torch.aten.mul.int %int4_3876, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3877 = torch.constant.int 4096
%6285 = torch.prim.ListConstruct %6284, %int4096_3877 : (!torch.int, !torch.int) -> !torch.list<int>
%6286 = torch.aten.view %6191, %6285 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6286, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6287 = torch.aten.mm %6286, %6265 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6287, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3878 = torch.constant.int 4
%int128_3879 = torch.constant.int 128
%6288 = torch.prim.ListConstruct %int4_3878, %2482, %int128_3879 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6289 = torch.aten.view %6287, %6288 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3880 = torch.constant.int 4
%6290 = torch.aten.mul.int %int4_3880, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3881 = torch.constant.int 4096
%6291 = torch.prim.ListConstruct %6290, %int4096_3881 : (!torch.int, !torch.int) -> !torch.list<int>
%6292 = torch.aten.view %6192, %6291 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6292, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6293 = torch.aten.mm %6292, %6267 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6293, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3882 = torch.constant.int 4
%int128_3883 = torch.constant.int 128
%6294 = torch.prim.ListConstruct %int4_3882, %2482, %int128_3883 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6295 = torch.aten.view %6293, %6294 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3884 = torch.constant.int 4
%6296 = torch.aten.mul.int %int4_3884, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3885 = torch.constant.int 4096
%6297 = torch.prim.ListConstruct %6296, %int4096_3885 : (!torch.int, !torch.int) -> !torch.list<int>
%6298 = torch.aten.view %6193, %6297 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6298, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6299 = torch.aten.mm %6298, %6269 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6299, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3886 = torch.constant.int 4
%int128_3887 = torch.constant.int 128
%6300 = torch.prim.ListConstruct %int4_3886, %2482, %int128_3887 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6301 = torch.aten.view %6299, %6300 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3888 = torch.constant.int 4
%6302 = torch.aten.mul.int %int4_3888, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3889 = torch.constant.int 4096
%6303 = torch.prim.ListConstruct %6302, %int4096_3889 : (!torch.int, !torch.int) -> !torch.list<int>
%6304 = torch.aten.view %6194, %6303 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6304, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6305 = torch.aten.mm %6304, %6271 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6305, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3890 = torch.constant.int 4
%int128_3891 = torch.constant.int 128
%6306 = torch.prim.ListConstruct %int4_3890, %2482, %int128_3891 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6307 = torch.aten.view %6305, %6306 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3892 = torch.constant.int 4
%6308 = torch.aten.mul.int %int4_3892, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3893 = torch.constant.int 4096
%6309 = torch.prim.ListConstruct %6308, %int4096_3893 : (!torch.int, !torch.int) -> !torch.list<int>
%6310 = torch.aten.view %6195, %6309 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6310, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6311 = torch.aten.mm %6310, %6273 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6311, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3894 = torch.constant.int 4
%int128_3895 = torch.constant.int 128
%6312 = torch.prim.ListConstruct %int4_3894, %2482, %int128_3895 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6313 = torch.aten.view %6311, %6312 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3896 = torch.constant.int 4
%6314 = torch.aten.mul.int %int4_3896, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3897 = torch.constant.int 4096
%6315 = torch.prim.ListConstruct %6314, %int4096_3897 : (!torch.int, !torch.int) -> !torch.list<int>
%6316 = torch.aten.view %6196, %6315 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6316, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6317 = torch.aten.mm %6316, %6275 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6317, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3898 = torch.constant.int 4
%int128_3899 = torch.constant.int 128
%6318 = torch.prim.ListConstruct %int4_3898, %2482, %int128_3899 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6319 = torch.aten.view %6317, %6318 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3900 = torch.constant.int 4
%6320 = torch.aten.mul.int %int4_3900, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3901 = torch.constant.int 4096
%6321 = torch.prim.ListConstruct %6320, %int4096_3901 : (!torch.int, !torch.int) -> !torch.list<int>
%6322 = torch.aten.view %6197, %6321 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6322, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6323 = torch.aten.mm %6322, %6277 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6323, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3902 = torch.constant.int 4
%int128_3903 = torch.constant.int 128
%6324 = torch.prim.ListConstruct %int4_3902, %2482, %int128_3903 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6325 = torch.aten.view %6323, %6324 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_3904 = torch.constant.int 1
%int0_3905 = torch.constant.int 0
%6326 = torch.prim.ListConstruct %int1_3904, %int0_3905 : (!torch.int, !torch.int) -> !torch.list<int>
%6327 = torch.aten.permute %176, %6326 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3906 = torch.constant.int 1
%int0_3907 = torch.constant.int 0
%6328 = torch.prim.ListConstruct %int1_3906, %int0_3907 : (!torch.int, !torch.int) -> !torch.list<int>
%6329 = torch.aten.permute %177, %6328 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3908 = torch.constant.int 1
%int0_3909 = torch.constant.int 0
%6330 = torch.prim.ListConstruct %int1_3908, %int0_3909 : (!torch.int, !torch.int) -> !torch.list<int>
%6331 = torch.aten.permute %178, %6330 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3910 = torch.constant.int 1
%int0_3911 = torch.constant.int 0
%6332 = torch.prim.ListConstruct %int1_3910, %int0_3911 : (!torch.int, !torch.int) -> !torch.list<int>
%6333 = torch.aten.permute %179, %6332 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3912 = torch.constant.int 1
%int0_3913 = torch.constant.int 0
%6334 = torch.prim.ListConstruct %int1_3912, %int0_3913 : (!torch.int, !torch.int) -> !torch.list<int>
%6335 = torch.aten.permute %180, %6334 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3914 = torch.constant.int 1
%int0_3915 = torch.constant.int 0
%6336 = torch.prim.ListConstruct %int1_3914, %int0_3915 : (!torch.int, !torch.int) -> !torch.list<int>
%6337 = torch.aten.permute %181, %6336 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3916 = torch.constant.int 1
%int0_3917 = torch.constant.int 0
%6338 = torch.prim.ListConstruct %int1_3916, %int0_3917 : (!torch.int, !torch.int) -> !torch.list<int>
%6339 = torch.aten.permute %182, %6338 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_3918 = torch.constant.int 1
%int0_3919 = torch.constant.int 0
%6340 = torch.prim.ListConstruct %int1_3918, %int0_3919 : (!torch.int, !torch.int) -> !torch.list<int>
%6341 = torch.aten.permute %183, %6340 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_3920 = torch.constant.int 4
%6342 = torch.aten.mul.int %int4_3920, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3921 = torch.constant.int 4096
%6343 = torch.prim.ListConstruct %6342, %int4096_3921 : (!torch.int, !torch.int) -> !torch.list<int>
%6344 = torch.aten.view %6190, %6343 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6344, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6345 = torch.aten.mm %6344, %6327 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6345, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3922 = torch.constant.int 4
%int128_3923 = torch.constant.int 128
%6346 = torch.prim.ListConstruct %int4_3922, %2482, %int128_3923 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6347 = torch.aten.view %6345, %6346 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3924 = torch.constant.int 4
%6348 = torch.aten.mul.int %int4_3924, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3925 = torch.constant.int 4096
%6349 = torch.prim.ListConstruct %6348, %int4096_3925 : (!torch.int, !torch.int) -> !torch.list<int>
%6350 = torch.aten.view %6191, %6349 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6350, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6351 = torch.aten.mm %6350, %6329 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6351, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3926 = torch.constant.int 4
%int128_3927 = torch.constant.int 128
%6352 = torch.prim.ListConstruct %int4_3926, %2482, %int128_3927 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6353 = torch.aten.view %6351, %6352 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3928 = torch.constant.int 4
%6354 = torch.aten.mul.int %int4_3928, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3929 = torch.constant.int 4096
%6355 = torch.prim.ListConstruct %6354, %int4096_3929 : (!torch.int, !torch.int) -> !torch.list<int>
%6356 = torch.aten.view %6192, %6355 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6356, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6357 = torch.aten.mm %6356, %6331 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6357, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3930 = torch.constant.int 4
%int128_3931 = torch.constant.int 128
%6358 = torch.prim.ListConstruct %int4_3930, %2482, %int128_3931 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6359 = torch.aten.view %6357, %6358 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3932 = torch.constant.int 4
%6360 = torch.aten.mul.int %int4_3932, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3933 = torch.constant.int 4096
%6361 = torch.prim.ListConstruct %6360, %int4096_3933 : (!torch.int, !torch.int) -> !torch.list<int>
%6362 = torch.aten.view %6193, %6361 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6362, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6363 = torch.aten.mm %6362, %6333 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6363, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3934 = torch.constant.int 4
%int128_3935 = torch.constant.int 128
%6364 = torch.prim.ListConstruct %int4_3934, %2482, %int128_3935 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6365 = torch.aten.view %6363, %6364 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3936 = torch.constant.int 4
%6366 = torch.aten.mul.int %int4_3936, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3937 = torch.constant.int 4096
%6367 = torch.prim.ListConstruct %6366, %int4096_3937 : (!torch.int, !torch.int) -> !torch.list<int>
%6368 = torch.aten.view %6194, %6367 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6368, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6369 = torch.aten.mm %6368, %6335 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6369, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3938 = torch.constant.int 4
%int128_3939 = torch.constant.int 128
%6370 = torch.prim.ListConstruct %int4_3938, %2482, %int128_3939 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6371 = torch.aten.view %6369, %6370 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3940 = torch.constant.int 4
%6372 = torch.aten.mul.int %int4_3940, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3941 = torch.constant.int 4096
%6373 = torch.prim.ListConstruct %6372, %int4096_3941 : (!torch.int, !torch.int) -> !torch.list<int>
%6374 = torch.aten.view %6195, %6373 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6374, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6375 = torch.aten.mm %6374, %6337 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6375, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3942 = torch.constant.int 4
%int128_3943 = torch.constant.int 128
%6376 = torch.prim.ListConstruct %int4_3942, %2482, %int128_3943 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6377 = torch.aten.view %6375, %6376 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3944 = torch.constant.int 4
%6378 = torch.aten.mul.int %int4_3944, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3945 = torch.constant.int 4096
%6379 = torch.prim.ListConstruct %6378, %int4096_3945 : (!torch.int, !torch.int) -> !torch.list<int>
%6380 = torch.aten.view %6196, %6379 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6380, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6381 = torch.aten.mm %6380, %6339 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6381, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3946 = torch.constant.int 4
%int128_3947 = torch.constant.int 128
%6382 = torch.prim.ListConstruct %int4_3946, %2482, %int128_3947 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6383 = torch.aten.view %6381, %6382 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3948 = torch.constant.int 4
%6384 = torch.aten.mul.int %int4_3948, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_3949 = torch.constant.int 4096
%6385 = torch.prim.ListConstruct %6384, %int4096_3949 : (!torch.int, !torch.int) -> !torch.list<int>
%6386 = torch.aten.view %6197, %6385 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %6386, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%6387 = torch.aten.mm %6386, %6341 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %6387, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_3950 = torch.constant.int 4
%int128_3951 = torch.constant.int 128
%6388 = torch.prim.ListConstruct %int4_3950, %2482, %int128_3951 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6389 = torch.aten.view %6387, %6388 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %6389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_3952 = torch.constant.int 4
%int4_3953 = torch.constant.int 4
%int128_3954 = torch.constant.int 128
%6390 = torch.prim.ListConstruct %int4_3952, %2482, %int4_3953, %int128_3954 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6391 = torch.aten.view %6219, %6390 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3955 = torch.constant.int 4
%int4_3956 = torch.constant.int 4
%int128_3957 = torch.constant.int 128
%6392 = torch.prim.ListConstruct %int4_3955, %2482, %int4_3956, %int128_3957 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6393 = torch.aten.view %6225, %6392 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3958 = torch.constant.int 4
%int4_3959 = torch.constant.int 4
%int128_3960 = torch.constant.int 128
%6394 = torch.prim.ListConstruct %int4_3958, %2482, %int4_3959, %int128_3960 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6395 = torch.aten.view %6231, %6394 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3961 = torch.constant.int 4
%int4_3962 = torch.constant.int 4
%int128_3963 = torch.constant.int 128
%6396 = torch.prim.ListConstruct %int4_3961, %2482, %int4_3962, %int128_3963 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6397 = torch.aten.view %6237, %6396 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3964 = torch.constant.int 4
%int4_3965 = torch.constant.int 4
%int128_3966 = torch.constant.int 128
%6398 = torch.prim.ListConstruct %int4_3964, %2482, %int4_3965, %int128_3966 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6399 = torch.aten.view %6243, %6398 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3967 = torch.constant.int 4
%int4_3968 = torch.constant.int 4
%int128_3969 = torch.constant.int 128
%6400 = torch.prim.ListConstruct %int4_3967, %2482, %int4_3968, %int128_3969 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6401 = torch.aten.view %6249, %6400 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3970 = torch.constant.int 4
%int4_3971 = torch.constant.int 4
%int128_3972 = torch.constant.int 128
%6402 = torch.prim.ListConstruct %int4_3970, %2482, %int4_3971, %int128_3972 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6403 = torch.aten.view %6255, %6402 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3973 = torch.constant.int 4
%int4_3974 = torch.constant.int 4
%int128_3975 = torch.constant.int 128
%6404 = torch.prim.ListConstruct %int4_3973, %2482, %int4_3974, %int128_3975 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6405 = torch.aten.view %6261, %6404 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_3976 = torch.constant.int 4
%int1_3977 = torch.constant.int 1
%int128_3978 = torch.constant.int 128
%6406 = torch.prim.ListConstruct %int4_3976, %2482, %int1_3977, %int128_3978 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6407 = torch.aten.view %6283, %6406 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3979 = torch.constant.int 4
%int1_3980 = torch.constant.int 1
%int128_3981 = torch.constant.int 128
%6408 = torch.prim.ListConstruct %int4_3979, %2482, %int1_3980, %int128_3981 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6409 = torch.aten.view %6289, %6408 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3982 = torch.constant.int 4
%int1_3983 = torch.constant.int 1
%int128_3984 = torch.constant.int 128
%6410 = torch.prim.ListConstruct %int4_3982, %2482, %int1_3983, %int128_3984 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6411 = torch.aten.view %6295, %6410 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3985 = torch.constant.int 4
%int1_3986 = torch.constant.int 1
%int128_3987 = torch.constant.int 128
%6412 = torch.prim.ListConstruct %int4_3985, %2482, %int1_3986, %int128_3987 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6413 = torch.aten.view %6301, %6412 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3988 = torch.constant.int 4
%int1_3989 = torch.constant.int 1
%int128_3990 = torch.constant.int 128
%6414 = torch.prim.ListConstruct %int4_3988, %2482, %int1_3989, %int128_3990 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6415 = torch.aten.view %6307, %6414 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3991 = torch.constant.int 4
%int1_3992 = torch.constant.int 1
%int128_3993 = torch.constant.int 128
%6416 = torch.prim.ListConstruct %int4_3991, %2482, %int1_3992, %int128_3993 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6417 = torch.aten.view %6313, %6416 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3994 = torch.constant.int 4
%int1_3995 = torch.constant.int 1
%int128_3996 = torch.constant.int 128
%6418 = torch.prim.ListConstruct %int4_3994, %2482, %int1_3995, %int128_3996 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6419 = torch.aten.view %6319, %6418 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_3997 = torch.constant.int 4
%int1_3998 = torch.constant.int 1
%int128_3999 = torch.constant.int 128
%6420 = torch.prim.ListConstruct %int4_3997, %2482, %int1_3998, %int128_3999 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6421 = torch.aten.view %6325, %6420 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4000 = torch.constant.int 4
%int1_4001 = torch.constant.int 1
%int128_4002 = torch.constant.int 128
%6422 = torch.prim.ListConstruct %int4_4000, %2482, %int1_4001, %int128_4002 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6423 = torch.aten.view %6347, %6422 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4003 = torch.constant.int 4
%int1_4004 = torch.constant.int 1
%int128_4005 = torch.constant.int 128
%6424 = torch.prim.ListConstruct %int4_4003, %2482, %int1_4004, %int128_4005 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6425 = torch.aten.view %6353, %6424 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4006 = torch.constant.int 4
%int1_4007 = torch.constant.int 1
%int128_4008 = torch.constant.int 128
%6426 = torch.prim.ListConstruct %int4_4006, %2482, %int1_4007, %int128_4008 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6427 = torch.aten.view %6359, %6426 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4009 = torch.constant.int 4
%int1_4010 = torch.constant.int 1
%int128_4011 = torch.constant.int 128
%6428 = torch.prim.ListConstruct %int4_4009, %2482, %int1_4010, %int128_4011 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6429 = torch.aten.view %6365, %6428 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4012 = torch.constant.int 4
%int1_4013 = torch.constant.int 1
%int128_4014 = torch.constant.int 128
%6430 = torch.prim.ListConstruct %int4_4012, %2482, %int1_4013, %int128_4014 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6431 = torch.aten.view %6371, %6430 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4015 = torch.constant.int 4
%int1_4016 = torch.constant.int 1
%int128_4017 = torch.constant.int 128
%6432 = torch.prim.ListConstruct %int4_4015, %2482, %int1_4016, %int128_4017 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6433 = torch.aten.view %6377, %6432 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4018 = torch.constant.int 4
%int1_4019 = torch.constant.int 1
%int128_4020 = torch.constant.int 128
%6434 = torch.prim.ListConstruct %int4_4018, %2482, %int1_4019, %int128_4020 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6435 = torch.aten.view %6383, %6434 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_4021 = torch.constant.int 4
%int1_4022 = torch.constant.int 1
%int128_4023 = torch.constant.int 128
%6436 = torch.prim.ListConstruct %int4_4021, %2482, %int1_4022, %int128_4023 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6437 = torch.aten.view %6389, %6436 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_4024 = torch.constant.int 131072
%none_4025 = torch.constant.none
%none_4026 = torch.constant.none
%cpu_4027 = torch.constant.device "cpu"
%false_4028 = torch.constant.bool false
%6438 = torch.aten.arange %int131072_4024, %none_4025, %none_4026, %cpu_4027, %false_4028 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_4029 = torch.constant.int 0
%int128_4030 = torch.constant.int 128
%int2_4031 = torch.constant.int 2
%none_4032 = torch.constant.none
%none_4033 = torch.constant.none
%cpu_4034 = torch.constant.device "cpu"
%false_4035 = torch.constant.bool false
%6439 = torch.aten.arange.start_step %int0_4029, %int128_4030, %int2_4031, %none_4032, %none_4033, %cpu_4034, %false_4035 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_4036 = torch.constant.int 0
%int0_4037 = torch.constant.int 0
%int64_4038 = torch.constant.int 64
%int1_4039 = torch.constant.int 1
%6440 = torch.aten.slice.Tensor %6439, %int0_4036, %int0_4037, %int64_4038, %int1_4039 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_4040 = torch.constant.int 6
%6441 = torch.prims.convert_element_type %6440, %int6_4040 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_4041 = torch.constant.int 128
%6442 = torch.aten.div.Scalar %6441, %int128_4041 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_4042 = torch.constant.float 5.000000e+05
%6443 = torch.aten.pow.Scalar %float5.000000e05_4042, %6442 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%6444 = torch.aten.reciprocal %6443 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_4043 = torch.constant.float 1.000000e+00
%6445 = torch.aten.mul.Scalar %6444, %float1.000000e00_4043 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_4044 = torch.constant.int 131072
%int1_4045 = torch.constant.int 1
%6446 = torch.prim.ListConstruct %int131072_4044, %int1_4045 : (!torch.int, !torch.int) -> !torch.list<int>
%6447 = torch.aten.view %6438, %6446 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%6448 = torch.aten.mul.Tensor %6447, %6445 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%6449 = torch.aten.cos %6448 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%6450 = torch.aten.sin %6448 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%6451 = torch.aten.complex %6449, %6450 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%6452 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6453 = flow.tensor.transfer %6452 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%6454 = torch_c.from_builtin_tensor %6453 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6455 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6456 = flow.tensor.transfer %6455 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%6457 = torch_c.from_builtin_tensor %6456 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6458 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6459 = flow.tensor.transfer %6458 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%6460 = torch_c.from_builtin_tensor %6459 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6461 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6462 = flow.tensor.transfer %6461 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%6463 = torch_c.from_builtin_tensor %6462 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6464 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6465 = flow.tensor.transfer %6464 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%6466 = torch_c.from_builtin_tensor %6465 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6467 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6468 = flow.tensor.transfer %6467 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%6469 = torch_c.from_builtin_tensor %6468 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6470 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6471 = flow.tensor.transfer %6470 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%6472 = torch_c.from_builtin_tensor %6471 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6473 = torch_c.to_builtin_tensor %6451 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6474 = flow.tensor.transfer %6473 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%6475 = torch_c.from_builtin_tensor %6474 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_4046 = torch.constant.int 1
%6476 = torch.aten.size.int %6219, %int1_4046 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4047 = torch.constant.int 0
%6477 = torch.aten.add.int %int0_4047, %6476 : !torch.int, !torch.int -> !torch.int
%int0_4048 = torch.constant.int 0
%int0_4049 = torch.constant.int 0
%int1_4050 = torch.constant.int 1
%6478 = torch.aten.slice.Tensor %6454, %int0_4048, %int0_4049, %6477, %int1_4050 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6478, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4051 = torch.constant.int 1
%int0_4052 = torch.constant.int 0
%int9223372036854775807_4053 = torch.constant.int 9223372036854775807
%int1_4054 = torch.constant.int 1
%6479 = torch.aten.slice.Tensor %6478, %int1_4051, %int0_4052, %int9223372036854775807_4053, %int1_4054 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6479, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4055 = torch.constant.int 0
%6480 = torch.aten.unsqueeze %6479, %int0_4055 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6480, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4056 = torch.constant.int 2
%6481 = torch.aten.unsqueeze %6480, %int2_4056 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6481, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4057 = torch.constant.int 3
%int0_4058 = torch.constant.int 0
%int9223372036854775807_4059 = torch.constant.int 9223372036854775807
%int1_4060 = torch.constant.int 1
%6482 = torch.aten.slice.Tensor %6481, %int3_4057, %int0_4058, %int9223372036854775807_4059, %int1_4060 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6482, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6483 = torch_c.to_builtin_tensor %6391 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4061 = arith.constant 1 : index
%dim_4062 = tensor.dim %6483, %c1_4061 : tensor<4x?x4x128xf16>
%6484 = flow.tensor.bitcast %6483 : tensor<4x?x4x128xf16>{%dim_4062} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4062}
%6485 = torch_c.from_builtin_tensor %6484 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6486 = torch.aten.mul.Tensor %6485, %6482 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6487 = torch_c.to_builtin_tensor %6486 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4063 = arith.constant 1 : index
%dim_4064 = tensor.dim %6487, %c1_4063 : tensor<4x?x4x64xcomplex<f32>>
%6488 = flow.tensor.bitcast %6487 : tensor<4x?x4x64xcomplex<f32>>{%dim_4064} -> tensor<4x?x4x128xf32>{%dim_4064}
%6489 = torch_c.from_builtin_tensor %6488 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4065 = torch.constant.int 5
%6490 = torch.prims.convert_element_type %6489, %int5_4065 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4066 = torch.constant.int 1
%6491 = torch.aten.size.int %6225, %int1_4066 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4067 = torch.constant.int 0
%6492 = torch.aten.add.int %int0_4067, %6491 : !torch.int, !torch.int -> !torch.int
%int0_4068 = torch.constant.int 0
%int0_4069 = torch.constant.int 0
%int1_4070 = torch.constant.int 1
%6493 = torch.aten.slice.Tensor %6457, %int0_4068, %int0_4069, %6492, %int1_4070 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6493, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4071 = torch.constant.int 1
%int0_4072 = torch.constant.int 0
%int9223372036854775807_4073 = torch.constant.int 9223372036854775807
%int1_4074 = torch.constant.int 1
%6494 = torch.aten.slice.Tensor %6493, %int1_4071, %int0_4072, %int9223372036854775807_4073, %int1_4074 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6494, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4075 = torch.constant.int 0
%6495 = torch.aten.unsqueeze %6494, %int0_4075 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6495, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4076 = torch.constant.int 2
%6496 = torch.aten.unsqueeze %6495, %int2_4076 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6496, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4077 = torch.constant.int 3
%int0_4078 = torch.constant.int 0
%int9223372036854775807_4079 = torch.constant.int 9223372036854775807
%int1_4080 = torch.constant.int 1
%6497 = torch.aten.slice.Tensor %6496, %int3_4077, %int0_4078, %int9223372036854775807_4079, %int1_4080 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6497, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6498 = torch_c.to_builtin_tensor %6393 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4081 = arith.constant 1 : index
%dim_4082 = tensor.dim %6498, %c1_4081 : tensor<4x?x4x128xf16>
%6499 = flow.tensor.bitcast %6498 : tensor<4x?x4x128xf16>{%dim_4082} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4082}
%6500 = torch_c.from_builtin_tensor %6499 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6501 = torch.aten.mul.Tensor %6500, %6497 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6502 = torch_c.to_builtin_tensor %6501 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4083 = arith.constant 1 : index
%dim_4084 = tensor.dim %6502, %c1_4083 : tensor<4x?x4x64xcomplex<f32>>
%6503 = flow.tensor.bitcast %6502 : tensor<4x?x4x64xcomplex<f32>>{%dim_4084} -> tensor<4x?x4x128xf32>{%dim_4084}
%6504 = torch_c.from_builtin_tensor %6503 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4085 = torch.constant.int 5
%6505 = torch.prims.convert_element_type %6504, %int5_4085 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4086 = torch.constant.int 1
%6506 = torch.aten.size.int %6231, %int1_4086 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4087 = torch.constant.int 0
%6507 = torch.aten.add.int %int0_4087, %6506 : !torch.int, !torch.int -> !torch.int
%int0_4088 = torch.constant.int 0
%int0_4089 = torch.constant.int 0
%int1_4090 = torch.constant.int 1
%6508 = torch.aten.slice.Tensor %6460, %int0_4088, %int0_4089, %6507, %int1_4090 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6508, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4091 = torch.constant.int 1
%int0_4092 = torch.constant.int 0
%int9223372036854775807_4093 = torch.constant.int 9223372036854775807
%int1_4094 = torch.constant.int 1
%6509 = torch.aten.slice.Tensor %6508, %int1_4091, %int0_4092, %int9223372036854775807_4093, %int1_4094 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6509, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4095 = torch.constant.int 0
%6510 = torch.aten.unsqueeze %6509, %int0_4095 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6510, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4096 = torch.constant.int 2
%6511 = torch.aten.unsqueeze %6510, %int2_4096 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6511, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4097 = torch.constant.int 3
%int0_4098 = torch.constant.int 0
%int9223372036854775807_4099 = torch.constant.int 9223372036854775807
%int1_4100 = torch.constant.int 1
%6512 = torch.aten.slice.Tensor %6511, %int3_4097, %int0_4098, %int9223372036854775807_4099, %int1_4100 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6512, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6513 = torch_c.to_builtin_tensor %6395 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4101 = arith.constant 1 : index
%dim_4102 = tensor.dim %6513, %c1_4101 : tensor<4x?x4x128xf16>
%6514 = flow.tensor.bitcast %6513 : tensor<4x?x4x128xf16>{%dim_4102} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4102}
%6515 = torch_c.from_builtin_tensor %6514 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6516 = torch.aten.mul.Tensor %6515, %6512 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6517 = torch_c.to_builtin_tensor %6516 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4103 = arith.constant 1 : index
%dim_4104 = tensor.dim %6517, %c1_4103 : tensor<4x?x4x64xcomplex<f32>>
%6518 = flow.tensor.bitcast %6517 : tensor<4x?x4x64xcomplex<f32>>{%dim_4104} -> tensor<4x?x4x128xf32>{%dim_4104}
%6519 = torch_c.from_builtin_tensor %6518 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4105 = torch.constant.int 5
%6520 = torch.prims.convert_element_type %6519, %int5_4105 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4106 = torch.constant.int 1
%6521 = torch.aten.size.int %6237, %int1_4106 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4107 = torch.constant.int 0
%6522 = torch.aten.add.int %int0_4107, %6521 : !torch.int, !torch.int -> !torch.int
%int0_4108 = torch.constant.int 0
%int0_4109 = torch.constant.int 0
%int1_4110 = torch.constant.int 1
%6523 = torch.aten.slice.Tensor %6463, %int0_4108, %int0_4109, %6522, %int1_4110 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6523, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4111 = torch.constant.int 1
%int0_4112 = torch.constant.int 0
%int9223372036854775807_4113 = torch.constant.int 9223372036854775807
%int1_4114 = torch.constant.int 1
%6524 = torch.aten.slice.Tensor %6523, %int1_4111, %int0_4112, %int9223372036854775807_4113, %int1_4114 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6524, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4115 = torch.constant.int 0
%6525 = torch.aten.unsqueeze %6524, %int0_4115 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6525, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4116 = torch.constant.int 2
%6526 = torch.aten.unsqueeze %6525, %int2_4116 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6526, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4117 = torch.constant.int 3
%int0_4118 = torch.constant.int 0
%int9223372036854775807_4119 = torch.constant.int 9223372036854775807
%int1_4120 = torch.constant.int 1
%6527 = torch.aten.slice.Tensor %6526, %int3_4117, %int0_4118, %int9223372036854775807_4119, %int1_4120 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6527, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6528 = torch_c.to_builtin_tensor %6397 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4121 = arith.constant 1 : index
%dim_4122 = tensor.dim %6528, %c1_4121 : tensor<4x?x4x128xf16>
%6529 = flow.tensor.bitcast %6528 : tensor<4x?x4x128xf16>{%dim_4122} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4122}
%6530 = torch_c.from_builtin_tensor %6529 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6531 = torch.aten.mul.Tensor %6530, %6527 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6532 = torch_c.to_builtin_tensor %6531 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4123 = arith.constant 1 : index
%dim_4124 = tensor.dim %6532, %c1_4123 : tensor<4x?x4x64xcomplex<f32>>
%6533 = flow.tensor.bitcast %6532 : tensor<4x?x4x64xcomplex<f32>>{%dim_4124} -> tensor<4x?x4x128xf32>{%dim_4124}
%6534 = torch_c.from_builtin_tensor %6533 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4125 = torch.constant.int 5
%6535 = torch.prims.convert_element_type %6534, %int5_4125 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4126 = torch.constant.int 1
%6536 = torch.aten.size.int %6243, %int1_4126 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4127 = torch.constant.int 0
%6537 = torch.aten.add.int %int0_4127, %6536 : !torch.int, !torch.int -> !torch.int
%int0_4128 = torch.constant.int 0
%int0_4129 = torch.constant.int 0
%int1_4130 = torch.constant.int 1
%6538 = torch.aten.slice.Tensor %6466, %int0_4128, %int0_4129, %6537, %int1_4130 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6538, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4131 = torch.constant.int 1
%int0_4132 = torch.constant.int 0
%int9223372036854775807_4133 = torch.constant.int 9223372036854775807
%int1_4134 = torch.constant.int 1
%6539 = torch.aten.slice.Tensor %6538, %int1_4131, %int0_4132, %int9223372036854775807_4133, %int1_4134 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6539, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4135 = torch.constant.int 0
%6540 = torch.aten.unsqueeze %6539, %int0_4135 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6540, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4136 = torch.constant.int 2
%6541 = torch.aten.unsqueeze %6540, %int2_4136 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6541, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4137 = torch.constant.int 3
%int0_4138 = torch.constant.int 0
%int9223372036854775807_4139 = torch.constant.int 9223372036854775807
%int1_4140 = torch.constant.int 1
%6542 = torch.aten.slice.Tensor %6541, %int3_4137, %int0_4138, %int9223372036854775807_4139, %int1_4140 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6542, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6543 = torch_c.to_builtin_tensor %6399 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4141 = arith.constant 1 : index
%dim_4142 = tensor.dim %6543, %c1_4141 : tensor<4x?x4x128xf16>
%6544 = flow.tensor.bitcast %6543 : tensor<4x?x4x128xf16>{%dim_4142} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4142}
%6545 = torch_c.from_builtin_tensor %6544 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6546 = torch.aten.mul.Tensor %6545, %6542 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6547 = torch_c.to_builtin_tensor %6546 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4143 = arith.constant 1 : index
%dim_4144 = tensor.dim %6547, %c1_4143 : tensor<4x?x4x64xcomplex<f32>>
%6548 = flow.tensor.bitcast %6547 : tensor<4x?x4x64xcomplex<f32>>{%dim_4144} -> tensor<4x?x4x128xf32>{%dim_4144}
%6549 = torch_c.from_builtin_tensor %6548 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4145 = torch.constant.int 5
%6550 = torch.prims.convert_element_type %6549, %int5_4145 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4146 = torch.constant.int 1
%6551 = torch.aten.size.int %6249, %int1_4146 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4147 = torch.constant.int 0
%6552 = torch.aten.add.int %int0_4147, %6551 : !torch.int, !torch.int -> !torch.int
%int0_4148 = torch.constant.int 0
%int0_4149 = torch.constant.int 0
%int1_4150 = torch.constant.int 1
%6553 = torch.aten.slice.Tensor %6469, %int0_4148, %int0_4149, %6552, %int1_4150 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6553, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4151 = torch.constant.int 1
%int0_4152 = torch.constant.int 0
%int9223372036854775807_4153 = torch.constant.int 9223372036854775807
%int1_4154 = torch.constant.int 1
%6554 = torch.aten.slice.Tensor %6553, %int1_4151, %int0_4152, %int9223372036854775807_4153, %int1_4154 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6554, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4155 = torch.constant.int 0
%6555 = torch.aten.unsqueeze %6554, %int0_4155 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6555, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4156 = torch.constant.int 2
%6556 = torch.aten.unsqueeze %6555, %int2_4156 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6556, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4157 = torch.constant.int 3
%int0_4158 = torch.constant.int 0
%int9223372036854775807_4159 = torch.constant.int 9223372036854775807
%int1_4160 = torch.constant.int 1
%6557 = torch.aten.slice.Tensor %6556, %int3_4157, %int0_4158, %int9223372036854775807_4159, %int1_4160 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6557, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6558 = torch_c.to_builtin_tensor %6401 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4161 = arith.constant 1 : index
%dim_4162 = tensor.dim %6558, %c1_4161 : tensor<4x?x4x128xf16>
%6559 = flow.tensor.bitcast %6558 : tensor<4x?x4x128xf16>{%dim_4162} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4162}
%6560 = torch_c.from_builtin_tensor %6559 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6561 = torch.aten.mul.Tensor %6560, %6557 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6562 = torch_c.to_builtin_tensor %6561 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4163 = arith.constant 1 : index
%dim_4164 = tensor.dim %6562, %c1_4163 : tensor<4x?x4x64xcomplex<f32>>
%6563 = flow.tensor.bitcast %6562 : tensor<4x?x4x64xcomplex<f32>>{%dim_4164} -> tensor<4x?x4x128xf32>{%dim_4164}
%6564 = torch_c.from_builtin_tensor %6563 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4165 = torch.constant.int 5
%6565 = torch.prims.convert_element_type %6564, %int5_4165 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4166 = torch.constant.int 1
%6566 = torch.aten.size.int %6255, %int1_4166 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4167 = torch.constant.int 0
%6567 = torch.aten.add.int %int0_4167, %6566 : !torch.int, !torch.int -> !torch.int
%int0_4168 = torch.constant.int 0
%int0_4169 = torch.constant.int 0
%int1_4170 = torch.constant.int 1
%6568 = torch.aten.slice.Tensor %6472, %int0_4168, %int0_4169, %6567, %int1_4170 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6568, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4171 = torch.constant.int 1
%int0_4172 = torch.constant.int 0
%int9223372036854775807_4173 = torch.constant.int 9223372036854775807
%int1_4174 = torch.constant.int 1
%6569 = torch.aten.slice.Tensor %6568, %int1_4171, %int0_4172, %int9223372036854775807_4173, %int1_4174 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6569, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4175 = torch.constant.int 0
%6570 = torch.aten.unsqueeze %6569, %int0_4175 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6570, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4176 = torch.constant.int 2
%6571 = torch.aten.unsqueeze %6570, %int2_4176 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6571, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4177 = torch.constant.int 3
%int0_4178 = torch.constant.int 0
%int9223372036854775807_4179 = torch.constant.int 9223372036854775807
%int1_4180 = torch.constant.int 1
%6572 = torch.aten.slice.Tensor %6571, %int3_4177, %int0_4178, %int9223372036854775807_4179, %int1_4180 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6572, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6573 = torch_c.to_builtin_tensor %6403 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4181 = arith.constant 1 : index
%dim_4182 = tensor.dim %6573, %c1_4181 : tensor<4x?x4x128xf16>
%6574 = flow.tensor.bitcast %6573 : tensor<4x?x4x128xf16>{%dim_4182} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4182}
%6575 = torch_c.from_builtin_tensor %6574 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6576 = torch.aten.mul.Tensor %6575, %6572 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6577 = torch_c.to_builtin_tensor %6576 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4183 = arith.constant 1 : index
%dim_4184 = tensor.dim %6577, %c1_4183 : tensor<4x?x4x64xcomplex<f32>>
%6578 = flow.tensor.bitcast %6577 : tensor<4x?x4x64xcomplex<f32>>{%dim_4184} -> tensor<4x?x4x128xf32>{%dim_4184}
%6579 = torch_c.from_builtin_tensor %6578 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4185 = torch.constant.int 5
%6580 = torch.prims.convert_element_type %6579, %int5_4185 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4186 = torch.constant.int 1
%6581 = torch.aten.size.int %6261, %int1_4186 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_4187 = torch.constant.int 0
%6582 = torch.aten.add.int %int0_4187, %6581 : !torch.int, !torch.int -> !torch.int
%int0_4188 = torch.constant.int 0
%int0_4189 = torch.constant.int 0
%int1_4190 = torch.constant.int 1
%6583 = torch.aten.slice.Tensor %6475, %int0_4188, %int0_4189, %6582, %int1_4190 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6583, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4191 = torch.constant.int 1
%int0_4192 = torch.constant.int 0
%int9223372036854775807_4193 = torch.constant.int 9223372036854775807
%int1_4194 = torch.constant.int 1
%6584 = torch.aten.slice.Tensor %6583, %int1_4191, %int0_4192, %int9223372036854775807_4193, %int1_4194 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6584, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4195 = torch.constant.int 0
%6585 = torch.aten.unsqueeze %6584, %int0_4195 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6585, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4196 = torch.constant.int 2
%6586 = torch.aten.unsqueeze %6585, %int2_4196 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6586, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4197 = torch.constant.int 3
%int0_4198 = torch.constant.int 0
%int9223372036854775807_4199 = torch.constant.int 9223372036854775807
%int1_4200 = torch.constant.int 1
%6587 = torch.aten.slice.Tensor %6586, %int3_4197, %int0_4198, %int9223372036854775807_4199, %int1_4200 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6587, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6588 = torch_c.to_builtin_tensor %6405 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_4201 = arith.constant 1 : index
%dim_4202 = tensor.dim %6588, %c1_4201 : tensor<4x?x4x128xf16>
%6589 = flow.tensor.bitcast %6588 : tensor<4x?x4x128xf16>{%dim_4202} -> tensor<4x?x4x64xcomplex<f16>>{%dim_4202}
%6590 = torch_c.from_builtin_tensor %6589 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %6590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%6591 = torch.aten.mul.Tensor %6590, %6587 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %6591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%6592 = torch_c.to_builtin_tensor %6591 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_4203 = arith.constant 1 : index
%dim_4204 = tensor.dim %6592, %c1_4203 : tensor<4x?x4x64xcomplex<f32>>
%6593 = flow.tensor.bitcast %6592 : tensor<4x?x4x64xcomplex<f32>>{%dim_4204} -> tensor<4x?x4x128xf32>{%dim_4204}
%6594 = torch_c.from_builtin_tensor %6593 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %6594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_4205 = torch.constant.int 5
%6595 = torch.prims.convert_element_type %6594, %int5_4205 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %6595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_4206 = torch.constant.int 131072
%none_4207 = torch.constant.none
%none_4208 = torch.constant.none
%cpu_4209 = torch.constant.device "cpu"
%false_4210 = torch.constant.bool false
%6596 = torch.aten.arange %int131072_4206, %none_4207, %none_4208, %cpu_4209, %false_4210 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_4211 = torch.constant.int 0
%int128_4212 = torch.constant.int 128
%int2_4213 = torch.constant.int 2
%none_4214 = torch.constant.none
%none_4215 = torch.constant.none
%cpu_4216 = torch.constant.device "cpu"
%false_4217 = torch.constant.bool false
%6597 = torch.aten.arange.start_step %int0_4211, %int128_4212, %int2_4213, %none_4214, %none_4215, %cpu_4216, %false_4217 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_4218 = torch.constant.int 0
%int0_4219 = torch.constant.int 0
%int64_4220 = torch.constant.int 64
%int1_4221 = torch.constant.int 1
%6598 = torch.aten.slice.Tensor %6597, %int0_4218, %int0_4219, %int64_4220, %int1_4221 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_4222 = torch.constant.int 6
%6599 = torch.prims.convert_element_type %6598, %int6_4222 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_4223 = torch.constant.int 128
%6600 = torch.aten.div.Scalar %6599, %int128_4223 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_4224 = torch.constant.float 5.000000e+05
%6601 = torch.aten.pow.Scalar %float5.000000e05_4224, %6600 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%6602 = torch.aten.reciprocal %6601 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_4225 = torch.constant.float 1.000000e+00
%6603 = torch.aten.mul.Scalar %6602, %float1.000000e00_4225 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_4226 = torch.constant.int 131072
%int1_4227 = torch.constant.int 1
%6604 = torch.prim.ListConstruct %int131072_4226, %int1_4227 : (!torch.int, !torch.int) -> !torch.list<int>
%6605 = torch.aten.view %6596, %6604 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%6606 = torch.aten.mul.Tensor %6605, %6603 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%6607 = torch.aten.cos %6606 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%6608 = torch.aten.sin %6606 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%6609 = torch.aten.complex %6607, %6608 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%6610 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6611 = flow.tensor.transfer %6610 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%6612 = torch_c.from_builtin_tensor %6611 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6613 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6614 = flow.tensor.transfer %6613 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%6615 = torch_c.from_builtin_tensor %6614 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6616 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6617 = flow.tensor.transfer %6616 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%6618 = torch_c.from_builtin_tensor %6617 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6619 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6620 = flow.tensor.transfer %6619 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%6621 = torch_c.from_builtin_tensor %6620 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6622 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6623 = flow.tensor.transfer %6622 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%6624 = torch_c.from_builtin_tensor %6623 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6625 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6626 = flow.tensor.transfer %6625 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%6627 = torch_c.from_builtin_tensor %6626 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6628 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6629 = flow.tensor.transfer %6628 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%6630 = torch_c.from_builtin_tensor %6629 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%6631 = torch_c.to_builtin_tensor %6609 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%6632 = flow.tensor.transfer %6631 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%6633 = torch_c.from_builtin_tensor %6632 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_4228 = torch.constant.int 1
%6634 = torch.aten.size.int %6283, %int1_4228 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4229 = torch.constant.int 0
%6635 = torch.aten.add.int %int0_4229, %6634 : !torch.int, !torch.int -> !torch.int
%int0_4230 = torch.constant.int 0
%int0_4231 = torch.constant.int 0
%int1_4232 = torch.constant.int 1
%6636 = torch.aten.slice.Tensor %6612, %int0_4230, %int0_4231, %6635, %int1_4232 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6636, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4233 = torch.constant.int 1
%int0_4234 = torch.constant.int 0
%int9223372036854775807_4235 = torch.constant.int 9223372036854775807
%int1_4236 = torch.constant.int 1
%6637 = torch.aten.slice.Tensor %6636, %int1_4233, %int0_4234, %int9223372036854775807_4235, %int1_4236 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6637, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4237 = torch.constant.int 0
%6638 = torch.aten.unsqueeze %6637, %int0_4237 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6638, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4238 = torch.constant.int 2
%6639 = torch.aten.unsqueeze %6638, %int2_4238 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6639, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4239 = torch.constant.int 3
%int0_4240 = torch.constant.int 0
%int9223372036854775807_4241 = torch.constant.int 9223372036854775807
%int1_4242 = torch.constant.int 1
%6640 = torch.aten.slice.Tensor %6639, %int3_4239, %int0_4240, %int9223372036854775807_4241, %int1_4242 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6640, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6641 = torch_c.to_builtin_tensor %6407 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4243 = arith.constant 1 : index
%dim_4244 = tensor.dim %6641, %c1_4243 : tensor<4x?x1x128xf16>
%6642 = flow.tensor.bitcast %6641 : tensor<4x?x1x128xf16>{%dim_4244} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4244}
%6643 = torch_c.from_builtin_tensor %6642 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6644 = torch.aten.mul.Tensor %6643, %6640 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6645 = torch_c.to_builtin_tensor %6644 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4245 = arith.constant 1 : index
%dim_4246 = tensor.dim %6645, %c1_4245 : tensor<4x?x1x64xcomplex<f32>>
%6646 = flow.tensor.bitcast %6645 : tensor<4x?x1x64xcomplex<f32>>{%dim_4246} -> tensor<4x?x1x128xf32>{%dim_4246}
%6647 = torch_c.from_builtin_tensor %6646 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4247 = torch.constant.int 5
%6648 = torch.prims.convert_element_type %6647, %int5_4247 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4248 = torch.constant.int 1
%6649 = torch.aten.size.int %6289, %int1_4248 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4249 = torch.constant.int 0
%6650 = torch.aten.add.int %int0_4249, %6649 : !torch.int, !torch.int -> !torch.int
%int0_4250 = torch.constant.int 0
%int0_4251 = torch.constant.int 0
%int1_4252 = torch.constant.int 1
%6651 = torch.aten.slice.Tensor %6615, %int0_4250, %int0_4251, %6650, %int1_4252 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6651, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4253 = torch.constant.int 1
%int0_4254 = torch.constant.int 0
%int9223372036854775807_4255 = torch.constant.int 9223372036854775807
%int1_4256 = torch.constant.int 1
%6652 = torch.aten.slice.Tensor %6651, %int1_4253, %int0_4254, %int9223372036854775807_4255, %int1_4256 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6652, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4257 = torch.constant.int 0
%6653 = torch.aten.unsqueeze %6652, %int0_4257 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6653, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4258 = torch.constant.int 2
%6654 = torch.aten.unsqueeze %6653, %int2_4258 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6654, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4259 = torch.constant.int 3
%int0_4260 = torch.constant.int 0
%int9223372036854775807_4261 = torch.constant.int 9223372036854775807
%int1_4262 = torch.constant.int 1
%6655 = torch.aten.slice.Tensor %6654, %int3_4259, %int0_4260, %int9223372036854775807_4261, %int1_4262 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6655, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6656 = torch_c.to_builtin_tensor %6409 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4263 = arith.constant 1 : index
%dim_4264 = tensor.dim %6656, %c1_4263 : tensor<4x?x1x128xf16>
%6657 = flow.tensor.bitcast %6656 : tensor<4x?x1x128xf16>{%dim_4264} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4264}
%6658 = torch_c.from_builtin_tensor %6657 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6659 = torch.aten.mul.Tensor %6658, %6655 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6660 = torch_c.to_builtin_tensor %6659 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4265 = arith.constant 1 : index
%dim_4266 = tensor.dim %6660, %c1_4265 : tensor<4x?x1x64xcomplex<f32>>
%6661 = flow.tensor.bitcast %6660 : tensor<4x?x1x64xcomplex<f32>>{%dim_4266} -> tensor<4x?x1x128xf32>{%dim_4266}
%6662 = torch_c.from_builtin_tensor %6661 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4267 = torch.constant.int 5
%6663 = torch.prims.convert_element_type %6662, %int5_4267 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4268 = torch.constant.int 1
%6664 = torch.aten.size.int %6295, %int1_4268 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4269 = torch.constant.int 0
%6665 = torch.aten.add.int %int0_4269, %6664 : !torch.int, !torch.int -> !torch.int
%int0_4270 = torch.constant.int 0
%int0_4271 = torch.constant.int 0
%int1_4272 = torch.constant.int 1
%6666 = torch.aten.slice.Tensor %6618, %int0_4270, %int0_4271, %6665, %int1_4272 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6666, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4273 = torch.constant.int 1
%int0_4274 = torch.constant.int 0
%int9223372036854775807_4275 = torch.constant.int 9223372036854775807
%int1_4276 = torch.constant.int 1
%6667 = torch.aten.slice.Tensor %6666, %int1_4273, %int0_4274, %int9223372036854775807_4275, %int1_4276 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6667, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4277 = torch.constant.int 0
%6668 = torch.aten.unsqueeze %6667, %int0_4277 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6668, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4278 = torch.constant.int 2
%6669 = torch.aten.unsqueeze %6668, %int2_4278 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6669, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4279 = torch.constant.int 3
%int0_4280 = torch.constant.int 0
%int9223372036854775807_4281 = torch.constant.int 9223372036854775807
%int1_4282 = torch.constant.int 1
%6670 = torch.aten.slice.Tensor %6669, %int3_4279, %int0_4280, %int9223372036854775807_4281, %int1_4282 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6670, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6671 = torch_c.to_builtin_tensor %6411 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4283 = arith.constant 1 : index
%dim_4284 = tensor.dim %6671, %c1_4283 : tensor<4x?x1x128xf16>
%6672 = flow.tensor.bitcast %6671 : tensor<4x?x1x128xf16>{%dim_4284} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4284}
%6673 = torch_c.from_builtin_tensor %6672 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6674 = torch.aten.mul.Tensor %6673, %6670 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6675 = torch_c.to_builtin_tensor %6674 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4285 = arith.constant 1 : index
%dim_4286 = tensor.dim %6675, %c1_4285 : tensor<4x?x1x64xcomplex<f32>>
%6676 = flow.tensor.bitcast %6675 : tensor<4x?x1x64xcomplex<f32>>{%dim_4286} -> tensor<4x?x1x128xf32>{%dim_4286}
%6677 = torch_c.from_builtin_tensor %6676 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4287 = torch.constant.int 5
%6678 = torch.prims.convert_element_type %6677, %int5_4287 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4288 = torch.constant.int 1
%6679 = torch.aten.size.int %6301, %int1_4288 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4289 = torch.constant.int 0
%6680 = torch.aten.add.int %int0_4289, %6679 : !torch.int, !torch.int -> !torch.int
%int0_4290 = torch.constant.int 0
%int0_4291 = torch.constant.int 0
%int1_4292 = torch.constant.int 1
%6681 = torch.aten.slice.Tensor %6621, %int0_4290, %int0_4291, %6680, %int1_4292 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6681, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4293 = torch.constant.int 1
%int0_4294 = torch.constant.int 0
%int9223372036854775807_4295 = torch.constant.int 9223372036854775807
%int1_4296 = torch.constant.int 1
%6682 = torch.aten.slice.Tensor %6681, %int1_4293, %int0_4294, %int9223372036854775807_4295, %int1_4296 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6682, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4297 = torch.constant.int 0
%6683 = torch.aten.unsqueeze %6682, %int0_4297 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6683, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4298 = torch.constant.int 2
%6684 = torch.aten.unsqueeze %6683, %int2_4298 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6684, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4299 = torch.constant.int 3
%int0_4300 = torch.constant.int 0
%int9223372036854775807_4301 = torch.constant.int 9223372036854775807
%int1_4302 = torch.constant.int 1
%6685 = torch.aten.slice.Tensor %6684, %int3_4299, %int0_4300, %int9223372036854775807_4301, %int1_4302 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6685, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6686 = torch_c.to_builtin_tensor %6413 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4303 = arith.constant 1 : index
%dim_4304 = tensor.dim %6686, %c1_4303 : tensor<4x?x1x128xf16>
%6687 = flow.tensor.bitcast %6686 : tensor<4x?x1x128xf16>{%dim_4304} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4304}
%6688 = torch_c.from_builtin_tensor %6687 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6689 = torch.aten.mul.Tensor %6688, %6685 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6690 = torch_c.to_builtin_tensor %6689 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4305 = arith.constant 1 : index
%dim_4306 = tensor.dim %6690, %c1_4305 : tensor<4x?x1x64xcomplex<f32>>
%6691 = flow.tensor.bitcast %6690 : tensor<4x?x1x64xcomplex<f32>>{%dim_4306} -> tensor<4x?x1x128xf32>{%dim_4306}
%6692 = torch_c.from_builtin_tensor %6691 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4307 = torch.constant.int 5
%6693 = torch.prims.convert_element_type %6692, %int5_4307 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4308 = torch.constant.int 1
%6694 = torch.aten.size.int %6307, %int1_4308 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4309 = torch.constant.int 0
%6695 = torch.aten.add.int %int0_4309, %6694 : !torch.int, !torch.int -> !torch.int
%int0_4310 = torch.constant.int 0
%int0_4311 = torch.constant.int 0
%int1_4312 = torch.constant.int 1
%6696 = torch.aten.slice.Tensor %6624, %int0_4310, %int0_4311, %6695, %int1_4312 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6696, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4313 = torch.constant.int 1
%int0_4314 = torch.constant.int 0
%int9223372036854775807_4315 = torch.constant.int 9223372036854775807
%int1_4316 = torch.constant.int 1
%6697 = torch.aten.slice.Tensor %6696, %int1_4313, %int0_4314, %int9223372036854775807_4315, %int1_4316 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6697, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4317 = torch.constant.int 0
%6698 = torch.aten.unsqueeze %6697, %int0_4317 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6698, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4318 = torch.constant.int 2
%6699 = torch.aten.unsqueeze %6698, %int2_4318 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6699, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4319 = torch.constant.int 3
%int0_4320 = torch.constant.int 0
%int9223372036854775807_4321 = torch.constant.int 9223372036854775807
%int1_4322 = torch.constant.int 1
%6700 = torch.aten.slice.Tensor %6699, %int3_4319, %int0_4320, %int9223372036854775807_4321, %int1_4322 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6700, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6701 = torch_c.to_builtin_tensor %6415 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4323 = arith.constant 1 : index
%dim_4324 = tensor.dim %6701, %c1_4323 : tensor<4x?x1x128xf16>
%6702 = flow.tensor.bitcast %6701 : tensor<4x?x1x128xf16>{%dim_4324} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4324}
%6703 = torch_c.from_builtin_tensor %6702 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6704 = torch.aten.mul.Tensor %6703, %6700 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6705 = torch_c.to_builtin_tensor %6704 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4325 = arith.constant 1 : index
%dim_4326 = tensor.dim %6705, %c1_4325 : tensor<4x?x1x64xcomplex<f32>>
%6706 = flow.tensor.bitcast %6705 : tensor<4x?x1x64xcomplex<f32>>{%dim_4326} -> tensor<4x?x1x128xf32>{%dim_4326}
%6707 = torch_c.from_builtin_tensor %6706 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4327 = torch.constant.int 5
%6708 = torch.prims.convert_element_type %6707, %int5_4327 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4328 = torch.constant.int 1
%6709 = torch.aten.size.int %6313, %int1_4328 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4329 = torch.constant.int 0
%6710 = torch.aten.add.int %int0_4329, %6709 : !torch.int, !torch.int -> !torch.int
%int0_4330 = torch.constant.int 0
%int0_4331 = torch.constant.int 0
%int1_4332 = torch.constant.int 1
%6711 = torch.aten.slice.Tensor %6627, %int0_4330, %int0_4331, %6710, %int1_4332 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6711, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4333 = torch.constant.int 1
%int0_4334 = torch.constant.int 0
%int9223372036854775807_4335 = torch.constant.int 9223372036854775807
%int1_4336 = torch.constant.int 1
%6712 = torch.aten.slice.Tensor %6711, %int1_4333, %int0_4334, %int9223372036854775807_4335, %int1_4336 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6712, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4337 = torch.constant.int 0
%6713 = torch.aten.unsqueeze %6712, %int0_4337 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6713, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4338 = torch.constant.int 2
%6714 = torch.aten.unsqueeze %6713, %int2_4338 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6714, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4339 = torch.constant.int 3
%int0_4340 = torch.constant.int 0
%int9223372036854775807_4341 = torch.constant.int 9223372036854775807
%int1_4342 = torch.constant.int 1
%6715 = torch.aten.slice.Tensor %6714, %int3_4339, %int0_4340, %int9223372036854775807_4341, %int1_4342 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6715, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6716 = torch_c.to_builtin_tensor %6417 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4343 = arith.constant 1 : index
%dim_4344 = tensor.dim %6716, %c1_4343 : tensor<4x?x1x128xf16>
%6717 = flow.tensor.bitcast %6716 : tensor<4x?x1x128xf16>{%dim_4344} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4344}
%6718 = torch_c.from_builtin_tensor %6717 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6719 = torch.aten.mul.Tensor %6718, %6715 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6720 = torch_c.to_builtin_tensor %6719 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4345 = arith.constant 1 : index
%dim_4346 = tensor.dim %6720, %c1_4345 : tensor<4x?x1x64xcomplex<f32>>
%6721 = flow.tensor.bitcast %6720 : tensor<4x?x1x64xcomplex<f32>>{%dim_4346} -> tensor<4x?x1x128xf32>{%dim_4346}
%6722 = torch_c.from_builtin_tensor %6721 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4347 = torch.constant.int 5
%6723 = torch.prims.convert_element_type %6722, %int5_4347 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4348 = torch.constant.int 1
%6724 = torch.aten.size.int %6319, %int1_4348 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4349 = torch.constant.int 0
%6725 = torch.aten.add.int %int0_4349, %6724 : !torch.int, !torch.int -> !torch.int
%int0_4350 = torch.constant.int 0
%int0_4351 = torch.constant.int 0
%int1_4352 = torch.constant.int 1
%6726 = torch.aten.slice.Tensor %6630, %int0_4350, %int0_4351, %6725, %int1_4352 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6726, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4353 = torch.constant.int 1
%int0_4354 = torch.constant.int 0
%int9223372036854775807_4355 = torch.constant.int 9223372036854775807
%int1_4356 = torch.constant.int 1
%6727 = torch.aten.slice.Tensor %6726, %int1_4353, %int0_4354, %int9223372036854775807_4355, %int1_4356 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6727, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4357 = torch.constant.int 0
%6728 = torch.aten.unsqueeze %6727, %int0_4357 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6728, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4358 = torch.constant.int 2
%6729 = torch.aten.unsqueeze %6728, %int2_4358 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6729, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4359 = torch.constant.int 3
%int0_4360 = torch.constant.int 0
%int9223372036854775807_4361 = torch.constant.int 9223372036854775807
%int1_4362 = torch.constant.int 1
%6730 = torch.aten.slice.Tensor %6729, %int3_4359, %int0_4360, %int9223372036854775807_4361, %int1_4362 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6730, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6731 = torch_c.to_builtin_tensor %6419 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4363 = arith.constant 1 : index
%dim_4364 = tensor.dim %6731, %c1_4363 : tensor<4x?x1x128xf16>
%6732 = flow.tensor.bitcast %6731 : tensor<4x?x1x128xf16>{%dim_4364} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4364}
%6733 = torch_c.from_builtin_tensor %6732 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6734 = torch.aten.mul.Tensor %6733, %6730 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6735 = torch_c.to_builtin_tensor %6734 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4365 = arith.constant 1 : index
%dim_4366 = tensor.dim %6735, %c1_4365 : tensor<4x?x1x64xcomplex<f32>>
%6736 = flow.tensor.bitcast %6735 : tensor<4x?x1x64xcomplex<f32>>{%dim_4366} -> tensor<4x?x1x128xf32>{%dim_4366}
%6737 = torch_c.from_builtin_tensor %6736 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4367 = torch.constant.int 5
%6738 = torch.prims.convert_element_type %6737, %int5_4367 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_4368 = torch.constant.int 1
%6739 = torch.aten.size.int %6325, %int1_4368 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_4369 = torch.constant.int 0
%6740 = torch.aten.add.int %int0_4369, %6739 : !torch.int, !torch.int -> !torch.int
%int0_4370 = torch.constant.int 0
%int0_4371 = torch.constant.int 0
%int1_4372 = torch.constant.int 1
%6741 = torch.aten.slice.Tensor %6633, %int0_4370, %int0_4371, %6740, %int1_4372 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6741, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_4373 = torch.constant.int 1
%int0_4374 = torch.constant.int 0
%int9223372036854775807_4375 = torch.constant.int 9223372036854775807
%int1_4376 = torch.constant.int 1
%6742 = torch.aten.slice.Tensor %6741, %int1_4373, %int0_4374, %int9223372036854775807_4375, %int1_4376 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %6742, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_4377 = torch.constant.int 0
%6743 = torch.aten.unsqueeze %6742, %int0_4377 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %6743, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_4378 = torch.constant.int 2
%6744 = torch.aten.unsqueeze %6743, %int2_4378 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6744, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_4379 = torch.constant.int 3
%int0_4380 = torch.constant.int 0
%int9223372036854775807_4381 = torch.constant.int 9223372036854775807
%int1_4382 = torch.constant.int 1
%6745 = torch.aten.slice.Tensor %6744, %int3_4379, %int0_4380, %int9223372036854775807_4381, %int1_4382 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6745, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%6746 = torch_c.to_builtin_tensor %6421 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_4383 = arith.constant 1 : index
%dim_4384 = tensor.dim %6746, %c1_4383 : tensor<4x?x1x128xf16>
%6747 = flow.tensor.bitcast %6746 : tensor<4x?x1x128xf16>{%dim_4384} -> tensor<4x?x1x64xcomplex<f16>>{%dim_4384}
%6748 = torch_c.from_builtin_tensor %6747 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %6748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%6749 = torch.aten.mul.Tensor %6748, %6745 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %6749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%6750 = torch_c.to_builtin_tensor %6749 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_4385 = arith.constant 1 : index
%dim_4386 = tensor.dim %6750, %c1_4385 : tensor<4x?x1x64xcomplex<f32>>
%6751 = flow.tensor.bitcast %6750 : tensor<4x?x1x64xcomplex<f32>>{%dim_4386} -> tensor<4x?x1x128xf32>{%dim_4386}
%6752 = torch_c.from_builtin_tensor %6751 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %6752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_4387 = torch.constant.int 5
%6753 = torch.prims.convert_element_type %6752, %int5_4387 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %6753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_4388 = torch.constant.int 64
%6754 = torch.aten.mul.Scalar %2364, %int64_4388 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6754, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4389 = torch.constant.int 64
%6755 = torch.aten.mul.Scalar %2367, %int64_4389 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6755, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4390 = torch.constant.int 64
%6756 = torch.aten.mul.Scalar %2370, %int64_4390 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6756, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4391 = torch.constant.int 64
%6757 = torch.aten.mul.Scalar %2373, %int64_4391 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6757, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4392 = torch.constant.int 64
%6758 = torch.aten.mul.Scalar %2376, %int64_4392 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6758, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4393 = torch.constant.int 64
%6759 = torch.aten.mul.Scalar %2379, %int64_4393 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6759, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4394 = torch.constant.int 64
%6760 = torch.aten.mul.Scalar %2382, %int64_4394 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6760, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_4395 = torch.constant.int 64
%6761 = torch.aten.mul.Scalar %2385, %int64_4395 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6761, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4396 = torch.constant.int 4
%int1_4397 = torch.constant.int 1
%6762 = torch.aten.add.Scalar %6754, %int4_4396, %int1_4397 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6762, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4398 = torch.constant.int 4
%int1_4399 = torch.constant.int 1
%6763 = torch.aten.add.Scalar %6755, %int4_4398, %int1_4399 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6763, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4400 = torch.constant.int 4
%int1_4401 = torch.constant.int 1
%6764 = torch.aten.add.Scalar %6756, %int4_4400, %int1_4401 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6764, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4402 = torch.constant.int 4
%int1_4403 = torch.constant.int 1
%6765 = torch.aten.add.Scalar %6757, %int4_4402, %int1_4403 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6765, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4404 = torch.constant.int 4
%int1_4405 = torch.constant.int 1
%6766 = torch.aten.add.Scalar %6758, %int4_4404, %int1_4405 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6766, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4406 = torch.constant.int 4
%int1_4407 = torch.constant.int 1
%6767 = torch.aten.add.Scalar %6759, %int4_4406, %int1_4407 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6767, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4408 = torch.constant.int 4
%int1_4409 = torch.constant.int 1
%6768 = torch.aten.add.Scalar %6760, %int4_4408, %int1_4409 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6768, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4410 = torch.constant.int 4
%int1_4411 = torch.constant.int 1
%6769 = torch.aten.add.Scalar %6761, %int4_4410, %int1_4411 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6769, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4412 = torch.constant.int 4
%int16_4413 = torch.constant.int 16
%int1_4414 = torch.constant.int 1
%int128_4415 = torch.constant.int 128
%6770 = torch.prim.ListConstruct %int4_4412, %3095, %int16_4413, %int1_4414, %int128_4415 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6771 = torch.aten.view %6648, %6770 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6771, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4416 = torch.constant.int 4
%int16_4417 = torch.constant.int 16
%int1_4418 = torch.constant.int 1
%int128_4419 = torch.constant.int 128
%6772 = torch.prim.ListConstruct %int4_4416, %3095, %int16_4417, %int1_4418, %int128_4419 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6773 = torch.aten.view %6663, %6772 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6773, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4420 = torch.constant.int 4
%int16_4421 = torch.constant.int 16
%int1_4422 = torch.constant.int 1
%int128_4423 = torch.constant.int 128
%6774 = torch.prim.ListConstruct %int4_4420, %3095, %int16_4421, %int1_4422, %int128_4423 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6775 = torch.aten.view %6678, %6774 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6775, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4424 = torch.constant.int 4
%int16_4425 = torch.constant.int 16
%int1_4426 = torch.constant.int 1
%int128_4427 = torch.constant.int 128
%6776 = torch.prim.ListConstruct %int4_4424, %3095, %int16_4425, %int1_4426, %int128_4427 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6777 = torch.aten.view %6693, %6776 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6777, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4428 = torch.constant.int 4
%int16_4429 = torch.constant.int 16
%int1_4430 = torch.constant.int 1
%int128_4431 = torch.constant.int 128
%6778 = torch.prim.ListConstruct %int4_4428, %3095, %int16_4429, %int1_4430, %int128_4431 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6779 = torch.aten.view %6708, %6778 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6779, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4432 = torch.constant.int 4
%int16_4433 = torch.constant.int 16
%int1_4434 = torch.constant.int 1
%int128_4435 = torch.constant.int 128
%6780 = torch.prim.ListConstruct %int4_4432, %3095, %int16_4433, %int1_4434, %int128_4435 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6781 = torch.aten.view %6723, %6780 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6781, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4436 = torch.constant.int 4
%int16_4437 = torch.constant.int 16
%int1_4438 = torch.constant.int 1
%int128_4439 = torch.constant.int 128
%6782 = torch.prim.ListConstruct %int4_4436, %3095, %int16_4437, %int1_4438, %int128_4439 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6783 = torch.aten.view %6738, %6782 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6783, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4440 = torch.constant.int 4
%int16_4441 = torch.constant.int 16
%int1_4442 = torch.constant.int 1
%int128_4443 = torch.constant.int 128
%6784 = torch.prim.ListConstruct %int4_4440, %3095, %int16_4441, %int1_4442, %int128_4443 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6785 = torch.aten.view %6753, %6784 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6785, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4444 = torch.constant.int 4
%6786 = torch.aten.mul.int %int4_4444, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4445 = torch.constant.int 16
%int1_4446 = torch.constant.int 1
%int128_4447 = torch.constant.int 128
%6787 = torch.prim.ListConstruct %6786, %int16_4445, %int1_4446, %int128_4447 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6788 = torch.aten.view %6771, %6787 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6788, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4448 = torch.constant.int 4
%6789 = torch.aten.mul.int %int4_4448, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4449 = torch.constant.int 16
%int1_4450 = torch.constant.int 1
%int128_4451 = torch.constant.int 128
%6790 = torch.prim.ListConstruct %6789, %int16_4449, %int1_4450, %int128_4451 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6791 = torch.aten.view %6773, %6790 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6791, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4452 = torch.constant.int 4
%6792 = torch.aten.mul.int %int4_4452, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4453 = torch.constant.int 16
%int1_4454 = torch.constant.int 1
%int128_4455 = torch.constant.int 128
%6793 = torch.prim.ListConstruct %6792, %int16_4453, %int1_4454, %int128_4455 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6794 = torch.aten.view %6775, %6793 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6794, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4456 = torch.constant.int 4
%6795 = torch.aten.mul.int %int4_4456, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4457 = torch.constant.int 16
%int1_4458 = torch.constant.int 1
%int128_4459 = torch.constant.int 128
%6796 = torch.prim.ListConstruct %6795, %int16_4457, %int1_4458, %int128_4459 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6797 = torch.aten.view %6777, %6796 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6797, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4460 = torch.constant.int 4
%6798 = torch.aten.mul.int %int4_4460, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4461 = torch.constant.int 16
%int1_4462 = torch.constant.int 1
%int128_4463 = torch.constant.int 128
%6799 = torch.prim.ListConstruct %6798, %int16_4461, %int1_4462, %int128_4463 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6800 = torch.aten.view %6779, %6799 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6800, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4464 = torch.constant.int 4
%6801 = torch.aten.mul.int %int4_4464, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4465 = torch.constant.int 16
%int1_4466 = torch.constant.int 1
%int128_4467 = torch.constant.int 128
%6802 = torch.prim.ListConstruct %6801, %int16_4465, %int1_4466, %int128_4467 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6803 = torch.aten.view %6781, %6802 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6803, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4468 = torch.constant.int 4
%6804 = torch.aten.mul.int %int4_4468, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4469 = torch.constant.int 16
%int1_4470 = torch.constant.int 1
%int128_4471 = torch.constant.int 128
%6805 = torch.prim.ListConstruct %6804, %int16_4469, %int1_4470, %int128_4471 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6806 = torch.aten.view %6783, %6805 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6806, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4472 = torch.constant.int 4
%6807 = torch.aten.mul.int %int4_4472, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4473 = torch.constant.int 16
%int1_4474 = torch.constant.int 1
%int128_4475 = torch.constant.int 128
%6808 = torch.prim.ListConstruct %6807, %int16_4473, %int1_4474, %int128_4475 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6809 = torch.aten.view %6785, %6808 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6809, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4476 = torch.constant.int 4
%6810 = torch.aten.mul.int %int4_4476, %3095 : !torch.int, !torch.int -> !torch.int
%6811 = torch.prim.ListConstruct %6810 : (!torch.int) -> !torch.list<int>
%6812 = torch.aten.view %6762, %6811 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6812, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4477 = torch.constant.int 4
%6813 = torch.aten.mul.int %int4_4477, %3095 : !torch.int, !torch.int -> !torch.int
%6814 = torch.prim.ListConstruct %6813 : (!torch.int) -> !torch.list<int>
%6815 = torch.aten.view %6763, %6814 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6815, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4478 = torch.constant.int 4
%6816 = torch.aten.mul.int %int4_4478, %3095 : !torch.int, !torch.int -> !torch.int
%6817 = torch.prim.ListConstruct %6816 : (!torch.int) -> !torch.list<int>
%6818 = torch.aten.view %6764, %6817 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6818, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4479 = torch.constant.int 4
%6819 = torch.aten.mul.int %int4_4479, %3095 : !torch.int, !torch.int -> !torch.int
%6820 = torch.prim.ListConstruct %6819 : (!torch.int) -> !torch.list<int>
%6821 = torch.aten.view %6765, %6820 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6821, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4480 = torch.constant.int 4
%6822 = torch.aten.mul.int %int4_4480, %3095 : !torch.int, !torch.int -> !torch.int
%6823 = torch.prim.ListConstruct %6822 : (!torch.int) -> !torch.list<int>
%6824 = torch.aten.view %6766, %6823 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6824, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4481 = torch.constant.int 4
%6825 = torch.aten.mul.int %int4_4481, %3095 : !torch.int, !torch.int -> !torch.int
%6826 = torch.prim.ListConstruct %6825 : (!torch.int) -> !torch.list<int>
%6827 = torch.aten.view %6767, %6826 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6827, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4482 = torch.constant.int 4
%6828 = torch.aten.mul.int %int4_4482, %3095 : !torch.int, !torch.int -> !torch.int
%6829 = torch.prim.ListConstruct %6828 : (!torch.int) -> !torch.list<int>
%6830 = torch.aten.view %6768, %6829 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6830, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4483 = torch.constant.int 4
%6831 = torch.aten.mul.int %int4_4483, %3095 : !torch.int, !torch.int -> !torch.int
%6832 = torch.prim.ListConstruct %6831 : (!torch.int) -> !torch.list<int>
%6833 = torch.aten.view %6769, %6832 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6833, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4484 = torch.constant.int 4
%int16_4485 = torch.constant.int 16
%int1_4486 = torch.constant.int 1
%int128_4487 = torch.constant.int 128
%6834 = torch.prim.ListConstruct %int4_4484, %3095, %int16_4485, %int1_4486, %int128_4487 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6835 = torch.aten.view %6423, %6834 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6835, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4488 = torch.constant.int 4
%int16_4489 = torch.constant.int 16
%int1_4490 = torch.constant.int 1
%int128_4491 = torch.constant.int 128
%6836 = torch.prim.ListConstruct %int4_4488, %3095, %int16_4489, %int1_4490, %int128_4491 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6837 = torch.aten.view %6425, %6836 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6837, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4492 = torch.constant.int 4
%int16_4493 = torch.constant.int 16
%int1_4494 = torch.constant.int 1
%int128_4495 = torch.constant.int 128
%6838 = torch.prim.ListConstruct %int4_4492, %3095, %int16_4493, %int1_4494, %int128_4495 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6839 = torch.aten.view %6427, %6838 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6839, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4496 = torch.constant.int 4
%int16_4497 = torch.constant.int 16
%int1_4498 = torch.constant.int 1
%int128_4499 = torch.constant.int 128
%6840 = torch.prim.ListConstruct %int4_4496, %3095, %int16_4497, %int1_4498, %int128_4499 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6841 = torch.aten.view %6429, %6840 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6841, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4500 = torch.constant.int 4
%int16_4501 = torch.constant.int 16
%int1_4502 = torch.constant.int 1
%int128_4503 = torch.constant.int 128
%6842 = torch.prim.ListConstruct %int4_4500, %3095, %int16_4501, %int1_4502, %int128_4503 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6843 = torch.aten.view %6431, %6842 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6843, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4504 = torch.constant.int 4
%int16_4505 = torch.constant.int 16
%int1_4506 = torch.constant.int 1
%int128_4507 = torch.constant.int 128
%6844 = torch.prim.ListConstruct %int4_4504, %3095, %int16_4505, %int1_4506, %int128_4507 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6845 = torch.aten.view %6433, %6844 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6845, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4508 = torch.constant.int 4
%int16_4509 = torch.constant.int 16
%int1_4510 = torch.constant.int 1
%int128_4511 = torch.constant.int 128
%6846 = torch.prim.ListConstruct %int4_4508, %3095, %int16_4509, %int1_4510, %int128_4511 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6847 = torch.aten.view %6435, %6846 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6847, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4512 = torch.constant.int 4
%int16_4513 = torch.constant.int 16
%int1_4514 = torch.constant.int 1
%int128_4515 = torch.constant.int 128
%6848 = torch.prim.ListConstruct %int4_4512, %3095, %int16_4513, %int1_4514, %int128_4515 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6849 = torch.aten.view %6437, %6848 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %6849, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_4516 = torch.constant.int 4
%6850 = torch.aten.mul.int %int4_4516, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4517 = torch.constant.int 16
%int1_4518 = torch.constant.int 1
%int128_4519 = torch.constant.int 128
%6851 = torch.prim.ListConstruct %6850, %int16_4517, %int1_4518, %int128_4519 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6852 = torch.aten.view %6835, %6851 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6852, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4520 = torch.constant.int 4
%6853 = torch.aten.mul.int %int4_4520, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4521 = torch.constant.int 16
%int1_4522 = torch.constant.int 1
%int128_4523 = torch.constant.int 128
%6854 = torch.prim.ListConstruct %6853, %int16_4521, %int1_4522, %int128_4523 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6855 = torch.aten.view %6837, %6854 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6855, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4524 = torch.constant.int 4
%6856 = torch.aten.mul.int %int4_4524, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4525 = torch.constant.int 16
%int1_4526 = torch.constant.int 1
%int128_4527 = torch.constant.int 128
%6857 = torch.prim.ListConstruct %6856, %int16_4525, %int1_4526, %int128_4527 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6858 = torch.aten.view %6839, %6857 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6858, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4528 = torch.constant.int 4
%6859 = torch.aten.mul.int %int4_4528, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4529 = torch.constant.int 16
%int1_4530 = torch.constant.int 1
%int128_4531 = torch.constant.int 128
%6860 = torch.prim.ListConstruct %6859, %int16_4529, %int1_4530, %int128_4531 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6861 = torch.aten.view %6841, %6860 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6861, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4532 = torch.constant.int 4
%6862 = torch.aten.mul.int %int4_4532, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4533 = torch.constant.int 16
%int1_4534 = torch.constant.int 1
%int128_4535 = torch.constant.int 128
%6863 = torch.prim.ListConstruct %6862, %int16_4533, %int1_4534, %int128_4535 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6864 = torch.aten.view %6843, %6863 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6864, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4536 = torch.constant.int 4
%6865 = torch.aten.mul.int %int4_4536, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4537 = torch.constant.int 16
%int1_4538 = torch.constant.int 1
%int128_4539 = torch.constant.int 128
%6866 = torch.prim.ListConstruct %6865, %int16_4537, %int1_4538, %int128_4539 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6867 = torch.aten.view %6845, %6866 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6867, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4540 = torch.constant.int 4
%6868 = torch.aten.mul.int %int4_4540, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4541 = torch.constant.int 16
%int1_4542 = torch.constant.int 1
%int128_4543 = torch.constant.int 128
%6869 = torch.prim.ListConstruct %6868, %int16_4541, %int1_4542, %int128_4543 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6870 = torch.aten.view %6847, %6869 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6870, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_4544 = torch.constant.int 4
%6871 = torch.aten.mul.int %int4_4544, %3095 : !torch.int, !torch.int -> !torch.int
%int16_4545 = torch.constant.int 16
%int1_4546 = torch.constant.int 1
%int128_4547 = torch.constant.int 128
%6872 = torch.prim.ListConstruct %6871, %int16_4545, %int1_4546, %int128_4547 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6873 = torch.aten.view %6849, %6872 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6873, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_4548 = torch.constant.int 1
%int1_4549 = torch.constant.int 1
%6874 = torch.aten.add.Scalar %6762, %int1_4548, %int1_4549 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6874, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4550 = torch.constant.int 1
%int1_4551 = torch.constant.int 1
%6875 = torch.aten.add.Scalar %6763, %int1_4550, %int1_4551 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6875, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4552 = torch.constant.int 1
%int1_4553 = torch.constant.int 1
%6876 = torch.aten.add.Scalar %6764, %int1_4552, %int1_4553 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6876, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4554 = torch.constant.int 1
%int1_4555 = torch.constant.int 1
%6877 = torch.aten.add.Scalar %6765, %int1_4554, %int1_4555 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6877, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4556 = torch.constant.int 1
%int1_4557 = torch.constant.int 1
%6878 = torch.aten.add.Scalar %6766, %int1_4556, %int1_4557 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6878, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4558 = torch.constant.int 1
%int1_4559 = torch.constant.int 1
%6879 = torch.aten.add.Scalar %6767, %int1_4558, %int1_4559 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6879, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4560 = torch.constant.int 1
%int1_4561 = torch.constant.int 1
%6880 = torch.aten.add.Scalar %6768, %int1_4560, %int1_4561 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6880, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_4562 = torch.constant.int 1
%int1_4563 = torch.constant.int 1
%6881 = torch.aten.add.Scalar %6769, %int1_4562, %int1_4563 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %6881, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_4564 = torch.constant.int 4
%6882 = torch.aten.mul.int %int4_4564, %3095 : !torch.int, !torch.int -> !torch.int
%6883 = torch.prim.ListConstruct %6882 : (!torch.int) -> !torch.list<int>
%6884 = torch.aten.view %6874, %6883 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6884, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4565 = torch.constant.int 4
%6885 = torch.aten.mul.int %int4_4565, %3095 : !torch.int, !torch.int -> !torch.int
%6886 = torch.prim.ListConstruct %6885 : (!torch.int) -> !torch.list<int>
%6887 = torch.aten.view %6875, %6886 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6887, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4566 = torch.constant.int 4
%6888 = torch.aten.mul.int %int4_4566, %3095 : !torch.int, !torch.int -> !torch.int
%6889 = torch.prim.ListConstruct %6888 : (!torch.int) -> !torch.list<int>
%6890 = torch.aten.view %6876, %6889 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6890, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4567 = torch.constant.int 4
%6891 = torch.aten.mul.int %int4_4567, %3095 : !torch.int, !torch.int -> !torch.int
%6892 = torch.prim.ListConstruct %6891 : (!torch.int) -> !torch.list<int>
%6893 = torch.aten.view %6877, %6892 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6893, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4568 = torch.constant.int 4
%6894 = torch.aten.mul.int %int4_4568, %3095 : !torch.int, !torch.int -> !torch.int
%6895 = torch.prim.ListConstruct %6894 : (!torch.int) -> !torch.list<int>
%6896 = torch.aten.view %6878, %6895 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6896, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4569 = torch.constant.int 4
%6897 = torch.aten.mul.int %int4_4569, %3095 : !torch.int, !torch.int -> !torch.int
%6898 = torch.prim.ListConstruct %6897 : (!torch.int) -> !torch.list<int>
%6899 = torch.aten.view %6879, %6898 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6899, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4570 = torch.constant.int 4
%6900 = torch.aten.mul.int %int4_4570, %3095 : !torch.int, !torch.int -> !torch.int
%6901 = torch.prim.ListConstruct %6900 : (!torch.int) -> !torch.list<int>
%6902 = torch.aten.view %6880, %6901 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6902, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_4571 = torch.constant.int 4
%6903 = torch.aten.mul.int %int4_4571, %3095 : !torch.int, !torch.int -> !torch.int
%6904 = torch.prim.ListConstruct %6903 : (!torch.int) -> !torch.list<int>
%6905 = torch.aten.view %6881, %6904 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6905, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%6906 = torch.prim.ListConstruct %6812, %6884 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4572 = torch.constant.int 0
%6907 = torch.aten.cat %6906, %int0_4572 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6907, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6908 = torch.prim.ListConstruct %6815, %6887 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4573 = torch.constant.int 0
%6909 = torch.aten.cat %6908, %int0_4573 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6909, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6910 = torch.prim.ListConstruct %6818, %6890 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4574 = torch.constant.int 0
%6911 = torch.aten.cat %6910, %int0_4574 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6911, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6912 = torch.prim.ListConstruct %6821, %6893 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4575 = torch.constant.int 0
%6913 = torch.aten.cat %6912, %int0_4575 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6913, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6914 = torch.prim.ListConstruct %6824, %6896 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4576 = torch.constant.int 0
%6915 = torch.aten.cat %6914, %int0_4576 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6915, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6916 = torch.prim.ListConstruct %6827, %6899 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4577 = torch.constant.int 0
%6917 = torch.aten.cat %6916, %int0_4577 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6917, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6918 = torch.prim.ListConstruct %6830, %6902 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4578 = torch.constant.int 0
%6919 = torch.aten.cat %6918, %int0_4578 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6919, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6920 = torch.prim.ListConstruct %6833, %6905 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_4579 = torch.constant.int 0
%6921 = torch.aten.cat %6920, %int0_4579 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %6921, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%6922 = torch.prim.ListConstruct %6788, %6852 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4580 = torch.constant.int 0
%6923 = torch.aten.cat %6922, %int0_4580 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6923, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6924 = torch.prim.ListConstruct %6791, %6855 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4581 = torch.constant.int 0
%6925 = torch.aten.cat %6924, %int0_4581 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6925, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6926 = torch.prim.ListConstruct %6794, %6858 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4582 = torch.constant.int 0
%6927 = torch.aten.cat %6926, %int0_4582 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6927, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6928 = torch.prim.ListConstruct %6797, %6861 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4583 = torch.constant.int 0
%6929 = torch.aten.cat %6928, %int0_4583 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6929, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6930 = torch.prim.ListConstruct %6800, %6864 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4584 = torch.constant.int 0
%6931 = torch.aten.cat %6930, %int0_4584 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6931, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6932 = torch.prim.ListConstruct %6803, %6867 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4585 = torch.constant.int 0
%6933 = torch.aten.cat %6932, %int0_4585 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6933, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6934 = torch.prim.ListConstruct %6806, %6870 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4586 = torch.constant.int 0
%6935 = torch.aten.cat %6934, %int0_4586 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6935, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6936 = torch.prim.ListConstruct %6809, %6873 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_4587 = torch.constant.int 0
%6937 = torch.aten.cat %6936, %int0_4587 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6937, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4588 = torch.constant.int 32
%int2_4589 = torch.constant.int 2
%int16_4590 = torch.constant.int 16
%int1_4591 = torch.constant.int 1
%int128_4592 = torch.constant.int 128
%6938 = torch.prim.ListConstruct %3023, %int32_4588, %int2_4589, %int16_4590, %int1_4591, %int128_4592 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6939 = torch.aten.view %5088, %6938 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6939, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4593 = torch.constant.int 32
%6940 = torch.aten.mul.int %3023, %int32_4593 : !torch.int, !torch.int -> !torch.int
%int2_4594 = torch.constant.int 2
%6941 = torch.aten.mul.int %6940, %int2_4594 : !torch.int, !torch.int -> !torch.int
%int16_4595 = torch.constant.int 16
%int1_4596 = torch.constant.int 1
%int128_4597 = torch.constant.int 128
%6942 = torch.prim.ListConstruct %6941, %int16_4595, %int1_4596, %int128_4597 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6943 = torch.aten.view %6939, %6942 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6943, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6944 = torch.prim.ListConstruct %6907 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4598 = torch.constant.bool false
%6945 = torch.aten.index_put %6943, %6944, %6923, %false_4598 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6945, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4599 = torch.constant.int 32
%int2_4600 = torch.constant.int 2
%int16_4601 = torch.constant.int 16
%int1_4602 = torch.constant.int 1
%int128_4603 = torch.constant.int 128
%6946 = torch.prim.ListConstruct %3023, %int32_4599, %int2_4600, %int16_4601, %int1_4602, %int128_4603 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6947 = torch.aten.view %6945, %6946 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6947, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4604 = torch.constant.int 131072
%6948 = torch.prim.ListConstruct %3023, %int131072_4604 : (!torch.int, !torch.int) -> !torch.list<int>
%6949 = torch.aten.view %6947, %6948 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %6949, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4605 = torch.constant.int 32
%int2_4606 = torch.constant.int 2
%int16_4607 = torch.constant.int 16
%int1_4608 = torch.constant.int 1
%int128_4609 = torch.constant.int 128
%6950 = torch.prim.ListConstruct %3026, %int32_4605, %int2_4606, %int16_4607, %int1_4608, %int128_4609 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6951 = torch.aten.view %5100, %6950 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6951, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4610 = torch.constant.int 32
%6952 = torch.aten.mul.int %3026, %int32_4610 : !torch.int, !torch.int -> !torch.int
%int2_4611 = torch.constant.int 2
%6953 = torch.aten.mul.int %6952, %int2_4611 : !torch.int, !torch.int -> !torch.int
%int16_4612 = torch.constant.int 16
%int1_4613 = torch.constant.int 1
%int128_4614 = torch.constant.int 128
%6954 = torch.prim.ListConstruct %6953, %int16_4612, %int1_4613, %int128_4614 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6955 = torch.aten.view %6951, %6954 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6955, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6956 = torch.prim.ListConstruct %6909 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4615 = torch.constant.bool false
%6957 = torch.aten.index_put %6955, %6956, %6925, %false_4615 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6957, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4616 = torch.constant.int 32
%int2_4617 = torch.constant.int 2
%int16_4618 = torch.constant.int 16
%int1_4619 = torch.constant.int 1
%int128_4620 = torch.constant.int 128
%6958 = torch.prim.ListConstruct %3026, %int32_4616, %int2_4617, %int16_4618, %int1_4619, %int128_4620 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6959 = torch.aten.view %6957, %6958 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6959, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4621 = torch.constant.int 131072
%6960 = torch.prim.ListConstruct %3026, %int131072_4621 : (!torch.int, !torch.int) -> !torch.list<int>
%6961 = torch.aten.view %6959, %6960 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %6961, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4622 = torch.constant.int 32
%int2_4623 = torch.constant.int 2
%int16_4624 = torch.constant.int 16
%int1_4625 = torch.constant.int 1
%int128_4626 = torch.constant.int 128
%6962 = torch.prim.ListConstruct %3029, %int32_4622, %int2_4623, %int16_4624, %int1_4625, %int128_4626 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6963 = torch.aten.view %5112, %6962 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6963, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4627 = torch.constant.int 32
%6964 = torch.aten.mul.int %3029, %int32_4627 : !torch.int, !torch.int -> !torch.int
%int2_4628 = torch.constant.int 2
%6965 = torch.aten.mul.int %6964, %int2_4628 : !torch.int, !torch.int -> !torch.int
%int16_4629 = torch.constant.int 16
%int1_4630 = torch.constant.int 1
%int128_4631 = torch.constant.int 128
%6966 = torch.prim.ListConstruct %6965, %int16_4629, %int1_4630, %int128_4631 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6967 = torch.aten.view %6963, %6966 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6967, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6968 = torch.prim.ListConstruct %6911 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4632 = torch.constant.bool false
%6969 = torch.aten.index_put %6967, %6968, %6927, %false_4632 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6969, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4633 = torch.constant.int 32
%int2_4634 = torch.constant.int 2
%int16_4635 = torch.constant.int 16
%int1_4636 = torch.constant.int 1
%int128_4637 = torch.constant.int 128
%6970 = torch.prim.ListConstruct %3029, %int32_4633, %int2_4634, %int16_4635, %int1_4636, %int128_4637 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6971 = torch.aten.view %6969, %6970 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6971, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4638 = torch.constant.int 131072
%6972 = torch.prim.ListConstruct %3029, %int131072_4638 : (!torch.int, !torch.int) -> !torch.list<int>
%6973 = torch.aten.view %6971, %6972 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %6973, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4639 = torch.constant.int 32
%int2_4640 = torch.constant.int 2
%int16_4641 = torch.constant.int 16
%int1_4642 = torch.constant.int 1
%int128_4643 = torch.constant.int 128
%6974 = torch.prim.ListConstruct %3032, %int32_4639, %int2_4640, %int16_4641, %int1_4642, %int128_4643 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6975 = torch.aten.view %5124, %6974 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6975, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4644 = torch.constant.int 32
%6976 = torch.aten.mul.int %3032, %int32_4644 : !torch.int, !torch.int -> !torch.int
%int2_4645 = torch.constant.int 2
%6977 = torch.aten.mul.int %6976, %int2_4645 : !torch.int, !torch.int -> !torch.int
%int16_4646 = torch.constant.int 16
%int1_4647 = torch.constant.int 1
%int128_4648 = torch.constant.int 128
%6978 = torch.prim.ListConstruct %6977, %int16_4646, %int1_4647, %int128_4648 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6979 = torch.aten.view %6975, %6978 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6979, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6980 = torch.prim.ListConstruct %6913 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4649 = torch.constant.bool false
%6981 = torch.aten.index_put %6979, %6980, %6929, %false_4649 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6981, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4650 = torch.constant.int 32
%int2_4651 = torch.constant.int 2
%int16_4652 = torch.constant.int 16
%int1_4653 = torch.constant.int 1
%int128_4654 = torch.constant.int 128
%6982 = torch.prim.ListConstruct %3032, %int32_4650, %int2_4651, %int16_4652, %int1_4653, %int128_4654 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6983 = torch.aten.view %6981, %6982 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6983, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4655 = torch.constant.int 131072
%6984 = torch.prim.ListConstruct %3032, %int131072_4655 : (!torch.int, !torch.int) -> !torch.list<int>
%6985 = torch.aten.view %6983, %6984 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %6985, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4656 = torch.constant.int 32
%int2_4657 = torch.constant.int 2
%int16_4658 = torch.constant.int 16
%int1_4659 = torch.constant.int 1
%int128_4660 = torch.constant.int 128
%6986 = torch.prim.ListConstruct %3035, %int32_4656, %int2_4657, %int16_4658, %int1_4659, %int128_4660 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6987 = torch.aten.view %5136, %6986 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6987, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4661 = torch.constant.int 32
%6988 = torch.aten.mul.int %3035, %int32_4661 : !torch.int, !torch.int -> !torch.int
%int2_4662 = torch.constant.int 2
%6989 = torch.aten.mul.int %6988, %int2_4662 : !torch.int, !torch.int -> !torch.int
%int16_4663 = torch.constant.int 16
%int1_4664 = torch.constant.int 1
%int128_4665 = torch.constant.int 128
%6990 = torch.prim.ListConstruct %6989, %int16_4663, %int1_4664, %int128_4665 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6991 = torch.aten.view %6987, %6990 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6991, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%6992 = torch.prim.ListConstruct %6915 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4666 = torch.constant.bool false
%6993 = torch.aten.index_put %6991, %6992, %6931, %false_4666 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %6993, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4667 = torch.constant.int 32
%int2_4668 = torch.constant.int 2
%int16_4669 = torch.constant.int 16
%int1_4670 = torch.constant.int 1
%int128_4671 = torch.constant.int 128
%6994 = torch.prim.ListConstruct %3035, %int32_4667, %int2_4668, %int16_4669, %int1_4670, %int128_4671 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6995 = torch.aten.view %6993, %6994 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6995, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4672 = torch.constant.int 131072
%6996 = torch.prim.ListConstruct %3035, %int131072_4672 : (!torch.int, !torch.int) -> !torch.list<int>
%6997 = torch.aten.view %6995, %6996 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %6997, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4673 = torch.constant.int 32
%int2_4674 = torch.constant.int 2
%int16_4675 = torch.constant.int 16
%int1_4676 = torch.constant.int 1
%int128_4677 = torch.constant.int 128
%6998 = torch.prim.ListConstruct %3038, %int32_4673, %int2_4674, %int16_4675, %int1_4676, %int128_4677 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%6999 = torch.aten.view %5148, %6998 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %6999, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4678 = torch.constant.int 32
%7000 = torch.aten.mul.int %3038, %int32_4678 : !torch.int, !torch.int -> !torch.int
%int2_4679 = torch.constant.int 2
%7001 = torch.aten.mul.int %7000, %int2_4679 : !torch.int, !torch.int -> !torch.int
%int16_4680 = torch.constant.int 16
%int1_4681 = torch.constant.int 1
%int128_4682 = torch.constant.int 128
%7002 = torch.prim.ListConstruct %7001, %int16_4680, %int1_4681, %int128_4682 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7003 = torch.aten.view %6999, %7002 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %7003, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%7004 = torch.prim.ListConstruct %6917 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4683 = torch.constant.bool false
%7005 = torch.aten.index_put %7003, %7004, %6933, %false_4683 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %7005, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4684 = torch.constant.int 32
%int2_4685 = torch.constant.int 2
%int16_4686 = torch.constant.int 16
%int1_4687 = torch.constant.int 1
%int128_4688 = torch.constant.int 128
%7006 = torch.prim.ListConstruct %3038, %int32_4684, %int2_4685, %int16_4686, %int1_4687, %int128_4688 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7007 = torch.aten.view %7005, %7006 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %7007, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4689 = torch.constant.int 131072
%7008 = torch.prim.ListConstruct %3038, %int131072_4689 : (!torch.int, !torch.int) -> !torch.list<int>
%7009 = torch.aten.view %7007, %7008 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %7009, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4690 = torch.constant.int 32
%int2_4691 = torch.constant.int 2
%int16_4692 = torch.constant.int 16
%int1_4693 = torch.constant.int 1
%int128_4694 = torch.constant.int 128
%7010 = torch.prim.ListConstruct %3041, %int32_4690, %int2_4691, %int16_4692, %int1_4693, %int128_4694 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7011 = torch.aten.view %5160, %7010 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %7011, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4695 = torch.constant.int 32
%7012 = torch.aten.mul.int %3041, %int32_4695 : !torch.int, !torch.int -> !torch.int
%int2_4696 = torch.constant.int 2
%7013 = torch.aten.mul.int %7012, %int2_4696 : !torch.int, !torch.int -> !torch.int
%int16_4697 = torch.constant.int 16
%int1_4698 = torch.constant.int 1
%int128_4699 = torch.constant.int 128
%7014 = torch.prim.ListConstruct %7013, %int16_4697, %int1_4698, %int128_4699 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7015 = torch.aten.view %7011, %7014 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %7015, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%7016 = torch.prim.ListConstruct %6919 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4700 = torch.constant.bool false
%7017 = torch.aten.index_put %7015, %7016, %6935, %false_4700 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %7017, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4701 = torch.constant.int 32
%int2_4702 = torch.constant.int 2
%int16_4703 = torch.constant.int 16
%int1_4704 = torch.constant.int 1
%int128_4705 = torch.constant.int 128
%7018 = torch.prim.ListConstruct %3041, %int32_4701, %int2_4702, %int16_4703, %int1_4704, %int128_4705 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7019 = torch.aten.view %7017, %7018 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %7019, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4706 = torch.constant.int 131072
%7020 = torch.prim.ListConstruct %3041, %int131072_4706 : (!torch.int, !torch.int) -> !torch.list<int>
%7021 = torch.aten.view %7019, %7020 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %7021, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_4707 = torch.constant.int 32
%int2_4708 = torch.constant.int 2
%int16_4709 = torch.constant.int 16
%int1_4710 = torch.constant.int 1
%int128_4711 = torch.constant.int 128
%7022 = torch.prim.ListConstruct %3044, %int32_4707, %int2_4708, %int16_4709, %int1_4710, %int128_4711 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7023 = torch.aten.view %5172, %7022 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %7023, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_4712 = torch.constant.int 32
%7024 = torch.aten.mul.int %3044, %int32_4712 : !torch.int, !torch.int -> !torch.int
%int2_4713 = torch.constant.int 2
%7025 = torch.aten.mul.int %7024, %int2_4713 : !torch.int, !torch.int -> !torch.int
%int16_4714 = torch.constant.int 16
%int1_4715 = torch.constant.int 1
%int128_4716 = torch.constant.int 128
%7026 = torch.prim.ListConstruct %7025, %int16_4714, %int1_4715, %int128_4716 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7027 = torch.aten.view %7023, %7026 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %7027, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%7028 = torch.prim.ListConstruct %6921 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_4717 = torch.constant.bool false
%7029 = torch.aten.index_put %7027, %7028, %6937, %false_4717 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %7029, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_4718 = torch.constant.int 32
%int2_4719 = torch.constant.int 2
%int16_4720 = torch.constant.int 16
%int1_4721 = torch.constant.int 1
%int128_4722 = torch.constant.int 128
%7030 = torch.prim.ListConstruct %3044, %int32_4718, %int2_4719, %int16_4720, %int1_4721, %int128_4722 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7031 = torch.aten.view %7029, %7030 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %7031, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_4723 = torch.constant.int 131072
%7032 = torch.prim.ListConstruct %3044, %int131072_4723 : (!torch.int, !torch.int) -> !torch.list<int>
%7033 = torch.aten.view %7031, %7032 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %7033, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_4724 = torch.constant.int -2
%7034 = torch.aten.unsqueeze %6648, %int-2_4724 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4725 = torch.constant.int -2
%7035 = torch.aten.unsqueeze %6663, %int-2_4725 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4726 = torch.constant.int -2
%7036 = torch.aten.unsqueeze %6678, %int-2_4726 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4727 = torch.constant.int -2
%7037 = torch.aten.unsqueeze %6693, %int-2_4727 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4728 = torch.constant.int -2
%7038 = torch.aten.unsqueeze %6708, %int-2_4728 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4729 = torch.constant.int -2
%7039 = torch.aten.unsqueeze %6723, %int-2_4729 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4730 = torch.constant.int -2
%7040 = torch.aten.unsqueeze %6738, %int-2_4730 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4731 = torch.constant.int -2
%7041 = torch.aten.unsqueeze %6753, %int-2_4731 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_4732 = torch.constant.int 4
%int1_4733 = torch.constant.int 1
%int4_4734 = torch.constant.int 4
%int128_4735 = torch.constant.int 128
%7042 = torch.prim.ListConstruct %int4_4732, %6634, %int1_4733, %int4_4734, %int128_4735 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4736 = torch.constant.bool false
%7043 = torch.aten.expand %7034, %7042, %false_4736 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4737 = torch.constant.int 4
%int1_4738 = torch.constant.int 1
%int4_4739 = torch.constant.int 4
%int128_4740 = torch.constant.int 128
%7044 = torch.prim.ListConstruct %int4_4737, %6634, %int1_4738, %int4_4739, %int128_4740 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4741 = torch.constant.bool false
%7045 = torch.aten.expand %7035, %7044, %false_4741 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4742 = torch.constant.int 4
%int1_4743 = torch.constant.int 1
%int4_4744 = torch.constant.int 4
%int128_4745 = torch.constant.int 128
%7046 = torch.prim.ListConstruct %int4_4742, %6634, %int1_4743, %int4_4744, %int128_4745 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4746 = torch.constant.bool false
%7047 = torch.aten.expand %7036, %7046, %false_4746 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4747 = torch.constant.int 4
%int1_4748 = torch.constant.int 1
%int4_4749 = torch.constant.int 4
%int128_4750 = torch.constant.int 128
%7048 = torch.prim.ListConstruct %int4_4747, %6634, %int1_4748, %int4_4749, %int128_4750 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4751 = torch.constant.bool false
%7049 = torch.aten.expand %7037, %7048, %false_4751 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4752 = torch.constant.int 4
%int1_4753 = torch.constant.int 1
%int4_4754 = torch.constant.int 4
%int128_4755 = torch.constant.int 128
%7050 = torch.prim.ListConstruct %int4_4752, %6634, %int1_4753, %int4_4754, %int128_4755 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4756 = torch.constant.bool false
%7051 = torch.aten.expand %7038, %7050, %false_4756 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4757 = torch.constant.int 4
%int1_4758 = torch.constant.int 1
%int4_4759 = torch.constant.int 4
%int128_4760 = torch.constant.int 128
%7052 = torch.prim.ListConstruct %int4_4757, %6634, %int1_4758, %int4_4759, %int128_4760 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4761 = torch.constant.bool false
%7053 = torch.aten.expand %7039, %7052, %false_4761 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4762 = torch.constant.int 4
%int1_4763 = torch.constant.int 1
%int4_4764 = torch.constant.int 4
%int128_4765 = torch.constant.int 128
%7054 = torch.prim.ListConstruct %int4_4762, %6634, %int1_4763, %int4_4764, %int128_4765 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4766 = torch.constant.bool false
%7055 = torch.aten.expand %7040, %7054, %false_4766 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4767 = torch.constant.int 4
%int1_4768 = torch.constant.int 1
%int4_4769 = torch.constant.int 4
%int128_4770 = torch.constant.int 128
%7056 = torch.prim.ListConstruct %int4_4767, %6634, %int1_4768, %int4_4769, %int128_4770 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4771 = torch.constant.bool false
%7057 = torch.aten.expand %7041, %7056, %false_4771 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4772 = torch.constant.int 4
%int4_4773 = torch.constant.int 4
%int128_4774 = torch.constant.int 128
%7058 = torch.prim.ListConstruct %int4_4772, %6634, %int4_4773, %int128_4774 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7059 = torch.aten.view %7043, %7058 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4775 = torch.constant.int 4
%int4_4776 = torch.constant.int 4
%int128_4777 = torch.constant.int 128
%7060 = torch.prim.ListConstruct %int4_4775, %6634, %int4_4776, %int128_4777 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7061 = torch.aten.view %7045, %7060 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4778 = torch.constant.int 4
%int4_4779 = torch.constant.int 4
%int128_4780 = torch.constant.int 128
%7062 = torch.prim.ListConstruct %int4_4778, %6634, %int4_4779, %int128_4780 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7063 = torch.aten.view %7047, %7062 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4781 = torch.constant.int 4
%int4_4782 = torch.constant.int 4
%int128_4783 = torch.constant.int 128
%7064 = torch.prim.ListConstruct %int4_4781, %6634, %int4_4782, %int128_4783 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7065 = torch.aten.view %7049, %7064 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4784 = torch.constant.int 4
%int4_4785 = torch.constant.int 4
%int128_4786 = torch.constant.int 128
%7066 = torch.prim.ListConstruct %int4_4784, %6634, %int4_4785, %int128_4786 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7067 = torch.aten.view %7051, %7066 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4787 = torch.constant.int 4
%int4_4788 = torch.constant.int 4
%int128_4789 = torch.constant.int 128
%7068 = torch.prim.ListConstruct %int4_4787, %6634, %int4_4788, %int128_4789 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7069 = torch.aten.view %7053, %7068 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4790 = torch.constant.int 4
%int4_4791 = torch.constant.int 4
%int128_4792 = torch.constant.int 128
%7070 = torch.prim.ListConstruct %int4_4790, %6634, %int4_4791, %int128_4792 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7071 = torch.aten.view %7055, %7070 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4793 = torch.constant.int 4
%int4_4794 = torch.constant.int 4
%int128_4795 = torch.constant.int 128
%7072 = torch.prim.ListConstruct %int4_4793, %6634, %int4_4794, %int128_4795 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7073 = torch.aten.view %7057, %7072 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_4796 = torch.constant.int -2
%7074 = torch.aten.unsqueeze %6423, %int-2_4796 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4797 = torch.constant.int -2
%7075 = torch.aten.unsqueeze %6425, %int-2_4797 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4798 = torch.constant.int -2
%7076 = torch.aten.unsqueeze %6427, %int-2_4798 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4799 = torch.constant.int -2
%7077 = torch.aten.unsqueeze %6429, %int-2_4799 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4800 = torch.constant.int -2
%7078 = torch.aten.unsqueeze %6431, %int-2_4800 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4801 = torch.constant.int -2
%7079 = torch.aten.unsqueeze %6433, %int-2_4801 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4802 = torch.constant.int -2
%7080 = torch.aten.unsqueeze %6435, %int-2_4802 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_4803 = torch.constant.int -2
%7081 = torch.aten.unsqueeze %6437, %int-2_4803 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %7081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_4804 = torch.constant.int 1
%7082 = torch.aten.size.int %6347, %int1_4804 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_4805 = torch.constant.int 4
%int1_4806 = torch.constant.int 1
%int4_4807 = torch.constant.int 4
%int128_4808 = torch.constant.int 128
%7083 = torch.prim.ListConstruct %int4_4805, %7082, %int1_4806, %int4_4807, %int128_4808 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4809 = torch.constant.bool false
%7084 = torch.aten.expand %7074, %7083, %false_4809 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4810 = torch.constant.int 4
%int1_4811 = torch.constant.int 1
%int4_4812 = torch.constant.int 4
%int128_4813 = torch.constant.int 128
%7085 = torch.prim.ListConstruct %int4_4810, %7082, %int1_4811, %int4_4812, %int128_4813 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4814 = torch.constant.bool false
%7086 = torch.aten.expand %7075, %7085, %false_4814 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4815 = torch.constant.int 4
%int1_4816 = torch.constant.int 1
%int4_4817 = torch.constant.int 4
%int128_4818 = torch.constant.int 128
%7087 = torch.prim.ListConstruct %int4_4815, %7082, %int1_4816, %int4_4817, %int128_4818 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4819 = torch.constant.bool false
%7088 = torch.aten.expand %7076, %7087, %false_4819 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4820 = torch.constant.int 4
%int1_4821 = torch.constant.int 1
%int4_4822 = torch.constant.int 4
%int128_4823 = torch.constant.int 128
%7089 = torch.prim.ListConstruct %int4_4820, %7082, %int1_4821, %int4_4822, %int128_4823 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4824 = torch.constant.bool false
%7090 = torch.aten.expand %7077, %7089, %false_4824 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4825 = torch.constant.int 4
%int1_4826 = torch.constant.int 1
%int4_4827 = torch.constant.int 4
%int128_4828 = torch.constant.int 128
%7091 = torch.prim.ListConstruct %int4_4825, %7082, %int1_4826, %int4_4827, %int128_4828 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4829 = torch.constant.bool false
%7092 = torch.aten.expand %7078, %7091, %false_4829 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4830 = torch.constant.int 4
%int1_4831 = torch.constant.int 1
%int4_4832 = torch.constant.int 4
%int128_4833 = torch.constant.int 128
%7093 = torch.prim.ListConstruct %int4_4830, %7082, %int1_4831, %int4_4832, %int128_4833 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4834 = torch.constant.bool false
%7094 = torch.aten.expand %7079, %7093, %false_4834 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4835 = torch.constant.int 4
%int1_4836 = torch.constant.int 1
%int4_4837 = torch.constant.int 4
%int128_4838 = torch.constant.int 128
%7095 = torch.prim.ListConstruct %int4_4835, %7082, %int1_4836, %int4_4837, %int128_4838 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4839 = torch.constant.bool false
%7096 = torch.aten.expand %7080, %7095, %false_4839 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4840 = torch.constant.int 4
%int1_4841 = torch.constant.int 1
%int4_4842 = torch.constant.int 4
%int128_4843 = torch.constant.int 128
%7097 = torch.prim.ListConstruct %int4_4840, %7082, %int1_4841, %int4_4842, %int128_4843 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_4844 = torch.constant.bool false
%7098 = torch.aten.expand %7081, %7097, %false_4844 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %7098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_4845 = torch.constant.int 4
%int4_4846 = torch.constant.int 4
%int128_4847 = torch.constant.int 128
%7099 = torch.prim.ListConstruct %int4_4845, %7082, %int4_4846, %int128_4847 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7100 = torch.aten.view %7084, %7099 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4848 = torch.constant.int 4
%int4_4849 = torch.constant.int 4
%int128_4850 = torch.constant.int 128
%7101 = torch.prim.ListConstruct %int4_4848, %7082, %int4_4849, %int128_4850 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7102 = torch.aten.view %7086, %7101 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4851 = torch.constant.int 4
%int4_4852 = torch.constant.int 4
%int128_4853 = torch.constant.int 128
%7103 = torch.prim.ListConstruct %int4_4851, %7082, %int4_4852, %int128_4853 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7104 = torch.aten.view %7088, %7103 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4854 = torch.constant.int 4
%int4_4855 = torch.constant.int 4
%int128_4856 = torch.constant.int 128
%7105 = torch.prim.ListConstruct %int4_4854, %7082, %int4_4855, %int128_4856 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7106 = torch.aten.view %7090, %7105 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4857 = torch.constant.int 4
%int4_4858 = torch.constant.int 4
%int128_4859 = torch.constant.int 128
%7107 = torch.prim.ListConstruct %int4_4857, %7082, %int4_4858, %int128_4859 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7108 = torch.aten.view %7092, %7107 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4860 = torch.constant.int 4
%int4_4861 = torch.constant.int 4
%int128_4862 = torch.constant.int 128
%7109 = torch.prim.ListConstruct %int4_4860, %7082, %int4_4861, %int128_4862 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7110 = torch.aten.view %7094, %7109 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4863 = torch.constant.int 4
%int4_4864 = torch.constant.int 4
%int128_4865 = torch.constant.int 128
%7111 = torch.prim.ListConstruct %int4_4863, %7082, %int4_4864, %int128_4865 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7112 = torch.aten.view %7096, %7111 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4866 = torch.constant.int 4
%int4_4867 = torch.constant.int 4
%int128_4868 = torch.constant.int 128
%7113 = torch.prim.ListConstruct %int4_4866, %7082, %int4_4867, %int128_4868 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7114 = torch.aten.view %7098, %7113 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4869 = torch.constant.int 1
%int2_4870 = torch.constant.int 2
%7115 = torch.aten.transpose.int %6490, %int1_4869, %int2_4870 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7115, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4871 = torch.constant.int 1
%int2_4872 = torch.constant.int 2
%7116 = torch.aten.transpose.int %6505, %int1_4871, %int2_4872 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7116, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4873 = torch.constant.int 1
%int2_4874 = torch.constant.int 2
%7117 = torch.aten.transpose.int %6520, %int1_4873, %int2_4874 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7117, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4875 = torch.constant.int 1
%int2_4876 = torch.constant.int 2
%7118 = torch.aten.transpose.int %6535, %int1_4875, %int2_4876 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7118, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4877 = torch.constant.int 1
%int2_4878 = torch.constant.int 2
%7119 = torch.aten.transpose.int %6550, %int1_4877, %int2_4878 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7119, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4879 = torch.constant.int 1
%int2_4880 = torch.constant.int 2
%7120 = torch.aten.transpose.int %6565, %int1_4879, %int2_4880 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7120, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4881 = torch.constant.int 1
%int2_4882 = torch.constant.int 2
%7121 = torch.aten.transpose.int %6580, %int1_4881, %int2_4882 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7121, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4883 = torch.constant.int 1
%int2_4884 = torch.constant.int 2
%7122 = torch.aten.transpose.int %6595, %int1_4883, %int2_4884 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7122, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4885 = torch.constant.int 1
%int2_4886 = torch.constant.int 2
%7123 = torch.aten.transpose.int %7059, %int1_4885, %int2_4886 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7123, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4887 = torch.constant.int 1
%int2_4888 = torch.constant.int 2
%7124 = torch.aten.transpose.int %7061, %int1_4887, %int2_4888 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7124, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4889 = torch.constant.int 1
%int2_4890 = torch.constant.int 2
%7125 = torch.aten.transpose.int %7063, %int1_4889, %int2_4890 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7125, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4891 = torch.constant.int 1
%int2_4892 = torch.constant.int 2
%7126 = torch.aten.transpose.int %7065, %int1_4891, %int2_4892 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7126, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4893 = torch.constant.int 1
%int2_4894 = torch.constant.int 2
%7127 = torch.aten.transpose.int %7067, %int1_4893, %int2_4894 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7127, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4895 = torch.constant.int 1
%int2_4896 = torch.constant.int 2
%7128 = torch.aten.transpose.int %7069, %int1_4895, %int2_4896 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7128, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4897 = torch.constant.int 1
%int2_4898 = torch.constant.int 2
%7129 = torch.aten.transpose.int %7071, %int1_4897, %int2_4898 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7129, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4899 = torch.constant.int 1
%int2_4900 = torch.constant.int 2
%7130 = torch.aten.transpose.int %7073, %int1_4899, %int2_4900 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7130, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4901 = torch.constant.int 1
%int2_4902 = torch.constant.int 2
%7131 = torch.aten.transpose.int %7100, %int1_4901, %int2_4902 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7131, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4903 = torch.constant.int 1
%int2_4904 = torch.constant.int 2
%7132 = torch.aten.transpose.int %7102, %int1_4903, %int2_4904 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7132, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4905 = torch.constant.int 1
%int2_4906 = torch.constant.int 2
%7133 = torch.aten.transpose.int %7104, %int1_4905, %int2_4906 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7133, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4907 = torch.constant.int 1
%int2_4908 = torch.constant.int 2
%7134 = torch.aten.transpose.int %7106, %int1_4907, %int2_4908 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7134, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4909 = torch.constant.int 1
%int2_4910 = torch.constant.int 2
%7135 = torch.aten.transpose.int %7108, %int1_4909, %int2_4910 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7135, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4911 = torch.constant.int 1
%int2_4912 = torch.constant.int 2
%7136 = torch.aten.transpose.int %7110, %int1_4911, %int2_4912 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7136, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4913 = torch.constant.int 1
%int2_4914 = torch.constant.int 2
%7137 = torch.aten.transpose.int %7112, %int1_4913, %int2_4914 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7137, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4915 = torch.constant.int 1
%int2_4916 = torch.constant.int 2
%7138 = torch.aten.transpose.int %7114, %int1_4915, %int2_4916 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %7138, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4917 = torch.constant.float 0.000000e+00
%true_4918 = torch.constant.bool true
%none_4919 = torch.constant.none
%none_4920 = torch.constant.none
%7139:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7115, %7123, %7131, %float0.000000e00_4917, %true_4918, %none_4919, %none_4920) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7139#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4921 = torch.constant.float 0.000000e+00
%true_4922 = torch.constant.bool true
%none_4923 = torch.constant.none
%none_4924 = torch.constant.none
%7140:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7116, %7124, %7132, %float0.000000e00_4921, %true_4922, %none_4923, %none_4924) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7140#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4925 = torch.constant.float 0.000000e+00
%true_4926 = torch.constant.bool true
%none_4927 = torch.constant.none
%none_4928 = torch.constant.none
%7141:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7117, %7125, %7133, %float0.000000e00_4925, %true_4926, %none_4927, %none_4928) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7141#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4929 = torch.constant.float 0.000000e+00
%true_4930 = torch.constant.bool true
%none_4931 = torch.constant.none
%none_4932 = torch.constant.none
%7142:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7118, %7126, %7134, %float0.000000e00_4929, %true_4930, %none_4931, %none_4932) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7142#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4933 = torch.constant.float 0.000000e+00
%true_4934 = torch.constant.bool true
%none_4935 = torch.constant.none
%none_4936 = torch.constant.none
%7143:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7119, %7127, %7135, %float0.000000e00_4933, %true_4934, %none_4935, %none_4936) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7143#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4937 = torch.constant.float 0.000000e+00
%true_4938 = torch.constant.bool true
%none_4939 = torch.constant.none
%none_4940 = torch.constant.none
%7144:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7120, %7128, %7136, %float0.000000e00_4937, %true_4938, %none_4939, %none_4940) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7144#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4941 = torch.constant.float 0.000000e+00
%true_4942 = torch.constant.bool true
%none_4943 = torch.constant.none
%none_4944 = torch.constant.none
%7145:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7121, %7129, %7137, %float0.000000e00_4941, %true_4942, %none_4943, %none_4944) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7145#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_4945 = torch.constant.float 0.000000e+00
%true_4946 = torch.constant.bool true
%none_4947 = torch.constant.none
%none_4948 = torch.constant.none
%7146:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%7122, %7130, %7138, %float0.000000e00_4945, %true_4946, %none_4947, %none_4948) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %7146#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_4949 = torch.constant.int 1
%int2_4950 = torch.constant.int 2
%7147 = torch.aten.transpose.int %7139#0, %int1_4949, %int2_4950 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4951 = torch.constant.int 1
%int2_4952 = torch.constant.int 2
%7148 = torch.aten.transpose.int %7140#0, %int1_4951, %int2_4952 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4953 = torch.constant.int 1
%int2_4954 = torch.constant.int 2
%7149 = torch.aten.transpose.int %7141#0, %int1_4953, %int2_4954 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4955 = torch.constant.int 1
%int2_4956 = torch.constant.int 2
%7150 = torch.aten.transpose.int %7142#0, %int1_4955, %int2_4956 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4957 = torch.constant.int 1
%int2_4958 = torch.constant.int 2
%7151 = torch.aten.transpose.int %7143#0, %int1_4957, %int2_4958 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4959 = torch.constant.int 1
%int2_4960 = torch.constant.int 2
%7152 = torch.aten.transpose.int %7144#0, %int1_4959, %int2_4960 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4961 = torch.constant.int 1
%int2_4962 = torch.constant.int 2
%7153 = torch.aten.transpose.int %7145#0, %int1_4961, %int2_4962 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_4963 = torch.constant.int 1
%int2_4964 = torch.constant.int 2
%7154 = torch.aten.transpose.int %7146#0, %int1_4963, %int2_4964 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %7154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_4965 = torch.constant.int 4
%int512_4966 = torch.constant.int 512
%7155 = torch.prim.ListConstruct %int4_4965, %6476, %int512_4966 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7156 = torch.aten.view %7147, %7155 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4967 = torch.constant.int 4
%int512_4968 = torch.constant.int 512
%7157 = torch.prim.ListConstruct %int4_4967, %6491, %int512_4968 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7158 = torch.aten.view %7148, %7157 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4969 = torch.constant.int 4
%int512_4970 = torch.constant.int 512
%7159 = torch.prim.ListConstruct %int4_4969, %6506, %int512_4970 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7160 = torch.aten.view %7149, %7159 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4971 = torch.constant.int 4
%int512_4972 = torch.constant.int 512
%7161 = torch.prim.ListConstruct %int4_4971, %6521, %int512_4972 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7162 = torch.aten.view %7150, %7161 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4973 = torch.constant.int 4
%int512_4974 = torch.constant.int 512
%7163 = torch.prim.ListConstruct %int4_4973, %6536, %int512_4974 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7164 = torch.aten.view %7151, %7163 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4975 = torch.constant.int 4
%int512_4976 = torch.constant.int 512
%7165 = torch.prim.ListConstruct %int4_4975, %6551, %int512_4976 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7166 = torch.aten.view %7152, %7165 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4977 = torch.constant.int 4
%int512_4978 = torch.constant.int 512
%7167 = torch.prim.ListConstruct %int4_4977, %6566, %int512_4978 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7168 = torch.aten.view %7153, %7167 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_4979 = torch.constant.int 4
%int512_4980 = torch.constant.int 512
%7169 = torch.prim.ListConstruct %int4_4979, %6581, %int512_4980 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7170 = torch.aten.view %7154, %7169 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %7170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_4981 = torch.constant.int 1
%int0_4982 = torch.constant.int 0
%7171 = torch.prim.ListConstruct %int1_4981, %int0_4982 : (!torch.int, !torch.int) -> !torch.list<int>
%7172 = torch.aten.permute %184, %7171 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4983 = torch.constant.int 1
%int0_4984 = torch.constant.int 0
%7173 = torch.prim.ListConstruct %int1_4983, %int0_4984 : (!torch.int, !torch.int) -> !torch.list<int>
%7174 = torch.aten.permute %185, %7173 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4985 = torch.constant.int 1
%int0_4986 = torch.constant.int 0
%7175 = torch.prim.ListConstruct %int1_4985, %int0_4986 : (!torch.int, !torch.int) -> !torch.list<int>
%7176 = torch.aten.permute %186, %7175 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4987 = torch.constant.int 1
%int0_4988 = torch.constant.int 0
%7177 = torch.prim.ListConstruct %int1_4987, %int0_4988 : (!torch.int, !torch.int) -> !torch.list<int>
%7178 = torch.aten.permute %187, %7177 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4989 = torch.constant.int 1
%int0_4990 = torch.constant.int 0
%7179 = torch.prim.ListConstruct %int1_4989, %int0_4990 : (!torch.int, !torch.int) -> !torch.list<int>
%7180 = torch.aten.permute %188, %7179 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4991 = torch.constant.int 1
%int0_4992 = torch.constant.int 0
%7181 = torch.prim.ListConstruct %int1_4991, %int0_4992 : (!torch.int, !torch.int) -> !torch.list<int>
%7182 = torch.aten.permute %189, %7181 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4993 = torch.constant.int 1
%int0_4994 = torch.constant.int 0
%7183 = torch.prim.ListConstruct %int1_4993, %int0_4994 : (!torch.int, !torch.int) -> !torch.list<int>
%7184 = torch.aten.permute %190, %7183 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_4995 = torch.constant.int 1
%int0_4996 = torch.constant.int 0
%7185 = torch.prim.ListConstruct %int1_4995, %int0_4996 : (!torch.int, !torch.int) -> !torch.list<int>
%7186 = torch.aten.permute %191, %7185 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_4997 = torch.constant.int 4
%7187 = torch.aten.mul.int %int4_4997, %6476 : !torch.int, !torch.int -> !torch.int
%int512_4998 = torch.constant.int 512
%7188 = torch.prim.ListConstruct %7187, %int512_4998 : (!torch.int, !torch.int) -> !torch.list<int>
%7189 = torch.aten.view %7156, %7188 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7189, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7190 = torch.aten.mm %7189, %7172 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7190, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_4999 = torch.constant.int 4
%int4096_5000 = torch.constant.int 4096
%7191 = torch.prim.ListConstruct %int4_4999, %6476, %int4096_5000 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7192 = torch.aten.view %7190, %7191 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5001 = torch.constant.int 4
%7193 = torch.aten.mul.int %int4_5001, %6491 : !torch.int, !torch.int -> !torch.int
%int512_5002 = torch.constant.int 512
%7194 = torch.prim.ListConstruct %7193, %int512_5002 : (!torch.int, !torch.int) -> !torch.list<int>
%7195 = torch.aten.view %7158, %7194 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7195, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7196 = torch.aten.mm %7195, %7174 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7196, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5003 = torch.constant.int 4
%int4096_5004 = torch.constant.int 4096
%7197 = torch.prim.ListConstruct %int4_5003, %6491, %int4096_5004 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7198 = torch.aten.view %7196, %7197 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5005 = torch.constant.int 4
%7199 = torch.aten.mul.int %int4_5005, %6506 : !torch.int, !torch.int -> !torch.int
%int512_5006 = torch.constant.int 512
%7200 = torch.prim.ListConstruct %7199, %int512_5006 : (!torch.int, !torch.int) -> !torch.list<int>
%7201 = torch.aten.view %7160, %7200 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7201, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7202 = torch.aten.mm %7201, %7176 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7202, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5007 = torch.constant.int 4
%int4096_5008 = torch.constant.int 4096
%7203 = torch.prim.ListConstruct %int4_5007, %6506, %int4096_5008 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7204 = torch.aten.view %7202, %7203 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5009 = torch.constant.int 4
%7205 = torch.aten.mul.int %int4_5009, %6521 : !torch.int, !torch.int -> !torch.int
%int512_5010 = torch.constant.int 512
%7206 = torch.prim.ListConstruct %7205, %int512_5010 : (!torch.int, !torch.int) -> !torch.list<int>
%7207 = torch.aten.view %7162, %7206 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7207, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7208 = torch.aten.mm %7207, %7178 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7208, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5011 = torch.constant.int 4
%int4096_5012 = torch.constant.int 4096
%7209 = torch.prim.ListConstruct %int4_5011, %6521, %int4096_5012 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7210 = torch.aten.view %7208, %7209 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5013 = torch.constant.int 4
%7211 = torch.aten.mul.int %int4_5013, %6536 : !torch.int, !torch.int -> !torch.int
%int512_5014 = torch.constant.int 512
%7212 = torch.prim.ListConstruct %7211, %int512_5014 : (!torch.int, !torch.int) -> !torch.list<int>
%7213 = torch.aten.view %7164, %7212 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7213, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7214 = torch.aten.mm %7213, %7180 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7214, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5015 = torch.constant.int 4
%int4096_5016 = torch.constant.int 4096
%7215 = torch.prim.ListConstruct %int4_5015, %6536, %int4096_5016 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7216 = torch.aten.view %7214, %7215 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5017 = torch.constant.int 4
%7217 = torch.aten.mul.int %int4_5017, %6551 : !torch.int, !torch.int -> !torch.int
%int512_5018 = torch.constant.int 512
%7218 = torch.prim.ListConstruct %7217, %int512_5018 : (!torch.int, !torch.int) -> !torch.list<int>
%7219 = torch.aten.view %7166, %7218 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7219, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7220 = torch.aten.mm %7219, %7182 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7220, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5019 = torch.constant.int 4
%int4096_5020 = torch.constant.int 4096
%7221 = torch.prim.ListConstruct %int4_5019, %6551, %int4096_5020 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7222 = torch.aten.view %7220, %7221 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5021 = torch.constant.int 4
%7223 = torch.aten.mul.int %int4_5021, %6566 : !torch.int, !torch.int -> !torch.int
%int512_5022 = torch.constant.int 512
%7224 = torch.prim.ListConstruct %7223, %int512_5022 : (!torch.int, !torch.int) -> !torch.list<int>
%7225 = torch.aten.view %7168, %7224 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7225, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7226 = torch.aten.mm %7225, %7184 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7226, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5023 = torch.constant.int 4
%int4096_5024 = torch.constant.int 4096
%7227 = torch.prim.ListConstruct %int4_5023, %6566, %int4096_5024 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7228 = torch.aten.view %7226, %7227 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_5025 = torch.constant.int 4
%7229 = torch.aten.mul.int %int4_5025, %6581 : !torch.int, !torch.int -> !torch.int
%int512_5026 = torch.constant.int 512
%7230 = torch.prim.ListConstruct %7229, %int512_5026 : (!torch.int, !torch.int) -> !torch.list<int>
%7231 = torch.aten.view %7170, %7230 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %7231, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%7232 = torch.aten.mm %7231, %7186 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7232, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5027 = torch.constant.int 4
%int4096_5028 = torch.constant.int 4096
%7233 = torch.prim.ListConstruct %int4_5027, %6581, %int4096_5028 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7234 = torch.aten.view %7232, %7233 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7235 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5029 = arith.constant 1 : index
%dim_5030 = tensor.dim %7235, %c1_5029 : tensor<4x?x4096xf16>
%7236 = flow.tensor.transfer %7235 : tensor<4x?x4096xf16>{%dim_5030} to #hal.device.promise<@__device_0>
%7237 = torch_c.from_builtin_tensor %7236 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7238 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5031 = arith.constant 1 : index
%dim_5032 = tensor.dim %7238, %c1_5031 : tensor<4x?x4096xf16>
%7239 = flow.tensor.transfer %7238 : tensor<4x?x4096xf16>{%dim_5032} to #hal.device.promise<@__device_0>
%7240 = torch_c.from_builtin_tensor %7239 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7241 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5033 = arith.constant 1 : index
%dim_5034 = tensor.dim %7241, %c1_5033 : tensor<4x?x4096xf16>
%7242 = flow.tensor.transfer %7241 : tensor<4x?x4096xf16>{%dim_5034} to #hal.device.promise<@__device_0>
%7243 = torch_c.from_builtin_tensor %7242 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7244 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5035 = arith.constant 1 : index
%dim_5036 = tensor.dim %7244, %c1_5035 : tensor<4x?x4096xf16>
%7245 = flow.tensor.transfer %7244 : tensor<4x?x4096xf16>{%dim_5036} to #hal.device.promise<@__device_0>
%7246 = torch_c.from_builtin_tensor %7245 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7247 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5037 = arith.constant 1 : index
%dim_5038 = tensor.dim %7247, %c1_5037 : tensor<4x?x4096xf16>
%7248 = flow.tensor.transfer %7247 : tensor<4x?x4096xf16>{%dim_5038} to #hal.device.promise<@__device_0>
%7249 = torch_c.from_builtin_tensor %7248 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7250 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5039 = arith.constant 1 : index
%dim_5040 = tensor.dim %7250, %c1_5039 : tensor<4x?x4096xf16>
%7251 = flow.tensor.transfer %7250 : tensor<4x?x4096xf16>{%dim_5040} to #hal.device.promise<@__device_0>
%7252 = torch_c.from_builtin_tensor %7251 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7253 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5041 = arith.constant 1 : index
%dim_5042 = tensor.dim %7253, %c1_5041 : tensor<4x?x4096xf16>
%7254 = flow.tensor.transfer %7253 : tensor<4x?x4096xf16>{%dim_5042} to #hal.device.promise<@__device_0>
%7255 = torch_c.from_builtin_tensor %7254 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5043 = torch.constant.int 1
%7256 = torch.aten.add.Tensor %7192, %7237, %int1_5043 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5044 = torch.constant.int 1
%7257 = torch.aten.add.Tensor %7256, %7240, %int1_5044 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5045 = torch.constant.int 1
%7258 = torch.aten.add.Tensor %7257, %7243, %int1_5045 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5046 = torch.constant.int 1
%7259 = torch.aten.add.Tensor %7258, %7246, %int1_5046 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5047 = torch.constant.int 1
%7260 = torch.aten.add.Tensor %7259, %7249, %int1_5047 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5048 = torch.constant.int 1
%7261 = torch.aten.add.Tensor %7260, %7252, %int1_5048 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5049 = torch.constant.int 1
%7262 = torch.aten.add.Tensor %7261, %7255, %int1_5049 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7263 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5050 = arith.constant 1 : index
%dim_5051 = tensor.dim %7263, %c1_5050 : tensor<4x?x4096xf16>
%7264 = flow.tensor.transfer %7263 : tensor<4x?x4096xf16>{%dim_5051} to #hal.device.promise<@__device_1>
%7265 = torch_c.from_builtin_tensor %7264 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7266 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5052 = arith.constant 1 : index
%dim_5053 = tensor.dim %7266, %c1_5052 : tensor<4x?x4096xf16>
%7267 = flow.tensor.transfer %7266 : tensor<4x?x4096xf16>{%dim_5053} to #hal.device.promise<@__device_1>
%7268 = torch_c.from_builtin_tensor %7267 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7269 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5054 = arith.constant 1 : index
%dim_5055 = tensor.dim %7269, %c1_5054 : tensor<4x?x4096xf16>
%7270 = flow.tensor.transfer %7269 : tensor<4x?x4096xf16>{%dim_5055} to #hal.device.promise<@__device_1>
%7271 = torch_c.from_builtin_tensor %7270 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7272 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5056 = arith.constant 1 : index
%dim_5057 = tensor.dim %7272, %c1_5056 : tensor<4x?x4096xf16>
%7273 = flow.tensor.transfer %7272 : tensor<4x?x4096xf16>{%dim_5057} to #hal.device.promise<@__device_1>
%7274 = torch_c.from_builtin_tensor %7273 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7275 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5058 = arith.constant 1 : index
%dim_5059 = tensor.dim %7275, %c1_5058 : tensor<4x?x4096xf16>
%7276 = flow.tensor.transfer %7275 : tensor<4x?x4096xf16>{%dim_5059} to #hal.device.promise<@__device_1>
%7277 = torch_c.from_builtin_tensor %7276 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7278 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5060 = arith.constant 1 : index
%dim_5061 = tensor.dim %7278, %c1_5060 : tensor<4x?x4096xf16>
%7279 = flow.tensor.transfer %7278 : tensor<4x?x4096xf16>{%dim_5061} to #hal.device.promise<@__device_1>
%7280 = torch_c.from_builtin_tensor %7279 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7281 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5062 = arith.constant 1 : index
%dim_5063 = tensor.dim %7281, %c1_5062 : tensor<4x?x4096xf16>
%7282 = flow.tensor.transfer %7281 : tensor<4x?x4096xf16>{%dim_5063} to #hal.device.promise<@__device_1>
%7283 = torch_c.from_builtin_tensor %7282 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5064 = torch.constant.int 1
%7284 = torch.aten.add.Tensor %7265, %7198, %int1_5064 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5065 = torch.constant.int 1
%7285 = torch.aten.add.Tensor %7284, %7268, %int1_5065 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5066 = torch.constant.int 1
%7286 = torch.aten.add.Tensor %7285, %7271, %int1_5066 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5067 = torch.constant.int 1
%7287 = torch.aten.add.Tensor %7286, %7274, %int1_5067 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5068 = torch.constant.int 1
%7288 = torch.aten.add.Tensor %7287, %7277, %int1_5068 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5069 = torch.constant.int 1
%7289 = torch.aten.add.Tensor %7288, %7280, %int1_5069 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5070 = torch.constant.int 1
%7290 = torch.aten.add.Tensor %7289, %7283, %int1_5070 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7291 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5071 = arith.constant 1 : index
%dim_5072 = tensor.dim %7291, %c1_5071 : tensor<4x?x4096xf16>
%7292 = flow.tensor.transfer %7291 : tensor<4x?x4096xf16>{%dim_5072} to #hal.device.promise<@__device_2>
%7293 = torch_c.from_builtin_tensor %7292 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7294 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5073 = arith.constant 1 : index
%dim_5074 = tensor.dim %7294, %c1_5073 : tensor<4x?x4096xf16>
%7295 = flow.tensor.transfer %7294 : tensor<4x?x4096xf16>{%dim_5074} to #hal.device.promise<@__device_2>
%7296 = torch_c.from_builtin_tensor %7295 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7297 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5075 = arith.constant 1 : index
%dim_5076 = tensor.dim %7297, %c1_5075 : tensor<4x?x4096xf16>
%7298 = flow.tensor.transfer %7297 : tensor<4x?x4096xf16>{%dim_5076} to #hal.device.promise<@__device_2>
%7299 = torch_c.from_builtin_tensor %7298 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7300 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5077 = arith.constant 1 : index
%dim_5078 = tensor.dim %7300, %c1_5077 : tensor<4x?x4096xf16>
%7301 = flow.tensor.transfer %7300 : tensor<4x?x4096xf16>{%dim_5078} to #hal.device.promise<@__device_2>
%7302 = torch_c.from_builtin_tensor %7301 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7303 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5079 = arith.constant 1 : index
%dim_5080 = tensor.dim %7303, %c1_5079 : tensor<4x?x4096xf16>
%7304 = flow.tensor.transfer %7303 : tensor<4x?x4096xf16>{%dim_5080} to #hal.device.promise<@__device_2>
%7305 = torch_c.from_builtin_tensor %7304 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7306 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5081 = arith.constant 1 : index
%dim_5082 = tensor.dim %7306, %c1_5081 : tensor<4x?x4096xf16>
%7307 = flow.tensor.transfer %7306 : tensor<4x?x4096xf16>{%dim_5082} to #hal.device.promise<@__device_2>
%7308 = torch_c.from_builtin_tensor %7307 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7309 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5083 = arith.constant 1 : index
%dim_5084 = tensor.dim %7309, %c1_5083 : tensor<4x?x4096xf16>
%7310 = flow.tensor.transfer %7309 : tensor<4x?x4096xf16>{%dim_5084} to #hal.device.promise<@__device_2>
%7311 = torch_c.from_builtin_tensor %7310 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5085 = torch.constant.int 1
%7312 = torch.aten.add.Tensor %7293, %7296, %int1_5085 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5086 = torch.constant.int 1
%7313 = torch.aten.add.Tensor %7312, %7204, %int1_5086 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5087 = torch.constant.int 1
%7314 = torch.aten.add.Tensor %7313, %7299, %int1_5087 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5088 = torch.constant.int 1
%7315 = torch.aten.add.Tensor %7314, %7302, %int1_5088 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5089 = torch.constant.int 1
%7316 = torch.aten.add.Tensor %7315, %7305, %int1_5089 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5090 = torch.constant.int 1
%7317 = torch.aten.add.Tensor %7316, %7308, %int1_5090 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5091 = torch.constant.int 1
%7318 = torch.aten.add.Tensor %7317, %7311, %int1_5091 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7319 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5092 = arith.constant 1 : index
%dim_5093 = tensor.dim %7319, %c1_5092 : tensor<4x?x4096xf16>
%7320 = flow.tensor.transfer %7319 : tensor<4x?x4096xf16>{%dim_5093} to #hal.device.promise<@__device_3>
%7321 = torch_c.from_builtin_tensor %7320 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7322 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5094 = arith.constant 1 : index
%dim_5095 = tensor.dim %7322, %c1_5094 : tensor<4x?x4096xf16>
%7323 = flow.tensor.transfer %7322 : tensor<4x?x4096xf16>{%dim_5095} to #hal.device.promise<@__device_3>
%7324 = torch_c.from_builtin_tensor %7323 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7325 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5096 = arith.constant 1 : index
%dim_5097 = tensor.dim %7325, %c1_5096 : tensor<4x?x4096xf16>
%7326 = flow.tensor.transfer %7325 : tensor<4x?x4096xf16>{%dim_5097} to #hal.device.promise<@__device_3>
%7327 = torch_c.from_builtin_tensor %7326 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7328 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5098 = arith.constant 1 : index
%dim_5099 = tensor.dim %7328, %c1_5098 : tensor<4x?x4096xf16>
%7329 = flow.tensor.transfer %7328 : tensor<4x?x4096xf16>{%dim_5099} to #hal.device.promise<@__device_3>
%7330 = torch_c.from_builtin_tensor %7329 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7331 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5100 = arith.constant 1 : index
%dim_5101 = tensor.dim %7331, %c1_5100 : tensor<4x?x4096xf16>
%7332 = flow.tensor.transfer %7331 : tensor<4x?x4096xf16>{%dim_5101} to #hal.device.promise<@__device_3>
%7333 = torch_c.from_builtin_tensor %7332 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7334 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5102 = arith.constant 1 : index
%dim_5103 = tensor.dim %7334, %c1_5102 : tensor<4x?x4096xf16>
%7335 = flow.tensor.transfer %7334 : tensor<4x?x4096xf16>{%dim_5103} to #hal.device.promise<@__device_3>
%7336 = torch_c.from_builtin_tensor %7335 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7337 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5104 = arith.constant 1 : index
%dim_5105 = tensor.dim %7337, %c1_5104 : tensor<4x?x4096xf16>
%7338 = flow.tensor.transfer %7337 : tensor<4x?x4096xf16>{%dim_5105} to #hal.device.promise<@__device_3>
%7339 = torch_c.from_builtin_tensor %7338 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5106 = torch.constant.int 1
%7340 = torch.aten.add.Tensor %7321, %7324, %int1_5106 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5107 = torch.constant.int 1
%7341 = torch.aten.add.Tensor %7340, %7327, %int1_5107 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5108 = torch.constant.int 1
%7342 = torch.aten.add.Tensor %7341, %7210, %int1_5108 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5109 = torch.constant.int 1
%7343 = torch.aten.add.Tensor %7342, %7330, %int1_5109 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5110 = torch.constant.int 1
%7344 = torch.aten.add.Tensor %7343, %7333, %int1_5110 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5111 = torch.constant.int 1
%7345 = torch.aten.add.Tensor %7344, %7336, %int1_5111 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5112 = torch.constant.int 1
%7346 = torch.aten.add.Tensor %7345, %7339, %int1_5112 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7347 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5113 = arith.constant 1 : index
%dim_5114 = tensor.dim %7347, %c1_5113 : tensor<4x?x4096xf16>
%7348 = flow.tensor.transfer %7347 : tensor<4x?x4096xf16>{%dim_5114} to #hal.device.promise<@__device_4>
%7349 = torch_c.from_builtin_tensor %7348 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7350 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5115 = arith.constant 1 : index
%dim_5116 = tensor.dim %7350, %c1_5115 : tensor<4x?x4096xf16>
%7351 = flow.tensor.transfer %7350 : tensor<4x?x4096xf16>{%dim_5116} to #hal.device.promise<@__device_4>
%7352 = torch_c.from_builtin_tensor %7351 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7353 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5117 = arith.constant 1 : index
%dim_5118 = tensor.dim %7353, %c1_5117 : tensor<4x?x4096xf16>
%7354 = flow.tensor.transfer %7353 : tensor<4x?x4096xf16>{%dim_5118} to #hal.device.promise<@__device_4>
%7355 = torch_c.from_builtin_tensor %7354 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7356 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5119 = arith.constant 1 : index
%dim_5120 = tensor.dim %7356, %c1_5119 : tensor<4x?x4096xf16>
%7357 = flow.tensor.transfer %7356 : tensor<4x?x4096xf16>{%dim_5120} to #hal.device.promise<@__device_4>
%7358 = torch_c.from_builtin_tensor %7357 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7359 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5121 = arith.constant 1 : index
%dim_5122 = tensor.dim %7359, %c1_5121 : tensor<4x?x4096xf16>
%7360 = flow.tensor.transfer %7359 : tensor<4x?x4096xf16>{%dim_5122} to #hal.device.promise<@__device_4>
%7361 = torch_c.from_builtin_tensor %7360 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7362 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5123 = arith.constant 1 : index
%dim_5124 = tensor.dim %7362, %c1_5123 : tensor<4x?x4096xf16>
%7363 = flow.tensor.transfer %7362 : tensor<4x?x4096xf16>{%dim_5124} to #hal.device.promise<@__device_4>
%7364 = torch_c.from_builtin_tensor %7363 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7365 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5125 = arith.constant 1 : index
%dim_5126 = tensor.dim %7365, %c1_5125 : tensor<4x?x4096xf16>
%7366 = flow.tensor.transfer %7365 : tensor<4x?x4096xf16>{%dim_5126} to #hal.device.promise<@__device_4>
%7367 = torch_c.from_builtin_tensor %7366 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5127 = torch.constant.int 1
%7368 = torch.aten.add.Tensor %7349, %7352, %int1_5127 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5128 = torch.constant.int 1
%7369 = torch.aten.add.Tensor %7368, %7355, %int1_5128 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5129 = torch.constant.int 1
%7370 = torch.aten.add.Tensor %7369, %7358, %int1_5129 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5130 = torch.constant.int 1
%7371 = torch.aten.add.Tensor %7370, %7216, %int1_5130 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5131 = torch.constant.int 1
%7372 = torch.aten.add.Tensor %7371, %7361, %int1_5131 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5132 = torch.constant.int 1
%7373 = torch.aten.add.Tensor %7372, %7364, %int1_5132 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5133 = torch.constant.int 1
%7374 = torch.aten.add.Tensor %7373, %7367, %int1_5133 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7375 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5134 = arith.constant 1 : index
%dim_5135 = tensor.dim %7375, %c1_5134 : tensor<4x?x4096xf16>
%7376 = flow.tensor.transfer %7375 : tensor<4x?x4096xf16>{%dim_5135} to #hal.device.promise<@__device_5>
%7377 = torch_c.from_builtin_tensor %7376 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7378 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5136 = arith.constant 1 : index
%dim_5137 = tensor.dim %7378, %c1_5136 : tensor<4x?x4096xf16>
%7379 = flow.tensor.transfer %7378 : tensor<4x?x4096xf16>{%dim_5137} to #hal.device.promise<@__device_5>
%7380 = torch_c.from_builtin_tensor %7379 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7381 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5138 = arith.constant 1 : index
%dim_5139 = tensor.dim %7381, %c1_5138 : tensor<4x?x4096xf16>
%7382 = flow.tensor.transfer %7381 : tensor<4x?x4096xf16>{%dim_5139} to #hal.device.promise<@__device_5>
%7383 = torch_c.from_builtin_tensor %7382 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7384 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5140 = arith.constant 1 : index
%dim_5141 = tensor.dim %7384, %c1_5140 : tensor<4x?x4096xf16>
%7385 = flow.tensor.transfer %7384 : tensor<4x?x4096xf16>{%dim_5141} to #hal.device.promise<@__device_5>
%7386 = torch_c.from_builtin_tensor %7385 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7387 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5142 = arith.constant 1 : index
%dim_5143 = tensor.dim %7387, %c1_5142 : tensor<4x?x4096xf16>
%7388 = flow.tensor.transfer %7387 : tensor<4x?x4096xf16>{%dim_5143} to #hal.device.promise<@__device_5>
%7389 = torch_c.from_builtin_tensor %7388 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7390 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5144 = arith.constant 1 : index
%dim_5145 = tensor.dim %7390, %c1_5144 : tensor<4x?x4096xf16>
%7391 = flow.tensor.transfer %7390 : tensor<4x?x4096xf16>{%dim_5145} to #hal.device.promise<@__device_5>
%7392 = torch_c.from_builtin_tensor %7391 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7393 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5146 = arith.constant 1 : index
%dim_5147 = tensor.dim %7393, %c1_5146 : tensor<4x?x4096xf16>
%7394 = flow.tensor.transfer %7393 : tensor<4x?x4096xf16>{%dim_5147} to #hal.device.promise<@__device_5>
%7395 = torch_c.from_builtin_tensor %7394 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5148 = torch.constant.int 1
%7396 = torch.aten.add.Tensor %7377, %7380, %int1_5148 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5149 = torch.constant.int 1
%7397 = torch.aten.add.Tensor %7396, %7383, %int1_5149 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5150 = torch.constant.int 1
%7398 = torch.aten.add.Tensor %7397, %7386, %int1_5150 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5151 = torch.constant.int 1
%7399 = torch.aten.add.Tensor %7398, %7389, %int1_5151 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5152 = torch.constant.int 1
%7400 = torch.aten.add.Tensor %7399, %7222, %int1_5152 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5153 = torch.constant.int 1
%7401 = torch.aten.add.Tensor %7400, %7392, %int1_5153 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5154 = torch.constant.int 1
%7402 = torch.aten.add.Tensor %7401, %7395, %int1_5154 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7402, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7403 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5155 = arith.constant 1 : index
%dim_5156 = tensor.dim %7403, %c1_5155 : tensor<4x?x4096xf16>
%7404 = flow.tensor.transfer %7403 : tensor<4x?x4096xf16>{%dim_5156} to #hal.device.promise<@__device_6>
%7405 = torch_c.from_builtin_tensor %7404 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7406 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5157 = arith.constant 1 : index
%dim_5158 = tensor.dim %7406, %c1_5157 : tensor<4x?x4096xf16>
%7407 = flow.tensor.transfer %7406 : tensor<4x?x4096xf16>{%dim_5158} to #hal.device.promise<@__device_6>
%7408 = torch_c.from_builtin_tensor %7407 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7409 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5159 = arith.constant 1 : index
%dim_5160 = tensor.dim %7409, %c1_5159 : tensor<4x?x4096xf16>
%7410 = flow.tensor.transfer %7409 : tensor<4x?x4096xf16>{%dim_5160} to #hal.device.promise<@__device_6>
%7411 = torch_c.from_builtin_tensor %7410 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7412 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5161 = arith.constant 1 : index
%dim_5162 = tensor.dim %7412, %c1_5161 : tensor<4x?x4096xf16>
%7413 = flow.tensor.transfer %7412 : tensor<4x?x4096xf16>{%dim_5162} to #hal.device.promise<@__device_6>
%7414 = torch_c.from_builtin_tensor %7413 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7415 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5163 = arith.constant 1 : index
%dim_5164 = tensor.dim %7415, %c1_5163 : tensor<4x?x4096xf16>
%7416 = flow.tensor.transfer %7415 : tensor<4x?x4096xf16>{%dim_5164} to #hal.device.promise<@__device_6>
%7417 = torch_c.from_builtin_tensor %7416 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7418 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5165 = arith.constant 1 : index
%dim_5166 = tensor.dim %7418, %c1_5165 : tensor<4x?x4096xf16>
%7419 = flow.tensor.transfer %7418 : tensor<4x?x4096xf16>{%dim_5166} to #hal.device.promise<@__device_6>
%7420 = torch_c.from_builtin_tensor %7419 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7421 = torch_c.to_builtin_tensor %7234 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5167 = arith.constant 1 : index
%dim_5168 = tensor.dim %7421, %c1_5167 : tensor<4x?x4096xf16>
%7422 = flow.tensor.transfer %7421 : tensor<4x?x4096xf16>{%dim_5168} to #hal.device.promise<@__device_6>
%7423 = torch_c.from_builtin_tensor %7422 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5169 = torch.constant.int 1
%7424 = torch.aten.add.Tensor %7405, %7408, %int1_5169 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5170 = torch.constant.int 1
%7425 = torch.aten.add.Tensor %7424, %7411, %int1_5170 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5171 = torch.constant.int 1
%7426 = torch.aten.add.Tensor %7425, %7414, %int1_5171 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5172 = torch.constant.int 1
%7427 = torch.aten.add.Tensor %7426, %7417, %int1_5172 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5173 = torch.constant.int 1
%7428 = torch.aten.add.Tensor %7427, %7420, %int1_5173 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5174 = torch.constant.int 1
%7429 = torch.aten.add.Tensor %7428, %7228, %int1_5174 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5175 = torch.constant.int 1
%7430 = torch.aten.add.Tensor %7429, %7423, %int1_5175 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7431 = torch_c.to_builtin_tensor %7192 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5176 = arith.constant 1 : index
%dim_5177 = tensor.dim %7431, %c1_5176 : tensor<4x?x4096xf16>
%7432 = flow.tensor.transfer %7431 : tensor<4x?x4096xf16>{%dim_5177} to #hal.device.promise<@__device_7>
%7433 = torch_c.from_builtin_tensor %7432 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7434 = torch_c.to_builtin_tensor %7198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5178 = arith.constant 1 : index
%dim_5179 = tensor.dim %7434, %c1_5178 : tensor<4x?x4096xf16>
%7435 = flow.tensor.transfer %7434 : tensor<4x?x4096xf16>{%dim_5179} to #hal.device.promise<@__device_7>
%7436 = torch_c.from_builtin_tensor %7435 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7437 = torch_c.to_builtin_tensor %7204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5180 = arith.constant 1 : index
%dim_5181 = tensor.dim %7437, %c1_5180 : tensor<4x?x4096xf16>
%7438 = flow.tensor.transfer %7437 : tensor<4x?x4096xf16>{%dim_5181} to #hal.device.promise<@__device_7>
%7439 = torch_c.from_builtin_tensor %7438 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7440 = torch_c.to_builtin_tensor %7210 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5182 = arith.constant 1 : index
%dim_5183 = tensor.dim %7440, %c1_5182 : tensor<4x?x4096xf16>
%7441 = flow.tensor.transfer %7440 : tensor<4x?x4096xf16>{%dim_5183} to #hal.device.promise<@__device_7>
%7442 = torch_c.from_builtin_tensor %7441 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7443 = torch_c.to_builtin_tensor %7216 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5184 = arith.constant 1 : index
%dim_5185 = tensor.dim %7443, %c1_5184 : tensor<4x?x4096xf16>
%7444 = flow.tensor.transfer %7443 : tensor<4x?x4096xf16>{%dim_5185} to #hal.device.promise<@__device_7>
%7445 = torch_c.from_builtin_tensor %7444 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7446 = torch_c.to_builtin_tensor %7222 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5186 = arith.constant 1 : index
%dim_5187 = tensor.dim %7446, %c1_5186 : tensor<4x?x4096xf16>
%7447 = flow.tensor.transfer %7446 : tensor<4x?x4096xf16>{%dim_5187} to #hal.device.promise<@__device_7>
%7448 = torch_c.from_builtin_tensor %7447 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7449 = torch_c.to_builtin_tensor %7228 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5188 = arith.constant 1 : index
%dim_5189 = tensor.dim %7449, %c1_5188 : tensor<4x?x4096xf16>
%7450 = flow.tensor.transfer %7449 : tensor<4x?x4096xf16>{%dim_5189} to #hal.device.promise<@__device_7>
%7451 = torch_c.from_builtin_tensor %7450 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5190 = torch.constant.int 1
%7452 = torch.aten.add.Tensor %7433, %7436, %int1_5190 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5191 = torch.constant.int 1
%7453 = torch.aten.add.Tensor %7452, %7439, %int1_5191 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5192 = torch.constant.int 1
%7454 = torch.aten.add.Tensor %7453, %7442, %int1_5192 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5193 = torch.constant.int 1
%7455 = torch.aten.add.Tensor %7454, %7445, %int1_5193 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5194 = torch.constant.int 1
%7456 = torch.aten.add.Tensor %7455, %7448, %int1_5194 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5195 = torch.constant.int 1
%7457 = torch.aten.add.Tensor %7456, %7451, %int1_5195 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5196 = torch.constant.int 1
%7458 = torch.aten.add.Tensor %7457, %7234, %int1_5196 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5197 = torch.constant.int 1
%7459 = torch.aten.add.Tensor %6118, %7262, %int1_5197 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5198 = torch.constant.int 1
%7460 = torch.aten.add.Tensor %6119, %7290, %int1_5198 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5199 = torch.constant.int 1
%7461 = torch.aten.add.Tensor %6120, %7318, %int1_5199 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5200 = torch.constant.int 1
%7462 = torch.aten.add.Tensor %6121, %7346, %int1_5200 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5201 = torch.constant.int 1
%7463 = torch.aten.add.Tensor %6122, %7374, %int1_5201 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5202 = torch.constant.int 1
%7464 = torch.aten.add.Tensor %6123, %7402, %int1_5202 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5203 = torch.constant.int 1
%7465 = torch.aten.add.Tensor %6124, %7430, %int1_5203 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5204 = torch.constant.int 1
%7466 = torch.aten.add.Tensor %6125, %7458, %int1_5204 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_5205 = torch.constant.int 6
%7467 = torch.prims.convert_element_type %7459, %int6_5205 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5206 = torch.constant.int 6
%7468 = torch.prims.convert_element_type %7460, %int6_5206 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5207 = torch.constant.int 6
%7469 = torch.prims.convert_element_type %7461, %int6_5207 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5208 = torch.constant.int 6
%7470 = torch.prims.convert_element_type %7462, %int6_5208 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5209 = torch.constant.int 6
%7471 = torch.prims.convert_element_type %7463, %int6_5209 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5210 = torch.constant.int 6
%7472 = torch.prims.convert_element_type %7464, %int6_5210 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5211 = torch.constant.int 6
%7473 = torch.prims.convert_element_type %7465, %int6_5211 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5212 = torch.constant.int 6
%7474 = torch.prims.convert_element_type %7466, %int6_5212 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5213 = torch.constant.int 2
%7475 = torch.aten.pow.Tensor_Scalar %7467, %int2_5213 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5214 = torch.constant.int 2
%7476 = torch.aten.pow.Tensor_Scalar %7468, %int2_5214 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5215 = torch.constant.int 2
%7477 = torch.aten.pow.Tensor_Scalar %7469, %int2_5215 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5216 = torch.constant.int 2
%7478 = torch.aten.pow.Tensor_Scalar %7470, %int2_5216 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5217 = torch.constant.int 2
%7479 = torch.aten.pow.Tensor_Scalar %7471, %int2_5217 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5218 = torch.constant.int 2
%7480 = torch.aten.pow.Tensor_Scalar %7472, %int2_5218 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5219 = torch.constant.int 2
%7481 = torch.aten.pow.Tensor_Scalar %7473, %int2_5219 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5220 = torch.constant.int 2
%7482 = torch.aten.pow.Tensor_Scalar %7474, %int2_5220 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_5221 = torch.constant.int -1
%7483 = torch.prim.ListConstruct %int-1_5221 : (!torch.int) -> !torch.list<int>
%true_5222 = torch.constant.bool true
%none_5223 = torch.constant.none
%7484 = torch.aten.mean.dim %7475, %7483, %true_5222, %none_5223 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5224 = torch.constant.int -1
%7485 = torch.prim.ListConstruct %int-1_5224 : (!torch.int) -> !torch.list<int>
%true_5225 = torch.constant.bool true
%none_5226 = torch.constant.none
%7486 = torch.aten.mean.dim %7476, %7485, %true_5225, %none_5226 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5227 = torch.constant.int -1
%7487 = torch.prim.ListConstruct %int-1_5227 : (!torch.int) -> !torch.list<int>
%true_5228 = torch.constant.bool true
%none_5229 = torch.constant.none
%7488 = torch.aten.mean.dim %7477, %7487, %true_5228, %none_5229 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5230 = torch.constant.int -1
%7489 = torch.prim.ListConstruct %int-1_5230 : (!torch.int) -> !torch.list<int>
%true_5231 = torch.constant.bool true
%none_5232 = torch.constant.none
%7490 = torch.aten.mean.dim %7478, %7489, %true_5231, %none_5232 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5233 = torch.constant.int -1
%7491 = torch.prim.ListConstruct %int-1_5233 : (!torch.int) -> !torch.list<int>
%true_5234 = torch.constant.bool true
%none_5235 = torch.constant.none
%7492 = torch.aten.mean.dim %7479, %7491, %true_5234, %none_5235 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5236 = torch.constant.int -1
%7493 = torch.prim.ListConstruct %int-1_5236 : (!torch.int) -> !torch.list<int>
%true_5237 = torch.constant.bool true
%none_5238 = torch.constant.none
%7494 = torch.aten.mean.dim %7480, %7493, %true_5237, %none_5238 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5239 = torch.constant.int -1
%7495 = torch.prim.ListConstruct %int-1_5239 : (!torch.int) -> !torch.list<int>
%true_5240 = torch.constant.bool true
%none_5241 = torch.constant.none
%7496 = torch.aten.mean.dim %7481, %7495, %true_5240, %none_5241 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5242 = torch.constant.int -1
%7497 = torch.prim.ListConstruct %int-1_5242 : (!torch.int) -> !torch.list<int>
%true_5243 = torch.constant.bool true
%none_5244 = torch.constant.none
%7498 = torch.aten.mean.dim %7482, %7497, %true_5243, %none_5244 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5245 = torch.constant.float 9.9999997473787516E-6
%int1_5246 = torch.constant.int 1
%7499 = torch.aten.add.Scalar %7484, %float9.999990e-06_5245, %int1_5246 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5247 = torch.constant.float 9.9999997473787516E-6
%int1_5248 = torch.constant.int 1
%7500 = torch.aten.add.Scalar %7486, %float9.999990e-06_5247, %int1_5248 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5249 = torch.constant.float 9.9999997473787516E-6
%int1_5250 = torch.constant.int 1
%7501 = torch.aten.add.Scalar %7488, %float9.999990e-06_5249, %int1_5250 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5251 = torch.constant.float 9.9999997473787516E-6
%int1_5252 = torch.constant.int 1
%7502 = torch.aten.add.Scalar %7490, %float9.999990e-06_5251, %int1_5252 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5253 = torch.constant.float 9.9999997473787516E-6
%int1_5254 = torch.constant.int 1
%7503 = torch.aten.add.Scalar %7492, %float9.999990e-06_5253, %int1_5254 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5255 = torch.constant.float 9.9999997473787516E-6
%int1_5256 = torch.constant.int 1
%7504 = torch.aten.add.Scalar %7494, %float9.999990e-06_5255, %int1_5256 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5257 = torch.constant.float 9.9999997473787516E-6
%int1_5258 = torch.constant.int 1
%7505 = torch.aten.add.Scalar %7496, %float9.999990e-06_5257, %int1_5258 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5259 = torch.constant.float 9.9999997473787516E-6
%int1_5260 = torch.constant.int 1
%7506 = torch.aten.add.Scalar %7498, %float9.999990e-06_5259, %int1_5260 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7507 = torch.aten.rsqrt %7499 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7508 = torch.aten.rsqrt %7500 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7509 = torch.aten.rsqrt %7501 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7510 = torch.aten.rsqrt %7502 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7511 = torch.aten.rsqrt %7503 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7512 = torch.aten.rsqrt %7504 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7513 = torch.aten.rsqrt %7505 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7514 = torch.aten.rsqrt %7506 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %7514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%7515 = torch.aten.mul.Tensor %7467, %7507 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7516 = torch.aten.mul.Tensor %7468, %7508 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7517 = torch.aten.mul.Tensor %7469, %7509 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7518 = torch.aten.mul.Tensor %7470, %7510 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7519 = torch.aten.mul.Tensor %7471, %7511 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7520 = torch.aten.mul.Tensor %7472, %7512 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7521 = torch.aten.mul.Tensor %7473, %7513 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7522 = torch.aten.mul.Tensor %7474, %7514 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7523 = torch.aten.mul.Tensor %192, %7515 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7524 = torch.aten.mul.Tensor %193, %7516 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7525 = torch.aten.mul.Tensor %194, %7517 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7526 = torch.aten.mul.Tensor %195, %7518 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7527 = torch.aten.mul.Tensor %196, %7519 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7528 = torch.aten.mul.Tensor %197, %7520 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7529 = torch.aten.mul.Tensor %198, %7521 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%7530 = torch.aten.mul.Tensor %199, %7522 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_5261 = torch.constant.int 5
%7531 = torch.prims.convert_element_type %7523, %int5_5261 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5262 = torch.constant.int 5
%7532 = torch.prims.convert_element_type %7524, %int5_5262 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5263 = torch.constant.int 5
%7533 = torch.prims.convert_element_type %7525, %int5_5263 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5264 = torch.constant.int 5
%7534 = torch.prims.convert_element_type %7526, %int5_5264 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5265 = torch.constant.int 5
%7535 = torch.prims.convert_element_type %7527, %int5_5265 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5266 = torch.constant.int 5
%7536 = torch.prims.convert_element_type %7528, %int5_5266 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5267 = torch.constant.int 5
%7537 = torch.prims.convert_element_type %7529, %int5_5267 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5268 = torch.constant.int 5
%7538 = torch.prims.convert_element_type %7530, %int5_5268 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5269 = torch.constant.int 1
%int0_5270 = torch.constant.int 0
%7539 = torch.prim.ListConstruct %int1_5269, %int0_5270 : (!torch.int, !torch.int) -> !torch.list<int>
%7540 = torch.aten.permute %200, %7539 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5271 = torch.constant.int 1
%int0_5272 = torch.constant.int 0
%7541 = torch.prim.ListConstruct %int1_5271, %int0_5272 : (!torch.int, !torch.int) -> !torch.list<int>
%7542 = torch.aten.permute %201, %7541 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5273 = torch.constant.int 1
%int0_5274 = torch.constant.int 0
%7543 = torch.prim.ListConstruct %int1_5273, %int0_5274 : (!torch.int, !torch.int) -> !torch.list<int>
%7544 = torch.aten.permute %202, %7543 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5275 = torch.constant.int 1
%int0_5276 = torch.constant.int 0
%7545 = torch.prim.ListConstruct %int1_5275, %int0_5276 : (!torch.int, !torch.int) -> !torch.list<int>
%7546 = torch.aten.permute %203, %7545 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5277 = torch.constant.int 1
%int0_5278 = torch.constant.int 0
%7547 = torch.prim.ListConstruct %int1_5277, %int0_5278 : (!torch.int, !torch.int) -> !torch.list<int>
%7548 = torch.aten.permute %204, %7547 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5279 = torch.constant.int 1
%int0_5280 = torch.constant.int 0
%7549 = torch.prim.ListConstruct %int1_5279, %int0_5280 : (!torch.int, !torch.int) -> !torch.list<int>
%7550 = torch.aten.permute %205, %7549 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5281 = torch.constant.int 1
%int0_5282 = torch.constant.int 0
%7551 = torch.prim.ListConstruct %int1_5281, %int0_5282 : (!torch.int, !torch.int) -> !torch.list<int>
%7552 = torch.aten.permute %206, %7551 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5283 = torch.constant.int 1
%int0_5284 = torch.constant.int 0
%7553 = torch.prim.ListConstruct %int1_5283, %int0_5284 : (!torch.int, !torch.int) -> !torch.list<int>
%7554 = torch.aten.permute %207, %7553 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_5285 = torch.constant.int 4
%7555 = torch.aten.mul.int %int4_5285, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5286 = torch.constant.int 4096
%7556 = torch.prim.ListConstruct %7555, %int4096_5286 : (!torch.int, !torch.int) -> !torch.list<int>
%7557 = torch.aten.view %7531, %7556 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7557, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7558 = torch.aten.mm %7557, %7540 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7558, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5287 = torch.constant.int 4
%int1792_5288 = torch.constant.int 1792
%7559 = torch.prim.ListConstruct %int4_5287, %2482, %int1792_5288 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7560 = torch.aten.view %7558, %7559 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5289 = torch.constant.int 4
%7561 = torch.aten.mul.int %int4_5289, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5290 = torch.constant.int 4096
%7562 = torch.prim.ListConstruct %7561, %int4096_5290 : (!torch.int, !torch.int) -> !torch.list<int>
%7563 = torch.aten.view %7532, %7562 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7563, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7564 = torch.aten.mm %7563, %7542 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7564, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5291 = torch.constant.int 4
%int1792_5292 = torch.constant.int 1792
%7565 = torch.prim.ListConstruct %int4_5291, %2482, %int1792_5292 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7566 = torch.aten.view %7564, %7565 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5293 = torch.constant.int 4
%7567 = torch.aten.mul.int %int4_5293, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5294 = torch.constant.int 4096
%7568 = torch.prim.ListConstruct %7567, %int4096_5294 : (!torch.int, !torch.int) -> !torch.list<int>
%7569 = torch.aten.view %7533, %7568 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7569, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7570 = torch.aten.mm %7569, %7544 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7570, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5295 = torch.constant.int 4
%int1792_5296 = torch.constant.int 1792
%7571 = torch.prim.ListConstruct %int4_5295, %2482, %int1792_5296 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7572 = torch.aten.view %7570, %7571 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5297 = torch.constant.int 4
%7573 = torch.aten.mul.int %int4_5297, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5298 = torch.constant.int 4096
%7574 = torch.prim.ListConstruct %7573, %int4096_5298 : (!torch.int, !torch.int) -> !torch.list<int>
%7575 = torch.aten.view %7534, %7574 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7575, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7576 = torch.aten.mm %7575, %7546 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7576, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5299 = torch.constant.int 4
%int1792_5300 = torch.constant.int 1792
%7577 = torch.prim.ListConstruct %int4_5299, %2482, %int1792_5300 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7578 = torch.aten.view %7576, %7577 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5301 = torch.constant.int 4
%7579 = torch.aten.mul.int %int4_5301, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5302 = torch.constant.int 4096
%7580 = torch.prim.ListConstruct %7579, %int4096_5302 : (!torch.int, !torch.int) -> !torch.list<int>
%7581 = torch.aten.view %7535, %7580 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7581, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7582 = torch.aten.mm %7581, %7548 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7582, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5303 = torch.constant.int 4
%int1792_5304 = torch.constant.int 1792
%7583 = torch.prim.ListConstruct %int4_5303, %2482, %int1792_5304 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7584 = torch.aten.view %7582, %7583 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5305 = torch.constant.int 4
%7585 = torch.aten.mul.int %int4_5305, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5306 = torch.constant.int 4096
%7586 = torch.prim.ListConstruct %7585, %int4096_5306 : (!torch.int, !torch.int) -> !torch.list<int>
%7587 = torch.aten.view %7536, %7586 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7587, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7588 = torch.aten.mm %7587, %7550 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7588, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5307 = torch.constant.int 4
%int1792_5308 = torch.constant.int 1792
%7589 = torch.prim.ListConstruct %int4_5307, %2482, %int1792_5308 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7590 = torch.aten.view %7588, %7589 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5309 = torch.constant.int 4
%7591 = torch.aten.mul.int %int4_5309, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5310 = torch.constant.int 4096
%7592 = torch.prim.ListConstruct %7591, %int4096_5310 : (!torch.int, !torch.int) -> !torch.list<int>
%7593 = torch.aten.view %7537, %7592 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7593, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7594 = torch.aten.mm %7593, %7552 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7594, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5311 = torch.constant.int 4
%int1792_5312 = torch.constant.int 1792
%7595 = torch.prim.ListConstruct %int4_5311, %2482, %int1792_5312 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7596 = torch.aten.view %7594, %7595 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5313 = torch.constant.int 4
%7597 = torch.aten.mul.int %int4_5313, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5314 = torch.constant.int 4096
%7598 = torch.prim.ListConstruct %7597, %int4096_5314 : (!torch.int, !torch.int) -> !torch.list<int>
%7599 = torch.aten.view %7538, %7598 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7599, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7600 = torch.aten.mm %7599, %7554 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7600, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5315 = torch.constant.int 4
%int1792_5316 = torch.constant.int 1792
%7601 = torch.prim.ListConstruct %int4_5315, %2482, %int1792_5316 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7602 = torch.aten.view %7600, %7601 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7603 = torch.aten.silu %7560 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7604 = torch.aten.silu %7566 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7605 = torch.aten.silu %7572 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7606 = torch.aten.silu %7578 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7607 = torch.aten.silu %7584 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7608 = torch.aten.silu %7590 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7609 = torch.aten.silu %7596 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7610 = torch.aten.silu %7602 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_5317 = torch.constant.int 1
%int0_5318 = torch.constant.int 0
%7611 = torch.prim.ListConstruct %int1_5317, %int0_5318 : (!torch.int, !torch.int) -> !torch.list<int>
%7612 = torch.aten.permute %208, %7611 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5319 = torch.constant.int 1
%int0_5320 = torch.constant.int 0
%7613 = torch.prim.ListConstruct %int1_5319, %int0_5320 : (!torch.int, !torch.int) -> !torch.list<int>
%7614 = torch.aten.permute %209, %7613 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5321 = torch.constant.int 1
%int0_5322 = torch.constant.int 0
%7615 = torch.prim.ListConstruct %int1_5321, %int0_5322 : (!torch.int, !torch.int) -> !torch.list<int>
%7616 = torch.aten.permute %210, %7615 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5323 = torch.constant.int 1
%int0_5324 = torch.constant.int 0
%7617 = torch.prim.ListConstruct %int1_5323, %int0_5324 : (!torch.int, !torch.int) -> !torch.list<int>
%7618 = torch.aten.permute %211, %7617 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5325 = torch.constant.int 1
%int0_5326 = torch.constant.int 0
%7619 = torch.prim.ListConstruct %int1_5325, %int0_5326 : (!torch.int, !torch.int) -> !torch.list<int>
%7620 = torch.aten.permute %212, %7619 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5327 = torch.constant.int 1
%int0_5328 = torch.constant.int 0
%7621 = torch.prim.ListConstruct %int1_5327, %int0_5328 : (!torch.int, !torch.int) -> !torch.list<int>
%7622 = torch.aten.permute %213, %7621 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5329 = torch.constant.int 1
%int0_5330 = torch.constant.int 0
%7623 = torch.prim.ListConstruct %int1_5329, %int0_5330 : (!torch.int, !torch.int) -> !torch.list<int>
%7624 = torch.aten.permute %214, %7623 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_5331 = torch.constant.int 1
%int0_5332 = torch.constant.int 0
%7625 = torch.prim.ListConstruct %int1_5331, %int0_5332 : (!torch.int, !torch.int) -> !torch.list<int>
%7626 = torch.aten.permute %215, %7625 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_5333 = torch.constant.int 4
%7627 = torch.aten.mul.int %int4_5333, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5334 = torch.constant.int 4096
%7628 = torch.prim.ListConstruct %7627, %int4096_5334 : (!torch.int, !torch.int) -> !torch.list<int>
%7629 = torch.aten.view %7531, %7628 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7629, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7630 = torch.aten.mm %7629, %7612 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7630, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5335 = torch.constant.int 4
%int1792_5336 = torch.constant.int 1792
%7631 = torch.prim.ListConstruct %int4_5335, %2482, %int1792_5336 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7632 = torch.aten.view %7630, %7631 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5337 = torch.constant.int 4
%7633 = torch.aten.mul.int %int4_5337, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5338 = torch.constant.int 4096
%7634 = torch.prim.ListConstruct %7633, %int4096_5338 : (!torch.int, !torch.int) -> !torch.list<int>
%7635 = torch.aten.view %7532, %7634 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7635, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7636 = torch.aten.mm %7635, %7614 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7636, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5339 = torch.constant.int 4
%int1792_5340 = torch.constant.int 1792
%7637 = torch.prim.ListConstruct %int4_5339, %2482, %int1792_5340 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7638 = torch.aten.view %7636, %7637 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5341 = torch.constant.int 4
%7639 = torch.aten.mul.int %int4_5341, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5342 = torch.constant.int 4096
%7640 = torch.prim.ListConstruct %7639, %int4096_5342 : (!torch.int, !torch.int) -> !torch.list<int>
%7641 = torch.aten.view %7533, %7640 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7641, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7642 = torch.aten.mm %7641, %7616 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7642, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5343 = torch.constant.int 4
%int1792_5344 = torch.constant.int 1792
%7643 = torch.prim.ListConstruct %int4_5343, %2482, %int1792_5344 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7644 = torch.aten.view %7642, %7643 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5345 = torch.constant.int 4
%7645 = torch.aten.mul.int %int4_5345, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5346 = torch.constant.int 4096
%7646 = torch.prim.ListConstruct %7645, %int4096_5346 : (!torch.int, !torch.int) -> !torch.list<int>
%7647 = torch.aten.view %7534, %7646 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7647, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7648 = torch.aten.mm %7647, %7618 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7648, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5347 = torch.constant.int 4
%int1792_5348 = torch.constant.int 1792
%7649 = torch.prim.ListConstruct %int4_5347, %2482, %int1792_5348 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7650 = torch.aten.view %7648, %7649 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5349 = torch.constant.int 4
%7651 = torch.aten.mul.int %int4_5349, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5350 = torch.constant.int 4096
%7652 = torch.prim.ListConstruct %7651, %int4096_5350 : (!torch.int, !torch.int) -> !torch.list<int>
%7653 = torch.aten.view %7535, %7652 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7653, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7654 = torch.aten.mm %7653, %7620 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7654, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5351 = torch.constant.int 4
%int1792_5352 = torch.constant.int 1792
%7655 = torch.prim.ListConstruct %int4_5351, %2482, %int1792_5352 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7656 = torch.aten.view %7654, %7655 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5353 = torch.constant.int 4
%7657 = torch.aten.mul.int %int4_5353, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5354 = torch.constant.int 4096
%7658 = torch.prim.ListConstruct %7657, %int4096_5354 : (!torch.int, !torch.int) -> !torch.list<int>
%7659 = torch.aten.view %7536, %7658 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7659, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7660 = torch.aten.mm %7659, %7622 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7660, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5355 = torch.constant.int 4
%int1792_5356 = torch.constant.int 1792
%7661 = torch.prim.ListConstruct %int4_5355, %2482, %int1792_5356 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7662 = torch.aten.view %7660, %7661 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5357 = torch.constant.int 4
%7663 = torch.aten.mul.int %int4_5357, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5358 = torch.constant.int 4096
%7664 = torch.prim.ListConstruct %7663, %int4096_5358 : (!torch.int, !torch.int) -> !torch.list<int>
%7665 = torch.aten.view %7537, %7664 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7665, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7666 = torch.aten.mm %7665, %7624 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7666, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5359 = torch.constant.int 4
%int1792_5360 = torch.constant.int 1792
%7667 = torch.prim.ListConstruct %int4_5359, %2482, %int1792_5360 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7668 = torch.aten.view %7666, %7667 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_5361 = torch.constant.int 4
%7669 = torch.aten.mul.int %int4_5361, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5362 = torch.constant.int 4096
%7670 = torch.prim.ListConstruct %7669, %int4096_5362 : (!torch.int, !torch.int) -> !torch.list<int>
%7671 = torch.aten.view %7538, %7670 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7671, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%7672 = torch.aten.mm %7671, %7626 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7672, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_5363 = torch.constant.int 4
%int1792_5364 = torch.constant.int 1792
%7673 = torch.prim.ListConstruct %int4_5363, %2482, %int1792_5364 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7674 = torch.aten.view %7672, %7673 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7675 = torch.aten.mul.Tensor %7603, %7632 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7676 = torch.aten.mul.Tensor %7604, %7638 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7677 = torch.aten.mul.Tensor %7605, %7644 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7678 = torch.aten.mul.Tensor %7606, %7650 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7679 = torch.aten.mul.Tensor %7607, %7656 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7680 = torch.aten.mul.Tensor %7608, %7662 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7681 = torch.aten.mul.Tensor %7609, %7668 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%7682 = torch.aten.mul.Tensor %7610, %7674 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %7682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_5365 = torch.constant.int 1
%int0_5366 = torch.constant.int 0
%7683 = torch.prim.ListConstruct %int1_5365, %int0_5366 : (!torch.int, !torch.int) -> !torch.list<int>
%7684 = torch.aten.permute %216, %7683 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5367 = torch.constant.int 1
%int0_5368 = torch.constant.int 0
%7685 = torch.prim.ListConstruct %int1_5367, %int0_5368 : (!torch.int, !torch.int) -> !torch.list<int>
%7686 = torch.aten.permute %217, %7685 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5369 = torch.constant.int 1
%int0_5370 = torch.constant.int 0
%7687 = torch.prim.ListConstruct %int1_5369, %int0_5370 : (!torch.int, !torch.int) -> !torch.list<int>
%7688 = torch.aten.permute %218, %7687 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5371 = torch.constant.int 1
%int0_5372 = torch.constant.int 0
%7689 = torch.prim.ListConstruct %int1_5371, %int0_5372 : (!torch.int, !torch.int) -> !torch.list<int>
%7690 = torch.aten.permute %219, %7689 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5373 = torch.constant.int 1
%int0_5374 = torch.constant.int 0
%7691 = torch.prim.ListConstruct %int1_5373, %int0_5374 : (!torch.int, !torch.int) -> !torch.list<int>
%7692 = torch.aten.permute %220, %7691 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5375 = torch.constant.int 1
%int0_5376 = torch.constant.int 0
%7693 = torch.prim.ListConstruct %int1_5375, %int0_5376 : (!torch.int, !torch.int) -> !torch.list<int>
%7694 = torch.aten.permute %221, %7693 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5377 = torch.constant.int 1
%int0_5378 = torch.constant.int 0
%7695 = torch.prim.ListConstruct %int1_5377, %int0_5378 : (!torch.int, !torch.int) -> !torch.list<int>
%7696 = torch.aten.permute %222, %7695 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5379 = torch.constant.int 1
%int0_5380 = torch.constant.int 0
%7697 = torch.prim.ListConstruct %int1_5379, %int0_5380 : (!torch.int, !torch.int) -> !torch.list<int>
%7698 = torch.aten.permute %223, %7697 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_5381 = torch.constant.int 1
%7699 = torch.aten.size.int %7560, %int1_5381 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5382 = torch.constant.int 4
%7700 = torch.aten.mul.int %int4_5382, %7699 : !torch.int, !torch.int -> !torch.int
%int1792_5383 = torch.constant.int 1792
%7701 = torch.prim.ListConstruct %7700, %int1792_5383 : (!torch.int, !torch.int) -> !torch.list<int>
%7702 = torch.aten.view %7675, %7701 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7702, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7703 = torch.aten.mm %7702, %7684 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7703, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5384 = torch.constant.int 4
%int4096_5385 = torch.constant.int 4096
%7704 = torch.prim.ListConstruct %int4_5384, %7699, %int4096_5385 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7705 = torch.aten.view %7703, %7704 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5386 = torch.constant.int 1
%7706 = torch.aten.size.int %7566, %int1_5386 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5387 = torch.constant.int 4
%7707 = torch.aten.mul.int %int4_5387, %7706 : !torch.int, !torch.int -> !torch.int
%int1792_5388 = torch.constant.int 1792
%7708 = torch.prim.ListConstruct %7707, %int1792_5388 : (!torch.int, !torch.int) -> !torch.list<int>
%7709 = torch.aten.view %7676, %7708 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7709, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7710 = torch.aten.mm %7709, %7686 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7710, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5389 = torch.constant.int 4
%int4096_5390 = torch.constant.int 4096
%7711 = torch.prim.ListConstruct %int4_5389, %7706, %int4096_5390 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7712 = torch.aten.view %7710, %7711 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5391 = torch.constant.int 1
%7713 = torch.aten.size.int %7572, %int1_5391 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5392 = torch.constant.int 4
%7714 = torch.aten.mul.int %int4_5392, %7713 : !torch.int, !torch.int -> !torch.int
%int1792_5393 = torch.constant.int 1792
%7715 = torch.prim.ListConstruct %7714, %int1792_5393 : (!torch.int, !torch.int) -> !torch.list<int>
%7716 = torch.aten.view %7677, %7715 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7716, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7717 = torch.aten.mm %7716, %7688 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7717, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5394 = torch.constant.int 4
%int4096_5395 = torch.constant.int 4096
%7718 = torch.prim.ListConstruct %int4_5394, %7713, %int4096_5395 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7719 = torch.aten.view %7717, %7718 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5396 = torch.constant.int 1
%7720 = torch.aten.size.int %7578, %int1_5396 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5397 = torch.constant.int 4
%7721 = torch.aten.mul.int %int4_5397, %7720 : !torch.int, !torch.int -> !torch.int
%int1792_5398 = torch.constant.int 1792
%7722 = torch.prim.ListConstruct %7721, %int1792_5398 : (!torch.int, !torch.int) -> !torch.list<int>
%7723 = torch.aten.view %7678, %7722 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7723, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7724 = torch.aten.mm %7723, %7690 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7724, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5399 = torch.constant.int 4
%int4096_5400 = torch.constant.int 4096
%7725 = torch.prim.ListConstruct %int4_5399, %7720, %int4096_5400 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7726 = torch.aten.view %7724, %7725 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5401 = torch.constant.int 1
%7727 = torch.aten.size.int %7584, %int1_5401 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5402 = torch.constant.int 4
%7728 = torch.aten.mul.int %int4_5402, %7727 : !torch.int, !torch.int -> !torch.int
%int1792_5403 = torch.constant.int 1792
%7729 = torch.prim.ListConstruct %7728, %int1792_5403 : (!torch.int, !torch.int) -> !torch.list<int>
%7730 = torch.aten.view %7679, %7729 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7730, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7731 = torch.aten.mm %7730, %7692 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7731, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5404 = torch.constant.int 4
%int4096_5405 = torch.constant.int 4096
%7732 = torch.prim.ListConstruct %int4_5404, %7727, %int4096_5405 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7733 = torch.aten.view %7731, %7732 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5406 = torch.constant.int 1
%7734 = torch.aten.size.int %7590, %int1_5406 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5407 = torch.constant.int 4
%7735 = torch.aten.mul.int %int4_5407, %7734 : !torch.int, !torch.int -> !torch.int
%int1792_5408 = torch.constant.int 1792
%7736 = torch.prim.ListConstruct %7735, %int1792_5408 : (!torch.int, !torch.int) -> !torch.list<int>
%7737 = torch.aten.view %7680, %7736 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7737, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7738 = torch.aten.mm %7737, %7694 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7738, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5409 = torch.constant.int 4
%int4096_5410 = torch.constant.int 4096
%7739 = torch.prim.ListConstruct %int4_5409, %7734, %int4096_5410 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7740 = torch.aten.view %7738, %7739 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5411 = torch.constant.int 1
%7741 = torch.aten.size.int %7596, %int1_5411 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5412 = torch.constant.int 4
%7742 = torch.aten.mul.int %int4_5412, %7741 : !torch.int, !torch.int -> !torch.int
%int1792_5413 = torch.constant.int 1792
%7743 = torch.prim.ListConstruct %7742, %int1792_5413 : (!torch.int, !torch.int) -> !torch.list<int>
%7744 = torch.aten.view %7681, %7743 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7744, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7745 = torch.aten.mm %7744, %7696 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7745, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5414 = torch.constant.int 4
%int4096_5415 = torch.constant.int 4096
%7746 = torch.prim.ListConstruct %int4_5414, %7741, %int4096_5415 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7747 = torch.aten.view %7745, %7746 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5416 = torch.constant.int 1
%7748 = torch.aten.size.int %7602, %int1_5416 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_5417 = torch.constant.int 4
%7749 = torch.aten.mul.int %int4_5417, %7748 : !torch.int, !torch.int -> !torch.int
%int1792_5418 = torch.constant.int 1792
%7750 = torch.prim.ListConstruct %7749, %int1792_5418 : (!torch.int, !torch.int) -> !torch.list<int>
%7751 = torch.aten.view %7682, %7750 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %7751, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%7752 = torch.aten.mm %7751, %7698 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %7752, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_5419 = torch.constant.int 4
%int4096_5420 = torch.constant.int 4096
%7753 = torch.prim.ListConstruct %int4_5419, %7748, %int4096_5420 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%7754 = torch.aten.view %7752, %7753 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7755 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5421 = arith.constant 1 : index
%dim_5422 = tensor.dim %7755, %c1_5421 : tensor<4x?x4096xf16>
%7756 = flow.tensor.transfer %7755 : tensor<4x?x4096xf16>{%dim_5422} to #hal.device.promise<@__device_0>
%7757 = torch_c.from_builtin_tensor %7756 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7758 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5423 = arith.constant 1 : index
%dim_5424 = tensor.dim %7758, %c1_5423 : tensor<4x?x4096xf16>
%7759 = flow.tensor.transfer %7758 : tensor<4x?x4096xf16>{%dim_5424} to #hal.device.promise<@__device_0>
%7760 = torch_c.from_builtin_tensor %7759 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7761 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5425 = arith.constant 1 : index
%dim_5426 = tensor.dim %7761, %c1_5425 : tensor<4x?x4096xf16>
%7762 = flow.tensor.transfer %7761 : tensor<4x?x4096xf16>{%dim_5426} to #hal.device.promise<@__device_0>
%7763 = torch_c.from_builtin_tensor %7762 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7764 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5427 = arith.constant 1 : index
%dim_5428 = tensor.dim %7764, %c1_5427 : tensor<4x?x4096xf16>
%7765 = flow.tensor.transfer %7764 : tensor<4x?x4096xf16>{%dim_5428} to #hal.device.promise<@__device_0>
%7766 = torch_c.from_builtin_tensor %7765 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7767 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5429 = arith.constant 1 : index
%dim_5430 = tensor.dim %7767, %c1_5429 : tensor<4x?x4096xf16>
%7768 = flow.tensor.transfer %7767 : tensor<4x?x4096xf16>{%dim_5430} to #hal.device.promise<@__device_0>
%7769 = torch_c.from_builtin_tensor %7768 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7770 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5431 = arith.constant 1 : index
%dim_5432 = tensor.dim %7770, %c1_5431 : tensor<4x?x4096xf16>
%7771 = flow.tensor.transfer %7770 : tensor<4x?x4096xf16>{%dim_5432} to #hal.device.promise<@__device_0>
%7772 = torch_c.from_builtin_tensor %7771 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7773 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5433 = arith.constant 1 : index
%dim_5434 = tensor.dim %7773, %c1_5433 : tensor<4x?x4096xf16>
%7774 = flow.tensor.transfer %7773 : tensor<4x?x4096xf16>{%dim_5434} to #hal.device.promise<@__device_0>
%7775 = torch_c.from_builtin_tensor %7774 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5435 = torch.constant.int 1
%7776 = torch.aten.add.Tensor %7705, %7757, %int1_5435 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5436 = torch.constant.int 1
%7777 = torch.aten.add.Tensor %7776, %7760, %int1_5436 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5437 = torch.constant.int 1
%7778 = torch.aten.add.Tensor %7777, %7763, %int1_5437 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5438 = torch.constant.int 1
%7779 = torch.aten.add.Tensor %7778, %7766, %int1_5438 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5439 = torch.constant.int 1
%7780 = torch.aten.add.Tensor %7779, %7769, %int1_5439 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5440 = torch.constant.int 1
%7781 = torch.aten.add.Tensor %7780, %7772, %int1_5440 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5441 = torch.constant.int 1
%7782 = torch.aten.add.Tensor %7781, %7775, %int1_5441 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7783 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5442 = arith.constant 1 : index
%dim_5443 = tensor.dim %7783, %c1_5442 : tensor<4x?x4096xf16>
%7784 = flow.tensor.transfer %7783 : tensor<4x?x4096xf16>{%dim_5443} to #hal.device.promise<@__device_1>
%7785 = torch_c.from_builtin_tensor %7784 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7786 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5444 = arith.constant 1 : index
%dim_5445 = tensor.dim %7786, %c1_5444 : tensor<4x?x4096xf16>
%7787 = flow.tensor.transfer %7786 : tensor<4x?x4096xf16>{%dim_5445} to #hal.device.promise<@__device_1>
%7788 = torch_c.from_builtin_tensor %7787 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7789 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5446 = arith.constant 1 : index
%dim_5447 = tensor.dim %7789, %c1_5446 : tensor<4x?x4096xf16>
%7790 = flow.tensor.transfer %7789 : tensor<4x?x4096xf16>{%dim_5447} to #hal.device.promise<@__device_1>
%7791 = torch_c.from_builtin_tensor %7790 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7792 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5448 = arith.constant 1 : index
%dim_5449 = tensor.dim %7792, %c1_5448 : tensor<4x?x4096xf16>
%7793 = flow.tensor.transfer %7792 : tensor<4x?x4096xf16>{%dim_5449} to #hal.device.promise<@__device_1>
%7794 = torch_c.from_builtin_tensor %7793 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7795 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5450 = arith.constant 1 : index
%dim_5451 = tensor.dim %7795, %c1_5450 : tensor<4x?x4096xf16>
%7796 = flow.tensor.transfer %7795 : tensor<4x?x4096xf16>{%dim_5451} to #hal.device.promise<@__device_1>
%7797 = torch_c.from_builtin_tensor %7796 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7798 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5452 = arith.constant 1 : index
%dim_5453 = tensor.dim %7798, %c1_5452 : tensor<4x?x4096xf16>
%7799 = flow.tensor.transfer %7798 : tensor<4x?x4096xf16>{%dim_5453} to #hal.device.promise<@__device_1>
%7800 = torch_c.from_builtin_tensor %7799 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7801 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5454 = arith.constant 1 : index
%dim_5455 = tensor.dim %7801, %c1_5454 : tensor<4x?x4096xf16>
%7802 = flow.tensor.transfer %7801 : tensor<4x?x4096xf16>{%dim_5455} to #hal.device.promise<@__device_1>
%7803 = torch_c.from_builtin_tensor %7802 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5456 = torch.constant.int 1
%7804 = torch.aten.add.Tensor %7785, %7712, %int1_5456 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5457 = torch.constant.int 1
%7805 = torch.aten.add.Tensor %7804, %7788, %int1_5457 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5458 = torch.constant.int 1
%7806 = torch.aten.add.Tensor %7805, %7791, %int1_5458 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5459 = torch.constant.int 1
%7807 = torch.aten.add.Tensor %7806, %7794, %int1_5459 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5460 = torch.constant.int 1
%7808 = torch.aten.add.Tensor %7807, %7797, %int1_5460 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5461 = torch.constant.int 1
%7809 = torch.aten.add.Tensor %7808, %7800, %int1_5461 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5462 = torch.constant.int 1
%7810 = torch.aten.add.Tensor %7809, %7803, %int1_5462 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7811 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5463 = arith.constant 1 : index
%dim_5464 = tensor.dim %7811, %c1_5463 : tensor<4x?x4096xf16>
%7812 = flow.tensor.transfer %7811 : tensor<4x?x4096xf16>{%dim_5464} to #hal.device.promise<@__device_2>
%7813 = torch_c.from_builtin_tensor %7812 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7814 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5465 = arith.constant 1 : index
%dim_5466 = tensor.dim %7814, %c1_5465 : tensor<4x?x4096xf16>
%7815 = flow.tensor.transfer %7814 : tensor<4x?x4096xf16>{%dim_5466} to #hal.device.promise<@__device_2>
%7816 = torch_c.from_builtin_tensor %7815 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7817 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5467 = arith.constant 1 : index
%dim_5468 = tensor.dim %7817, %c1_5467 : tensor<4x?x4096xf16>
%7818 = flow.tensor.transfer %7817 : tensor<4x?x4096xf16>{%dim_5468} to #hal.device.promise<@__device_2>
%7819 = torch_c.from_builtin_tensor %7818 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7820 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5469 = arith.constant 1 : index
%dim_5470 = tensor.dim %7820, %c1_5469 : tensor<4x?x4096xf16>
%7821 = flow.tensor.transfer %7820 : tensor<4x?x4096xf16>{%dim_5470} to #hal.device.promise<@__device_2>
%7822 = torch_c.from_builtin_tensor %7821 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7823 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5471 = arith.constant 1 : index
%dim_5472 = tensor.dim %7823, %c1_5471 : tensor<4x?x4096xf16>
%7824 = flow.tensor.transfer %7823 : tensor<4x?x4096xf16>{%dim_5472} to #hal.device.promise<@__device_2>
%7825 = torch_c.from_builtin_tensor %7824 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7826 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5473 = arith.constant 1 : index
%dim_5474 = tensor.dim %7826, %c1_5473 : tensor<4x?x4096xf16>
%7827 = flow.tensor.transfer %7826 : tensor<4x?x4096xf16>{%dim_5474} to #hal.device.promise<@__device_2>
%7828 = torch_c.from_builtin_tensor %7827 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7829 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5475 = arith.constant 1 : index
%dim_5476 = tensor.dim %7829, %c1_5475 : tensor<4x?x4096xf16>
%7830 = flow.tensor.transfer %7829 : tensor<4x?x4096xf16>{%dim_5476} to #hal.device.promise<@__device_2>
%7831 = torch_c.from_builtin_tensor %7830 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5477 = torch.constant.int 1
%7832 = torch.aten.add.Tensor %7813, %7816, %int1_5477 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5478 = torch.constant.int 1
%7833 = torch.aten.add.Tensor %7832, %7719, %int1_5478 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5479 = torch.constant.int 1
%7834 = torch.aten.add.Tensor %7833, %7819, %int1_5479 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5480 = torch.constant.int 1
%7835 = torch.aten.add.Tensor %7834, %7822, %int1_5480 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5481 = torch.constant.int 1
%7836 = torch.aten.add.Tensor %7835, %7825, %int1_5481 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5482 = torch.constant.int 1
%7837 = torch.aten.add.Tensor %7836, %7828, %int1_5482 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5483 = torch.constant.int 1
%7838 = torch.aten.add.Tensor %7837, %7831, %int1_5483 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7839 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5484 = arith.constant 1 : index
%dim_5485 = tensor.dim %7839, %c1_5484 : tensor<4x?x4096xf16>
%7840 = flow.tensor.transfer %7839 : tensor<4x?x4096xf16>{%dim_5485} to #hal.device.promise<@__device_3>
%7841 = torch_c.from_builtin_tensor %7840 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7842 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5486 = arith.constant 1 : index
%dim_5487 = tensor.dim %7842, %c1_5486 : tensor<4x?x4096xf16>
%7843 = flow.tensor.transfer %7842 : tensor<4x?x4096xf16>{%dim_5487} to #hal.device.promise<@__device_3>
%7844 = torch_c.from_builtin_tensor %7843 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7845 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5488 = arith.constant 1 : index
%dim_5489 = tensor.dim %7845, %c1_5488 : tensor<4x?x4096xf16>
%7846 = flow.tensor.transfer %7845 : tensor<4x?x4096xf16>{%dim_5489} to #hal.device.promise<@__device_3>
%7847 = torch_c.from_builtin_tensor %7846 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7848 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5490 = arith.constant 1 : index
%dim_5491 = tensor.dim %7848, %c1_5490 : tensor<4x?x4096xf16>
%7849 = flow.tensor.transfer %7848 : tensor<4x?x4096xf16>{%dim_5491} to #hal.device.promise<@__device_3>
%7850 = torch_c.from_builtin_tensor %7849 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7851 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5492 = arith.constant 1 : index
%dim_5493 = tensor.dim %7851, %c1_5492 : tensor<4x?x4096xf16>
%7852 = flow.tensor.transfer %7851 : tensor<4x?x4096xf16>{%dim_5493} to #hal.device.promise<@__device_3>
%7853 = torch_c.from_builtin_tensor %7852 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7854 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5494 = arith.constant 1 : index
%dim_5495 = tensor.dim %7854, %c1_5494 : tensor<4x?x4096xf16>
%7855 = flow.tensor.transfer %7854 : tensor<4x?x4096xf16>{%dim_5495} to #hal.device.promise<@__device_3>
%7856 = torch_c.from_builtin_tensor %7855 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7857 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5496 = arith.constant 1 : index
%dim_5497 = tensor.dim %7857, %c1_5496 : tensor<4x?x4096xf16>
%7858 = flow.tensor.transfer %7857 : tensor<4x?x4096xf16>{%dim_5497} to #hal.device.promise<@__device_3>
%7859 = torch_c.from_builtin_tensor %7858 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5498 = torch.constant.int 1
%7860 = torch.aten.add.Tensor %7841, %7844, %int1_5498 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5499 = torch.constant.int 1
%7861 = torch.aten.add.Tensor %7860, %7847, %int1_5499 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5500 = torch.constant.int 1
%7862 = torch.aten.add.Tensor %7861, %7726, %int1_5500 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5501 = torch.constant.int 1
%7863 = torch.aten.add.Tensor %7862, %7850, %int1_5501 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5502 = torch.constant.int 1
%7864 = torch.aten.add.Tensor %7863, %7853, %int1_5502 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5503 = torch.constant.int 1
%7865 = torch.aten.add.Tensor %7864, %7856, %int1_5503 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5504 = torch.constant.int 1
%7866 = torch.aten.add.Tensor %7865, %7859, %int1_5504 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7867 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5505 = arith.constant 1 : index
%dim_5506 = tensor.dim %7867, %c1_5505 : tensor<4x?x4096xf16>
%7868 = flow.tensor.transfer %7867 : tensor<4x?x4096xf16>{%dim_5506} to #hal.device.promise<@__device_4>
%7869 = torch_c.from_builtin_tensor %7868 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7870 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5507 = arith.constant 1 : index
%dim_5508 = tensor.dim %7870, %c1_5507 : tensor<4x?x4096xf16>
%7871 = flow.tensor.transfer %7870 : tensor<4x?x4096xf16>{%dim_5508} to #hal.device.promise<@__device_4>
%7872 = torch_c.from_builtin_tensor %7871 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7873 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5509 = arith.constant 1 : index
%dim_5510 = tensor.dim %7873, %c1_5509 : tensor<4x?x4096xf16>
%7874 = flow.tensor.transfer %7873 : tensor<4x?x4096xf16>{%dim_5510} to #hal.device.promise<@__device_4>
%7875 = torch_c.from_builtin_tensor %7874 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7876 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5511 = arith.constant 1 : index
%dim_5512 = tensor.dim %7876, %c1_5511 : tensor<4x?x4096xf16>
%7877 = flow.tensor.transfer %7876 : tensor<4x?x4096xf16>{%dim_5512} to #hal.device.promise<@__device_4>
%7878 = torch_c.from_builtin_tensor %7877 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7879 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5513 = arith.constant 1 : index
%dim_5514 = tensor.dim %7879, %c1_5513 : tensor<4x?x4096xf16>
%7880 = flow.tensor.transfer %7879 : tensor<4x?x4096xf16>{%dim_5514} to #hal.device.promise<@__device_4>
%7881 = torch_c.from_builtin_tensor %7880 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7882 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5515 = arith.constant 1 : index
%dim_5516 = tensor.dim %7882, %c1_5515 : tensor<4x?x4096xf16>
%7883 = flow.tensor.transfer %7882 : tensor<4x?x4096xf16>{%dim_5516} to #hal.device.promise<@__device_4>
%7884 = torch_c.from_builtin_tensor %7883 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7885 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5517 = arith.constant 1 : index
%dim_5518 = tensor.dim %7885, %c1_5517 : tensor<4x?x4096xf16>
%7886 = flow.tensor.transfer %7885 : tensor<4x?x4096xf16>{%dim_5518} to #hal.device.promise<@__device_4>
%7887 = torch_c.from_builtin_tensor %7886 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5519 = torch.constant.int 1
%7888 = torch.aten.add.Tensor %7869, %7872, %int1_5519 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5520 = torch.constant.int 1
%7889 = torch.aten.add.Tensor %7888, %7875, %int1_5520 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5521 = torch.constant.int 1
%7890 = torch.aten.add.Tensor %7889, %7878, %int1_5521 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5522 = torch.constant.int 1
%7891 = torch.aten.add.Tensor %7890, %7733, %int1_5522 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5523 = torch.constant.int 1
%7892 = torch.aten.add.Tensor %7891, %7881, %int1_5523 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5524 = torch.constant.int 1
%7893 = torch.aten.add.Tensor %7892, %7884, %int1_5524 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5525 = torch.constant.int 1
%7894 = torch.aten.add.Tensor %7893, %7887, %int1_5525 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7895 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5526 = arith.constant 1 : index
%dim_5527 = tensor.dim %7895, %c1_5526 : tensor<4x?x4096xf16>
%7896 = flow.tensor.transfer %7895 : tensor<4x?x4096xf16>{%dim_5527} to #hal.device.promise<@__device_5>
%7897 = torch_c.from_builtin_tensor %7896 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7898 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5528 = arith.constant 1 : index
%dim_5529 = tensor.dim %7898, %c1_5528 : tensor<4x?x4096xf16>
%7899 = flow.tensor.transfer %7898 : tensor<4x?x4096xf16>{%dim_5529} to #hal.device.promise<@__device_5>
%7900 = torch_c.from_builtin_tensor %7899 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7901 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5530 = arith.constant 1 : index
%dim_5531 = tensor.dim %7901, %c1_5530 : tensor<4x?x4096xf16>
%7902 = flow.tensor.transfer %7901 : tensor<4x?x4096xf16>{%dim_5531} to #hal.device.promise<@__device_5>
%7903 = torch_c.from_builtin_tensor %7902 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7904 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5532 = arith.constant 1 : index
%dim_5533 = tensor.dim %7904, %c1_5532 : tensor<4x?x4096xf16>
%7905 = flow.tensor.transfer %7904 : tensor<4x?x4096xf16>{%dim_5533} to #hal.device.promise<@__device_5>
%7906 = torch_c.from_builtin_tensor %7905 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7907 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5534 = arith.constant 1 : index
%dim_5535 = tensor.dim %7907, %c1_5534 : tensor<4x?x4096xf16>
%7908 = flow.tensor.transfer %7907 : tensor<4x?x4096xf16>{%dim_5535} to #hal.device.promise<@__device_5>
%7909 = torch_c.from_builtin_tensor %7908 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7910 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5536 = arith.constant 1 : index
%dim_5537 = tensor.dim %7910, %c1_5536 : tensor<4x?x4096xf16>
%7911 = flow.tensor.transfer %7910 : tensor<4x?x4096xf16>{%dim_5537} to #hal.device.promise<@__device_5>
%7912 = torch_c.from_builtin_tensor %7911 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7913 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5538 = arith.constant 1 : index
%dim_5539 = tensor.dim %7913, %c1_5538 : tensor<4x?x4096xf16>
%7914 = flow.tensor.transfer %7913 : tensor<4x?x4096xf16>{%dim_5539} to #hal.device.promise<@__device_5>
%7915 = torch_c.from_builtin_tensor %7914 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5540 = torch.constant.int 1
%7916 = torch.aten.add.Tensor %7897, %7900, %int1_5540 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5541 = torch.constant.int 1
%7917 = torch.aten.add.Tensor %7916, %7903, %int1_5541 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5542 = torch.constant.int 1
%7918 = torch.aten.add.Tensor %7917, %7906, %int1_5542 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5543 = torch.constant.int 1
%7919 = torch.aten.add.Tensor %7918, %7909, %int1_5543 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5544 = torch.constant.int 1
%7920 = torch.aten.add.Tensor %7919, %7740, %int1_5544 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5545 = torch.constant.int 1
%7921 = torch.aten.add.Tensor %7920, %7912, %int1_5545 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5546 = torch.constant.int 1
%7922 = torch.aten.add.Tensor %7921, %7915, %int1_5546 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7923 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5547 = arith.constant 1 : index
%dim_5548 = tensor.dim %7923, %c1_5547 : tensor<4x?x4096xf16>
%7924 = flow.tensor.transfer %7923 : tensor<4x?x4096xf16>{%dim_5548} to #hal.device.promise<@__device_6>
%7925 = torch_c.from_builtin_tensor %7924 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7926 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5549 = arith.constant 1 : index
%dim_5550 = tensor.dim %7926, %c1_5549 : tensor<4x?x4096xf16>
%7927 = flow.tensor.transfer %7926 : tensor<4x?x4096xf16>{%dim_5550} to #hal.device.promise<@__device_6>
%7928 = torch_c.from_builtin_tensor %7927 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7929 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5551 = arith.constant 1 : index
%dim_5552 = tensor.dim %7929, %c1_5551 : tensor<4x?x4096xf16>
%7930 = flow.tensor.transfer %7929 : tensor<4x?x4096xf16>{%dim_5552} to #hal.device.promise<@__device_6>
%7931 = torch_c.from_builtin_tensor %7930 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7932 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5553 = arith.constant 1 : index
%dim_5554 = tensor.dim %7932, %c1_5553 : tensor<4x?x4096xf16>
%7933 = flow.tensor.transfer %7932 : tensor<4x?x4096xf16>{%dim_5554} to #hal.device.promise<@__device_6>
%7934 = torch_c.from_builtin_tensor %7933 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7935 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5555 = arith.constant 1 : index
%dim_5556 = tensor.dim %7935, %c1_5555 : tensor<4x?x4096xf16>
%7936 = flow.tensor.transfer %7935 : tensor<4x?x4096xf16>{%dim_5556} to #hal.device.promise<@__device_6>
%7937 = torch_c.from_builtin_tensor %7936 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7938 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5557 = arith.constant 1 : index
%dim_5558 = tensor.dim %7938, %c1_5557 : tensor<4x?x4096xf16>
%7939 = flow.tensor.transfer %7938 : tensor<4x?x4096xf16>{%dim_5558} to #hal.device.promise<@__device_6>
%7940 = torch_c.from_builtin_tensor %7939 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7941 = torch_c.to_builtin_tensor %7754 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5559 = arith.constant 1 : index
%dim_5560 = tensor.dim %7941, %c1_5559 : tensor<4x?x4096xf16>
%7942 = flow.tensor.transfer %7941 : tensor<4x?x4096xf16>{%dim_5560} to #hal.device.promise<@__device_6>
%7943 = torch_c.from_builtin_tensor %7942 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5561 = torch.constant.int 1
%7944 = torch.aten.add.Tensor %7925, %7928, %int1_5561 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5562 = torch.constant.int 1
%7945 = torch.aten.add.Tensor %7944, %7931, %int1_5562 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5563 = torch.constant.int 1
%7946 = torch.aten.add.Tensor %7945, %7934, %int1_5563 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5564 = torch.constant.int 1
%7947 = torch.aten.add.Tensor %7946, %7937, %int1_5564 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5565 = torch.constant.int 1
%7948 = torch.aten.add.Tensor %7947, %7940, %int1_5565 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5566 = torch.constant.int 1
%7949 = torch.aten.add.Tensor %7948, %7747, %int1_5566 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5567 = torch.constant.int 1
%7950 = torch.aten.add.Tensor %7949, %7943, %int1_5567 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7951 = torch_c.to_builtin_tensor %7705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5568 = arith.constant 1 : index
%dim_5569 = tensor.dim %7951, %c1_5568 : tensor<4x?x4096xf16>
%7952 = flow.tensor.transfer %7951 : tensor<4x?x4096xf16>{%dim_5569} to #hal.device.promise<@__device_7>
%7953 = torch_c.from_builtin_tensor %7952 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7954 = torch_c.to_builtin_tensor %7712 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5570 = arith.constant 1 : index
%dim_5571 = tensor.dim %7954, %c1_5570 : tensor<4x?x4096xf16>
%7955 = flow.tensor.transfer %7954 : tensor<4x?x4096xf16>{%dim_5571} to #hal.device.promise<@__device_7>
%7956 = torch_c.from_builtin_tensor %7955 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7957 = torch_c.to_builtin_tensor %7719 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5572 = arith.constant 1 : index
%dim_5573 = tensor.dim %7957, %c1_5572 : tensor<4x?x4096xf16>
%7958 = flow.tensor.transfer %7957 : tensor<4x?x4096xf16>{%dim_5573} to #hal.device.promise<@__device_7>
%7959 = torch_c.from_builtin_tensor %7958 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7960 = torch_c.to_builtin_tensor %7726 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5574 = arith.constant 1 : index
%dim_5575 = tensor.dim %7960, %c1_5574 : tensor<4x?x4096xf16>
%7961 = flow.tensor.transfer %7960 : tensor<4x?x4096xf16>{%dim_5575} to #hal.device.promise<@__device_7>
%7962 = torch_c.from_builtin_tensor %7961 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7963 = torch_c.to_builtin_tensor %7733 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5576 = arith.constant 1 : index
%dim_5577 = tensor.dim %7963, %c1_5576 : tensor<4x?x4096xf16>
%7964 = flow.tensor.transfer %7963 : tensor<4x?x4096xf16>{%dim_5577} to #hal.device.promise<@__device_7>
%7965 = torch_c.from_builtin_tensor %7964 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7966 = torch_c.to_builtin_tensor %7740 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5578 = arith.constant 1 : index
%dim_5579 = tensor.dim %7966, %c1_5578 : tensor<4x?x4096xf16>
%7967 = flow.tensor.transfer %7966 : tensor<4x?x4096xf16>{%dim_5579} to #hal.device.promise<@__device_7>
%7968 = torch_c.from_builtin_tensor %7967 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%7969 = torch_c.to_builtin_tensor %7747 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_5580 = arith.constant 1 : index
%dim_5581 = tensor.dim %7969, %c1_5580 : tensor<4x?x4096xf16>
%7970 = flow.tensor.transfer %7969 : tensor<4x?x4096xf16>{%dim_5581} to #hal.device.promise<@__device_7>
%7971 = torch_c.from_builtin_tensor %7970 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5582 = torch.constant.int 1
%7972 = torch.aten.add.Tensor %7953, %7956, %int1_5582 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5583 = torch.constant.int 1
%7973 = torch.aten.add.Tensor %7972, %7959, %int1_5583 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5584 = torch.constant.int 1
%7974 = torch.aten.add.Tensor %7973, %7962, %int1_5584 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5585 = torch.constant.int 1
%7975 = torch.aten.add.Tensor %7974, %7965, %int1_5585 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5586 = torch.constant.int 1
%7976 = torch.aten.add.Tensor %7975, %7968, %int1_5586 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5587 = torch.constant.int 1
%7977 = torch.aten.add.Tensor %7976, %7971, %int1_5587 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5588 = torch.constant.int 1
%7978 = torch.aten.add.Tensor %7977, %7754, %int1_5588 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5589 = torch.constant.int 1
%7979 = torch.aten.add.Tensor %7459, %7782, %int1_5589 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5590 = torch.constant.int 1
%7980 = torch.aten.add.Tensor %7460, %7810, %int1_5590 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5591 = torch.constant.int 1
%7981 = torch.aten.add.Tensor %7461, %7838, %int1_5591 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5592 = torch.constant.int 1
%7982 = torch.aten.add.Tensor %7462, %7866, %int1_5592 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5593 = torch.constant.int 1
%7983 = torch.aten.add.Tensor %7463, %7894, %int1_5593 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5594 = torch.constant.int 1
%7984 = torch.aten.add.Tensor %7464, %7922, %int1_5594 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5595 = torch.constant.int 1
%7985 = torch.aten.add.Tensor %7465, %7950, %int1_5595 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7985, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5596 = torch.constant.int 1
%7986 = torch.aten.add.Tensor %7466, %7978, %int1_5596 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %7986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_5597 = torch.constant.int 6
%7987 = torch.prims.convert_element_type %7979, %int6_5597 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5598 = torch.constant.int 6
%7988 = torch.prims.convert_element_type %7980, %int6_5598 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5599 = torch.constant.int 6
%7989 = torch.prims.convert_element_type %7981, %int6_5599 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5600 = torch.constant.int 6
%7990 = torch.prims.convert_element_type %7982, %int6_5600 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5601 = torch.constant.int 6
%7991 = torch.prims.convert_element_type %7983, %int6_5601 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5602 = torch.constant.int 6
%7992 = torch.prims.convert_element_type %7984, %int6_5602 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5603 = torch.constant.int 6
%7993 = torch.prims.convert_element_type %7985, %int6_5603 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_5604 = torch.constant.int 6
%7994 = torch.prims.convert_element_type %7986, %int6_5604 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5605 = torch.constant.int 2
%7995 = torch.aten.pow.Tensor_Scalar %7987, %int2_5605 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5606 = torch.constant.int 2
%7996 = torch.aten.pow.Tensor_Scalar %7988, %int2_5606 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5607 = torch.constant.int 2
%7997 = torch.aten.pow.Tensor_Scalar %7989, %int2_5607 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5608 = torch.constant.int 2
%7998 = torch.aten.pow.Tensor_Scalar %7990, %int2_5608 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5609 = torch.constant.int 2
%7999 = torch.aten.pow.Tensor_Scalar %7991, %int2_5609 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %7999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5610 = torch.constant.int 2
%8000 = torch.aten.pow.Tensor_Scalar %7992, %int2_5610 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5611 = torch.constant.int 2
%8001 = torch.aten.pow.Tensor_Scalar %7993, %int2_5611 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_5612 = torch.constant.int 2
%8002 = torch.aten.pow.Tensor_Scalar %7994, %int2_5612 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_5613 = torch.constant.int -1
%8003 = torch.prim.ListConstruct %int-1_5613 : (!torch.int) -> !torch.list<int>
%true_5614 = torch.constant.bool true
%none_5615 = torch.constant.none
%8004 = torch.aten.mean.dim %7995, %8003, %true_5614, %none_5615 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5616 = torch.constant.int -1
%8005 = torch.prim.ListConstruct %int-1_5616 : (!torch.int) -> !torch.list<int>
%true_5617 = torch.constant.bool true
%none_5618 = torch.constant.none
%8006 = torch.aten.mean.dim %7996, %8005, %true_5617, %none_5618 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5619 = torch.constant.int -1
%8007 = torch.prim.ListConstruct %int-1_5619 : (!torch.int) -> !torch.list<int>
%true_5620 = torch.constant.bool true
%none_5621 = torch.constant.none
%8008 = torch.aten.mean.dim %7997, %8007, %true_5620, %none_5621 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5622 = torch.constant.int -1
%8009 = torch.prim.ListConstruct %int-1_5622 : (!torch.int) -> !torch.list<int>
%true_5623 = torch.constant.bool true
%none_5624 = torch.constant.none
%8010 = torch.aten.mean.dim %7998, %8009, %true_5623, %none_5624 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5625 = torch.constant.int -1
%8011 = torch.prim.ListConstruct %int-1_5625 : (!torch.int) -> !torch.list<int>
%true_5626 = torch.constant.bool true
%none_5627 = torch.constant.none
%8012 = torch.aten.mean.dim %7999, %8011, %true_5626, %none_5627 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5628 = torch.constant.int -1
%8013 = torch.prim.ListConstruct %int-1_5628 : (!torch.int) -> !torch.list<int>
%true_5629 = torch.constant.bool true
%none_5630 = torch.constant.none
%8014 = torch.aten.mean.dim %8000, %8013, %true_5629, %none_5630 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5631 = torch.constant.int -1
%8015 = torch.prim.ListConstruct %int-1_5631 : (!torch.int) -> !torch.list<int>
%true_5632 = torch.constant.bool true
%none_5633 = torch.constant.none
%8016 = torch.aten.mean.dim %8001, %8015, %true_5632, %none_5633 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_5634 = torch.constant.int -1
%8017 = torch.prim.ListConstruct %int-1_5634 : (!torch.int) -> !torch.list<int>
%true_5635 = torch.constant.bool true
%none_5636 = torch.constant.none
%8018 = torch.aten.mean.dim %8002, %8017, %true_5635, %none_5636 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5637 = torch.constant.float 9.9999997473787516E-6
%int1_5638 = torch.constant.int 1
%8019 = torch.aten.add.Scalar %8004, %float9.999990e-06_5637, %int1_5638 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5639 = torch.constant.float 9.9999997473787516E-6
%int1_5640 = torch.constant.int 1
%8020 = torch.aten.add.Scalar %8006, %float9.999990e-06_5639, %int1_5640 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5641 = torch.constant.float 9.9999997473787516E-6
%int1_5642 = torch.constant.int 1
%8021 = torch.aten.add.Scalar %8008, %float9.999990e-06_5641, %int1_5642 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5643 = torch.constant.float 9.9999997473787516E-6
%int1_5644 = torch.constant.int 1
%8022 = torch.aten.add.Scalar %8010, %float9.999990e-06_5643, %int1_5644 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5645 = torch.constant.float 9.9999997473787516E-6
%int1_5646 = torch.constant.int 1
%8023 = torch.aten.add.Scalar %8012, %float9.999990e-06_5645, %int1_5646 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5647 = torch.constant.float 9.9999997473787516E-6
%int1_5648 = torch.constant.int 1
%8024 = torch.aten.add.Scalar %8014, %float9.999990e-06_5647, %int1_5648 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5649 = torch.constant.float 9.9999997473787516E-6
%int1_5650 = torch.constant.int 1
%8025 = torch.aten.add.Scalar %8016, %float9.999990e-06_5649, %int1_5650 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_5651 = torch.constant.float 9.9999997473787516E-6
%int1_5652 = torch.constant.int 1
%8026 = torch.aten.add.Scalar %8018, %float9.999990e-06_5651, %int1_5652 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8027 = torch.aten.rsqrt %8019 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8028 = torch.aten.rsqrt %8020 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8029 = torch.aten.rsqrt %8021 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8030 = torch.aten.rsqrt %8022 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8031 = torch.aten.rsqrt %8023 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8032 = torch.aten.rsqrt %8024 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8033 = torch.aten.rsqrt %8025 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8034 = torch.aten.rsqrt %8026 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %8034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%8035 = torch.aten.mul.Tensor %7987, %8027 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8036 = torch.aten.mul.Tensor %7988, %8028 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8037 = torch.aten.mul.Tensor %7989, %8029 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8038 = torch.aten.mul.Tensor %7990, %8030 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8039 = torch.aten.mul.Tensor %7991, %8031 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8040 = torch.aten.mul.Tensor %7992, %8032 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8041 = torch.aten.mul.Tensor %7993, %8033 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8042 = torch.aten.mul.Tensor %7994, %8034 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8043 = torch.aten.mul.Tensor %224, %8035 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8044 = torch.aten.mul.Tensor %225, %8036 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8045 = torch.aten.mul.Tensor %226, %8037 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8046 = torch.aten.mul.Tensor %227, %8038 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8047 = torch.aten.mul.Tensor %228, %8039 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8048 = torch.aten.mul.Tensor %229, %8040 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8049 = torch.aten.mul.Tensor %230, %8041 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%8050 = torch.aten.mul.Tensor %231, %8042 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %8050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_5653 = torch.constant.int 5
%8051 = torch.prims.convert_element_type %8043, %int5_5653 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5654 = torch.constant.int 5
%8052 = torch.prims.convert_element_type %8044, %int5_5654 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5655 = torch.constant.int 5
%8053 = torch.prims.convert_element_type %8045, %int5_5655 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5656 = torch.constant.int 5
%8054 = torch.prims.convert_element_type %8046, %int5_5656 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5657 = torch.constant.int 5
%8055 = torch.prims.convert_element_type %8047, %int5_5657 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5658 = torch.constant.int 5
%8056 = torch.prims.convert_element_type %8048, %int5_5658 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5659 = torch.constant.int 5
%8057 = torch.prims.convert_element_type %8049, %int5_5659 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_5660 = torch.constant.int 5
%8058 = torch.prims.convert_element_type %8050, %int5_5660 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %8058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_5661 = torch.constant.int 1
%int0_5662 = torch.constant.int 0
%8059 = torch.prim.ListConstruct %int1_5661, %int0_5662 : (!torch.int, !torch.int) -> !torch.list<int>
%8060 = torch.aten.permute %232, %8059 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5663 = torch.constant.int 1
%int0_5664 = torch.constant.int 0
%8061 = torch.prim.ListConstruct %int1_5663, %int0_5664 : (!torch.int, !torch.int) -> !torch.list<int>
%8062 = torch.aten.permute %233, %8061 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5665 = torch.constant.int 1
%int0_5666 = torch.constant.int 0
%8063 = torch.prim.ListConstruct %int1_5665, %int0_5666 : (!torch.int, !torch.int) -> !torch.list<int>
%8064 = torch.aten.permute %234, %8063 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5667 = torch.constant.int 1
%int0_5668 = torch.constant.int 0
%8065 = torch.prim.ListConstruct %int1_5667, %int0_5668 : (!torch.int, !torch.int) -> !torch.list<int>
%8066 = torch.aten.permute %235, %8065 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5669 = torch.constant.int 1
%int0_5670 = torch.constant.int 0
%8067 = torch.prim.ListConstruct %int1_5669, %int0_5670 : (!torch.int, !torch.int) -> !torch.list<int>
%8068 = torch.aten.permute %236, %8067 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5671 = torch.constant.int 1
%int0_5672 = torch.constant.int 0
%8069 = torch.prim.ListConstruct %int1_5671, %int0_5672 : (!torch.int, !torch.int) -> !torch.list<int>
%8070 = torch.aten.permute %237, %8069 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5673 = torch.constant.int 1
%int0_5674 = torch.constant.int 0
%8071 = torch.prim.ListConstruct %int1_5673, %int0_5674 : (!torch.int, !torch.int) -> !torch.list<int>
%8072 = torch.aten.permute %238, %8071 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_5675 = torch.constant.int 1
%int0_5676 = torch.constant.int 0
%8073 = torch.prim.ListConstruct %int1_5675, %int0_5676 : (!torch.int, !torch.int) -> !torch.list<int>
%8074 = torch.aten.permute %239, %8073 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_5677 = torch.constant.int 4
%8075 = torch.aten.mul.int %int4_5677, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5678 = torch.constant.int 4096
%8076 = torch.prim.ListConstruct %8075, %int4096_5678 : (!torch.int, !torch.int) -> !torch.list<int>
%8077 = torch.aten.view %8051, %8076 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8077, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8078 = torch.aten.mm %8077, %8060 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8078, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5679 = torch.constant.int 4
%int512_5680 = torch.constant.int 512
%8079 = torch.prim.ListConstruct %int4_5679, %2482, %int512_5680 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8080 = torch.aten.view %8078, %8079 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5681 = torch.constant.int 4
%8081 = torch.aten.mul.int %int4_5681, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5682 = torch.constant.int 4096
%8082 = torch.prim.ListConstruct %8081, %int4096_5682 : (!torch.int, !torch.int) -> !torch.list<int>
%8083 = torch.aten.view %8052, %8082 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8083, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8084 = torch.aten.mm %8083, %8062 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8084, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5683 = torch.constant.int 4
%int512_5684 = torch.constant.int 512
%8085 = torch.prim.ListConstruct %int4_5683, %2482, %int512_5684 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8086 = torch.aten.view %8084, %8085 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5685 = torch.constant.int 4
%8087 = torch.aten.mul.int %int4_5685, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5686 = torch.constant.int 4096
%8088 = torch.prim.ListConstruct %8087, %int4096_5686 : (!torch.int, !torch.int) -> !torch.list<int>
%8089 = torch.aten.view %8053, %8088 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8089, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8090 = torch.aten.mm %8089, %8064 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8090, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5687 = torch.constant.int 4
%int512_5688 = torch.constant.int 512
%8091 = torch.prim.ListConstruct %int4_5687, %2482, %int512_5688 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8092 = torch.aten.view %8090, %8091 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5689 = torch.constant.int 4
%8093 = torch.aten.mul.int %int4_5689, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5690 = torch.constant.int 4096
%8094 = torch.prim.ListConstruct %8093, %int4096_5690 : (!torch.int, !torch.int) -> !torch.list<int>
%8095 = torch.aten.view %8054, %8094 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8095, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8096 = torch.aten.mm %8095, %8066 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8096, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5691 = torch.constant.int 4
%int512_5692 = torch.constant.int 512
%8097 = torch.prim.ListConstruct %int4_5691, %2482, %int512_5692 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8098 = torch.aten.view %8096, %8097 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5693 = torch.constant.int 4
%8099 = torch.aten.mul.int %int4_5693, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5694 = torch.constant.int 4096
%8100 = torch.prim.ListConstruct %8099, %int4096_5694 : (!torch.int, !torch.int) -> !torch.list<int>
%8101 = torch.aten.view %8055, %8100 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8101, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8102 = torch.aten.mm %8101, %8068 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8102, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5695 = torch.constant.int 4
%int512_5696 = torch.constant.int 512
%8103 = torch.prim.ListConstruct %int4_5695, %2482, %int512_5696 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8104 = torch.aten.view %8102, %8103 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5697 = torch.constant.int 4
%8105 = torch.aten.mul.int %int4_5697, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5698 = torch.constant.int 4096
%8106 = torch.prim.ListConstruct %8105, %int4096_5698 : (!torch.int, !torch.int) -> !torch.list<int>
%8107 = torch.aten.view %8056, %8106 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8107, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8108 = torch.aten.mm %8107, %8070 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8108, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5699 = torch.constant.int 4
%int512_5700 = torch.constant.int 512
%8109 = torch.prim.ListConstruct %int4_5699, %2482, %int512_5700 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8110 = torch.aten.view %8108, %8109 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5701 = torch.constant.int 4
%8111 = torch.aten.mul.int %int4_5701, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5702 = torch.constant.int 4096
%8112 = torch.prim.ListConstruct %8111, %int4096_5702 : (!torch.int, !torch.int) -> !torch.list<int>
%8113 = torch.aten.view %8057, %8112 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8113, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8114 = torch.aten.mm %8113, %8072 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8114, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5703 = torch.constant.int 4
%int512_5704 = torch.constant.int 512
%8115 = torch.prim.ListConstruct %int4_5703, %2482, %int512_5704 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8116 = torch.aten.view %8114, %8115 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_5705 = torch.constant.int 4
%8117 = torch.aten.mul.int %int4_5705, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5706 = torch.constant.int 4096
%8118 = torch.prim.ListConstruct %8117, %int4096_5706 : (!torch.int, !torch.int) -> !torch.list<int>
%8119 = torch.aten.view %8058, %8118 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8119, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8120 = torch.aten.mm %8119, %8074 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %8120, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_5707 = torch.constant.int 4
%int512_5708 = torch.constant.int 512
%8121 = torch.prim.ListConstruct %int4_5707, %2482, %int512_5708 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8122 = torch.aten.view %8120, %8121 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %8122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_5709 = torch.constant.int 1
%int0_5710 = torch.constant.int 0
%8123 = torch.prim.ListConstruct %int1_5709, %int0_5710 : (!torch.int, !torch.int) -> !torch.list<int>
%8124 = torch.aten.permute %240, %8123 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5711 = torch.constant.int 1
%int0_5712 = torch.constant.int 0
%8125 = torch.prim.ListConstruct %int1_5711, %int0_5712 : (!torch.int, !torch.int) -> !torch.list<int>
%8126 = torch.aten.permute %241, %8125 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5713 = torch.constant.int 1
%int0_5714 = torch.constant.int 0
%8127 = torch.prim.ListConstruct %int1_5713, %int0_5714 : (!torch.int, !torch.int) -> !torch.list<int>
%8128 = torch.aten.permute %242, %8127 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5715 = torch.constant.int 1
%int0_5716 = torch.constant.int 0
%8129 = torch.prim.ListConstruct %int1_5715, %int0_5716 : (!torch.int, !torch.int) -> !torch.list<int>
%8130 = torch.aten.permute %243, %8129 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5717 = torch.constant.int 1
%int0_5718 = torch.constant.int 0
%8131 = torch.prim.ListConstruct %int1_5717, %int0_5718 : (!torch.int, !torch.int) -> !torch.list<int>
%8132 = torch.aten.permute %244, %8131 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5719 = torch.constant.int 1
%int0_5720 = torch.constant.int 0
%8133 = torch.prim.ListConstruct %int1_5719, %int0_5720 : (!torch.int, !torch.int) -> !torch.list<int>
%8134 = torch.aten.permute %245, %8133 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5721 = torch.constant.int 1
%int0_5722 = torch.constant.int 0
%8135 = torch.prim.ListConstruct %int1_5721, %int0_5722 : (!torch.int, !torch.int) -> !torch.list<int>
%8136 = torch.aten.permute %246, %8135 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5723 = torch.constant.int 1
%int0_5724 = torch.constant.int 0
%8137 = torch.prim.ListConstruct %int1_5723, %int0_5724 : (!torch.int, !torch.int) -> !torch.list<int>
%8138 = torch.aten.permute %247, %8137 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_5725 = torch.constant.int 4
%8139 = torch.aten.mul.int %int4_5725, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5726 = torch.constant.int 4096
%8140 = torch.prim.ListConstruct %8139, %int4096_5726 : (!torch.int, !torch.int) -> !torch.list<int>
%8141 = torch.aten.view %8051, %8140 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8141, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8142 = torch.aten.mm %8141, %8124 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8142, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5727 = torch.constant.int 4
%int128_5728 = torch.constant.int 128
%8143 = torch.prim.ListConstruct %int4_5727, %2482, %int128_5728 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8144 = torch.aten.view %8142, %8143 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5729 = torch.constant.int 4
%8145 = torch.aten.mul.int %int4_5729, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5730 = torch.constant.int 4096
%8146 = torch.prim.ListConstruct %8145, %int4096_5730 : (!torch.int, !torch.int) -> !torch.list<int>
%8147 = torch.aten.view %8052, %8146 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8147, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8148 = torch.aten.mm %8147, %8126 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8148, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5731 = torch.constant.int 4
%int128_5732 = torch.constant.int 128
%8149 = torch.prim.ListConstruct %int4_5731, %2482, %int128_5732 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8150 = torch.aten.view %8148, %8149 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5733 = torch.constant.int 4
%8151 = torch.aten.mul.int %int4_5733, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5734 = torch.constant.int 4096
%8152 = torch.prim.ListConstruct %8151, %int4096_5734 : (!torch.int, !torch.int) -> !torch.list<int>
%8153 = torch.aten.view %8053, %8152 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8153, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8154 = torch.aten.mm %8153, %8128 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8154, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5735 = torch.constant.int 4
%int128_5736 = torch.constant.int 128
%8155 = torch.prim.ListConstruct %int4_5735, %2482, %int128_5736 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8156 = torch.aten.view %8154, %8155 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5737 = torch.constant.int 4
%8157 = torch.aten.mul.int %int4_5737, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5738 = torch.constant.int 4096
%8158 = torch.prim.ListConstruct %8157, %int4096_5738 : (!torch.int, !torch.int) -> !torch.list<int>
%8159 = torch.aten.view %8054, %8158 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8159, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8160 = torch.aten.mm %8159, %8130 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8160, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5739 = torch.constant.int 4
%int128_5740 = torch.constant.int 128
%8161 = torch.prim.ListConstruct %int4_5739, %2482, %int128_5740 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8162 = torch.aten.view %8160, %8161 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5741 = torch.constant.int 4
%8163 = torch.aten.mul.int %int4_5741, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5742 = torch.constant.int 4096
%8164 = torch.prim.ListConstruct %8163, %int4096_5742 : (!torch.int, !torch.int) -> !torch.list<int>
%8165 = torch.aten.view %8055, %8164 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8165, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8166 = torch.aten.mm %8165, %8132 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8166, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5743 = torch.constant.int 4
%int128_5744 = torch.constant.int 128
%8167 = torch.prim.ListConstruct %int4_5743, %2482, %int128_5744 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8168 = torch.aten.view %8166, %8167 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5745 = torch.constant.int 4
%8169 = torch.aten.mul.int %int4_5745, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5746 = torch.constant.int 4096
%8170 = torch.prim.ListConstruct %8169, %int4096_5746 : (!torch.int, !torch.int) -> !torch.list<int>
%8171 = torch.aten.view %8056, %8170 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8171, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8172 = torch.aten.mm %8171, %8134 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8172, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5747 = torch.constant.int 4
%int128_5748 = torch.constant.int 128
%8173 = torch.prim.ListConstruct %int4_5747, %2482, %int128_5748 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8174 = torch.aten.view %8172, %8173 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5749 = torch.constant.int 4
%8175 = torch.aten.mul.int %int4_5749, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5750 = torch.constant.int 4096
%8176 = torch.prim.ListConstruct %8175, %int4096_5750 : (!torch.int, !torch.int) -> !torch.list<int>
%8177 = torch.aten.view %8057, %8176 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8177, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8178 = torch.aten.mm %8177, %8136 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8178, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5751 = torch.constant.int 4
%int128_5752 = torch.constant.int 128
%8179 = torch.prim.ListConstruct %int4_5751, %2482, %int128_5752 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8180 = torch.aten.view %8178, %8179 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5753 = torch.constant.int 4
%8181 = torch.aten.mul.int %int4_5753, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5754 = torch.constant.int 4096
%8182 = torch.prim.ListConstruct %8181, %int4096_5754 : (!torch.int, !torch.int) -> !torch.list<int>
%8183 = torch.aten.view %8058, %8182 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8183, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8184 = torch.aten.mm %8183, %8138 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8184, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5755 = torch.constant.int 4
%int128_5756 = torch.constant.int 128
%8185 = torch.prim.ListConstruct %int4_5755, %2482, %int128_5756 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8186 = torch.aten.view %8184, %8185 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_5757 = torch.constant.int 1
%int0_5758 = torch.constant.int 0
%8187 = torch.prim.ListConstruct %int1_5757, %int0_5758 : (!torch.int, !torch.int) -> !torch.list<int>
%8188 = torch.aten.permute %248, %8187 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5759 = torch.constant.int 1
%int0_5760 = torch.constant.int 0
%8189 = torch.prim.ListConstruct %int1_5759, %int0_5760 : (!torch.int, !torch.int) -> !torch.list<int>
%8190 = torch.aten.permute %249, %8189 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5761 = torch.constant.int 1
%int0_5762 = torch.constant.int 0
%8191 = torch.prim.ListConstruct %int1_5761, %int0_5762 : (!torch.int, !torch.int) -> !torch.list<int>
%8192 = torch.aten.permute %250, %8191 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5763 = torch.constant.int 1
%int0_5764 = torch.constant.int 0
%8193 = torch.prim.ListConstruct %int1_5763, %int0_5764 : (!torch.int, !torch.int) -> !torch.list<int>
%8194 = torch.aten.permute %251, %8193 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5765 = torch.constant.int 1
%int0_5766 = torch.constant.int 0
%8195 = torch.prim.ListConstruct %int1_5765, %int0_5766 : (!torch.int, !torch.int) -> !torch.list<int>
%8196 = torch.aten.permute %252, %8195 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5767 = torch.constant.int 1
%int0_5768 = torch.constant.int 0
%8197 = torch.prim.ListConstruct %int1_5767, %int0_5768 : (!torch.int, !torch.int) -> !torch.list<int>
%8198 = torch.aten.permute %253, %8197 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5769 = torch.constant.int 1
%int0_5770 = torch.constant.int 0
%8199 = torch.prim.ListConstruct %int1_5769, %int0_5770 : (!torch.int, !torch.int) -> !torch.list<int>
%8200 = torch.aten.permute %254, %8199 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_5771 = torch.constant.int 1
%int0_5772 = torch.constant.int 0
%8201 = torch.prim.ListConstruct %int1_5771, %int0_5772 : (!torch.int, !torch.int) -> !torch.list<int>
%8202 = torch.aten.permute %255, %8201 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_5773 = torch.constant.int 4
%8203 = torch.aten.mul.int %int4_5773, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5774 = torch.constant.int 4096
%8204 = torch.prim.ListConstruct %8203, %int4096_5774 : (!torch.int, !torch.int) -> !torch.list<int>
%8205 = torch.aten.view %8051, %8204 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8205, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8206 = torch.aten.mm %8205, %8188 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8206, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5775 = torch.constant.int 4
%int128_5776 = torch.constant.int 128
%8207 = torch.prim.ListConstruct %int4_5775, %2482, %int128_5776 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8208 = torch.aten.view %8206, %8207 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5777 = torch.constant.int 4
%8209 = torch.aten.mul.int %int4_5777, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5778 = torch.constant.int 4096
%8210 = torch.prim.ListConstruct %8209, %int4096_5778 : (!torch.int, !torch.int) -> !torch.list<int>
%8211 = torch.aten.view %8052, %8210 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8211, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8212 = torch.aten.mm %8211, %8190 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8212, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5779 = torch.constant.int 4
%int128_5780 = torch.constant.int 128
%8213 = torch.prim.ListConstruct %int4_5779, %2482, %int128_5780 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8214 = torch.aten.view %8212, %8213 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5781 = torch.constant.int 4
%8215 = torch.aten.mul.int %int4_5781, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5782 = torch.constant.int 4096
%8216 = torch.prim.ListConstruct %8215, %int4096_5782 : (!torch.int, !torch.int) -> !torch.list<int>
%8217 = torch.aten.view %8053, %8216 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8217, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8218 = torch.aten.mm %8217, %8192 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8218, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5783 = torch.constant.int 4
%int128_5784 = torch.constant.int 128
%8219 = torch.prim.ListConstruct %int4_5783, %2482, %int128_5784 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8220 = torch.aten.view %8218, %8219 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5785 = torch.constant.int 4
%8221 = torch.aten.mul.int %int4_5785, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5786 = torch.constant.int 4096
%8222 = torch.prim.ListConstruct %8221, %int4096_5786 : (!torch.int, !torch.int) -> !torch.list<int>
%8223 = torch.aten.view %8054, %8222 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8223, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8224 = torch.aten.mm %8223, %8194 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8224, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5787 = torch.constant.int 4
%int128_5788 = torch.constant.int 128
%8225 = torch.prim.ListConstruct %int4_5787, %2482, %int128_5788 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8226 = torch.aten.view %8224, %8225 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5789 = torch.constant.int 4
%8227 = torch.aten.mul.int %int4_5789, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5790 = torch.constant.int 4096
%8228 = torch.prim.ListConstruct %8227, %int4096_5790 : (!torch.int, !torch.int) -> !torch.list<int>
%8229 = torch.aten.view %8055, %8228 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8229, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8230 = torch.aten.mm %8229, %8196 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8230, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5791 = torch.constant.int 4
%int128_5792 = torch.constant.int 128
%8231 = torch.prim.ListConstruct %int4_5791, %2482, %int128_5792 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8232 = torch.aten.view %8230, %8231 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5793 = torch.constant.int 4
%8233 = torch.aten.mul.int %int4_5793, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5794 = torch.constant.int 4096
%8234 = torch.prim.ListConstruct %8233, %int4096_5794 : (!torch.int, !torch.int) -> !torch.list<int>
%8235 = torch.aten.view %8056, %8234 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8235, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8236 = torch.aten.mm %8235, %8198 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8236, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5795 = torch.constant.int 4
%int128_5796 = torch.constant.int 128
%8237 = torch.prim.ListConstruct %int4_5795, %2482, %int128_5796 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8238 = torch.aten.view %8236, %8237 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5797 = torch.constant.int 4
%8239 = torch.aten.mul.int %int4_5797, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5798 = torch.constant.int 4096
%8240 = torch.prim.ListConstruct %8239, %int4096_5798 : (!torch.int, !torch.int) -> !torch.list<int>
%8241 = torch.aten.view %8057, %8240 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8241, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8242 = torch.aten.mm %8241, %8200 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8242, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5799 = torch.constant.int 4
%int128_5800 = torch.constant.int 128
%8243 = torch.prim.ListConstruct %int4_5799, %2482, %int128_5800 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8244 = torch.aten.view %8242, %8243 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5801 = torch.constant.int 4
%8245 = torch.aten.mul.int %int4_5801, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_5802 = torch.constant.int 4096
%8246 = torch.prim.ListConstruct %8245, %int4096_5802 : (!torch.int, !torch.int) -> !torch.list<int>
%8247 = torch.aten.view %8058, %8246 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %8247, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%8248 = torch.aten.mm %8247, %8202 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %8248, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_5803 = torch.constant.int 4
%int128_5804 = torch.constant.int 128
%8249 = torch.prim.ListConstruct %int4_5803, %2482, %int128_5804 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8250 = torch.aten.view %8248, %8249 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %8250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_5805 = torch.constant.int 4
%int4_5806 = torch.constant.int 4
%int128_5807 = torch.constant.int 128
%8251 = torch.prim.ListConstruct %int4_5805, %2482, %int4_5806, %int128_5807 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8252 = torch.aten.view %8080, %8251 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5808 = torch.constant.int 4
%int4_5809 = torch.constant.int 4
%int128_5810 = torch.constant.int 128
%8253 = torch.prim.ListConstruct %int4_5808, %2482, %int4_5809, %int128_5810 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8254 = torch.aten.view %8086, %8253 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5811 = torch.constant.int 4
%int4_5812 = torch.constant.int 4
%int128_5813 = torch.constant.int 128
%8255 = torch.prim.ListConstruct %int4_5811, %2482, %int4_5812, %int128_5813 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8256 = torch.aten.view %8092, %8255 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5814 = torch.constant.int 4
%int4_5815 = torch.constant.int 4
%int128_5816 = torch.constant.int 128
%8257 = torch.prim.ListConstruct %int4_5814, %2482, %int4_5815, %int128_5816 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8258 = torch.aten.view %8098, %8257 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5817 = torch.constant.int 4
%int4_5818 = torch.constant.int 4
%int128_5819 = torch.constant.int 128
%8259 = torch.prim.ListConstruct %int4_5817, %2482, %int4_5818, %int128_5819 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8260 = torch.aten.view %8104, %8259 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5820 = torch.constant.int 4
%int4_5821 = torch.constant.int 4
%int128_5822 = torch.constant.int 128
%8261 = torch.prim.ListConstruct %int4_5820, %2482, %int4_5821, %int128_5822 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8262 = torch.aten.view %8110, %8261 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5823 = torch.constant.int 4
%int4_5824 = torch.constant.int 4
%int128_5825 = torch.constant.int 128
%8263 = torch.prim.ListConstruct %int4_5823, %2482, %int4_5824, %int128_5825 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8264 = torch.aten.view %8116, %8263 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5826 = torch.constant.int 4
%int4_5827 = torch.constant.int 4
%int128_5828 = torch.constant.int 128
%8265 = torch.prim.ListConstruct %int4_5826, %2482, %int4_5827, %int128_5828 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8266 = torch.aten.view %8122, %8265 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_5829 = torch.constant.int 4
%int1_5830 = torch.constant.int 1
%int128_5831 = torch.constant.int 128
%8267 = torch.prim.ListConstruct %int4_5829, %2482, %int1_5830, %int128_5831 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8268 = torch.aten.view %8144, %8267 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5832 = torch.constant.int 4
%int1_5833 = torch.constant.int 1
%int128_5834 = torch.constant.int 128
%8269 = torch.prim.ListConstruct %int4_5832, %2482, %int1_5833, %int128_5834 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8270 = torch.aten.view %8150, %8269 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5835 = torch.constant.int 4
%int1_5836 = torch.constant.int 1
%int128_5837 = torch.constant.int 128
%8271 = torch.prim.ListConstruct %int4_5835, %2482, %int1_5836, %int128_5837 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8272 = torch.aten.view %8156, %8271 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5838 = torch.constant.int 4
%int1_5839 = torch.constant.int 1
%int128_5840 = torch.constant.int 128
%8273 = torch.prim.ListConstruct %int4_5838, %2482, %int1_5839, %int128_5840 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8274 = torch.aten.view %8162, %8273 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5841 = torch.constant.int 4
%int1_5842 = torch.constant.int 1
%int128_5843 = torch.constant.int 128
%8275 = torch.prim.ListConstruct %int4_5841, %2482, %int1_5842, %int128_5843 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8276 = torch.aten.view %8168, %8275 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5844 = torch.constant.int 4
%int1_5845 = torch.constant.int 1
%int128_5846 = torch.constant.int 128
%8277 = torch.prim.ListConstruct %int4_5844, %2482, %int1_5845, %int128_5846 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8278 = torch.aten.view %8174, %8277 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5847 = torch.constant.int 4
%int1_5848 = torch.constant.int 1
%int128_5849 = torch.constant.int 128
%8279 = torch.prim.ListConstruct %int4_5847, %2482, %int1_5848, %int128_5849 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8280 = torch.aten.view %8180, %8279 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5850 = torch.constant.int 4
%int1_5851 = torch.constant.int 1
%int128_5852 = torch.constant.int 128
%8281 = torch.prim.ListConstruct %int4_5850, %2482, %int1_5851, %int128_5852 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8282 = torch.aten.view %8186, %8281 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5853 = torch.constant.int 4
%int1_5854 = torch.constant.int 1
%int128_5855 = torch.constant.int 128
%8283 = torch.prim.ListConstruct %int4_5853, %2482, %int1_5854, %int128_5855 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8284 = torch.aten.view %8208, %8283 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5856 = torch.constant.int 4
%int1_5857 = torch.constant.int 1
%int128_5858 = torch.constant.int 128
%8285 = torch.prim.ListConstruct %int4_5856, %2482, %int1_5857, %int128_5858 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8286 = torch.aten.view %8214, %8285 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5859 = torch.constant.int 4
%int1_5860 = torch.constant.int 1
%int128_5861 = torch.constant.int 128
%8287 = torch.prim.ListConstruct %int4_5859, %2482, %int1_5860, %int128_5861 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8288 = torch.aten.view %8220, %8287 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5862 = torch.constant.int 4
%int1_5863 = torch.constant.int 1
%int128_5864 = torch.constant.int 128
%8289 = torch.prim.ListConstruct %int4_5862, %2482, %int1_5863, %int128_5864 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8290 = torch.aten.view %8226, %8289 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5865 = torch.constant.int 4
%int1_5866 = torch.constant.int 1
%int128_5867 = torch.constant.int 128
%8291 = torch.prim.ListConstruct %int4_5865, %2482, %int1_5866, %int128_5867 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8292 = torch.aten.view %8232, %8291 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5868 = torch.constant.int 4
%int1_5869 = torch.constant.int 1
%int128_5870 = torch.constant.int 128
%8293 = torch.prim.ListConstruct %int4_5868, %2482, %int1_5869, %int128_5870 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8294 = torch.aten.view %8238, %8293 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5871 = torch.constant.int 4
%int1_5872 = torch.constant.int 1
%int128_5873 = torch.constant.int 128
%8295 = torch.prim.ListConstruct %int4_5871, %2482, %int1_5872, %int128_5873 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8296 = torch.aten.view %8244, %8295 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_5874 = torch.constant.int 4
%int1_5875 = torch.constant.int 1
%int128_5876 = torch.constant.int 128
%8297 = torch.prim.ListConstruct %int4_5874, %2482, %int1_5875, %int128_5876 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8298 = torch.aten.view %8250, %8297 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_5877 = torch.constant.int 131072
%none_5878 = torch.constant.none
%none_5879 = torch.constant.none
%cpu_5880 = torch.constant.device "cpu"
%false_5881 = torch.constant.bool false
%8299 = torch.aten.arange %int131072_5877, %none_5878, %none_5879, %cpu_5880, %false_5881 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_5882 = torch.constant.int 0
%int128_5883 = torch.constant.int 128
%int2_5884 = torch.constant.int 2
%none_5885 = torch.constant.none
%none_5886 = torch.constant.none
%cpu_5887 = torch.constant.device "cpu"
%false_5888 = torch.constant.bool false
%8300 = torch.aten.arange.start_step %int0_5882, %int128_5883, %int2_5884, %none_5885, %none_5886, %cpu_5887, %false_5888 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_5889 = torch.constant.int 0
%int0_5890 = torch.constant.int 0
%int64_5891 = torch.constant.int 64
%int1_5892 = torch.constant.int 1
%8301 = torch.aten.slice.Tensor %8300, %int0_5889, %int0_5890, %int64_5891, %int1_5892 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_5893 = torch.constant.int 6
%8302 = torch.prims.convert_element_type %8301, %int6_5893 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_5894 = torch.constant.int 128
%8303 = torch.aten.div.Scalar %8302, %int128_5894 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_5895 = torch.constant.float 5.000000e+05
%8304 = torch.aten.pow.Scalar %float5.000000e05_5895, %8303 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%8305 = torch.aten.reciprocal %8304 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_5896 = torch.constant.float 1.000000e+00
%8306 = torch.aten.mul.Scalar %8305, %float1.000000e00_5896 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_5897 = torch.constant.int 131072
%int1_5898 = torch.constant.int 1
%8307 = torch.prim.ListConstruct %int131072_5897, %int1_5898 : (!torch.int, !torch.int) -> !torch.list<int>
%8308 = torch.aten.view %8299, %8307 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%8309 = torch.aten.mul.Tensor %8308, %8306 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%8310 = torch.aten.cos %8309 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%8311 = torch.aten.sin %8309 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%8312 = torch.aten.complex %8310, %8311 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%8313 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8314 = flow.tensor.transfer %8313 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%8315 = torch_c.from_builtin_tensor %8314 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8316 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8317 = flow.tensor.transfer %8316 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%8318 = torch_c.from_builtin_tensor %8317 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8319 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8320 = flow.tensor.transfer %8319 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%8321 = torch_c.from_builtin_tensor %8320 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8322 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8323 = flow.tensor.transfer %8322 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%8324 = torch_c.from_builtin_tensor %8323 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8325 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8326 = flow.tensor.transfer %8325 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%8327 = torch_c.from_builtin_tensor %8326 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8328 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8329 = flow.tensor.transfer %8328 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%8330 = torch_c.from_builtin_tensor %8329 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8331 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8332 = flow.tensor.transfer %8331 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%8333 = torch_c.from_builtin_tensor %8332 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8334 = torch_c.to_builtin_tensor %8312 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8335 = flow.tensor.transfer %8334 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%8336 = torch_c.from_builtin_tensor %8335 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_5899 = torch.constant.int 1
%8337 = torch.aten.size.int %8080, %int1_5899 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_5900 = torch.constant.int 0
%8338 = torch.aten.add.int %int0_5900, %8337 : !torch.int, !torch.int -> !torch.int
%int0_5901 = torch.constant.int 0
%int0_5902 = torch.constant.int 0
%int1_5903 = torch.constant.int 1
%8339 = torch.aten.slice.Tensor %8315, %int0_5901, %int0_5902, %8338, %int1_5903 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8339, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_5904 = torch.constant.int 1
%int0_5905 = torch.constant.int 0
%int9223372036854775807_5906 = torch.constant.int 9223372036854775807
%int1_5907 = torch.constant.int 1
%8340 = torch.aten.slice.Tensor %8339, %int1_5904, %int0_5905, %int9223372036854775807_5906, %int1_5907 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8340, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_5908 = torch.constant.int 0
%8341 = torch.aten.unsqueeze %8340, %int0_5908 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8341, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_5909 = torch.constant.int 2
%8342 = torch.aten.unsqueeze %8341, %int2_5909 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8342, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_5910 = torch.constant.int 3
%int0_5911 = torch.constant.int 0
%int9223372036854775807_5912 = torch.constant.int 9223372036854775807
%int1_5913 = torch.constant.int 1
%8343 = torch.aten.slice.Tensor %8342, %int3_5910, %int0_5911, %int9223372036854775807_5912, %int1_5913 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8343, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8344 = torch_c.to_builtin_tensor %8252 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_5914 = arith.constant 1 : index
%dim_5915 = tensor.dim %8344, %c1_5914 : tensor<4x?x4x128xf16>
%8345 = flow.tensor.bitcast %8344 : tensor<4x?x4x128xf16>{%dim_5915} -> tensor<4x?x4x64xcomplex<f16>>{%dim_5915}
%8346 = torch_c.from_builtin_tensor %8345 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8347 = torch.aten.mul.Tensor %8346, %8343 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8348 = torch_c.to_builtin_tensor %8347 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_5916 = arith.constant 1 : index
%dim_5917 = tensor.dim %8348, %c1_5916 : tensor<4x?x4x64xcomplex<f32>>
%8349 = flow.tensor.bitcast %8348 : tensor<4x?x4x64xcomplex<f32>>{%dim_5917} -> tensor<4x?x4x128xf32>{%dim_5917}
%8350 = torch_c.from_builtin_tensor %8349 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_5918 = torch.constant.int 5
%8351 = torch.prims.convert_element_type %8350, %int5_5918 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_5919 = torch.constant.int 1
%8352 = torch.aten.size.int %8086, %int1_5919 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_5920 = torch.constant.int 0
%8353 = torch.aten.add.int %int0_5920, %8352 : !torch.int, !torch.int -> !torch.int
%int0_5921 = torch.constant.int 0
%int0_5922 = torch.constant.int 0
%int1_5923 = torch.constant.int 1
%8354 = torch.aten.slice.Tensor %8318, %int0_5921, %int0_5922, %8353, %int1_5923 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8354, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_5924 = torch.constant.int 1
%int0_5925 = torch.constant.int 0
%int9223372036854775807_5926 = torch.constant.int 9223372036854775807
%int1_5927 = torch.constant.int 1
%8355 = torch.aten.slice.Tensor %8354, %int1_5924, %int0_5925, %int9223372036854775807_5926, %int1_5927 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8355, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_5928 = torch.constant.int 0
%8356 = torch.aten.unsqueeze %8355, %int0_5928 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8356, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_5929 = torch.constant.int 2
%8357 = torch.aten.unsqueeze %8356, %int2_5929 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8357, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_5930 = torch.constant.int 3
%int0_5931 = torch.constant.int 0
%int9223372036854775807_5932 = torch.constant.int 9223372036854775807
%int1_5933 = torch.constant.int 1
%8358 = torch.aten.slice.Tensor %8357, %int3_5930, %int0_5931, %int9223372036854775807_5932, %int1_5933 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8358, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8359 = torch_c.to_builtin_tensor %8254 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_5934 = arith.constant 1 : index
%dim_5935 = tensor.dim %8359, %c1_5934 : tensor<4x?x4x128xf16>
%8360 = flow.tensor.bitcast %8359 : tensor<4x?x4x128xf16>{%dim_5935} -> tensor<4x?x4x64xcomplex<f16>>{%dim_5935}
%8361 = torch_c.from_builtin_tensor %8360 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8362 = torch.aten.mul.Tensor %8361, %8358 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8363 = torch_c.to_builtin_tensor %8362 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_5936 = arith.constant 1 : index
%dim_5937 = tensor.dim %8363, %c1_5936 : tensor<4x?x4x64xcomplex<f32>>
%8364 = flow.tensor.bitcast %8363 : tensor<4x?x4x64xcomplex<f32>>{%dim_5937} -> tensor<4x?x4x128xf32>{%dim_5937}
%8365 = torch_c.from_builtin_tensor %8364 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_5938 = torch.constant.int 5
%8366 = torch.prims.convert_element_type %8365, %int5_5938 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_5939 = torch.constant.int 1
%8367 = torch.aten.size.int %8092, %int1_5939 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_5940 = torch.constant.int 0
%8368 = torch.aten.add.int %int0_5940, %8367 : !torch.int, !torch.int -> !torch.int
%int0_5941 = torch.constant.int 0
%int0_5942 = torch.constant.int 0
%int1_5943 = torch.constant.int 1
%8369 = torch.aten.slice.Tensor %8321, %int0_5941, %int0_5942, %8368, %int1_5943 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8369, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_5944 = torch.constant.int 1
%int0_5945 = torch.constant.int 0
%int9223372036854775807_5946 = torch.constant.int 9223372036854775807
%int1_5947 = torch.constant.int 1
%8370 = torch.aten.slice.Tensor %8369, %int1_5944, %int0_5945, %int9223372036854775807_5946, %int1_5947 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8370, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_5948 = torch.constant.int 0
%8371 = torch.aten.unsqueeze %8370, %int0_5948 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8371, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_5949 = torch.constant.int 2
%8372 = torch.aten.unsqueeze %8371, %int2_5949 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8372, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_5950 = torch.constant.int 3
%int0_5951 = torch.constant.int 0
%int9223372036854775807_5952 = torch.constant.int 9223372036854775807
%int1_5953 = torch.constant.int 1
%8373 = torch.aten.slice.Tensor %8372, %int3_5950, %int0_5951, %int9223372036854775807_5952, %int1_5953 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8373, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8374 = torch_c.to_builtin_tensor %8256 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_5954 = arith.constant 1 : index
%dim_5955 = tensor.dim %8374, %c1_5954 : tensor<4x?x4x128xf16>
%8375 = flow.tensor.bitcast %8374 : tensor<4x?x4x128xf16>{%dim_5955} -> tensor<4x?x4x64xcomplex<f16>>{%dim_5955}
%8376 = torch_c.from_builtin_tensor %8375 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8377 = torch.aten.mul.Tensor %8376, %8373 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8378 = torch_c.to_builtin_tensor %8377 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_5956 = arith.constant 1 : index
%dim_5957 = tensor.dim %8378, %c1_5956 : tensor<4x?x4x64xcomplex<f32>>
%8379 = flow.tensor.bitcast %8378 : tensor<4x?x4x64xcomplex<f32>>{%dim_5957} -> tensor<4x?x4x128xf32>{%dim_5957}
%8380 = torch_c.from_builtin_tensor %8379 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_5958 = torch.constant.int 5
%8381 = torch.prims.convert_element_type %8380, %int5_5958 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_5959 = torch.constant.int 1
%8382 = torch.aten.size.int %8098, %int1_5959 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_5960 = torch.constant.int 0
%8383 = torch.aten.add.int %int0_5960, %8382 : !torch.int, !torch.int -> !torch.int
%int0_5961 = torch.constant.int 0
%int0_5962 = torch.constant.int 0
%int1_5963 = torch.constant.int 1
%8384 = torch.aten.slice.Tensor %8324, %int0_5961, %int0_5962, %8383, %int1_5963 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8384, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_5964 = torch.constant.int 1
%int0_5965 = torch.constant.int 0
%int9223372036854775807_5966 = torch.constant.int 9223372036854775807
%int1_5967 = torch.constant.int 1
%8385 = torch.aten.slice.Tensor %8384, %int1_5964, %int0_5965, %int9223372036854775807_5966, %int1_5967 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8385, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_5968 = torch.constant.int 0
%8386 = torch.aten.unsqueeze %8385, %int0_5968 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8386, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_5969 = torch.constant.int 2
%8387 = torch.aten.unsqueeze %8386, %int2_5969 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8387, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_5970 = torch.constant.int 3
%int0_5971 = torch.constant.int 0
%int9223372036854775807_5972 = torch.constant.int 9223372036854775807
%int1_5973 = torch.constant.int 1
%8388 = torch.aten.slice.Tensor %8387, %int3_5970, %int0_5971, %int9223372036854775807_5972, %int1_5973 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8388, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8389 = torch_c.to_builtin_tensor %8258 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_5974 = arith.constant 1 : index
%dim_5975 = tensor.dim %8389, %c1_5974 : tensor<4x?x4x128xf16>
%8390 = flow.tensor.bitcast %8389 : tensor<4x?x4x128xf16>{%dim_5975} -> tensor<4x?x4x64xcomplex<f16>>{%dim_5975}
%8391 = torch_c.from_builtin_tensor %8390 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8392 = torch.aten.mul.Tensor %8391, %8388 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8393 = torch_c.to_builtin_tensor %8392 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_5976 = arith.constant 1 : index
%dim_5977 = tensor.dim %8393, %c1_5976 : tensor<4x?x4x64xcomplex<f32>>
%8394 = flow.tensor.bitcast %8393 : tensor<4x?x4x64xcomplex<f32>>{%dim_5977} -> tensor<4x?x4x128xf32>{%dim_5977}
%8395 = torch_c.from_builtin_tensor %8394 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_5978 = torch.constant.int 5
%8396 = torch.prims.convert_element_type %8395, %int5_5978 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_5979 = torch.constant.int 1
%8397 = torch.aten.size.int %8104, %int1_5979 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_5980 = torch.constant.int 0
%8398 = torch.aten.add.int %int0_5980, %8397 : !torch.int, !torch.int -> !torch.int
%int0_5981 = torch.constant.int 0
%int0_5982 = torch.constant.int 0
%int1_5983 = torch.constant.int 1
%8399 = torch.aten.slice.Tensor %8327, %int0_5981, %int0_5982, %8398, %int1_5983 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8399, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_5984 = torch.constant.int 1
%int0_5985 = torch.constant.int 0
%int9223372036854775807_5986 = torch.constant.int 9223372036854775807
%int1_5987 = torch.constant.int 1
%8400 = torch.aten.slice.Tensor %8399, %int1_5984, %int0_5985, %int9223372036854775807_5986, %int1_5987 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8400, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_5988 = torch.constant.int 0
%8401 = torch.aten.unsqueeze %8400, %int0_5988 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8401, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_5989 = torch.constant.int 2
%8402 = torch.aten.unsqueeze %8401, %int2_5989 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8402, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_5990 = torch.constant.int 3
%int0_5991 = torch.constant.int 0
%int9223372036854775807_5992 = torch.constant.int 9223372036854775807
%int1_5993 = torch.constant.int 1
%8403 = torch.aten.slice.Tensor %8402, %int3_5990, %int0_5991, %int9223372036854775807_5992, %int1_5993 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8403, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8404 = torch_c.to_builtin_tensor %8260 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_5994 = arith.constant 1 : index
%dim_5995 = tensor.dim %8404, %c1_5994 : tensor<4x?x4x128xf16>
%8405 = flow.tensor.bitcast %8404 : tensor<4x?x4x128xf16>{%dim_5995} -> tensor<4x?x4x64xcomplex<f16>>{%dim_5995}
%8406 = torch_c.from_builtin_tensor %8405 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8407 = torch.aten.mul.Tensor %8406, %8403 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8408 = torch_c.to_builtin_tensor %8407 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_5996 = arith.constant 1 : index
%dim_5997 = tensor.dim %8408, %c1_5996 : tensor<4x?x4x64xcomplex<f32>>
%8409 = flow.tensor.bitcast %8408 : tensor<4x?x4x64xcomplex<f32>>{%dim_5997} -> tensor<4x?x4x128xf32>{%dim_5997}
%8410 = torch_c.from_builtin_tensor %8409 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_5998 = torch.constant.int 5
%8411 = torch.prims.convert_element_type %8410, %int5_5998 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_5999 = torch.constant.int 1
%8412 = torch.aten.size.int %8110, %int1_5999 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_6000 = torch.constant.int 0
%8413 = torch.aten.add.int %int0_6000, %8412 : !torch.int, !torch.int -> !torch.int
%int0_6001 = torch.constant.int 0
%int0_6002 = torch.constant.int 0
%int1_6003 = torch.constant.int 1
%8414 = torch.aten.slice.Tensor %8330, %int0_6001, %int0_6002, %8413, %int1_6003 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8414, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6004 = torch.constant.int 1
%int0_6005 = torch.constant.int 0
%int9223372036854775807_6006 = torch.constant.int 9223372036854775807
%int1_6007 = torch.constant.int 1
%8415 = torch.aten.slice.Tensor %8414, %int1_6004, %int0_6005, %int9223372036854775807_6006, %int1_6007 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8415, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6008 = torch.constant.int 0
%8416 = torch.aten.unsqueeze %8415, %int0_6008 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8416, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6009 = torch.constant.int 2
%8417 = torch.aten.unsqueeze %8416, %int2_6009 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8417, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6010 = torch.constant.int 3
%int0_6011 = torch.constant.int 0
%int9223372036854775807_6012 = torch.constant.int 9223372036854775807
%int1_6013 = torch.constant.int 1
%8418 = torch.aten.slice.Tensor %8417, %int3_6010, %int0_6011, %int9223372036854775807_6012, %int1_6013 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8418, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8419 = torch_c.to_builtin_tensor %8262 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_6014 = arith.constant 1 : index
%dim_6015 = tensor.dim %8419, %c1_6014 : tensor<4x?x4x128xf16>
%8420 = flow.tensor.bitcast %8419 : tensor<4x?x4x128xf16>{%dim_6015} -> tensor<4x?x4x64xcomplex<f16>>{%dim_6015}
%8421 = torch_c.from_builtin_tensor %8420 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8422 = torch.aten.mul.Tensor %8421, %8418 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8423 = torch_c.to_builtin_tensor %8422 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_6016 = arith.constant 1 : index
%dim_6017 = tensor.dim %8423, %c1_6016 : tensor<4x?x4x64xcomplex<f32>>
%8424 = flow.tensor.bitcast %8423 : tensor<4x?x4x64xcomplex<f32>>{%dim_6017} -> tensor<4x?x4x128xf32>{%dim_6017}
%8425 = torch_c.from_builtin_tensor %8424 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_6018 = torch.constant.int 5
%8426 = torch.prims.convert_element_type %8425, %int5_6018 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6019 = torch.constant.int 1
%8427 = torch.aten.size.int %8116, %int1_6019 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_6020 = torch.constant.int 0
%8428 = torch.aten.add.int %int0_6020, %8427 : !torch.int, !torch.int -> !torch.int
%int0_6021 = torch.constant.int 0
%int0_6022 = torch.constant.int 0
%int1_6023 = torch.constant.int 1
%8429 = torch.aten.slice.Tensor %8333, %int0_6021, %int0_6022, %8428, %int1_6023 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8429, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6024 = torch.constant.int 1
%int0_6025 = torch.constant.int 0
%int9223372036854775807_6026 = torch.constant.int 9223372036854775807
%int1_6027 = torch.constant.int 1
%8430 = torch.aten.slice.Tensor %8429, %int1_6024, %int0_6025, %int9223372036854775807_6026, %int1_6027 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8430, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6028 = torch.constant.int 0
%8431 = torch.aten.unsqueeze %8430, %int0_6028 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8431, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6029 = torch.constant.int 2
%8432 = torch.aten.unsqueeze %8431, %int2_6029 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8432, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6030 = torch.constant.int 3
%int0_6031 = torch.constant.int 0
%int9223372036854775807_6032 = torch.constant.int 9223372036854775807
%int1_6033 = torch.constant.int 1
%8433 = torch.aten.slice.Tensor %8432, %int3_6030, %int0_6031, %int9223372036854775807_6032, %int1_6033 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8433, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8434 = torch_c.to_builtin_tensor %8264 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_6034 = arith.constant 1 : index
%dim_6035 = tensor.dim %8434, %c1_6034 : tensor<4x?x4x128xf16>
%8435 = flow.tensor.bitcast %8434 : tensor<4x?x4x128xf16>{%dim_6035} -> tensor<4x?x4x64xcomplex<f16>>{%dim_6035}
%8436 = torch_c.from_builtin_tensor %8435 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8437 = torch.aten.mul.Tensor %8436, %8433 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8438 = torch_c.to_builtin_tensor %8437 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_6036 = arith.constant 1 : index
%dim_6037 = tensor.dim %8438, %c1_6036 : tensor<4x?x4x64xcomplex<f32>>
%8439 = flow.tensor.bitcast %8438 : tensor<4x?x4x64xcomplex<f32>>{%dim_6037} -> tensor<4x?x4x128xf32>{%dim_6037}
%8440 = torch_c.from_builtin_tensor %8439 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_6038 = torch.constant.int 5
%8441 = torch.prims.convert_element_type %8440, %int5_6038 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6039 = torch.constant.int 1
%8442 = torch.aten.size.int %8122, %int1_6039 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_6040 = torch.constant.int 0
%8443 = torch.aten.add.int %int0_6040, %8442 : !torch.int, !torch.int -> !torch.int
%int0_6041 = torch.constant.int 0
%int0_6042 = torch.constant.int 0
%int1_6043 = torch.constant.int 1
%8444 = torch.aten.slice.Tensor %8336, %int0_6041, %int0_6042, %8443, %int1_6043 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8444, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6044 = torch.constant.int 1
%int0_6045 = torch.constant.int 0
%int9223372036854775807_6046 = torch.constant.int 9223372036854775807
%int1_6047 = torch.constant.int 1
%8445 = torch.aten.slice.Tensor %8444, %int1_6044, %int0_6045, %int9223372036854775807_6046, %int1_6047 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8445, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6048 = torch.constant.int 0
%8446 = torch.aten.unsqueeze %8445, %int0_6048 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8446, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6049 = torch.constant.int 2
%8447 = torch.aten.unsqueeze %8446, %int2_6049 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8447, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6050 = torch.constant.int 3
%int0_6051 = torch.constant.int 0
%int9223372036854775807_6052 = torch.constant.int 9223372036854775807
%int1_6053 = torch.constant.int 1
%8448 = torch.aten.slice.Tensor %8447, %int3_6050, %int0_6051, %int9223372036854775807_6052, %int1_6053 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8448, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8449 = torch_c.to_builtin_tensor %8266 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_6054 = arith.constant 1 : index
%dim_6055 = tensor.dim %8449, %c1_6054 : tensor<4x?x4x128xf16>
%8450 = flow.tensor.bitcast %8449 : tensor<4x?x4x128xf16>{%dim_6055} -> tensor<4x?x4x64xcomplex<f16>>{%dim_6055}
%8451 = torch_c.from_builtin_tensor %8450 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %8451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%8452 = torch.aten.mul.Tensor %8451, %8448 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %8452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%8453 = torch_c.to_builtin_tensor %8452 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_6056 = arith.constant 1 : index
%dim_6057 = tensor.dim %8453, %c1_6056 : tensor<4x?x4x64xcomplex<f32>>
%8454 = flow.tensor.bitcast %8453 : tensor<4x?x4x64xcomplex<f32>>{%dim_6057} -> tensor<4x?x4x128xf32>{%dim_6057}
%8455 = torch_c.from_builtin_tensor %8454 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %8455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_6058 = torch.constant.int 5
%8456 = torch.prims.convert_element_type %8455, %int5_6058 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_6059 = torch.constant.int 131072
%none_6060 = torch.constant.none
%none_6061 = torch.constant.none
%cpu_6062 = torch.constant.device "cpu"
%false_6063 = torch.constant.bool false
%8457 = torch.aten.arange %int131072_6059, %none_6060, %none_6061, %cpu_6062, %false_6063 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_6064 = torch.constant.int 0
%int128_6065 = torch.constant.int 128
%int2_6066 = torch.constant.int 2
%none_6067 = torch.constant.none
%none_6068 = torch.constant.none
%cpu_6069 = torch.constant.device "cpu"
%false_6070 = torch.constant.bool false
%8458 = torch.aten.arange.start_step %int0_6064, %int128_6065, %int2_6066, %none_6067, %none_6068, %cpu_6069, %false_6070 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_6071 = torch.constant.int 0
%int0_6072 = torch.constant.int 0
%int64_6073 = torch.constant.int 64
%int1_6074 = torch.constant.int 1
%8459 = torch.aten.slice.Tensor %8458, %int0_6071, %int0_6072, %int64_6073, %int1_6074 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_6075 = torch.constant.int 6
%8460 = torch.prims.convert_element_type %8459, %int6_6075 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_6076 = torch.constant.int 128
%8461 = torch.aten.div.Scalar %8460, %int128_6076 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_6077 = torch.constant.float 5.000000e+05
%8462 = torch.aten.pow.Scalar %float5.000000e05_6077, %8461 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%8463 = torch.aten.reciprocal %8462 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_6078 = torch.constant.float 1.000000e+00
%8464 = torch.aten.mul.Scalar %8463, %float1.000000e00_6078 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_6079 = torch.constant.int 131072
%int1_6080 = torch.constant.int 1
%8465 = torch.prim.ListConstruct %int131072_6079, %int1_6080 : (!torch.int, !torch.int) -> !torch.list<int>
%8466 = torch.aten.view %8457, %8465 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%8467 = torch.aten.mul.Tensor %8466, %8464 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%8468 = torch.aten.cos %8467 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%8469 = torch.aten.sin %8467 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%8470 = torch.aten.complex %8468, %8469 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%8471 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8472 = flow.tensor.transfer %8471 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%8473 = torch_c.from_builtin_tensor %8472 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8474 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8475 = flow.tensor.transfer %8474 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%8476 = torch_c.from_builtin_tensor %8475 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8477 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8478 = flow.tensor.transfer %8477 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%8479 = torch_c.from_builtin_tensor %8478 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8480 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8481 = flow.tensor.transfer %8480 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%8482 = torch_c.from_builtin_tensor %8481 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8483 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8484 = flow.tensor.transfer %8483 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%8485 = torch_c.from_builtin_tensor %8484 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8486 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8487 = flow.tensor.transfer %8486 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%8488 = torch_c.from_builtin_tensor %8487 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8489 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8490 = flow.tensor.transfer %8489 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%8491 = torch_c.from_builtin_tensor %8490 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%8492 = torch_c.to_builtin_tensor %8470 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%8493 = flow.tensor.transfer %8492 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%8494 = torch_c.from_builtin_tensor %8493 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_6081 = torch.constant.int 1
%8495 = torch.aten.size.int %8144, %int1_6081 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6082 = torch.constant.int 0
%8496 = torch.aten.add.int %int0_6082, %8495 : !torch.int, !torch.int -> !torch.int
%int0_6083 = torch.constant.int 0
%int0_6084 = torch.constant.int 0
%int1_6085 = torch.constant.int 1
%8497 = torch.aten.slice.Tensor %8473, %int0_6083, %int0_6084, %8496, %int1_6085 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8497, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6086 = torch.constant.int 1
%int0_6087 = torch.constant.int 0
%int9223372036854775807_6088 = torch.constant.int 9223372036854775807
%int1_6089 = torch.constant.int 1
%8498 = torch.aten.slice.Tensor %8497, %int1_6086, %int0_6087, %int9223372036854775807_6088, %int1_6089 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8498, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6090 = torch.constant.int 0
%8499 = torch.aten.unsqueeze %8498, %int0_6090 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8499, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6091 = torch.constant.int 2
%8500 = torch.aten.unsqueeze %8499, %int2_6091 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8500, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6092 = torch.constant.int 3
%int0_6093 = torch.constant.int 0
%int9223372036854775807_6094 = torch.constant.int 9223372036854775807
%int1_6095 = torch.constant.int 1
%8501 = torch.aten.slice.Tensor %8500, %int3_6092, %int0_6093, %int9223372036854775807_6094, %int1_6095 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8501, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8502 = torch_c.to_builtin_tensor %8268 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6096 = arith.constant 1 : index
%dim_6097 = tensor.dim %8502, %c1_6096 : tensor<4x?x1x128xf16>
%8503 = flow.tensor.bitcast %8502 : tensor<4x?x1x128xf16>{%dim_6097} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6097}
%8504 = torch_c.from_builtin_tensor %8503 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8505 = torch.aten.mul.Tensor %8504, %8501 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8506 = torch_c.to_builtin_tensor %8505 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6098 = arith.constant 1 : index
%dim_6099 = tensor.dim %8506, %c1_6098 : tensor<4x?x1x64xcomplex<f32>>
%8507 = flow.tensor.bitcast %8506 : tensor<4x?x1x64xcomplex<f32>>{%dim_6099} -> tensor<4x?x1x128xf32>{%dim_6099}
%8508 = torch_c.from_builtin_tensor %8507 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6100 = torch.constant.int 5
%8509 = torch.prims.convert_element_type %8508, %int5_6100 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6101 = torch.constant.int 1
%8510 = torch.aten.size.int %8150, %int1_6101 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6102 = torch.constant.int 0
%8511 = torch.aten.add.int %int0_6102, %8510 : !torch.int, !torch.int -> !torch.int
%int0_6103 = torch.constant.int 0
%int0_6104 = torch.constant.int 0
%int1_6105 = torch.constant.int 1
%8512 = torch.aten.slice.Tensor %8476, %int0_6103, %int0_6104, %8511, %int1_6105 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8512, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6106 = torch.constant.int 1
%int0_6107 = torch.constant.int 0
%int9223372036854775807_6108 = torch.constant.int 9223372036854775807
%int1_6109 = torch.constant.int 1
%8513 = torch.aten.slice.Tensor %8512, %int1_6106, %int0_6107, %int9223372036854775807_6108, %int1_6109 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8513, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6110 = torch.constant.int 0
%8514 = torch.aten.unsqueeze %8513, %int0_6110 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8514, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6111 = torch.constant.int 2
%8515 = torch.aten.unsqueeze %8514, %int2_6111 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8515, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6112 = torch.constant.int 3
%int0_6113 = torch.constant.int 0
%int9223372036854775807_6114 = torch.constant.int 9223372036854775807
%int1_6115 = torch.constant.int 1
%8516 = torch.aten.slice.Tensor %8515, %int3_6112, %int0_6113, %int9223372036854775807_6114, %int1_6115 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8516, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8517 = torch_c.to_builtin_tensor %8270 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6116 = arith.constant 1 : index
%dim_6117 = tensor.dim %8517, %c1_6116 : tensor<4x?x1x128xf16>
%8518 = flow.tensor.bitcast %8517 : tensor<4x?x1x128xf16>{%dim_6117} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6117}
%8519 = torch_c.from_builtin_tensor %8518 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8520 = torch.aten.mul.Tensor %8519, %8516 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8521 = torch_c.to_builtin_tensor %8520 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6118 = arith.constant 1 : index
%dim_6119 = tensor.dim %8521, %c1_6118 : tensor<4x?x1x64xcomplex<f32>>
%8522 = flow.tensor.bitcast %8521 : tensor<4x?x1x64xcomplex<f32>>{%dim_6119} -> tensor<4x?x1x128xf32>{%dim_6119}
%8523 = torch_c.from_builtin_tensor %8522 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6120 = torch.constant.int 5
%8524 = torch.prims.convert_element_type %8523, %int5_6120 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6121 = torch.constant.int 1
%8525 = torch.aten.size.int %8156, %int1_6121 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6122 = torch.constant.int 0
%8526 = torch.aten.add.int %int0_6122, %8525 : !torch.int, !torch.int -> !torch.int
%int0_6123 = torch.constant.int 0
%int0_6124 = torch.constant.int 0
%int1_6125 = torch.constant.int 1
%8527 = torch.aten.slice.Tensor %8479, %int0_6123, %int0_6124, %8526, %int1_6125 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8527, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6126 = torch.constant.int 1
%int0_6127 = torch.constant.int 0
%int9223372036854775807_6128 = torch.constant.int 9223372036854775807
%int1_6129 = torch.constant.int 1
%8528 = torch.aten.slice.Tensor %8527, %int1_6126, %int0_6127, %int9223372036854775807_6128, %int1_6129 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8528, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6130 = torch.constant.int 0
%8529 = torch.aten.unsqueeze %8528, %int0_6130 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8529, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6131 = torch.constant.int 2
%8530 = torch.aten.unsqueeze %8529, %int2_6131 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8530, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6132 = torch.constant.int 3
%int0_6133 = torch.constant.int 0
%int9223372036854775807_6134 = torch.constant.int 9223372036854775807
%int1_6135 = torch.constant.int 1
%8531 = torch.aten.slice.Tensor %8530, %int3_6132, %int0_6133, %int9223372036854775807_6134, %int1_6135 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8531, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8532 = torch_c.to_builtin_tensor %8272 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6136 = arith.constant 1 : index
%dim_6137 = tensor.dim %8532, %c1_6136 : tensor<4x?x1x128xf16>
%8533 = flow.tensor.bitcast %8532 : tensor<4x?x1x128xf16>{%dim_6137} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6137}
%8534 = torch_c.from_builtin_tensor %8533 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8535 = torch.aten.mul.Tensor %8534, %8531 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8536 = torch_c.to_builtin_tensor %8535 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6138 = arith.constant 1 : index
%dim_6139 = tensor.dim %8536, %c1_6138 : tensor<4x?x1x64xcomplex<f32>>
%8537 = flow.tensor.bitcast %8536 : tensor<4x?x1x64xcomplex<f32>>{%dim_6139} -> tensor<4x?x1x128xf32>{%dim_6139}
%8538 = torch_c.from_builtin_tensor %8537 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6140 = torch.constant.int 5
%8539 = torch.prims.convert_element_type %8538, %int5_6140 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6141 = torch.constant.int 1
%8540 = torch.aten.size.int %8162, %int1_6141 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6142 = torch.constant.int 0
%8541 = torch.aten.add.int %int0_6142, %8540 : !torch.int, !torch.int -> !torch.int
%int0_6143 = torch.constant.int 0
%int0_6144 = torch.constant.int 0
%int1_6145 = torch.constant.int 1
%8542 = torch.aten.slice.Tensor %8482, %int0_6143, %int0_6144, %8541, %int1_6145 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8542, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6146 = torch.constant.int 1
%int0_6147 = torch.constant.int 0
%int9223372036854775807_6148 = torch.constant.int 9223372036854775807
%int1_6149 = torch.constant.int 1
%8543 = torch.aten.slice.Tensor %8542, %int1_6146, %int0_6147, %int9223372036854775807_6148, %int1_6149 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8543, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6150 = torch.constant.int 0
%8544 = torch.aten.unsqueeze %8543, %int0_6150 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8544, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6151 = torch.constant.int 2
%8545 = torch.aten.unsqueeze %8544, %int2_6151 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8545, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6152 = torch.constant.int 3
%int0_6153 = torch.constant.int 0
%int9223372036854775807_6154 = torch.constant.int 9223372036854775807
%int1_6155 = torch.constant.int 1
%8546 = torch.aten.slice.Tensor %8545, %int3_6152, %int0_6153, %int9223372036854775807_6154, %int1_6155 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8546, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8547 = torch_c.to_builtin_tensor %8274 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6156 = arith.constant 1 : index
%dim_6157 = tensor.dim %8547, %c1_6156 : tensor<4x?x1x128xf16>
%8548 = flow.tensor.bitcast %8547 : tensor<4x?x1x128xf16>{%dim_6157} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6157}
%8549 = torch_c.from_builtin_tensor %8548 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8550 = torch.aten.mul.Tensor %8549, %8546 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8551 = torch_c.to_builtin_tensor %8550 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6158 = arith.constant 1 : index
%dim_6159 = tensor.dim %8551, %c1_6158 : tensor<4x?x1x64xcomplex<f32>>
%8552 = flow.tensor.bitcast %8551 : tensor<4x?x1x64xcomplex<f32>>{%dim_6159} -> tensor<4x?x1x128xf32>{%dim_6159}
%8553 = torch_c.from_builtin_tensor %8552 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6160 = torch.constant.int 5
%8554 = torch.prims.convert_element_type %8553, %int5_6160 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6161 = torch.constant.int 1
%8555 = torch.aten.size.int %8168, %int1_6161 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6162 = torch.constant.int 0
%8556 = torch.aten.add.int %int0_6162, %8555 : !torch.int, !torch.int -> !torch.int
%int0_6163 = torch.constant.int 0
%int0_6164 = torch.constant.int 0
%int1_6165 = torch.constant.int 1
%8557 = torch.aten.slice.Tensor %8485, %int0_6163, %int0_6164, %8556, %int1_6165 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8557, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6166 = torch.constant.int 1
%int0_6167 = torch.constant.int 0
%int9223372036854775807_6168 = torch.constant.int 9223372036854775807
%int1_6169 = torch.constant.int 1
%8558 = torch.aten.slice.Tensor %8557, %int1_6166, %int0_6167, %int9223372036854775807_6168, %int1_6169 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8558, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6170 = torch.constant.int 0
%8559 = torch.aten.unsqueeze %8558, %int0_6170 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8559, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6171 = torch.constant.int 2
%8560 = torch.aten.unsqueeze %8559, %int2_6171 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8560, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6172 = torch.constant.int 3
%int0_6173 = torch.constant.int 0
%int9223372036854775807_6174 = torch.constant.int 9223372036854775807
%int1_6175 = torch.constant.int 1
%8561 = torch.aten.slice.Tensor %8560, %int3_6172, %int0_6173, %int9223372036854775807_6174, %int1_6175 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8561, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8562 = torch_c.to_builtin_tensor %8276 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6176 = arith.constant 1 : index
%dim_6177 = tensor.dim %8562, %c1_6176 : tensor<4x?x1x128xf16>
%8563 = flow.tensor.bitcast %8562 : tensor<4x?x1x128xf16>{%dim_6177} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6177}
%8564 = torch_c.from_builtin_tensor %8563 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8565 = torch.aten.mul.Tensor %8564, %8561 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8566 = torch_c.to_builtin_tensor %8565 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6178 = arith.constant 1 : index
%dim_6179 = tensor.dim %8566, %c1_6178 : tensor<4x?x1x64xcomplex<f32>>
%8567 = flow.tensor.bitcast %8566 : tensor<4x?x1x64xcomplex<f32>>{%dim_6179} -> tensor<4x?x1x128xf32>{%dim_6179}
%8568 = torch_c.from_builtin_tensor %8567 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6180 = torch.constant.int 5
%8569 = torch.prims.convert_element_type %8568, %int5_6180 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6181 = torch.constant.int 1
%8570 = torch.aten.size.int %8174, %int1_6181 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6182 = torch.constant.int 0
%8571 = torch.aten.add.int %int0_6182, %8570 : !torch.int, !torch.int -> !torch.int
%int0_6183 = torch.constant.int 0
%int0_6184 = torch.constant.int 0
%int1_6185 = torch.constant.int 1
%8572 = torch.aten.slice.Tensor %8488, %int0_6183, %int0_6184, %8571, %int1_6185 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8572, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6186 = torch.constant.int 1
%int0_6187 = torch.constant.int 0
%int9223372036854775807_6188 = torch.constant.int 9223372036854775807
%int1_6189 = torch.constant.int 1
%8573 = torch.aten.slice.Tensor %8572, %int1_6186, %int0_6187, %int9223372036854775807_6188, %int1_6189 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8573, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6190 = torch.constant.int 0
%8574 = torch.aten.unsqueeze %8573, %int0_6190 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8574, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6191 = torch.constant.int 2
%8575 = torch.aten.unsqueeze %8574, %int2_6191 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8575, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6192 = torch.constant.int 3
%int0_6193 = torch.constant.int 0
%int9223372036854775807_6194 = torch.constant.int 9223372036854775807
%int1_6195 = torch.constant.int 1
%8576 = torch.aten.slice.Tensor %8575, %int3_6192, %int0_6193, %int9223372036854775807_6194, %int1_6195 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8576, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8577 = torch_c.to_builtin_tensor %8278 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6196 = arith.constant 1 : index
%dim_6197 = tensor.dim %8577, %c1_6196 : tensor<4x?x1x128xf16>
%8578 = flow.tensor.bitcast %8577 : tensor<4x?x1x128xf16>{%dim_6197} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6197}
%8579 = torch_c.from_builtin_tensor %8578 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8580 = torch.aten.mul.Tensor %8579, %8576 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8581 = torch_c.to_builtin_tensor %8580 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6198 = arith.constant 1 : index
%dim_6199 = tensor.dim %8581, %c1_6198 : tensor<4x?x1x64xcomplex<f32>>
%8582 = flow.tensor.bitcast %8581 : tensor<4x?x1x64xcomplex<f32>>{%dim_6199} -> tensor<4x?x1x128xf32>{%dim_6199}
%8583 = torch_c.from_builtin_tensor %8582 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6200 = torch.constant.int 5
%8584 = torch.prims.convert_element_type %8583, %int5_6200 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6201 = torch.constant.int 1
%8585 = torch.aten.size.int %8180, %int1_6201 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6202 = torch.constant.int 0
%8586 = torch.aten.add.int %int0_6202, %8585 : !torch.int, !torch.int -> !torch.int
%int0_6203 = torch.constant.int 0
%int0_6204 = torch.constant.int 0
%int1_6205 = torch.constant.int 1
%8587 = torch.aten.slice.Tensor %8491, %int0_6203, %int0_6204, %8586, %int1_6205 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8587, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6206 = torch.constant.int 1
%int0_6207 = torch.constant.int 0
%int9223372036854775807_6208 = torch.constant.int 9223372036854775807
%int1_6209 = torch.constant.int 1
%8588 = torch.aten.slice.Tensor %8587, %int1_6206, %int0_6207, %int9223372036854775807_6208, %int1_6209 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8588, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6210 = torch.constant.int 0
%8589 = torch.aten.unsqueeze %8588, %int0_6210 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8589, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6211 = torch.constant.int 2
%8590 = torch.aten.unsqueeze %8589, %int2_6211 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8590, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6212 = torch.constant.int 3
%int0_6213 = torch.constant.int 0
%int9223372036854775807_6214 = torch.constant.int 9223372036854775807
%int1_6215 = torch.constant.int 1
%8591 = torch.aten.slice.Tensor %8590, %int3_6212, %int0_6213, %int9223372036854775807_6214, %int1_6215 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8591, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8592 = torch_c.to_builtin_tensor %8280 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6216 = arith.constant 1 : index
%dim_6217 = tensor.dim %8592, %c1_6216 : tensor<4x?x1x128xf16>
%8593 = flow.tensor.bitcast %8592 : tensor<4x?x1x128xf16>{%dim_6217} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6217}
%8594 = torch_c.from_builtin_tensor %8593 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8595 = torch.aten.mul.Tensor %8594, %8591 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8596 = torch_c.to_builtin_tensor %8595 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6218 = arith.constant 1 : index
%dim_6219 = tensor.dim %8596, %c1_6218 : tensor<4x?x1x64xcomplex<f32>>
%8597 = flow.tensor.bitcast %8596 : tensor<4x?x1x64xcomplex<f32>>{%dim_6219} -> tensor<4x?x1x128xf32>{%dim_6219}
%8598 = torch_c.from_builtin_tensor %8597 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6220 = torch.constant.int 5
%8599 = torch.prims.convert_element_type %8598, %int5_6220 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_6221 = torch.constant.int 1
%8600 = torch.aten.size.int %8186, %int1_6221 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_6222 = torch.constant.int 0
%8601 = torch.aten.add.int %int0_6222, %8600 : !torch.int, !torch.int -> !torch.int
%int0_6223 = torch.constant.int 0
%int0_6224 = torch.constant.int 0
%int1_6225 = torch.constant.int 1
%8602 = torch.aten.slice.Tensor %8494, %int0_6223, %int0_6224, %8601, %int1_6225 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8602, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_6226 = torch.constant.int 1
%int0_6227 = torch.constant.int 0
%int9223372036854775807_6228 = torch.constant.int 9223372036854775807
%int1_6229 = torch.constant.int 1
%8603 = torch.aten.slice.Tensor %8602, %int1_6226, %int0_6227, %int9223372036854775807_6228, %int1_6229 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %8603, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_6230 = torch.constant.int 0
%8604 = torch.aten.unsqueeze %8603, %int0_6230 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %8604, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_6231 = torch.constant.int 2
%8605 = torch.aten.unsqueeze %8604, %int2_6231 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8605, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_6232 = torch.constant.int 3
%int0_6233 = torch.constant.int 0
%int9223372036854775807_6234 = torch.constant.int 9223372036854775807
%int1_6235 = torch.constant.int 1
%8606 = torch.aten.slice.Tensor %8605, %int3_6232, %int0_6233, %int9223372036854775807_6234, %int1_6235 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8606, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%8607 = torch_c.to_builtin_tensor %8282 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_6236 = arith.constant 1 : index
%dim_6237 = tensor.dim %8607, %c1_6236 : tensor<4x?x1x128xf16>
%8608 = flow.tensor.bitcast %8607 : tensor<4x?x1x128xf16>{%dim_6237} -> tensor<4x?x1x64xcomplex<f16>>{%dim_6237}
%8609 = torch_c.from_builtin_tensor %8608 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %8609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%8610 = torch.aten.mul.Tensor %8609, %8606 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %8610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%8611 = torch_c.to_builtin_tensor %8610 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_6238 = arith.constant 1 : index
%dim_6239 = tensor.dim %8611, %c1_6238 : tensor<4x?x1x64xcomplex<f32>>
%8612 = flow.tensor.bitcast %8611 : tensor<4x?x1x64xcomplex<f32>>{%dim_6239} -> tensor<4x?x1x128xf32>{%dim_6239}
%8613 = torch_c.from_builtin_tensor %8612 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %8613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_6240 = torch.constant.int 5
%8614 = torch.prims.convert_element_type %8613, %int5_6240 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %8614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_6241 = torch.constant.int 64
%8615 = torch.aten.mul.Scalar %2364, %int64_6241 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8615, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6242 = torch.constant.int 64
%8616 = torch.aten.mul.Scalar %2367, %int64_6242 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8616, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6243 = torch.constant.int 64
%8617 = torch.aten.mul.Scalar %2370, %int64_6243 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8617, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6244 = torch.constant.int 64
%8618 = torch.aten.mul.Scalar %2373, %int64_6244 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8618, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6245 = torch.constant.int 64
%8619 = torch.aten.mul.Scalar %2376, %int64_6245 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8619, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6246 = torch.constant.int 64
%8620 = torch.aten.mul.Scalar %2379, %int64_6246 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8620, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6247 = torch.constant.int 64
%8621 = torch.aten.mul.Scalar %2382, %int64_6247 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8621, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_6248 = torch.constant.int 64
%8622 = torch.aten.mul.Scalar %2385, %int64_6248 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8622, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6249 = torch.constant.int 6
%int1_6250 = torch.constant.int 1
%8623 = torch.aten.add.Scalar %8615, %int6_6249, %int1_6250 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8623, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6251 = torch.constant.int 6
%int1_6252 = torch.constant.int 1
%8624 = torch.aten.add.Scalar %8616, %int6_6251, %int1_6252 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8624, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6253 = torch.constant.int 6
%int1_6254 = torch.constant.int 1
%8625 = torch.aten.add.Scalar %8617, %int6_6253, %int1_6254 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8625, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6255 = torch.constant.int 6
%int1_6256 = torch.constant.int 1
%8626 = torch.aten.add.Scalar %8618, %int6_6255, %int1_6256 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8626, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6257 = torch.constant.int 6
%int1_6258 = torch.constant.int 1
%8627 = torch.aten.add.Scalar %8619, %int6_6257, %int1_6258 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8627, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6259 = torch.constant.int 6
%int1_6260 = torch.constant.int 1
%8628 = torch.aten.add.Scalar %8620, %int6_6259, %int1_6260 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8628, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6261 = torch.constant.int 6
%int1_6262 = torch.constant.int 1
%8629 = torch.aten.add.Scalar %8621, %int6_6261, %int1_6262 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8629, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int6_6263 = torch.constant.int 6
%int1_6264 = torch.constant.int 1
%8630 = torch.aten.add.Scalar %8622, %int6_6263, %int1_6264 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8630, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_6265 = torch.constant.int 4
%int16_6266 = torch.constant.int 16
%int1_6267 = torch.constant.int 1
%int128_6268 = torch.constant.int 128
%8631 = torch.prim.ListConstruct %int4_6265, %3095, %int16_6266, %int1_6267, %int128_6268 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8632 = torch.aten.view %8509, %8631 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8632, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6269 = torch.constant.int 4
%int16_6270 = torch.constant.int 16
%int1_6271 = torch.constant.int 1
%int128_6272 = torch.constant.int 128
%8633 = torch.prim.ListConstruct %int4_6269, %3095, %int16_6270, %int1_6271, %int128_6272 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8634 = torch.aten.view %8524, %8633 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8634, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6273 = torch.constant.int 4
%int16_6274 = torch.constant.int 16
%int1_6275 = torch.constant.int 1
%int128_6276 = torch.constant.int 128
%8635 = torch.prim.ListConstruct %int4_6273, %3095, %int16_6274, %int1_6275, %int128_6276 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8636 = torch.aten.view %8539, %8635 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8636, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6277 = torch.constant.int 4
%int16_6278 = torch.constant.int 16
%int1_6279 = torch.constant.int 1
%int128_6280 = torch.constant.int 128
%8637 = torch.prim.ListConstruct %int4_6277, %3095, %int16_6278, %int1_6279, %int128_6280 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8638 = torch.aten.view %8554, %8637 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8638, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6281 = torch.constant.int 4
%int16_6282 = torch.constant.int 16
%int1_6283 = torch.constant.int 1
%int128_6284 = torch.constant.int 128
%8639 = torch.prim.ListConstruct %int4_6281, %3095, %int16_6282, %int1_6283, %int128_6284 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8640 = torch.aten.view %8569, %8639 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8640, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6285 = torch.constant.int 4
%int16_6286 = torch.constant.int 16
%int1_6287 = torch.constant.int 1
%int128_6288 = torch.constant.int 128
%8641 = torch.prim.ListConstruct %int4_6285, %3095, %int16_6286, %int1_6287, %int128_6288 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8642 = torch.aten.view %8584, %8641 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8642, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6289 = torch.constant.int 4
%int16_6290 = torch.constant.int 16
%int1_6291 = torch.constant.int 1
%int128_6292 = torch.constant.int 128
%8643 = torch.prim.ListConstruct %int4_6289, %3095, %int16_6290, %int1_6291, %int128_6292 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8644 = torch.aten.view %8599, %8643 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8644, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6293 = torch.constant.int 4
%int16_6294 = torch.constant.int 16
%int1_6295 = torch.constant.int 1
%int128_6296 = torch.constant.int 128
%8645 = torch.prim.ListConstruct %int4_6293, %3095, %int16_6294, %int1_6295, %int128_6296 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8646 = torch.aten.view %8614, %8645 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8646, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6297 = torch.constant.int 4
%8647 = torch.aten.mul.int %int4_6297, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6298 = torch.constant.int 16
%int1_6299 = torch.constant.int 1
%int128_6300 = torch.constant.int 128
%8648 = torch.prim.ListConstruct %8647, %int16_6298, %int1_6299, %int128_6300 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8649 = torch.aten.view %8632, %8648 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8649, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6301 = torch.constant.int 4
%8650 = torch.aten.mul.int %int4_6301, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6302 = torch.constant.int 16
%int1_6303 = torch.constant.int 1
%int128_6304 = torch.constant.int 128
%8651 = torch.prim.ListConstruct %8650, %int16_6302, %int1_6303, %int128_6304 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8652 = torch.aten.view %8634, %8651 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8652, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6305 = torch.constant.int 4
%8653 = torch.aten.mul.int %int4_6305, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6306 = torch.constant.int 16
%int1_6307 = torch.constant.int 1
%int128_6308 = torch.constant.int 128
%8654 = torch.prim.ListConstruct %8653, %int16_6306, %int1_6307, %int128_6308 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8655 = torch.aten.view %8636, %8654 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8655, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6309 = torch.constant.int 4
%8656 = torch.aten.mul.int %int4_6309, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6310 = torch.constant.int 16
%int1_6311 = torch.constant.int 1
%int128_6312 = torch.constant.int 128
%8657 = torch.prim.ListConstruct %8656, %int16_6310, %int1_6311, %int128_6312 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8658 = torch.aten.view %8638, %8657 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8658, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6313 = torch.constant.int 4
%8659 = torch.aten.mul.int %int4_6313, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6314 = torch.constant.int 16
%int1_6315 = torch.constant.int 1
%int128_6316 = torch.constant.int 128
%8660 = torch.prim.ListConstruct %8659, %int16_6314, %int1_6315, %int128_6316 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8661 = torch.aten.view %8640, %8660 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8661, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6317 = torch.constant.int 4
%8662 = torch.aten.mul.int %int4_6317, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6318 = torch.constant.int 16
%int1_6319 = torch.constant.int 1
%int128_6320 = torch.constant.int 128
%8663 = torch.prim.ListConstruct %8662, %int16_6318, %int1_6319, %int128_6320 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8664 = torch.aten.view %8642, %8663 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8664, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6321 = torch.constant.int 4
%8665 = torch.aten.mul.int %int4_6321, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6322 = torch.constant.int 16
%int1_6323 = torch.constant.int 1
%int128_6324 = torch.constant.int 128
%8666 = torch.prim.ListConstruct %8665, %int16_6322, %int1_6323, %int128_6324 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8667 = torch.aten.view %8644, %8666 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8667, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6325 = torch.constant.int 4
%8668 = torch.aten.mul.int %int4_6325, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6326 = torch.constant.int 16
%int1_6327 = torch.constant.int 1
%int128_6328 = torch.constant.int 128
%8669 = torch.prim.ListConstruct %8668, %int16_6326, %int1_6327, %int128_6328 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8670 = torch.aten.view %8646, %8669 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8670, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6329 = torch.constant.int 4
%8671 = torch.aten.mul.int %int4_6329, %3095 : !torch.int, !torch.int -> !torch.int
%8672 = torch.prim.ListConstruct %8671 : (!torch.int) -> !torch.list<int>
%8673 = torch.aten.view %8623, %8672 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8673, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6330 = torch.constant.int 4
%8674 = torch.aten.mul.int %int4_6330, %3095 : !torch.int, !torch.int -> !torch.int
%8675 = torch.prim.ListConstruct %8674 : (!torch.int) -> !torch.list<int>
%8676 = torch.aten.view %8624, %8675 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8676, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6331 = torch.constant.int 4
%8677 = torch.aten.mul.int %int4_6331, %3095 : !torch.int, !torch.int -> !torch.int
%8678 = torch.prim.ListConstruct %8677 : (!torch.int) -> !torch.list<int>
%8679 = torch.aten.view %8625, %8678 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8679, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6332 = torch.constant.int 4
%8680 = torch.aten.mul.int %int4_6332, %3095 : !torch.int, !torch.int -> !torch.int
%8681 = torch.prim.ListConstruct %8680 : (!torch.int) -> !torch.list<int>
%8682 = torch.aten.view %8626, %8681 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8682, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6333 = torch.constant.int 4
%8683 = torch.aten.mul.int %int4_6333, %3095 : !torch.int, !torch.int -> !torch.int
%8684 = torch.prim.ListConstruct %8683 : (!torch.int) -> !torch.list<int>
%8685 = torch.aten.view %8627, %8684 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8685, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6334 = torch.constant.int 4
%8686 = torch.aten.mul.int %int4_6334, %3095 : !torch.int, !torch.int -> !torch.int
%8687 = torch.prim.ListConstruct %8686 : (!torch.int) -> !torch.list<int>
%8688 = torch.aten.view %8628, %8687 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8688, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6335 = torch.constant.int 4
%8689 = torch.aten.mul.int %int4_6335, %3095 : !torch.int, !torch.int -> !torch.int
%8690 = torch.prim.ListConstruct %8689 : (!torch.int) -> !torch.list<int>
%8691 = torch.aten.view %8629, %8690 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8691, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6336 = torch.constant.int 4
%8692 = torch.aten.mul.int %int4_6336, %3095 : !torch.int, !torch.int -> !torch.int
%8693 = torch.prim.ListConstruct %8692 : (!torch.int) -> !torch.list<int>
%8694 = torch.aten.view %8630, %8693 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8694, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6337 = torch.constant.int 4
%int16_6338 = torch.constant.int 16
%int1_6339 = torch.constant.int 1
%int128_6340 = torch.constant.int 128
%8695 = torch.prim.ListConstruct %int4_6337, %3095, %int16_6338, %int1_6339, %int128_6340 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8696 = torch.aten.view %8284, %8695 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8696, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6341 = torch.constant.int 4
%int16_6342 = torch.constant.int 16
%int1_6343 = torch.constant.int 1
%int128_6344 = torch.constant.int 128
%8697 = torch.prim.ListConstruct %int4_6341, %3095, %int16_6342, %int1_6343, %int128_6344 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8698 = torch.aten.view %8286, %8697 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8698, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6345 = torch.constant.int 4
%int16_6346 = torch.constant.int 16
%int1_6347 = torch.constant.int 1
%int128_6348 = torch.constant.int 128
%8699 = torch.prim.ListConstruct %int4_6345, %3095, %int16_6346, %int1_6347, %int128_6348 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8700 = torch.aten.view %8288, %8699 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8700, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6349 = torch.constant.int 4
%int16_6350 = torch.constant.int 16
%int1_6351 = torch.constant.int 1
%int128_6352 = torch.constant.int 128
%8701 = torch.prim.ListConstruct %int4_6349, %3095, %int16_6350, %int1_6351, %int128_6352 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8702 = torch.aten.view %8290, %8701 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8702, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6353 = torch.constant.int 4
%int16_6354 = torch.constant.int 16
%int1_6355 = torch.constant.int 1
%int128_6356 = torch.constant.int 128
%8703 = torch.prim.ListConstruct %int4_6353, %3095, %int16_6354, %int1_6355, %int128_6356 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8704 = torch.aten.view %8292, %8703 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8704, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6357 = torch.constant.int 4
%int16_6358 = torch.constant.int 16
%int1_6359 = torch.constant.int 1
%int128_6360 = torch.constant.int 128
%8705 = torch.prim.ListConstruct %int4_6357, %3095, %int16_6358, %int1_6359, %int128_6360 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8706 = torch.aten.view %8294, %8705 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8706, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6361 = torch.constant.int 4
%int16_6362 = torch.constant.int 16
%int1_6363 = torch.constant.int 1
%int128_6364 = torch.constant.int 128
%8707 = torch.prim.ListConstruct %int4_6361, %3095, %int16_6362, %int1_6363, %int128_6364 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8708 = torch.aten.view %8296, %8707 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8708, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6365 = torch.constant.int 4
%int16_6366 = torch.constant.int 16
%int1_6367 = torch.constant.int 1
%int128_6368 = torch.constant.int 128
%8709 = torch.prim.ListConstruct %int4_6365, %3095, %int16_6366, %int1_6367, %int128_6368 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8710 = torch.aten.view %8298, %8709 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %8710, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_6369 = torch.constant.int 4
%8711 = torch.aten.mul.int %int4_6369, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6370 = torch.constant.int 16
%int1_6371 = torch.constant.int 1
%int128_6372 = torch.constant.int 128
%8712 = torch.prim.ListConstruct %8711, %int16_6370, %int1_6371, %int128_6372 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8713 = torch.aten.view %8696, %8712 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8713, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6373 = torch.constant.int 4
%8714 = torch.aten.mul.int %int4_6373, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6374 = torch.constant.int 16
%int1_6375 = torch.constant.int 1
%int128_6376 = torch.constant.int 128
%8715 = torch.prim.ListConstruct %8714, %int16_6374, %int1_6375, %int128_6376 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8716 = torch.aten.view %8698, %8715 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8716, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6377 = torch.constant.int 4
%8717 = torch.aten.mul.int %int4_6377, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6378 = torch.constant.int 16
%int1_6379 = torch.constant.int 1
%int128_6380 = torch.constant.int 128
%8718 = torch.prim.ListConstruct %8717, %int16_6378, %int1_6379, %int128_6380 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8719 = torch.aten.view %8700, %8718 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8719, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6381 = torch.constant.int 4
%8720 = torch.aten.mul.int %int4_6381, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6382 = torch.constant.int 16
%int1_6383 = torch.constant.int 1
%int128_6384 = torch.constant.int 128
%8721 = torch.prim.ListConstruct %8720, %int16_6382, %int1_6383, %int128_6384 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8722 = torch.aten.view %8702, %8721 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8722, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6385 = torch.constant.int 4
%8723 = torch.aten.mul.int %int4_6385, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6386 = torch.constant.int 16
%int1_6387 = torch.constant.int 1
%int128_6388 = torch.constant.int 128
%8724 = torch.prim.ListConstruct %8723, %int16_6386, %int1_6387, %int128_6388 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8725 = torch.aten.view %8704, %8724 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8725, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6389 = torch.constant.int 4
%8726 = torch.aten.mul.int %int4_6389, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6390 = torch.constant.int 16
%int1_6391 = torch.constant.int 1
%int128_6392 = torch.constant.int 128
%8727 = torch.prim.ListConstruct %8726, %int16_6390, %int1_6391, %int128_6392 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8728 = torch.aten.view %8706, %8727 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8728, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6393 = torch.constant.int 4
%8729 = torch.aten.mul.int %int4_6393, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6394 = torch.constant.int 16
%int1_6395 = torch.constant.int 1
%int128_6396 = torch.constant.int 128
%8730 = torch.prim.ListConstruct %8729, %int16_6394, %int1_6395, %int128_6396 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8731 = torch.aten.view %8708, %8730 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8731, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_6397 = torch.constant.int 4
%8732 = torch.aten.mul.int %int4_6397, %3095 : !torch.int, !torch.int -> !torch.int
%int16_6398 = torch.constant.int 16
%int1_6399 = torch.constant.int 1
%int128_6400 = torch.constant.int 128
%8733 = torch.prim.ListConstruct %8732, %int16_6398, %int1_6399, %int128_6400 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8734 = torch.aten.view %8710, %8733 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8734, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_6401 = torch.constant.int 1
%int1_6402 = torch.constant.int 1
%8735 = torch.aten.add.Scalar %8623, %int1_6401, %int1_6402 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8735, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6403 = torch.constant.int 1
%int1_6404 = torch.constant.int 1
%8736 = torch.aten.add.Scalar %8624, %int1_6403, %int1_6404 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8736, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6405 = torch.constant.int 1
%int1_6406 = torch.constant.int 1
%8737 = torch.aten.add.Scalar %8625, %int1_6405, %int1_6406 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8737, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6407 = torch.constant.int 1
%int1_6408 = torch.constant.int 1
%8738 = torch.aten.add.Scalar %8626, %int1_6407, %int1_6408 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8738, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6409 = torch.constant.int 1
%int1_6410 = torch.constant.int 1
%8739 = torch.aten.add.Scalar %8627, %int1_6409, %int1_6410 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8739, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6411 = torch.constant.int 1
%int1_6412 = torch.constant.int 1
%8740 = torch.aten.add.Scalar %8628, %int1_6411, %int1_6412 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8740, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6413 = torch.constant.int 1
%int1_6414 = torch.constant.int 1
%8741 = torch.aten.add.Scalar %8629, %int1_6413, %int1_6414 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8741, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_6415 = torch.constant.int 1
%int1_6416 = torch.constant.int 1
%8742 = torch.aten.add.Scalar %8630, %int1_6415, %int1_6416 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %8742, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_6417 = torch.constant.int 4
%8743 = torch.aten.mul.int %int4_6417, %3095 : !torch.int, !torch.int -> !torch.int
%8744 = torch.prim.ListConstruct %8743 : (!torch.int) -> !torch.list<int>
%8745 = torch.aten.view %8735, %8744 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8745, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6418 = torch.constant.int 4
%8746 = torch.aten.mul.int %int4_6418, %3095 : !torch.int, !torch.int -> !torch.int
%8747 = torch.prim.ListConstruct %8746 : (!torch.int) -> !torch.list<int>
%8748 = torch.aten.view %8736, %8747 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8748, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6419 = torch.constant.int 4
%8749 = torch.aten.mul.int %int4_6419, %3095 : !torch.int, !torch.int -> !torch.int
%8750 = torch.prim.ListConstruct %8749 : (!torch.int) -> !torch.list<int>
%8751 = torch.aten.view %8737, %8750 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8751, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6420 = torch.constant.int 4
%8752 = torch.aten.mul.int %int4_6420, %3095 : !torch.int, !torch.int -> !torch.int
%8753 = torch.prim.ListConstruct %8752 : (!torch.int) -> !torch.list<int>
%8754 = torch.aten.view %8738, %8753 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8754, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6421 = torch.constant.int 4
%8755 = torch.aten.mul.int %int4_6421, %3095 : !torch.int, !torch.int -> !torch.int
%8756 = torch.prim.ListConstruct %8755 : (!torch.int) -> !torch.list<int>
%8757 = torch.aten.view %8739, %8756 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8757, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6422 = torch.constant.int 4
%8758 = torch.aten.mul.int %int4_6422, %3095 : !torch.int, !torch.int -> !torch.int
%8759 = torch.prim.ListConstruct %8758 : (!torch.int) -> !torch.list<int>
%8760 = torch.aten.view %8740, %8759 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8760, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6423 = torch.constant.int 4
%8761 = torch.aten.mul.int %int4_6423, %3095 : !torch.int, !torch.int -> !torch.int
%8762 = torch.prim.ListConstruct %8761 : (!torch.int) -> !torch.list<int>
%8763 = torch.aten.view %8741, %8762 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8763, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_6424 = torch.constant.int 4
%8764 = torch.aten.mul.int %int4_6424, %3095 : !torch.int, !torch.int -> !torch.int
%8765 = torch.prim.ListConstruct %8764 : (!torch.int) -> !torch.list<int>
%8766 = torch.aten.view %8742, %8765 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8766, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%8767 = torch.prim.ListConstruct %8673, %8745 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6425 = torch.constant.int 0
%8768 = torch.aten.cat %8767, %int0_6425 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8768, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8769 = torch.prim.ListConstruct %8676, %8748 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6426 = torch.constant.int 0
%8770 = torch.aten.cat %8769, %int0_6426 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8770, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8771 = torch.prim.ListConstruct %8679, %8751 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6427 = torch.constant.int 0
%8772 = torch.aten.cat %8771, %int0_6427 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8772, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8773 = torch.prim.ListConstruct %8682, %8754 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6428 = torch.constant.int 0
%8774 = torch.aten.cat %8773, %int0_6428 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8774, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8775 = torch.prim.ListConstruct %8685, %8757 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6429 = torch.constant.int 0
%8776 = torch.aten.cat %8775, %int0_6429 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8776, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8777 = torch.prim.ListConstruct %8688, %8760 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6430 = torch.constant.int 0
%8778 = torch.aten.cat %8777, %int0_6430 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8778, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8779 = torch.prim.ListConstruct %8691, %8763 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6431 = torch.constant.int 0
%8780 = torch.aten.cat %8779, %int0_6431 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8780, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8781 = torch.prim.ListConstruct %8694, %8766 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_6432 = torch.constant.int 0
%8782 = torch.aten.cat %8781, %int0_6432 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %8782, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%8783 = torch.prim.ListConstruct %8649, %8713 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6433 = torch.constant.int 0
%8784 = torch.aten.cat %8783, %int0_6433 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8784, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8785 = torch.prim.ListConstruct %8652, %8716 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6434 = torch.constant.int 0
%8786 = torch.aten.cat %8785, %int0_6434 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8786, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8787 = torch.prim.ListConstruct %8655, %8719 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6435 = torch.constant.int 0
%8788 = torch.aten.cat %8787, %int0_6435 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8788, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8789 = torch.prim.ListConstruct %8658, %8722 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6436 = torch.constant.int 0
%8790 = torch.aten.cat %8789, %int0_6436 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8790, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8791 = torch.prim.ListConstruct %8661, %8725 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6437 = torch.constant.int 0
%8792 = torch.aten.cat %8791, %int0_6437 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8792, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8793 = torch.prim.ListConstruct %8664, %8728 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6438 = torch.constant.int 0
%8794 = torch.aten.cat %8793, %int0_6438 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8794, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8795 = torch.prim.ListConstruct %8667, %8731 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6439 = torch.constant.int 0
%8796 = torch.aten.cat %8795, %int0_6439 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8796, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8797 = torch.prim.ListConstruct %8670, %8734 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_6440 = torch.constant.int 0
%8798 = torch.aten.cat %8797, %int0_6440 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8798, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6441 = torch.constant.int 32
%int2_6442 = torch.constant.int 2
%int16_6443 = torch.constant.int 16
%int1_6444 = torch.constant.int 1
%int128_6445 = torch.constant.int 128
%8799 = torch.prim.ListConstruct %3023, %int32_6441, %int2_6442, %int16_6443, %int1_6444, %int128_6445 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8800 = torch.aten.view %6949, %8799 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8800, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6446 = torch.constant.int 32
%8801 = torch.aten.mul.int %3023, %int32_6446 : !torch.int, !torch.int -> !torch.int
%int2_6447 = torch.constant.int 2
%8802 = torch.aten.mul.int %8801, %int2_6447 : !torch.int, !torch.int -> !torch.int
%int16_6448 = torch.constant.int 16
%int1_6449 = torch.constant.int 1
%int128_6450 = torch.constant.int 128
%8803 = torch.prim.ListConstruct %8802, %int16_6448, %int1_6449, %int128_6450 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8804 = torch.aten.view %8800, %8803 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8804, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8805 = torch.prim.ListConstruct %8768 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6451 = torch.constant.bool false
%8806 = torch.aten.index_put %8804, %8805, %8784, %false_6451 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8806, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6452 = torch.constant.int 32
%int2_6453 = torch.constant.int 2
%int16_6454 = torch.constant.int 16
%int1_6455 = torch.constant.int 1
%int128_6456 = torch.constant.int 128
%8807 = torch.prim.ListConstruct %3023, %int32_6452, %int2_6453, %int16_6454, %int1_6455, %int128_6456 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8808 = torch.aten.view %8806, %8807 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8808, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6457 = torch.constant.int 131072
%8809 = torch.prim.ListConstruct %3023, %int131072_6457 : (!torch.int, !torch.int) -> !torch.list<int>
%8810 = torch.aten.view %8808, %8809 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8810, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6458 = torch.constant.int 32
%int2_6459 = torch.constant.int 2
%int16_6460 = torch.constant.int 16
%int1_6461 = torch.constant.int 1
%int128_6462 = torch.constant.int 128
%8811 = torch.prim.ListConstruct %3026, %int32_6458, %int2_6459, %int16_6460, %int1_6461, %int128_6462 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8812 = torch.aten.view %6961, %8811 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8812, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6463 = torch.constant.int 32
%8813 = torch.aten.mul.int %3026, %int32_6463 : !torch.int, !torch.int -> !torch.int
%int2_6464 = torch.constant.int 2
%8814 = torch.aten.mul.int %8813, %int2_6464 : !torch.int, !torch.int -> !torch.int
%int16_6465 = torch.constant.int 16
%int1_6466 = torch.constant.int 1
%int128_6467 = torch.constant.int 128
%8815 = torch.prim.ListConstruct %8814, %int16_6465, %int1_6466, %int128_6467 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8816 = torch.aten.view %8812, %8815 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8816, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8817 = torch.prim.ListConstruct %8770 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6468 = torch.constant.bool false
%8818 = torch.aten.index_put %8816, %8817, %8786, %false_6468 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8818, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6469 = torch.constant.int 32
%int2_6470 = torch.constant.int 2
%int16_6471 = torch.constant.int 16
%int1_6472 = torch.constant.int 1
%int128_6473 = torch.constant.int 128
%8819 = torch.prim.ListConstruct %3026, %int32_6469, %int2_6470, %int16_6471, %int1_6472, %int128_6473 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8820 = torch.aten.view %8818, %8819 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8820, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6474 = torch.constant.int 131072
%8821 = torch.prim.ListConstruct %3026, %int131072_6474 : (!torch.int, !torch.int) -> !torch.list<int>
%8822 = torch.aten.view %8820, %8821 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8822, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6475 = torch.constant.int 32
%int2_6476 = torch.constant.int 2
%int16_6477 = torch.constant.int 16
%int1_6478 = torch.constant.int 1
%int128_6479 = torch.constant.int 128
%8823 = torch.prim.ListConstruct %3029, %int32_6475, %int2_6476, %int16_6477, %int1_6478, %int128_6479 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8824 = torch.aten.view %6973, %8823 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8824, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6480 = torch.constant.int 32
%8825 = torch.aten.mul.int %3029, %int32_6480 : !torch.int, !torch.int -> !torch.int
%int2_6481 = torch.constant.int 2
%8826 = torch.aten.mul.int %8825, %int2_6481 : !torch.int, !torch.int -> !torch.int
%int16_6482 = torch.constant.int 16
%int1_6483 = torch.constant.int 1
%int128_6484 = torch.constant.int 128
%8827 = torch.prim.ListConstruct %8826, %int16_6482, %int1_6483, %int128_6484 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8828 = torch.aten.view %8824, %8827 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8828, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8829 = torch.prim.ListConstruct %8772 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6485 = torch.constant.bool false
%8830 = torch.aten.index_put %8828, %8829, %8788, %false_6485 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8830, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6486 = torch.constant.int 32
%int2_6487 = torch.constant.int 2
%int16_6488 = torch.constant.int 16
%int1_6489 = torch.constant.int 1
%int128_6490 = torch.constant.int 128
%8831 = torch.prim.ListConstruct %3029, %int32_6486, %int2_6487, %int16_6488, %int1_6489, %int128_6490 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8832 = torch.aten.view %8830, %8831 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8832, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6491 = torch.constant.int 131072
%8833 = torch.prim.ListConstruct %3029, %int131072_6491 : (!torch.int, !torch.int) -> !torch.list<int>
%8834 = torch.aten.view %8832, %8833 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8834, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6492 = torch.constant.int 32
%int2_6493 = torch.constant.int 2
%int16_6494 = torch.constant.int 16
%int1_6495 = torch.constant.int 1
%int128_6496 = torch.constant.int 128
%8835 = torch.prim.ListConstruct %3032, %int32_6492, %int2_6493, %int16_6494, %int1_6495, %int128_6496 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8836 = torch.aten.view %6985, %8835 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8836, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6497 = torch.constant.int 32
%8837 = torch.aten.mul.int %3032, %int32_6497 : !torch.int, !torch.int -> !torch.int
%int2_6498 = torch.constant.int 2
%8838 = torch.aten.mul.int %8837, %int2_6498 : !torch.int, !torch.int -> !torch.int
%int16_6499 = torch.constant.int 16
%int1_6500 = torch.constant.int 1
%int128_6501 = torch.constant.int 128
%8839 = torch.prim.ListConstruct %8838, %int16_6499, %int1_6500, %int128_6501 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8840 = torch.aten.view %8836, %8839 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8840, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8841 = torch.prim.ListConstruct %8774 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6502 = torch.constant.bool false
%8842 = torch.aten.index_put %8840, %8841, %8790, %false_6502 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8842, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6503 = torch.constant.int 32
%int2_6504 = torch.constant.int 2
%int16_6505 = torch.constant.int 16
%int1_6506 = torch.constant.int 1
%int128_6507 = torch.constant.int 128
%8843 = torch.prim.ListConstruct %3032, %int32_6503, %int2_6504, %int16_6505, %int1_6506, %int128_6507 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8844 = torch.aten.view %8842, %8843 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8844, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6508 = torch.constant.int 131072
%8845 = torch.prim.ListConstruct %3032, %int131072_6508 : (!torch.int, !torch.int) -> !torch.list<int>
%8846 = torch.aten.view %8844, %8845 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8846, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6509 = torch.constant.int 32
%int2_6510 = torch.constant.int 2
%int16_6511 = torch.constant.int 16
%int1_6512 = torch.constant.int 1
%int128_6513 = torch.constant.int 128
%8847 = torch.prim.ListConstruct %3035, %int32_6509, %int2_6510, %int16_6511, %int1_6512, %int128_6513 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8848 = torch.aten.view %6997, %8847 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8848, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6514 = torch.constant.int 32
%8849 = torch.aten.mul.int %3035, %int32_6514 : !torch.int, !torch.int -> !torch.int
%int2_6515 = torch.constant.int 2
%8850 = torch.aten.mul.int %8849, %int2_6515 : !torch.int, !torch.int -> !torch.int
%int16_6516 = torch.constant.int 16
%int1_6517 = torch.constant.int 1
%int128_6518 = torch.constant.int 128
%8851 = torch.prim.ListConstruct %8850, %int16_6516, %int1_6517, %int128_6518 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8852 = torch.aten.view %8848, %8851 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8852, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8853 = torch.prim.ListConstruct %8776 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6519 = torch.constant.bool false
%8854 = torch.aten.index_put %8852, %8853, %8792, %false_6519 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8854, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6520 = torch.constant.int 32
%int2_6521 = torch.constant.int 2
%int16_6522 = torch.constant.int 16
%int1_6523 = torch.constant.int 1
%int128_6524 = torch.constant.int 128
%8855 = torch.prim.ListConstruct %3035, %int32_6520, %int2_6521, %int16_6522, %int1_6523, %int128_6524 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8856 = torch.aten.view %8854, %8855 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8856, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6525 = torch.constant.int 131072
%8857 = torch.prim.ListConstruct %3035, %int131072_6525 : (!torch.int, !torch.int) -> !torch.list<int>
%8858 = torch.aten.view %8856, %8857 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8858, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6526 = torch.constant.int 32
%int2_6527 = torch.constant.int 2
%int16_6528 = torch.constant.int 16
%int1_6529 = torch.constant.int 1
%int128_6530 = torch.constant.int 128
%8859 = torch.prim.ListConstruct %3038, %int32_6526, %int2_6527, %int16_6528, %int1_6529, %int128_6530 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8860 = torch.aten.view %7009, %8859 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8860, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6531 = torch.constant.int 32
%8861 = torch.aten.mul.int %3038, %int32_6531 : !torch.int, !torch.int -> !torch.int
%int2_6532 = torch.constant.int 2
%8862 = torch.aten.mul.int %8861, %int2_6532 : !torch.int, !torch.int -> !torch.int
%int16_6533 = torch.constant.int 16
%int1_6534 = torch.constant.int 1
%int128_6535 = torch.constant.int 128
%8863 = torch.prim.ListConstruct %8862, %int16_6533, %int1_6534, %int128_6535 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8864 = torch.aten.view %8860, %8863 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8864, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8865 = torch.prim.ListConstruct %8778 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6536 = torch.constant.bool false
%8866 = torch.aten.index_put %8864, %8865, %8794, %false_6536 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8866, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6537 = torch.constant.int 32
%int2_6538 = torch.constant.int 2
%int16_6539 = torch.constant.int 16
%int1_6540 = torch.constant.int 1
%int128_6541 = torch.constant.int 128
%8867 = torch.prim.ListConstruct %3038, %int32_6537, %int2_6538, %int16_6539, %int1_6540, %int128_6541 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8868 = torch.aten.view %8866, %8867 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8868, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6542 = torch.constant.int 131072
%8869 = torch.prim.ListConstruct %3038, %int131072_6542 : (!torch.int, !torch.int) -> !torch.list<int>
%8870 = torch.aten.view %8868, %8869 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8870, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6543 = torch.constant.int 32
%int2_6544 = torch.constant.int 2
%int16_6545 = torch.constant.int 16
%int1_6546 = torch.constant.int 1
%int128_6547 = torch.constant.int 128
%8871 = torch.prim.ListConstruct %3041, %int32_6543, %int2_6544, %int16_6545, %int1_6546, %int128_6547 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8872 = torch.aten.view %7021, %8871 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8872, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6548 = torch.constant.int 32
%8873 = torch.aten.mul.int %3041, %int32_6548 : !torch.int, !torch.int -> !torch.int
%int2_6549 = torch.constant.int 2
%8874 = torch.aten.mul.int %8873, %int2_6549 : !torch.int, !torch.int -> !torch.int
%int16_6550 = torch.constant.int 16
%int1_6551 = torch.constant.int 1
%int128_6552 = torch.constant.int 128
%8875 = torch.prim.ListConstruct %8874, %int16_6550, %int1_6551, %int128_6552 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8876 = torch.aten.view %8872, %8875 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8876, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8877 = torch.prim.ListConstruct %8780 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6553 = torch.constant.bool false
%8878 = torch.aten.index_put %8876, %8877, %8796, %false_6553 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8878, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6554 = torch.constant.int 32
%int2_6555 = torch.constant.int 2
%int16_6556 = torch.constant.int 16
%int1_6557 = torch.constant.int 1
%int128_6558 = torch.constant.int 128
%8879 = torch.prim.ListConstruct %3041, %int32_6554, %int2_6555, %int16_6556, %int1_6557, %int128_6558 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8880 = torch.aten.view %8878, %8879 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8880, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6559 = torch.constant.int 131072
%8881 = torch.prim.ListConstruct %3041, %int131072_6559 : (!torch.int, !torch.int) -> !torch.list<int>
%8882 = torch.aten.view %8880, %8881 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8882, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_6560 = torch.constant.int 32
%int2_6561 = torch.constant.int 2
%int16_6562 = torch.constant.int 16
%int1_6563 = torch.constant.int 1
%int128_6564 = torch.constant.int 128
%8883 = torch.prim.ListConstruct %3044, %int32_6560, %int2_6561, %int16_6562, %int1_6563, %int128_6564 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8884 = torch.aten.view %7033, %8883 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8884, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_6565 = torch.constant.int 32
%8885 = torch.aten.mul.int %3044, %int32_6565 : !torch.int, !torch.int -> !torch.int
%int2_6566 = torch.constant.int 2
%8886 = torch.aten.mul.int %8885, %int2_6566 : !torch.int, !torch.int -> !torch.int
%int16_6567 = torch.constant.int 16
%int1_6568 = torch.constant.int 1
%int128_6569 = torch.constant.int 128
%8887 = torch.prim.ListConstruct %8886, %int16_6567, %int1_6568, %int128_6569 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8888 = torch.aten.view %8884, %8887 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8888, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%8889 = torch.prim.ListConstruct %8782 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_6570 = torch.constant.bool false
%8890 = torch.aten.index_put %8888, %8889, %8798, %false_6570 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %8890, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_6571 = torch.constant.int 32
%int2_6572 = torch.constant.int 2
%int16_6573 = torch.constant.int 16
%int1_6574 = torch.constant.int 1
%int128_6575 = torch.constant.int 128
%8891 = torch.prim.ListConstruct %3044, %int32_6571, %int2_6572, %int16_6573, %int1_6574, %int128_6575 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8892 = torch.aten.view %8890, %8891 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %8892, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_6576 = torch.constant.int 131072
%8893 = torch.prim.ListConstruct %3044, %int131072_6576 : (!torch.int, !torch.int) -> !torch.list<int>
%8894 = torch.aten.view %8892, %8893 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %8894, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_6577 = torch.constant.int -2
%8895 = torch.aten.unsqueeze %8509, %int-2_6577 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6578 = torch.constant.int -2
%8896 = torch.aten.unsqueeze %8524, %int-2_6578 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6579 = torch.constant.int -2
%8897 = torch.aten.unsqueeze %8539, %int-2_6579 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6580 = torch.constant.int -2
%8898 = torch.aten.unsqueeze %8554, %int-2_6580 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6581 = torch.constant.int -2
%8899 = torch.aten.unsqueeze %8569, %int-2_6581 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6582 = torch.constant.int -2
%8900 = torch.aten.unsqueeze %8584, %int-2_6582 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6583 = torch.constant.int -2
%8901 = torch.aten.unsqueeze %8599, %int-2_6583 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6584 = torch.constant.int -2
%8902 = torch.aten.unsqueeze %8614, %int-2_6584 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_6585 = torch.constant.int 4
%int1_6586 = torch.constant.int 1
%int4_6587 = torch.constant.int 4
%int128_6588 = torch.constant.int 128
%8903 = torch.prim.ListConstruct %int4_6585, %8495, %int1_6586, %int4_6587, %int128_6588 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6589 = torch.constant.bool false
%8904 = torch.aten.expand %8895, %8903, %false_6589 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6590 = torch.constant.int 4
%int1_6591 = torch.constant.int 1
%int4_6592 = torch.constant.int 4
%int128_6593 = torch.constant.int 128
%8905 = torch.prim.ListConstruct %int4_6590, %8495, %int1_6591, %int4_6592, %int128_6593 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6594 = torch.constant.bool false
%8906 = torch.aten.expand %8896, %8905, %false_6594 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6595 = torch.constant.int 4
%int1_6596 = torch.constant.int 1
%int4_6597 = torch.constant.int 4
%int128_6598 = torch.constant.int 128
%8907 = torch.prim.ListConstruct %int4_6595, %8495, %int1_6596, %int4_6597, %int128_6598 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6599 = torch.constant.bool false
%8908 = torch.aten.expand %8897, %8907, %false_6599 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6600 = torch.constant.int 4
%int1_6601 = torch.constant.int 1
%int4_6602 = torch.constant.int 4
%int128_6603 = torch.constant.int 128
%8909 = torch.prim.ListConstruct %int4_6600, %8495, %int1_6601, %int4_6602, %int128_6603 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6604 = torch.constant.bool false
%8910 = torch.aten.expand %8898, %8909, %false_6604 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6605 = torch.constant.int 4
%int1_6606 = torch.constant.int 1
%int4_6607 = torch.constant.int 4
%int128_6608 = torch.constant.int 128
%8911 = torch.prim.ListConstruct %int4_6605, %8495, %int1_6606, %int4_6607, %int128_6608 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6609 = torch.constant.bool false
%8912 = torch.aten.expand %8899, %8911, %false_6609 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6610 = torch.constant.int 4
%int1_6611 = torch.constant.int 1
%int4_6612 = torch.constant.int 4
%int128_6613 = torch.constant.int 128
%8913 = torch.prim.ListConstruct %int4_6610, %8495, %int1_6611, %int4_6612, %int128_6613 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6614 = torch.constant.bool false
%8914 = torch.aten.expand %8900, %8913, %false_6614 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6615 = torch.constant.int 4
%int1_6616 = torch.constant.int 1
%int4_6617 = torch.constant.int 4
%int128_6618 = torch.constant.int 128
%8915 = torch.prim.ListConstruct %int4_6615, %8495, %int1_6616, %int4_6617, %int128_6618 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6619 = torch.constant.bool false
%8916 = torch.aten.expand %8901, %8915, %false_6619 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6620 = torch.constant.int 4
%int1_6621 = torch.constant.int 1
%int4_6622 = torch.constant.int 4
%int128_6623 = torch.constant.int 128
%8917 = torch.prim.ListConstruct %int4_6620, %8495, %int1_6621, %int4_6622, %int128_6623 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6624 = torch.constant.bool false
%8918 = torch.aten.expand %8902, %8917, %false_6624 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6625 = torch.constant.int 4
%int4_6626 = torch.constant.int 4
%int128_6627 = torch.constant.int 128
%8919 = torch.prim.ListConstruct %int4_6625, %8495, %int4_6626, %int128_6627 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8920 = torch.aten.view %8904, %8919 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6628 = torch.constant.int 4
%int4_6629 = torch.constant.int 4
%int128_6630 = torch.constant.int 128
%8921 = torch.prim.ListConstruct %int4_6628, %8495, %int4_6629, %int128_6630 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8922 = torch.aten.view %8906, %8921 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6631 = torch.constant.int 4
%int4_6632 = torch.constant.int 4
%int128_6633 = torch.constant.int 128
%8923 = torch.prim.ListConstruct %int4_6631, %8495, %int4_6632, %int128_6633 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8924 = torch.aten.view %8908, %8923 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6634 = torch.constant.int 4
%int4_6635 = torch.constant.int 4
%int128_6636 = torch.constant.int 128
%8925 = torch.prim.ListConstruct %int4_6634, %8495, %int4_6635, %int128_6636 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8926 = torch.aten.view %8910, %8925 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6637 = torch.constant.int 4
%int4_6638 = torch.constant.int 4
%int128_6639 = torch.constant.int 128
%8927 = torch.prim.ListConstruct %int4_6637, %8495, %int4_6638, %int128_6639 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8928 = torch.aten.view %8912, %8927 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6640 = torch.constant.int 4
%int4_6641 = torch.constant.int 4
%int128_6642 = torch.constant.int 128
%8929 = torch.prim.ListConstruct %int4_6640, %8495, %int4_6641, %int128_6642 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8930 = torch.aten.view %8914, %8929 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6643 = torch.constant.int 4
%int4_6644 = torch.constant.int 4
%int128_6645 = torch.constant.int 128
%8931 = torch.prim.ListConstruct %int4_6643, %8495, %int4_6644, %int128_6645 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8932 = torch.aten.view %8916, %8931 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6646 = torch.constant.int 4
%int4_6647 = torch.constant.int 4
%int128_6648 = torch.constant.int 128
%8933 = torch.prim.ListConstruct %int4_6646, %8495, %int4_6647, %int128_6648 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8934 = torch.aten.view %8918, %8933 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_6649 = torch.constant.int -2
%8935 = torch.aten.unsqueeze %8284, %int-2_6649 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6650 = torch.constant.int -2
%8936 = torch.aten.unsqueeze %8286, %int-2_6650 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6651 = torch.constant.int -2
%8937 = torch.aten.unsqueeze %8288, %int-2_6651 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6652 = torch.constant.int -2
%8938 = torch.aten.unsqueeze %8290, %int-2_6652 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6653 = torch.constant.int -2
%8939 = torch.aten.unsqueeze %8292, %int-2_6653 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6654 = torch.constant.int -2
%8940 = torch.aten.unsqueeze %8294, %int-2_6654 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6655 = torch.constant.int -2
%8941 = torch.aten.unsqueeze %8296, %int-2_6655 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_6656 = torch.constant.int -2
%8942 = torch.aten.unsqueeze %8298, %int-2_6656 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %8942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_6657 = torch.constant.int 1
%8943 = torch.aten.size.int %8208, %int1_6657 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_6658 = torch.constant.int 4
%int1_6659 = torch.constant.int 1
%int4_6660 = torch.constant.int 4
%int128_6661 = torch.constant.int 128
%8944 = torch.prim.ListConstruct %int4_6658, %8943, %int1_6659, %int4_6660, %int128_6661 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6662 = torch.constant.bool false
%8945 = torch.aten.expand %8935, %8944, %false_6662 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6663 = torch.constant.int 4
%int1_6664 = torch.constant.int 1
%int4_6665 = torch.constant.int 4
%int128_6666 = torch.constant.int 128
%8946 = torch.prim.ListConstruct %int4_6663, %8943, %int1_6664, %int4_6665, %int128_6666 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6667 = torch.constant.bool false
%8947 = torch.aten.expand %8936, %8946, %false_6667 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6668 = torch.constant.int 4
%int1_6669 = torch.constant.int 1
%int4_6670 = torch.constant.int 4
%int128_6671 = torch.constant.int 128
%8948 = torch.prim.ListConstruct %int4_6668, %8943, %int1_6669, %int4_6670, %int128_6671 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6672 = torch.constant.bool false
%8949 = torch.aten.expand %8937, %8948, %false_6672 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6673 = torch.constant.int 4
%int1_6674 = torch.constant.int 1
%int4_6675 = torch.constant.int 4
%int128_6676 = torch.constant.int 128
%8950 = torch.prim.ListConstruct %int4_6673, %8943, %int1_6674, %int4_6675, %int128_6676 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6677 = torch.constant.bool false
%8951 = torch.aten.expand %8938, %8950, %false_6677 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6678 = torch.constant.int 4
%int1_6679 = torch.constant.int 1
%int4_6680 = torch.constant.int 4
%int128_6681 = torch.constant.int 128
%8952 = torch.prim.ListConstruct %int4_6678, %8943, %int1_6679, %int4_6680, %int128_6681 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6682 = torch.constant.bool false
%8953 = torch.aten.expand %8939, %8952, %false_6682 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6683 = torch.constant.int 4
%int1_6684 = torch.constant.int 1
%int4_6685 = torch.constant.int 4
%int128_6686 = torch.constant.int 128
%8954 = torch.prim.ListConstruct %int4_6683, %8943, %int1_6684, %int4_6685, %int128_6686 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6687 = torch.constant.bool false
%8955 = torch.aten.expand %8940, %8954, %false_6687 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6688 = torch.constant.int 4
%int1_6689 = torch.constant.int 1
%int4_6690 = torch.constant.int 4
%int128_6691 = torch.constant.int 128
%8956 = torch.prim.ListConstruct %int4_6688, %8943, %int1_6689, %int4_6690, %int128_6691 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6692 = torch.constant.bool false
%8957 = torch.aten.expand %8941, %8956, %false_6692 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6693 = torch.constant.int 4
%int1_6694 = torch.constant.int 1
%int4_6695 = torch.constant.int 4
%int128_6696 = torch.constant.int 128
%8958 = torch.prim.ListConstruct %int4_6693, %8943, %int1_6694, %int4_6695, %int128_6696 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_6697 = torch.constant.bool false
%8959 = torch.aten.expand %8942, %8958, %false_6697 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %8959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_6698 = torch.constant.int 4
%int4_6699 = torch.constant.int 4
%int128_6700 = torch.constant.int 128
%8960 = torch.prim.ListConstruct %int4_6698, %8943, %int4_6699, %int128_6700 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8961 = torch.aten.view %8945, %8960 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6701 = torch.constant.int 4
%int4_6702 = torch.constant.int 4
%int128_6703 = torch.constant.int 128
%8962 = torch.prim.ListConstruct %int4_6701, %8943, %int4_6702, %int128_6703 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8963 = torch.aten.view %8947, %8962 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6704 = torch.constant.int 4
%int4_6705 = torch.constant.int 4
%int128_6706 = torch.constant.int 128
%8964 = torch.prim.ListConstruct %int4_6704, %8943, %int4_6705, %int128_6706 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8965 = torch.aten.view %8949, %8964 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6707 = torch.constant.int 4
%int4_6708 = torch.constant.int 4
%int128_6709 = torch.constant.int 128
%8966 = torch.prim.ListConstruct %int4_6707, %8943, %int4_6708, %int128_6709 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8967 = torch.aten.view %8951, %8966 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6710 = torch.constant.int 4
%int4_6711 = torch.constant.int 4
%int128_6712 = torch.constant.int 128
%8968 = torch.prim.ListConstruct %int4_6710, %8943, %int4_6711, %int128_6712 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8969 = torch.aten.view %8953, %8968 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6713 = torch.constant.int 4
%int4_6714 = torch.constant.int 4
%int128_6715 = torch.constant.int 128
%8970 = torch.prim.ListConstruct %int4_6713, %8943, %int4_6714, %int128_6715 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8971 = torch.aten.view %8955, %8970 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6716 = torch.constant.int 4
%int4_6717 = torch.constant.int 4
%int128_6718 = torch.constant.int 128
%8972 = torch.prim.ListConstruct %int4_6716, %8943, %int4_6717, %int128_6718 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8973 = torch.aten.view %8957, %8972 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6719 = torch.constant.int 4
%int4_6720 = torch.constant.int 4
%int128_6721 = torch.constant.int 128
%8974 = torch.prim.ListConstruct %int4_6719, %8943, %int4_6720, %int128_6721 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%8975 = torch.aten.view %8959, %8974 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %8975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6722 = torch.constant.int 1
%int2_6723 = torch.constant.int 2
%8976 = torch.aten.transpose.int %8351, %int1_6722, %int2_6723 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8976, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6724 = torch.constant.int 1
%int2_6725 = torch.constant.int 2
%8977 = torch.aten.transpose.int %8366, %int1_6724, %int2_6725 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8977, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6726 = torch.constant.int 1
%int2_6727 = torch.constant.int 2
%8978 = torch.aten.transpose.int %8381, %int1_6726, %int2_6727 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8978, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6728 = torch.constant.int 1
%int2_6729 = torch.constant.int 2
%8979 = torch.aten.transpose.int %8396, %int1_6728, %int2_6729 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8979, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6730 = torch.constant.int 1
%int2_6731 = torch.constant.int 2
%8980 = torch.aten.transpose.int %8411, %int1_6730, %int2_6731 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8980, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6732 = torch.constant.int 1
%int2_6733 = torch.constant.int 2
%8981 = torch.aten.transpose.int %8426, %int1_6732, %int2_6733 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8981, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6734 = torch.constant.int 1
%int2_6735 = torch.constant.int 2
%8982 = torch.aten.transpose.int %8441, %int1_6734, %int2_6735 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8982, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6736 = torch.constant.int 1
%int2_6737 = torch.constant.int 2
%8983 = torch.aten.transpose.int %8456, %int1_6736, %int2_6737 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8983, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6738 = torch.constant.int 1
%int2_6739 = torch.constant.int 2
%8984 = torch.aten.transpose.int %8920, %int1_6738, %int2_6739 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8984, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6740 = torch.constant.int 1
%int2_6741 = torch.constant.int 2
%8985 = torch.aten.transpose.int %8922, %int1_6740, %int2_6741 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8985, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6742 = torch.constant.int 1
%int2_6743 = torch.constant.int 2
%8986 = torch.aten.transpose.int %8924, %int1_6742, %int2_6743 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8986, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6744 = torch.constant.int 1
%int2_6745 = torch.constant.int 2
%8987 = torch.aten.transpose.int %8926, %int1_6744, %int2_6745 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8987, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6746 = torch.constant.int 1
%int2_6747 = torch.constant.int 2
%8988 = torch.aten.transpose.int %8928, %int1_6746, %int2_6747 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8988, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6748 = torch.constant.int 1
%int2_6749 = torch.constant.int 2
%8989 = torch.aten.transpose.int %8930, %int1_6748, %int2_6749 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8989, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6750 = torch.constant.int 1
%int2_6751 = torch.constant.int 2
%8990 = torch.aten.transpose.int %8932, %int1_6750, %int2_6751 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8990, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6752 = torch.constant.int 1
%int2_6753 = torch.constant.int 2
%8991 = torch.aten.transpose.int %8934, %int1_6752, %int2_6753 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8991, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6754 = torch.constant.int 1
%int2_6755 = torch.constant.int 2
%8992 = torch.aten.transpose.int %8961, %int1_6754, %int2_6755 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8992, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6756 = torch.constant.int 1
%int2_6757 = torch.constant.int 2
%8993 = torch.aten.transpose.int %8963, %int1_6756, %int2_6757 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8993, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6758 = torch.constant.int 1
%int2_6759 = torch.constant.int 2
%8994 = torch.aten.transpose.int %8965, %int1_6758, %int2_6759 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8994, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6760 = torch.constant.int 1
%int2_6761 = torch.constant.int 2
%8995 = torch.aten.transpose.int %8967, %int1_6760, %int2_6761 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8995, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6762 = torch.constant.int 1
%int2_6763 = torch.constant.int 2
%8996 = torch.aten.transpose.int %8969, %int1_6762, %int2_6763 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8996, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6764 = torch.constant.int 1
%int2_6765 = torch.constant.int 2
%8997 = torch.aten.transpose.int %8971, %int1_6764, %int2_6765 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8997, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6766 = torch.constant.int 1
%int2_6767 = torch.constant.int 2
%8998 = torch.aten.transpose.int %8973, %int1_6766, %int2_6767 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8998, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6768 = torch.constant.int 1
%int2_6769 = torch.constant.int 2
%8999 = torch.aten.transpose.int %8975, %int1_6768, %int2_6769 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %8999, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6770 = torch.constant.float 0.000000e+00
%true_6771 = torch.constant.bool true
%none_6772 = torch.constant.none
%none_6773 = torch.constant.none
%9000:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8976, %8984, %8992, %float0.000000e00_6770, %true_6771, %none_6772, %none_6773) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9000#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6774 = torch.constant.float 0.000000e+00
%true_6775 = torch.constant.bool true
%none_6776 = torch.constant.none
%none_6777 = torch.constant.none
%9001:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8977, %8985, %8993, %float0.000000e00_6774, %true_6775, %none_6776, %none_6777) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9001#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6778 = torch.constant.float 0.000000e+00
%true_6779 = torch.constant.bool true
%none_6780 = torch.constant.none
%none_6781 = torch.constant.none
%9002:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8978, %8986, %8994, %float0.000000e00_6778, %true_6779, %none_6780, %none_6781) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9002#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6782 = torch.constant.float 0.000000e+00
%true_6783 = torch.constant.bool true
%none_6784 = torch.constant.none
%none_6785 = torch.constant.none
%9003:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8979, %8987, %8995, %float0.000000e00_6782, %true_6783, %none_6784, %none_6785) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9003#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6786 = torch.constant.float 0.000000e+00
%true_6787 = torch.constant.bool true
%none_6788 = torch.constant.none
%none_6789 = torch.constant.none
%9004:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8980, %8988, %8996, %float0.000000e00_6786, %true_6787, %none_6788, %none_6789) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9004#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6790 = torch.constant.float 0.000000e+00
%true_6791 = torch.constant.bool true
%none_6792 = torch.constant.none
%none_6793 = torch.constant.none
%9005:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8981, %8989, %8997, %float0.000000e00_6790, %true_6791, %none_6792, %none_6793) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9005#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6794 = torch.constant.float 0.000000e+00
%true_6795 = torch.constant.bool true
%none_6796 = torch.constant.none
%none_6797 = torch.constant.none
%9006:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8982, %8990, %8998, %float0.000000e00_6794, %true_6795, %none_6796, %none_6797) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9006#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_6798 = torch.constant.float 0.000000e+00
%true_6799 = torch.constant.bool true
%none_6800 = torch.constant.none
%none_6801 = torch.constant.none
%9007:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%8983, %8991, %8999, %float0.000000e00_6798, %true_6799, %none_6800, %none_6801) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %9007#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_6802 = torch.constant.int 1
%int2_6803 = torch.constant.int 2
%9008 = torch.aten.transpose.int %9000#0, %int1_6802, %int2_6803 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6804 = torch.constant.int 1
%int2_6805 = torch.constant.int 2
%9009 = torch.aten.transpose.int %9001#0, %int1_6804, %int2_6805 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6806 = torch.constant.int 1
%int2_6807 = torch.constant.int 2
%9010 = torch.aten.transpose.int %9002#0, %int1_6806, %int2_6807 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6808 = torch.constant.int 1
%int2_6809 = torch.constant.int 2
%9011 = torch.aten.transpose.int %9003#0, %int1_6808, %int2_6809 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6810 = torch.constant.int 1
%int2_6811 = torch.constant.int 2
%9012 = torch.aten.transpose.int %9004#0, %int1_6810, %int2_6811 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6812 = torch.constant.int 1
%int2_6813 = torch.constant.int 2
%9013 = torch.aten.transpose.int %9005#0, %int1_6812, %int2_6813 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6814 = torch.constant.int 1
%int2_6815 = torch.constant.int 2
%9014 = torch.aten.transpose.int %9006#0, %int1_6814, %int2_6815 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_6816 = torch.constant.int 1
%int2_6817 = torch.constant.int 2
%9015 = torch.aten.transpose.int %9007#0, %int1_6816, %int2_6817 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %9015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_6818 = torch.constant.int 4
%int512_6819 = torch.constant.int 512
%9016 = torch.prim.ListConstruct %int4_6818, %8337, %int512_6819 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9017 = torch.aten.view %9008, %9016 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6820 = torch.constant.int 4
%int512_6821 = torch.constant.int 512
%9018 = torch.prim.ListConstruct %int4_6820, %8352, %int512_6821 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9019 = torch.aten.view %9009, %9018 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6822 = torch.constant.int 4
%int512_6823 = torch.constant.int 512
%9020 = torch.prim.ListConstruct %int4_6822, %8367, %int512_6823 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9021 = torch.aten.view %9010, %9020 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6824 = torch.constant.int 4
%int512_6825 = torch.constant.int 512
%9022 = torch.prim.ListConstruct %int4_6824, %8382, %int512_6825 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9023 = torch.aten.view %9011, %9022 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6826 = torch.constant.int 4
%int512_6827 = torch.constant.int 512
%9024 = torch.prim.ListConstruct %int4_6826, %8397, %int512_6827 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9025 = torch.aten.view %9012, %9024 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6828 = torch.constant.int 4
%int512_6829 = torch.constant.int 512
%9026 = torch.prim.ListConstruct %int4_6828, %8412, %int512_6829 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9027 = torch.aten.view %9013, %9026 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6830 = torch.constant.int 4
%int512_6831 = torch.constant.int 512
%9028 = torch.prim.ListConstruct %int4_6830, %8427, %int512_6831 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9029 = torch.aten.view %9014, %9028 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_6832 = torch.constant.int 4
%int512_6833 = torch.constant.int 512
%9030 = torch.prim.ListConstruct %int4_6832, %8442, %int512_6833 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9031 = torch.aten.view %9015, %9030 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_6834 = torch.constant.int 1
%int0_6835 = torch.constant.int 0
%9032 = torch.prim.ListConstruct %int1_6834, %int0_6835 : (!torch.int, !torch.int) -> !torch.list<int>
%9033 = torch.aten.permute %256, %9032 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6836 = torch.constant.int 1
%int0_6837 = torch.constant.int 0
%9034 = torch.prim.ListConstruct %int1_6836, %int0_6837 : (!torch.int, !torch.int) -> !torch.list<int>
%9035 = torch.aten.permute %257, %9034 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6838 = torch.constant.int 1
%int0_6839 = torch.constant.int 0
%9036 = torch.prim.ListConstruct %int1_6838, %int0_6839 : (!torch.int, !torch.int) -> !torch.list<int>
%9037 = torch.aten.permute %258, %9036 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6840 = torch.constant.int 1
%int0_6841 = torch.constant.int 0
%9038 = torch.prim.ListConstruct %int1_6840, %int0_6841 : (!torch.int, !torch.int) -> !torch.list<int>
%9039 = torch.aten.permute %259, %9038 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6842 = torch.constant.int 1
%int0_6843 = torch.constant.int 0
%9040 = torch.prim.ListConstruct %int1_6842, %int0_6843 : (!torch.int, !torch.int) -> !torch.list<int>
%9041 = torch.aten.permute %260, %9040 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6844 = torch.constant.int 1
%int0_6845 = torch.constant.int 0
%9042 = torch.prim.ListConstruct %int1_6844, %int0_6845 : (!torch.int, !torch.int) -> !torch.list<int>
%9043 = torch.aten.permute %261, %9042 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6846 = torch.constant.int 1
%int0_6847 = torch.constant.int 0
%9044 = torch.prim.ListConstruct %int1_6846, %int0_6847 : (!torch.int, !torch.int) -> !torch.list<int>
%9045 = torch.aten.permute %262, %9044 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_6848 = torch.constant.int 1
%int0_6849 = torch.constant.int 0
%9046 = torch.prim.ListConstruct %int1_6848, %int0_6849 : (!torch.int, !torch.int) -> !torch.list<int>
%9047 = torch.aten.permute %263, %9046 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_6850 = torch.constant.int 4
%9048 = torch.aten.mul.int %int4_6850, %8337 : !torch.int, !torch.int -> !torch.int
%int512_6851 = torch.constant.int 512
%9049 = torch.prim.ListConstruct %9048, %int512_6851 : (!torch.int, !torch.int) -> !torch.list<int>
%9050 = torch.aten.view %9017, %9049 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9050, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9051 = torch.aten.mm %9050, %9033 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9051, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6852 = torch.constant.int 4
%int4096_6853 = torch.constant.int 4096
%9052 = torch.prim.ListConstruct %int4_6852, %8337, %int4096_6853 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9053 = torch.aten.view %9051, %9052 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6854 = torch.constant.int 4
%9054 = torch.aten.mul.int %int4_6854, %8352 : !torch.int, !torch.int -> !torch.int
%int512_6855 = torch.constant.int 512
%9055 = torch.prim.ListConstruct %9054, %int512_6855 : (!torch.int, !torch.int) -> !torch.list<int>
%9056 = torch.aten.view %9019, %9055 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9056, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9057 = torch.aten.mm %9056, %9035 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9057, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6856 = torch.constant.int 4
%int4096_6857 = torch.constant.int 4096
%9058 = torch.prim.ListConstruct %int4_6856, %8352, %int4096_6857 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9059 = torch.aten.view %9057, %9058 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6858 = torch.constant.int 4
%9060 = torch.aten.mul.int %int4_6858, %8367 : !torch.int, !torch.int -> !torch.int
%int512_6859 = torch.constant.int 512
%9061 = torch.prim.ListConstruct %9060, %int512_6859 : (!torch.int, !torch.int) -> !torch.list<int>
%9062 = torch.aten.view %9021, %9061 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9062, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9063 = torch.aten.mm %9062, %9037 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9063, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6860 = torch.constant.int 4
%int4096_6861 = torch.constant.int 4096
%9064 = torch.prim.ListConstruct %int4_6860, %8367, %int4096_6861 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9065 = torch.aten.view %9063, %9064 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6862 = torch.constant.int 4
%9066 = torch.aten.mul.int %int4_6862, %8382 : !torch.int, !torch.int -> !torch.int
%int512_6863 = torch.constant.int 512
%9067 = torch.prim.ListConstruct %9066, %int512_6863 : (!torch.int, !torch.int) -> !torch.list<int>
%9068 = torch.aten.view %9023, %9067 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9068, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9069 = torch.aten.mm %9068, %9039 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9069, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6864 = torch.constant.int 4
%int4096_6865 = torch.constant.int 4096
%9070 = torch.prim.ListConstruct %int4_6864, %8382, %int4096_6865 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9071 = torch.aten.view %9069, %9070 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6866 = torch.constant.int 4
%9072 = torch.aten.mul.int %int4_6866, %8397 : !torch.int, !torch.int -> !torch.int
%int512_6867 = torch.constant.int 512
%9073 = torch.prim.ListConstruct %9072, %int512_6867 : (!torch.int, !torch.int) -> !torch.list<int>
%9074 = torch.aten.view %9025, %9073 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9074, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9075 = torch.aten.mm %9074, %9041 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9075, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6868 = torch.constant.int 4
%int4096_6869 = torch.constant.int 4096
%9076 = torch.prim.ListConstruct %int4_6868, %8397, %int4096_6869 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9077 = torch.aten.view %9075, %9076 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6870 = torch.constant.int 4
%9078 = torch.aten.mul.int %int4_6870, %8412 : !torch.int, !torch.int -> !torch.int
%int512_6871 = torch.constant.int 512
%9079 = torch.prim.ListConstruct %9078, %int512_6871 : (!torch.int, !torch.int) -> !torch.list<int>
%9080 = torch.aten.view %9027, %9079 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9080, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9081 = torch.aten.mm %9080, %9043 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9081, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6872 = torch.constant.int 4
%int4096_6873 = torch.constant.int 4096
%9082 = torch.prim.ListConstruct %int4_6872, %8412, %int4096_6873 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9083 = torch.aten.view %9081, %9082 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6874 = torch.constant.int 4
%9084 = torch.aten.mul.int %int4_6874, %8427 : !torch.int, !torch.int -> !torch.int
%int512_6875 = torch.constant.int 512
%9085 = torch.prim.ListConstruct %9084, %int512_6875 : (!torch.int, !torch.int) -> !torch.list<int>
%9086 = torch.aten.view %9029, %9085 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9086, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9087 = torch.aten.mm %9086, %9045 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9087, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6876 = torch.constant.int 4
%int4096_6877 = torch.constant.int 4096
%9088 = torch.prim.ListConstruct %int4_6876, %8427, %int4096_6877 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9089 = torch.aten.view %9087, %9088 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_6878 = torch.constant.int 4
%9090 = torch.aten.mul.int %int4_6878, %8442 : !torch.int, !torch.int -> !torch.int
%int512_6879 = torch.constant.int 512
%9091 = torch.prim.ListConstruct %9090, %int512_6879 : (!torch.int, !torch.int) -> !torch.list<int>
%9092 = torch.aten.view %9031, %9091 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9092, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%9093 = torch.aten.mm %9092, %9047 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9093, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_6880 = torch.constant.int 4
%int4096_6881 = torch.constant.int 4096
%9094 = torch.prim.ListConstruct %int4_6880, %8442, %int4096_6881 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9095 = torch.aten.view %9093, %9094 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9096 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6882 = arith.constant 1 : index
%dim_6883 = tensor.dim %9096, %c1_6882 : tensor<4x?x4096xf16>
%9097 = flow.tensor.transfer %9096 : tensor<4x?x4096xf16>{%dim_6883} to #hal.device.promise<@__device_0>
%9098 = torch_c.from_builtin_tensor %9097 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9099 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6884 = arith.constant 1 : index
%dim_6885 = tensor.dim %9099, %c1_6884 : tensor<4x?x4096xf16>
%9100 = flow.tensor.transfer %9099 : tensor<4x?x4096xf16>{%dim_6885} to #hal.device.promise<@__device_0>
%9101 = torch_c.from_builtin_tensor %9100 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9102 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6886 = arith.constant 1 : index
%dim_6887 = tensor.dim %9102, %c1_6886 : tensor<4x?x4096xf16>
%9103 = flow.tensor.transfer %9102 : tensor<4x?x4096xf16>{%dim_6887} to #hal.device.promise<@__device_0>
%9104 = torch_c.from_builtin_tensor %9103 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9105 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6888 = arith.constant 1 : index
%dim_6889 = tensor.dim %9105, %c1_6888 : tensor<4x?x4096xf16>
%9106 = flow.tensor.transfer %9105 : tensor<4x?x4096xf16>{%dim_6889} to #hal.device.promise<@__device_0>
%9107 = torch_c.from_builtin_tensor %9106 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9108 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6890 = arith.constant 1 : index
%dim_6891 = tensor.dim %9108, %c1_6890 : tensor<4x?x4096xf16>
%9109 = flow.tensor.transfer %9108 : tensor<4x?x4096xf16>{%dim_6891} to #hal.device.promise<@__device_0>
%9110 = torch_c.from_builtin_tensor %9109 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9111 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6892 = arith.constant 1 : index
%dim_6893 = tensor.dim %9111, %c1_6892 : tensor<4x?x4096xf16>
%9112 = flow.tensor.transfer %9111 : tensor<4x?x4096xf16>{%dim_6893} to #hal.device.promise<@__device_0>
%9113 = torch_c.from_builtin_tensor %9112 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9114 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6894 = arith.constant 1 : index
%dim_6895 = tensor.dim %9114, %c1_6894 : tensor<4x?x4096xf16>
%9115 = flow.tensor.transfer %9114 : tensor<4x?x4096xf16>{%dim_6895} to #hal.device.promise<@__device_0>
%9116 = torch_c.from_builtin_tensor %9115 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6896 = torch.constant.int 1
%9117 = torch.aten.add.Tensor %9053, %9098, %int1_6896 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6897 = torch.constant.int 1
%9118 = torch.aten.add.Tensor %9117, %9101, %int1_6897 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6898 = torch.constant.int 1
%9119 = torch.aten.add.Tensor %9118, %9104, %int1_6898 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6899 = torch.constant.int 1
%9120 = torch.aten.add.Tensor %9119, %9107, %int1_6899 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6900 = torch.constant.int 1
%9121 = torch.aten.add.Tensor %9120, %9110, %int1_6900 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6901 = torch.constant.int 1
%9122 = torch.aten.add.Tensor %9121, %9113, %int1_6901 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6902 = torch.constant.int 1
%9123 = torch.aten.add.Tensor %9122, %9116, %int1_6902 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9124 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6903 = arith.constant 1 : index
%dim_6904 = tensor.dim %9124, %c1_6903 : tensor<4x?x4096xf16>
%9125 = flow.tensor.transfer %9124 : tensor<4x?x4096xf16>{%dim_6904} to #hal.device.promise<@__device_1>
%9126 = torch_c.from_builtin_tensor %9125 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9127 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6905 = arith.constant 1 : index
%dim_6906 = tensor.dim %9127, %c1_6905 : tensor<4x?x4096xf16>
%9128 = flow.tensor.transfer %9127 : tensor<4x?x4096xf16>{%dim_6906} to #hal.device.promise<@__device_1>
%9129 = torch_c.from_builtin_tensor %9128 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9130 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6907 = arith.constant 1 : index
%dim_6908 = tensor.dim %9130, %c1_6907 : tensor<4x?x4096xf16>
%9131 = flow.tensor.transfer %9130 : tensor<4x?x4096xf16>{%dim_6908} to #hal.device.promise<@__device_1>
%9132 = torch_c.from_builtin_tensor %9131 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9133 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6909 = arith.constant 1 : index
%dim_6910 = tensor.dim %9133, %c1_6909 : tensor<4x?x4096xf16>
%9134 = flow.tensor.transfer %9133 : tensor<4x?x4096xf16>{%dim_6910} to #hal.device.promise<@__device_1>
%9135 = torch_c.from_builtin_tensor %9134 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9136 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6911 = arith.constant 1 : index
%dim_6912 = tensor.dim %9136, %c1_6911 : tensor<4x?x4096xf16>
%9137 = flow.tensor.transfer %9136 : tensor<4x?x4096xf16>{%dim_6912} to #hal.device.promise<@__device_1>
%9138 = torch_c.from_builtin_tensor %9137 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9139 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6913 = arith.constant 1 : index
%dim_6914 = tensor.dim %9139, %c1_6913 : tensor<4x?x4096xf16>
%9140 = flow.tensor.transfer %9139 : tensor<4x?x4096xf16>{%dim_6914} to #hal.device.promise<@__device_1>
%9141 = torch_c.from_builtin_tensor %9140 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9142 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6915 = arith.constant 1 : index
%dim_6916 = tensor.dim %9142, %c1_6915 : tensor<4x?x4096xf16>
%9143 = flow.tensor.transfer %9142 : tensor<4x?x4096xf16>{%dim_6916} to #hal.device.promise<@__device_1>
%9144 = torch_c.from_builtin_tensor %9143 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6917 = torch.constant.int 1
%9145 = torch.aten.add.Tensor %9126, %9059, %int1_6917 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6918 = torch.constant.int 1
%9146 = torch.aten.add.Tensor %9145, %9129, %int1_6918 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6919 = torch.constant.int 1
%9147 = torch.aten.add.Tensor %9146, %9132, %int1_6919 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6920 = torch.constant.int 1
%9148 = torch.aten.add.Tensor %9147, %9135, %int1_6920 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6921 = torch.constant.int 1
%9149 = torch.aten.add.Tensor %9148, %9138, %int1_6921 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6922 = torch.constant.int 1
%9150 = torch.aten.add.Tensor %9149, %9141, %int1_6922 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6923 = torch.constant.int 1
%9151 = torch.aten.add.Tensor %9150, %9144, %int1_6923 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9152 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6924 = arith.constant 1 : index
%dim_6925 = tensor.dim %9152, %c1_6924 : tensor<4x?x4096xf16>
%9153 = flow.tensor.transfer %9152 : tensor<4x?x4096xf16>{%dim_6925} to #hal.device.promise<@__device_2>
%9154 = torch_c.from_builtin_tensor %9153 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9155 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6926 = arith.constant 1 : index
%dim_6927 = tensor.dim %9155, %c1_6926 : tensor<4x?x4096xf16>
%9156 = flow.tensor.transfer %9155 : tensor<4x?x4096xf16>{%dim_6927} to #hal.device.promise<@__device_2>
%9157 = torch_c.from_builtin_tensor %9156 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9158 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6928 = arith.constant 1 : index
%dim_6929 = tensor.dim %9158, %c1_6928 : tensor<4x?x4096xf16>
%9159 = flow.tensor.transfer %9158 : tensor<4x?x4096xf16>{%dim_6929} to #hal.device.promise<@__device_2>
%9160 = torch_c.from_builtin_tensor %9159 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9161 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6930 = arith.constant 1 : index
%dim_6931 = tensor.dim %9161, %c1_6930 : tensor<4x?x4096xf16>
%9162 = flow.tensor.transfer %9161 : tensor<4x?x4096xf16>{%dim_6931} to #hal.device.promise<@__device_2>
%9163 = torch_c.from_builtin_tensor %9162 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9164 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6932 = arith.constant 1 : index
%dim_6933 = tensor.dim %9164, %c1_6932 : tensor<4x?x4096xf16>
%9165 = flow.tensor.transfer %9164 : tensor<4x?x4096xf16>{%dim_6933} to #hal.device.promise<@__device_2>
%9166 = torch_c.from_builtin_tensor %9165 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9167 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6934 = arith.constant 1 : index
%dim_6935 = tensor.dim %9167, %c1_6934 : tensor<4x?x4096xf16>
%9168 = flow.tensor.transfer %9167 : tensor<4x?x4096xf16>{%dim_6935} to #hal.device.promise<@__device_2>
%9169 = torch_c.from_builtin_tensor %9168 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9170 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6936 = arith.constant 1 : index
%dim_6937 = tensor.dim %9170, %c1_6936 : tensor<4x?x4096xf16>
%9171 = flow.tensor.transfer %9170 : tensor<4x?x4096xf16>{%dim_6937} to #hal.device.promise<@__device_2>
%9172 = torch_c.from_builtin_tensor %9171 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6938 = torch.constant.int 1
%9173 = torch.aten.add.Tensor %9154, %9157, %int1_6938 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6939 = torch.constant.int 1
%9174 = torch.aten.add.Tensor %9173, %9065, %int1_6939 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6940 = torch.constant.int 1
%9175 = torch.aten.add.Tensor %9174, %9160, %int1_6940 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6941 = torch.constant.int 1
%9176 = torch.aten.add.Tensor %9175, %9163, %int1_6941 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6942 = torch.constant.int 1
%9177 = torch.aten.add.Tensor %9176, %9166, %int1_6942 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6943 = torch.constant.int 1
%9178 = torch.aten.add.Tensor %9177, %9169, %int1_6943 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6944 = torch.constant.int 1
%9179 = torch.aten.add.Tensor %9178, %9172, %int1_6944 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9180 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6945 = arith.constant 1 : index
%dim_6946 = tensor.dim %9180, %c1_6945 : tensor<4x?x4096xf16>
%9181 = flow.tensor.transfer %9180 : tensor<4x?x4096xf16>{%dim_6946} to #hal.device.promise<@__device_3>
%9182 = torch_c.from_builtin_tensor %9181 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9183 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6947 = arith.constant 1 : index
%dim_6948 = tensor.dim %9183, %c1_6947 : tensor<4x?x4096xf16>
%9184 = flow.tensor.transfer %9183 : tensor<4x?x4096xf16>{%dim_6948} to #hal.device.promise<@__device_3>
%9185 = torch_c.from_builtin_tensor %9184 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9186 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6949 = arith.constant 1 : index
%dim_6950 = tensor.dim %9186, %c1_6949 : tensor<4x?x4096xf16>
%9187 = flow.tensor.transfer %9186 : tensor<4x?x4096xf16>{%dim_6950} to #hal.device.promise<@__device_3>
%9188 = torch_c.from_builtin_tensor %9187 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9189 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6951 = arith.constant 1 : index
%dim_6952 = tensor.dim %9189, %c1_6951 : tensor<4x?x4096xf16>
%9190 = flow.tensor.transfer %9189 : tensor<4x?x4096xf16>{%dim_6952} to #hal.device.promise<@__device_3>
%9191 = torch_c.from_builtin_tensor %9190 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9192 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6953 = arith.constant 1 : index
%dim_6954 = tensor.dim %9192, %c1_6953 : tensor<4x?x4096xf16>
%9193 = flow.tensor.transfer %9192 : tensor<4x?x4096xf16>{%dim_6954} to #hal.device.promise<@__device_3>
%9194 = torch_c.from_builtin_tensor %9193 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9195 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6955 = arith.constant 1 : index
%dim_6956 = tensor.dim %9195, %c1_6955 : tensor<4x?x4096xf16>
%9196 = flow.tensor.transfer %9195 : tensor<4x?x4096xf16>{%dim_6956} to #hal.device.promise<@__device_3>
%9197 = torch_c.from_builtin_tensor %9196 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9198 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6957 = arith.constant 1 : index
%dim_6958 = tensor.dim %9198, %c1_6957 : tensor<4x?x4096xf16>
%9199 = flow.tensor.transfer %9198 : tensor<4x?x4096xf16>{%dim_6958} to #hal.device.promise<@__device_3>
%9200 = torch_c.from_builtin_tensor %9199 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6959 = torch.constant.int 1
%9201 = torch.aten.add.Tensor %9182, %9185, %int1_6959 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6960 = torch.constant.int 1
%9202 = torch.aten.add.Tensor %9201, %9188, %int1_6960 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6961 = torch.constant.int 1
%9203 = torch.aten.add.Tensor %9202, %9071, %int1_6961 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6962 = torch.constant.int 1
%9204 = torch.aten.add.Tensor %9203, %9191, %int1_6962 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6963 = torch.constant.int 1
%9205 = torch.aten.add.Tensor %9204, %9194, %int1_6963 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6964 = torch.constant.int 1
%9206 = torch.aten.add.Tensor %9205, %9197, %int1_6964 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6965 = torch.constant.int 1
%9207 = torch.aten.add.Tensor %9206, %9200, %int1_6965 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9208 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6966 = arith.constant 1 : index
%dim_6967 = tensor.dim %9208, %c1_6966 : tensor<4x?x4096xf16>
%9209 = flow.tensor.transfer %9208 : tensor<4x?x4096xf16>{%dim_6967} to #hal.device.promise<@__device_4>
%9210 = torch_c.from_builtin_tensor %9209 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9211 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6968 = arith.constant 1 : index
%dim_6969 = tensor.dim %9211, %c1_6968 : tensor<4x?x4096xf16>
%9212 = flow.tensor.transfer %9211 : tensor<4x?x4096xf16>{%dim_6969} to #hal.device.promise<@__device_4>
%9213 = torch_c.from_builtin_tensor %9212 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9214 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6970 = arith.constant 1 : index
%dim_6971 = tensor.dim %9214, %c1_6970 : tensor<4x?x4096xf16>
%9215 = flow.tensor.transfer %9214 : tensor<4x?x4096xf16>{%dim_6971} to #hal.device.promise<@__device_4>
%9216 = torch_c.from_builtin_tensor %9215 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9217 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6972 = arith.constant 1 : index
%dim_6973 = tensor.dim %9217, %c1_6972 : tensor<4x?x4096xf16>
%9218 = flow.tensor.transfer %9217 : tensor<4x?x4096xf16>{%dim_6973} to #hal.device.promise<@__device_4>
%9219 = torch_c.from_builtin_tensor %9218 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9220 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6974 = arith.constant 1 : index
%dim_6975 = tensor.dim %9220, %c1_6974 : tensor<4x?x4096xf16>
%9221 = flow.tensor.transfer %9220 : tensor<4x?x4096xf16>{%dim_6975} to #hal.device.promise<@__device_4>
%9222 = torch_c.from_builtin_tensor %9221 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9223 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6976 = arith.constant 1 : index
%dim_6977 = tensor.dim %9223, %c1_6976 : tensor<4x?x4096xf16>
%9224 = flow.tensor.transfer %9223 : tensor<4x?x4096xf16>{%dim_6977} to #hal.device.promise<@__device_4>
%9225 = torch_c.from_builtin_tensor %9224 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9226 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6978 = arith.constant 1 : index
%dim_6979 = tensor.dim %9226, %c1_6978 : tensor<4x?x4096xf16>
%9227 = flow.tensor.transfer %9226 : tensor<4x?x4096xf16>{%dim_6979} to #hal.device.promise<@__device_4>
%9228 = torch_c.from_builtin_tensor %9227 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6980 = torch.constant.int 1
%9229 = torch.aten.add.Tensor %9210, %9213, %int1_6980 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6981 = torch.constant.int 1
%9230 = torch.aten.add.Tensor %9229, %9216, %int1_6981 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6982 = torch.constant.int 1
%9231 = torch.aten.add.Tensor %9230, %9219, %int1_6982 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6983 = torch.constant.int 1
%9232 = torch.aten.add.Tensor %9231, %9077, %int1_6983 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6984 = torch.constant.int 1
%9233 = torch.aten.add.Tensor %9232, %9222, %int1_6984 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6985 = torch.constant.int 1
%9234 = torch.aten.add.Tensor %9233, %9225, %int1_6985 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_6986 = torch.constant.int 1
%9235 = torch.aten.add.Tensor %9234, %9228, %int1_6986 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9236 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6987 = arith.constant 1 : index
%dim_6988 = tensor.dim %9236, %c1_6987 : tensor<4x?x4096xf16>
%9237 = flow.tensor.transfer %9236 : tensor<4x?x4096xf16>{%dim_6988} to #hal.device.promise<@__device_5>
%9238 = torch_c.from_builtin_tensor %9237 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9239 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6989 = arith.constant 1 : index
%dim_6990 = tensor.dim %9239, %c1_6989 : tensor<4x?x4096xf16>
%9240 = flow.tensor.transfer %9239 : tensor<4x?x4096xf16>{%dim_6990} to #hal.device.promise<@__device_5>
%9241 = torch_c.from_builtin_tensor %9240 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9242 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6991 = arith.constant 1 : index
%dim_6992 = tensor.dim %9242, %c1_6991 : tensor<4x?x4096xf16>
%9243 = flow.tensor.transfer %9242 : tensor<4x?x4096xf16>{%dim_6992} to #hal.device.promise<@__device_5>
%9244 = torch_c.from_builtin_tensor %9243 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9245 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6993 = arith.constant 1 : index
%dim_6994 = tensor.dim %9245, %c1_6993 : tensor<4x?x4096xf16>
%9246 = flow.tensor.transfer %9245 : tensor<4x?x4096xf16>{%dim_6994} to #hal.device.promise<@__device_5>
%9247 = torch_c.from_builtin_tensor %9246 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9248 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6995 = arith.constant 1 : index
%dim_6996 = tensor.dim %9248, %c1_6995 : tensor<4x?x4096xf16>
%9249 = flow.tensor.transfer %9248 : tensor<4x?x4096xf16>{%dim_6996} to #hal.device.promise<@__device_5>
%9250 = torch_c.from_builtin_tensor %9249 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9251 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6997 = arith.constant 1 : index
%dim_6998 = tensor.dim %9251, %c1_6997 : tensor<4x?x4096xf16>
%9252 = flow.tensor.transfer %9251 : tensor<4x?x4096xf16>{%dim_6998} to #hal.device.promise<@__device_5>
%9253 = torch_c.from_builtin_tensor %9252 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9254 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_6999 = arith.constant 1 : index
%dim_7000 = tensor.dim %9254, %c1_6999 : tensor<4x?x4096xf16>
%9255 = flow.tensor.transfer %9254 : tensor<4x?x4096xf16>{%dim_7000} to #hal.device.promise<@__device_5>
%9256 = torch_c.from_builtin_tensor %9255 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7001 = torch.constant.int 1
%9257 = torch.aten.add.Tensor %9238, %9241, %int1_7001 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7002 = torch.constant.int 1
%9258 = torch.aten.add.Tensor %9257, %9244, %int1_7002 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7003 = torch.constant.int 1
%9259 = torch.aten.add.Tensor %9258, %9247, %int1_7003 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7004 = torch.constant.int 1
%9260 = torch.aten.add.Tensor %9259, %9250, %int1_7004 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7005 = torch.constant.int 1
%9261 = torch.aten.add.Tensor %9260, %9083, %int1_7005 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7006 = torch.constant.int 1
%9262 = torch.aten.add.Tensor %9261, %9253, %int1_7006 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7007 = torch.constant.int 1
%9263 = torch.aten.add.Tensor %9262, %9256, %int1_7007 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9263, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9264 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7008 = arith.constant 1 : index
%dim_7009 = tensor.dim %9264, %c1_7008 : tensor<4x?x4096xf16>
%9265 = flow.tensor.transfer %9264 : tensor<4x?x4096xf16>{%dim_7009} to #hal.device.promise<@__device_6>
%9266 = torch_c.from_builtin_tensor %9265 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9267 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7010 = arith.constant 1 : index
%dim_7011 = tensor.dim %9267, %c1_7010 : tensor<4x?x4096xf16>
%9268 = flow.tensor.transfer %9267 : tensor<4x?x4096xf16>{%dim_7011} to #hal.device.promise<@__device_6>
%9269 = torch_c.from_builtin_tensor %9268 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9270 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7012 = arith.constant 1 : index
%dim_7013 = tensor.dim %9270, %c1_7012 : tensor<4x?x4096xf16>
%9271 = flow.tensor.transfer %9270 : tensor<4x?x4096xf16>{%dim_7013} to #hal.device.promise<@__device_6>
%9272 = torch_c.from_builtin_tensor %9271 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9273 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7014 = arith.constant 1 : index
%dim_7015 = tensor.dim %9273, %c1_7014 : tensor<4x?x4096xf16>
%9274 = flow.tensor.transfer %9273 : tensor<4x?x4096xf16>{%dim_7015} to #hal.device.promise<@__device_6>
%9275 = torch_c.from_builtin_tensor %9274 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9276 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7016 = arith.constant 1 : index
%dim_7017 = tensor.dim %9276, %c1_7016 : tensor<4x?x4096xf16>
%9277 = flow.tensor.transfer %9276 : tensor<4x?x4096xf16>{%dim_7017} to #hal.device.promise<@__device_6>
%9278 = torch_c.from_builtin_tensor %9277 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9279 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7018 = arith.constant 1 : index
%dim_7019 = tensor.dim %9279, %c1_7018 : tensor<4x?x4096xf16>
%9280 = flow.tensor.transfer %9279 : tensor<4x?x4096xf16>{%dim_7019} to #hal.device.promise<@__device_6>
%9281 = torch_c.from_builtin_tensor %9280 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9282 = torch_c.to_builtin_tensor %9095 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7020 = arith.constant 1 : index
%dim_7021 = tensor.dim %9282, %c1_7020 : tensor<4x?x4096xf16>
%9283 = flow.tensor.transfer %9282 : tensor<4x?x4096xf16>{%dim_7021} to #hal.device.promise<@__device_6>
%9284 = torch_c.from_builtin_tensor %9283 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7022 = torch.constant.int 1
%9285 = torch.aten.add.Tensor %9266, %9269, %int1_7022 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7023 = torch.constant.int 1
%9286 = torch.aten.add.Tensor %9285, %9272, %int1_7023 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7024 = torch.constant.int 1
%9287 = torch.aten.add.Tensor %9286, %9275, %int1_7024 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7025 = torch.constant.int 1
%9288 = torch.aten.add.Tensor %9287, %9278, %int1_7025 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7026 = torch.constant.int 1
%9289 = torch.aten.add.Tensor %9288, %9281, %int1_7026 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7027 = torch.constant.int 1
%9290 = torch.aten.add.Tensor %9289, %9089, %int1_7027 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7028 = torch.constant.int 1
%9291 = torch.aten.add.Tensor %9290, %9284, %int1_7028 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9292 = torch_c.to_builtin_tensor %9053 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7029 = arith.constant 1 : index
%dim_7030 = tensor.dim %9292, %c1_7029 : tensor<4x?x4096xf16>
%9293 = flow.tensor.transfer %9292 : tensor<4x?x4096xf16>{%dim_7030} to #hal.device.promise<@__device_7>
%9294 = torch_c.from_builtin_tensor %9293 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9295 = torch_c.to_builtin_tensor %9059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7031 = arith.constant 1 : index
%dim_7032 = tensor.dim %9295, %c1_7031 : tensor<4x?x4096xf16>
%9296 = flow.tensor.transfer %9295 : tensor<4x?x4096xf16>{%dim_7032} to #hal.device.promise<@__device_7>
%9297 = torch_c.from_builtin_tensor %9296 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9298 = torch_c.to_builtin_tensor %9065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7033 = arith.constant 1 : index
%dim_7034 = tensor.dim %9298, %c1_7033 : tensor<4x?x4096xf16>
%9299 = flow.tensor.transfer %9298 : tensor<4x?x4096xf16>{%dim_7034} to #hal.device.promise<@__device_7>
%9300 = torch_c.from_builtin_tensor %9299 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9301 = torch_c.to_builtin_tensor %9071 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7035 = arith.constant 1 : index
%dim_7036 = tensor.dim %9301, %c1_7035 : tensor<4x?x4096xf16>
%9302 = flow.tensor.transfer %9301 : tensor<4x?x4096xf16>{%dim_7036} to #hal.device.promise<@__device_7>
%9303 = torch_c.from_builtin_tensor %9302 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9304 = torch_c.to_builtin_tensor %9077 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7037 = arith.constant 1 : index
%dim_7038 = tensor.dim %9304, %c1_7037 : tensor<4x?x4096xf16>
%9305 = flow.tensor.transfer %9304 : tensor<4x?x4096xf16>{%dim_7038} to #hal.device.promise<@__device_7>
%9306 = torch_c.from_builtin_tensor %9305 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9307 = torch_c.to_builtin_tensor %9083 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7039 = arith.constant 1 : index
%dim_7040 = tensor.dim %9307, %c1_7039 : tensor<4x?x4096xf16>
%9308 = flow.tensor.transfer %9307 : tensor<4x?x4096xf16>{%dim_7040} to #hal.device.promise<@__device_7>
%9309 = torch_c.from_builtin_tensor %9308 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9310 = torch_c.to_builtin_tensor %9089 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7041 = arith.constant 1 : index
%dim_7042 = tensor.dim %9310, %c1_7041 : tensor<4x?x4096xf16>
%9311 = flow.tensor.transfer %9310 : tensor<4x?x4096xf16>{%dim_7042} to #hal.device.promise<@__device_7>
%9312 = torch_c.from_builtin_tensor %9311 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7043 = torch.constant.int 1
%9313 = torch.aten.add.Tensor %9294, %9297, %int1_7043 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7044 = torch.constant.int 1
%9314 = torch.aten.add.Tensor %9313, %9300, %int1_7044 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7045 = torch.constant.int 1
%9315 = torch.aten.add.Tensor %9314, %9303, %int1_7045 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7046 = torch.constant.int 1
%9316 = torch.aten.add.Tensor %9315, %9306, %int1_7046 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7047 = torch.constant.int 1
%9317 = torch.aten.add.Tensor %9316, %9309, %int1_7047 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7048 = torch.constant.int 1
%9318 = torch.aten.add.Tensor %9317, %9312, %int1_7048 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7049 = torch.constant.int 1
%9319 = torch.aten.add.Tensor %9318, %9095, %int1_7049 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7050 = torch.constant.int 1
%9320 = torch.aten.add.Tensor %7979, %9123, %int1_7050 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7051 = torch.constant.int 1
%9321 = torch.aten.add.Tensor %7980, %9151, %int1_7051 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7052 = torch.constant.int 1
%9322 = torch.aten.add.Tensor %7981, %9179, %int1_7052 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7053 = torch.constant.int 1
%9323 = torch.aten.add.Tensor %7982, %9207, %int1_7053 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7054 = torch.constant.int 1
%9324 = torch.aten.add.Tensor %7983, %9235, %int1_7054 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7055 = torch.constant.int 1
%9325 = torch.aten.add.Tensor %7984, %9263, %int1_7055 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7056 = torch.constant.int 1
%9326 = torch.aten.add.Tensor %7985, %9291, %int1_7056 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7057 = torch.constant.int 1
%9327 = torch.aten.add.Tensor %7986, %9319, %int1_7057 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_7058 = torch.constant.int 6
%9328 = torch.prims.convert_element_type %9320, %int6_7058 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7059 = torch.constant.int 6
%9329 = torch.prims.convert_element_type %9321, %int6_7059 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7060 = torch.constant.int 6
%9330 = torch.prims.convert_element_type %9322, %int6_7060 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7061 = torch.constant.int 6
%9331 = torch.prims.convert_element_type %9323, %int6_7061 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7062 = torch.constant.int 6
%9332 = torch.prims.convert_element_type %9324, %int6_7062 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7063 = torch.constant.int 6
%9333 = torch.prims.convert_element_type %9325, %int6_7063 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7064 = torch.constant.int 6
%9334 = torch.prims.convert_element_type %9326, %int6_7064 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7065 = torch.constant.int 6
%9335 = torch.prims.convert_element_type %9327, %int6_7065 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7066 = torch.constant.int 2
%9336 = torch.aten.pow.Tensor_Scalar %9328, %int2_7066 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7067 = torch.constant.int 2
%9337 = torch.aten.pow.Tensor_Scalar %9329, %int2_7067 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7068 = torch.constant.int 2
%9338 = torch.aten.pow.Tensor_Scalar %9330, %int2_7068 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7069 = torch.constant.int 2
%9339 = torch.aten.pow.Tensor_Scalar %9331, %int2_7069 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7070 = torch.constant.int 2
%9340 = torch.aten.pow.Tensor_Scalar %9332, %int2_7070 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7071 = torch.constant.int 2
%9341 = torch.aten.pow.Tensor_Scalar %9333, %int2_7071 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7072 = torch.constant.int 2
%9342 = torch.aten.pow.Tensor_Scalar %9334, %int2_7072 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7073 = torch.constant.int 2
%9343 = torch.aten.pow.Tensor_Scalar %9335, %int2_7073 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_7074 = torch.constant.int -1
%9344 = torch.prim.ListConstruct %int-1_7074 : (!torch.int) -> !torch.list<int>
%true_7075 = torch.constant.bool true
%none_7076 = torch.constant.none
%9345 = torch.aten.mean.dim %9336, %9344, %true_7075, %none_7076 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7077 = torch.constant.int -1
%9346 = torch.prim.ListConstruct %int-1_7077 : (!torch.int) -> !torch.list<int>
%true_7078 = torch.constant.bool true
%none_7079 = torch.constant.none
%9347 = torch.aten.mean.dim %9337, %9346, %true_7078, %none_7079 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7080 = torch.constant.int -1
%9348 = torch.prim.ListConstruct %int-1_7080 : (!torch.int) -> !torch.list<int>
%true_7081 = torch.constant.bool true
%none_7082 = torch.constant.none
%9349 = torch.aten.mean.dim %9338, %9348, %true_7081, %none_7082 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7083 = torch.constant.int -1
%9350 = torch.prim.ListConstruct %int-1_7083 : (!torch.int) -> !torch.list<int>
%true_7084 = torch.constant.bool true
%none_7085 = torch.constant.none
%9351 = torch.aten.mean.dim %9339, %9350, %true_7084, %none_7085 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7086 = torch.constant.int -1
%9352 = torch.prim.ListConstruct %int-1_7086 : (!torch.int) -> !torch.list<int>
%true_7087 = torch.constant.bool true
%none_7088 = torch.constant.none
%9353 = torch.aten.mean.dim %9340, %9352, %true_7087, %none_7088 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7089 = torch.constant.int -1
%9354 = torch.prim.ListConstruct %int-1_7089 : (!torch.int) -> !torch.list<int>
%true_7090 = torch.constant.bool true
%none_7091 = torch.constant.none
%9355 = torch.aten.mean.dim %9341, %9354, %true_7090, %none_7091 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7092 = torch.constant.int -1
%9356 = torch.prim.ListConstruct %int-1_7092 : (!torch.int) -> !torch.list<int>
%true_7093 = torch.constant.bool true
%none_7094 = torch.constant.none
%9357 = torch.aten.mean.dim %9342, %9356, %true_7093, %none_7094 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7095 = torch.constant.int -1
%9358 = torch.prim.ListConstruct %int-1_7095 : (!torch.int) -> !torch.list<int>
%true_7096 = torch.constant.bool true
%none_7097 = torch.constant.none
%9359 = torch.aten.mean.dim %9343, %9358, %true_7096, %none_7097 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7098 = torch.constant.float 9.9999997473787516E-6
%int1_7099 = torch.constant.int 1
%9360 = torch.aten.add.Scalar %9345, %float9.999990e-06_7098, %int1_7099 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7100 = torch.constant.float 9.9999997473787516E-6
%int1_7101 = torch.constant.int 1
%9361 = torch.aten.add.Scalar %9347, %float9.999990e-06_7100, %int1_7101 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7102 = torch.constant.float 9.9999997473787516E-6
%int1_7103 = torch.constant.int 1
%9362 = torch.aten.add.Scalar %9349, %float9.999990e-06_7102, %int1_7103 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7104 = torch.constant.float 9.9999997473787516E-6
%int1_7105 = torch.constant.int 1
%9363 = torch.aten.add.Scalar %9351, %float9.999990e-06_7104, %int1_7105 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7106 = torch.constant.float 9.9999997473787516E-6
%int1_7107 = torch.constant.int 1
%9364 = torch.aten.add.Scalar %9353, %float9.999990e-06_7106, %int1_7107 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7108 = torch.constant.float 9.9999997473787516E-6
%int1_7109 = torch.constant.int 1
%9365 = torch.aten.add.Scalar %9355, %float9.999990e-06_7108, %int1_7109 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7110 = torch.constant.float 9.9999997473787516E-6
%int1_7111 = torch.constant.int 1
%9366 = torch.aten.add.Scalar %9357, %float9.999990e-06_7110, %int1_7111 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7112 = torch.constant.float 9.9999997473787516E-6
%int1_7113 = torch.constant.int 1
%9367 = torch.aten.add.Scalar %9359, %float9.999990e-06_7112, %int1_7113 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9368 = torch.aten.rsqrt %9360 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9369 = torch.aten.rsqrt %9361 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9370 = torch.aten.rsqrt %9362 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9371 = torch.aten.rsqrt %9363 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9372 = torch.aten.rsqrt %9364 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9373 = torch.aten.rsqrt %9365 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9374 = torch.aten.rsqrt %9366 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9375 = torch.aten.rsqrt %9367 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9376 = torch.aten.mul.Tensor %9328, %9368 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9377 = torch.aten.mul.Tensor %9329, %9369 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9378 = torch.aten.mul.Tensor %9330, %9370 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9379 = torch.aten.mul.Tensor %9331, %9371 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9380 = torch.aten.mul.Tensor %9332, %9372 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9381 = torch.aten.mul.Tensor %9333, %9373 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9382 = torch.aten.mul.Tensor %9334, %9374 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9383 = torch.aten.mul.Tensor %9335, %9375 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9384 = torch.aten.mul.Tensor %264, %9376 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9385 = torch.aten.mul.Tensor %265, %9377 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9386 = torch.aten.mul.Tensor %266, %9378 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9387 = torch.aten.mul.Tensor %267, %9379 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9388 = torch.aten.mul.Tensor %268, %9380 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9389 = torch.aten.mul.Tensor %269, %9381 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9390 = torch.aten.mul.Tensor %270, %9382 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9391 = torch.aten.mul.Tensor %271, %9383 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_7114 = torch.constant.int 5
%9392 = torch.prims.convert_element_type %9384, %int5_7114 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7115 = torch.constant.int 5
%9393 = torch.prims.convert_element_type %9385, %int5_7115 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7116 = torch.constant.int 5
%9394 = torch.prims.convert_element_type %9386, %int5_7116 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7117 = torch.constant.int 5
%9395 = torch.prims.convert_element_type %9387, %int5_7117 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7118 = torch.constant.int 5
%9396 = torch.prims.convert_element_type %9388, %int5_7118 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7119 = torch.constant.int 5
%9397 = torch.prims.convert_element_type %9389, %int5_7119 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7120 = torch.constant.int 5
%9398 = torch.prims.convert_element_type %9390, %int5_7120 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7121 = torch.constant.int 5
%9399 = torch.prims.convert_element_type %9391, %int5_7121 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7122 = torch.constant.int 1
%int0_7123 = torch.constant.int 0
%9400 = torch.prim.ListConstruct %int1_7122, %int0_7123 : (!torch.int, !torch.int) -> !torch.list<int>
%9401 = torch.aten.permute %272, %9400 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7124 = torch.constant.int 1
%int0_7125 = torch.constant.int 0
%9402 = torch.prim.ListConstruct %int1_7124, %int0_7125 : (!torch.int, !torch.int) -> !torch.list<int>
%9403 = torch.aten.permute %273, %9402 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7126 = torch.constant.int 1
%int0_7127 = torch.constant.int 0
%9404 = torch.prim.ListConstruct %int1_7126, %int0_7127 : (!torch.int, !torch.int) -> !torch.list<int>
%9405 = torch.aten.permute %274, %9404 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7128 = torch.constant.int 1
%int0_7129 = torch.constant.int 0
%9406 = torch.prim.ListConstruct %int1_7128, %int0_7129 : (!torch.int, !torch.int) -> !torch.list<int>
%9407 = torch.aten.permute %275, %9406 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7130 = torch.constant.int 1
%int0_7131 = torch.constant.int 0
%9408 = torch.prim.ListConstruct %int1_7130, %int0_7131 : (!torch.int, !torch.int) -> !torch.list<int>
%9409 = torch.aten.permute %276, %9408 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7132 = torch.constant.int 1
%int0_7133 = torch.constant.int 0
%9410 = torch.prim.ListConstruct %int1_7132, %int0_7133 : (!torch.int, !torch.int) -> !torch.list<int>
%9411 = torch.aten.permute %277, %9410 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7134 = torch.constant.int 1
%int0_7135 = torch.constant.int 0
%9412 = torch.prim.ListConstruct %int1_7134, %int0_7135 : (!torch.int, !torch.int) -> !torch.list<int>
%9413 = torch.aten.permute %278, %9412 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7136 = torch.constant.int 1
%int0_7137 = torch.constant.int 0
%9414 = torch.prim.ListConstruct %int1_7136, %int0_7137 : (!torch.int, !torch.int) -> !torch.list<int>
%9415 = torch.aten.permute %279, %9414 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_7138 = torch.constant.int 4
%9416 = torch.aten.mul.int %int4_7138, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7139 = torch.constant.int 4096
%9417 = torch.prim.ListConstruct %9416, %int4096_7139 : (!torch.int, !torch.int) -> !torch.list<int>
%9418 = torch.aten.view %9392, %9417 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9418, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9419 = torch.aten.mm %9418, %9401 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9419, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7140 = torch.constant.int 4
%int1792_7141 = torch.constant.int 1792
%9420 = torch.prim.ListConstruct %int4_7140, %2482, %int1792_7141 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9421 = torch.aten.view %9419, %9420 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7142 = torch.constant.int 4
%9422 = torch.aten.mul.int %int4_7142, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7143 = torch.constant.int 4096
%9423 = torch.prim.ListConstruct %9422, %int4096_7143 : (!torch.int, !torch.int) -> !torch.list<int>
%9424 = torch.aten.view %9393, %9423 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9424, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9425 = torch.aten.mm %9424, %9403 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9425, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7144 = torch.constant.int 4
%int1792_7145 = torch.constant.int 1792
%9426 = torch.prim.ListConstruct %int4_7144, %2482, %int1792_7145 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9427 = torch.aten.view %9425, %9426 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7146 = torch.constant.int 4
%9428 = torch.aten.mul.int %int4_7146, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7147 = torch.constant.int 4096
%9429 = torch.prim.ListConstruct %9428, %int4096_7147 : (!torch.int, !torch.int) -> !torch.list<int>
%9430 = torch.aten.view %9394, %9429 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9430, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9431 = torch.aten.mm %9430, %9405 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9431, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7148 = torch.constant.int 4
%int1792_7149 = torch.constant.int 1792
%9432 = torch.prim.ListConstruct %int4_7148, %2482, %int1792_7149 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9433 = torch.aten.view %9431, %9432 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7150 = torch.constant.int 4
%9434 = torch.aten.mul.int %int4_7150, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7151 = torch.constant.int 4096
%9435 = torch.prim.ListConstruct %9434, %int4096_7151 : (!torch.int, !torch.int) -> !torch.list<int>
%9436 = torch.aten.view %9395, %9435 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9436, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9437 = torch.aten.mm %9436, %9407 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9437, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7152 = torch.constant.int 4
%int1792_7153 = torch.constant.int 1792
%9438 = torch.prim.ListConstruct %int4_7152, %2482, %int1792_7153 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9439 = torch.aten.view %9437, %9438 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7154 = torch.constant.int 4
%9440 = torch.aten.mul.int %int4_7154, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7155 = torch.constant.int 4096
%9441 = torch.prim.ListConstruct %9440, %int4096_7155 : (!torch.int, !torch.int) -> !torch.list<int>
%9442 = torch.aten.view %9396, %9441 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9442, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9443 = torch.aten.mm %9442, %9409 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9443, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7156 = torch.constant.int 4
%int1792_7157 = torch.constant.int 1792
%9444 = torch.prim.ListConstruct %int4_7156, %2482, %int1792_7157 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9445 = torch.aten.view %9443, %9444 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7158 = torch.constant.int 4
%9446 = torch.aten.mul.int %int4_7158, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7159 = torch.constant.int 4096
%9447 = torch.prim.ListConstruct %9446, %int4096_7159 : (!torch.int, !torch.int) -> !torch.list<int>
%9448 = torch.aten.view %9397, %9447 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9448, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9449 = torch.aten.mm %9448, %9411 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9449, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7160 = torch.constant.int 4
%int1792_7161 = torch.constant.int 1792
%9450 = torch.prim.ListConstruct %int4_7160, %2482, %int1792_7161 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9451 = torch.aten.view %9449, %9450 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7162 = torch.constant.int 4
%9452 = torch.aten.mul.int %int4_7162, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7163 = torch.constant.int 4096
%9453 = torch.prim.ListConstruct %9452, %int4096_7163 : (!torch.int, !torch.int) -> !torch.list<int>
%9454 = torch.aten.view %9398, %9453 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9454, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9455 = torch.aten.mm %9454, %9413 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9455, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7164 = torch.constant.int 4
%int1792_7165 = torch.constant.int 1792
%9456 = torch.prim.ListConstruct %int4_7164, %2482, %int1792_7165 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9457 = torch.aten.view %9455, %9456 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7166 = torch.constant.int 4
%9458 = torch.aten.mul.int %int4_7166, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7167 = torch.constant.int 4096
%9459 = torch.prim.ListConstruct %9458, %int4096_7167 : (!torch.int, !torch.int) -> !torch.list<int>
%9460 = torch.aten.view %9399, %9459 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9460, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9461 = torch.aten.mm %9460, %9415 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9461, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7168 = torch.constant.int 4
%int1792_7169 = torch.constant.int 1792
%9462 = torch.prim.ListConstruct %int4_7168, %2482, %int1792_7169 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9463 = torch.aten.view %9461, %9462 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9464 = torch.aten.silu %9421 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9465 = torch.aten.silu %9427 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9466 = torch.aten.silu %9433 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9467 = torch.aten.silu %9439 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9468 = torch.aten.silu %9445 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9469 = torch.aten.silu %9451 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9470 = torch.aten.silu %9457 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9471 = torch.aten.silu %9463 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_7170 = torch.constant.int 1
%int0_7171 = torch.constant.int 0
%9472 = torch.prim.ListConstruct %int1_7170, %int0_7171 : (!torch.int, !torch.int) -> !torch.list<int>
%9473 = torch.aten.permute %280, %9472 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7172 = torch.constant.int 1
%int0_7173 = torch.constant.int 0
%9474 = torch.prim.ListConstruct %int1_7172, %int0_7173 : (!torch.int, !torch.int) -> !torch.list<int>
%9475 = torch.aten.permute %281, %9474 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7174 = torch.constant.int 1
%int0_7175 = torch.constant.int 0
%9476 = torch.prim.ListConstruct %int1_7174, %int0_7175 : (!torch.int, !torch.int) -> !torch.list<int>
%9477 = torch.aten.permute %282, %9476 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7176 = torch.constant.int 1
%int0_7177 = torch.constant.int 0
%9478 = torch.prim.ListConstruct %int1_7176, %int0_7177 : (!torch.int, !torch.int) -> !torch.list<int>
%9479 = torch.aten.permute %283, %9478 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7178 = torch.constant.int 1
%int0_7179 = torch.constant.int 0
%9480 = torch.prim.ListConstruct %int1_7178, %int0_7179 : (!torch.int, !torch.int) -> !torch.list<int>
%9481 = torch.aten.permute %284, %9480 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7180 = torch.constant.int 1
%int0_7181 = torch.constant.int 0
%9482 = torch.prim.ListConstruct %int1_7180, %int0_7181 : (!torch.int, !torch.int) -> !torch.list<int>
%9483 = torch.aten.permute %285, %9482 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7182 = torch.constant.int 1
%int0_7183 = torch.constant.int 0
%9484 = torch.prim.ListConstruct %int1_7182, %int0_7183 : (!torch.int, !torch.int) -> !torch.list<int>
%9485 = torch.aten.permute %286, %9484 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_7184 = torch.constant.int 1
%int0_7185 = torch.constant.int 0
%9486 = torch.prim.ListConstruct %int1_7184, %int0_7185 : (!torch.int, !torch.int) -> !torch.list<int>
%9487 = torch.aten.permute %287, %9486 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_7186 = torch.constant.int 4
%9488 = torch.aten.mul.int %int4_7186, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7187 = torch.constant.int 4096
%9489 = torch.prim.ListConstruct %9488, %int4096_7187 : (!torch.int, !torch.int) -> !torch.list<int>
%9490 = torch.aten.view %9392, %9489 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9490, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9491 = torch.aten.mm %9490, %9473 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9491, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7188 = torch.constant.int 4
%int1792_7189 = torch.constant.int 1792
%9492 = torch.prim.ListConstruct %int4_7188, %2482, %int1792_7189 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9493 = torch.aten.view %9491, %9492 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7190 = torch.constant.int 4
%9494 = torch.aten.mul.int %int4_7190, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7191 = torch.constant.int 4096
%9495 = torch.prim.ListConstruct %9494, %int4096_7191 : (!torch.int, !torch.int) -> !torch.list<int>
%9496 = torch.aten.view %9393, %9495 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9496, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9497 = torch.aten.mm %9496, %9475 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9497, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7192 = torch.constant.int 4
%int1792_7193 = torch.constant.int 1792
%9498 = torch.prim.ListConstruct %int4_7192, %2482, %int1792_7193 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9499 = torch.aten.view %9497, %9498 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7194 = torch.constant.int 4
%9500 = torch.aten.mul.int %int4_7194, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7195 = torch.constant.int 4096
%9501 = torch.prim.ListConstruct %9500, %int4096_7195 : (!torch.int, !torch.int) -> !torch.list<int>
%9502 = torch.aten.view %9394, %9501 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9502, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9503 = torch.aten.mm %9502, %9477 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9503, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7196 = torch.constant.int 4
%int1792_7197 = torch.constant.int 1792
%9504 = torch.prim.ListConstruct %int4_7196, %2482, %int1792_7197 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9505 = torch.aten.view %9503, %9504 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7198 = torch.constant.int 4
%9506 = torch.aten.mul.int %int4_7198, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7199 = torch.constant.int 4096
%9507 = torch.prim.ListConstruct %9506, %int4096_7199 : (!torch.int, !torch.int) -> !torch.list<int>
%9508 = torch.aten.view %9395, %9507 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9508, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9509 = torch.aten.mm %9508, %9479 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9509, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7200 = torch.constant.int 4
%int1792_7201 = torch.constant.int 1792
%9510 = torch.prim.ListConstruct %int4_7200, %2482, %int1792_7201 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9511 = torch.aten.view %9509, %9510 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7202 = torch.constant.int 4
%9512 = torch.aten.mul.int %int4_7202, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7203 = torch.constant.int 4096
%9513 = torch.prim.ListConstruct %9512, %int4096_7203 : (!torch.int, !torch.int) -> !torch.list<int>
%9514 = torch.aten.view %9396, %9513 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9514, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9515 = torch.aten.mm %9514, %9481 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9515, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7204 = torch.constant.int 4
%int1792_7205 = torch.constant.int 1792
%9516 = torch.prim.ListConstruct %int4_7204, %2482, %int1792_7205 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9517 = torch.aten.view %9515, %9516 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7206 = torch.constant.int 4
%9518 = torch.aten.mul.int %int4_7206, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7207 = torch.constant.int 4096
%9519 = torch.prim.ListConstruct %9518, %int4096_7207 : (!torch.int, !torch.int) -> !torch.list<int>
%9520 = torch.aten.view %9397, %9519 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9520, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9521 = torch.aten.mm %9520, %9483 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9521, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7208 = torch.constant.int 4
%int1792_7209 = torch.constant.int 1792
%9522 = torch.prim.ListConstruct %int4_7208, %2482, %int1792_7209 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9523 = torch.aten.view %9521, %9522 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7210 = torch.constant.int 4
%9524 = torch.aten.mul.int %int4_7210, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7211 = torch.constant.int 4096
%9525 = torch.prim.ListConstruct %9524, %int4096_7211 : (!torch.int, !torch.int) -> !torch.list<int>
%9526 = torch.aten.view %9398, %9525 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9526, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9527 = torch.aten.mm %9526, %9485 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9527, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7212 = torch.constant.int 4
%int1792_7213 = torch.constant.int 1792
%9528 = torch.prim.ListConstruct %int4_7212, %2482, %int1792_7213 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9529 = torch.aten.view %9527, %9528 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_7214 = torch.constant.int 4
%9530 = torch.aten.mul.int %int4_7214, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7215 = torch.constant.int 4096
%9531 = torch.prim.ListConstruct %9530, %int4096_7215 : (!torch.int, !torch.int) -> !torch.list<int>
%9532 = torch.aten.view %9399, %9531 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9532, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9533 = torch.aten.mm %9532, %9487 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9533, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_7216 = torch.constant.int 4
%int1792_7217 = torch.constant.int 1792
%9534 = torch.prim.ListConstruct %int4_7216, %2482, %int1792_7217 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9535 = torch.aten.view %9533, %9534 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9536 = torch.aten.mul.Tensor %9464, %9493 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9537 = torch.aten.mul.Tensor %9465, %9499 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9538 = torch.aten.mul.Tensor %9466, %9505 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9539 = torch.aten.mul.Tensor %9467, %9511 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9540 = torch.aten.mul.Tensor %9468, %9517 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9541 = torch.aten.mul.Tensor %9469, %9523 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9541, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9542 = torch.aten.mul.Tensor %9470, %9529 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%9543 = torch.aten.mul.Tensor %9471, %9535 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %9543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_7218 = torch.constant.int 1
%int0_7219 = torch.constant.int 0
%9544 = torch.prim.ListConstruct %int1_7218, %int0_7219 : (!torch.int, !torch.int) -> !torch.list<int>
%9545 = torch.aten.permute %288, %9544 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7220 = torch.constant.int 1
%int0_7221 = torch.constant.int 0
%9546 = torch.prim.ListConstruct %int1_7220, %int0_7221 : (!torch.int, !torch.int) -> !torch.list<int>
%9547 = torch.aten.permute %289, %9546 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7222 = torch.constant.int 1
%int0_7223 = torch.constant.int 0
%9548 = torch.prim.ListConstruct %int1_7222, %int0_7223 : (!torch.int, !torch.int) -> !torch.list<int>
%9549 = torch.aten.permute %290, %9548 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7224 = torch.constant.int 1
%int0_7225 = torch.constant.int 0
%9550 = torch.prim.ListConstruct %int1_7224, %int0_7225 : (!torch.int, !torch.int) -> !torch.list<int>
%9551 = torch.aten.permute %291, %9550 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7226 = torch.constant.int 1
%int0_7227 = torch.constant.int 0
%9552 = torch.prim.ListConstruct %int1_7226, %int0_7227 : (!torch.int, !torch.int) -> !torch.list<int>
%9553 = torch.aten.permute %292, %9552 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7228 = torch.constant.int 1
%int0_7229 = torch.constant.int 0
%9554 = torch.prim.ListConstruct %int1_7228, %int0_7229 : (!torch.int, !torch.int) -> !torch.list<int>
%9555 = torch.aten.permute %293, %9554 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7230 = torch.constant.int 1
%int0_7231 = torch.constant.int 0
%9556 = torch.prim.ListConstruct %int1_7230, %int0_7231 : (!torch.int, !torch.int) -> !torch.list<int>
%9557 = torch.aten.permute %294, %9556 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7232 = torch.constant.int 1
%int0_7233 = torch.constant.int 0
%9558 = torch.prim.ListConstruct %int1_7232, %int0_7233 : (!torch.int, !torch.int) -> !torch.list<int>
%9559 = torch.aten.permute %295, %9558 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_7234 = torch.constant.int 1
%9560 = torch.aten.size.int %9421, %int1_7234 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7235 = torch.constant.int 4
%9561 = torch.aten.mul.int %int4_7235, %9560 : !torch.int, !torch.int -> !torch.int
%int1792_7236 = torch.constant.int 1792
%9562 = torch.prim.ListConstruct %9561, %int1792_7236 : (!torch.int, !torch.int) -> !torch.list<int>
%9563 = torch.aten.view %9536, %9562 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9563, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9564 = torch.aten.mm %9563, %9545 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9564, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7237 = torch.constant.int 4
%int4096_7238 = torch.constant.int 4096
%9565 = torch.prim.ListConstruct %int4_7237, %9560, %int4096_7238 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9566 = torch.aten.view %9564, %9565 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7239 = torch.constant.int 1
%9567 = torch.aten.size.int %9427, %int1_7239 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7240 = torch.constant.int 4
%9568 = torch.aten.mul.int %int4_7240, %9567 : !torch.int, !torch.int -> !torch.int
%int1792_7241 = torch.constant.int 1792
%9569 = torch.prim.ListConstruct %9568, %int1792_7241 : (!torch.int, !torch.int) -> !torch.list<int>
%9570 = torch.aten.view %9537, %9569 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9570, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9571 = torch.aten.mm %9570, %9547 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9571, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7242 = torch.constant.int 4
%int4096_7243 = torch.constant.int 4096
%9572 = torch.prim.ListConstruct %int4_7242, %9567, %int4096_7243 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9573 = torch.aten.view %9571, %9572 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7244 = torch.constant.int 1
%9574 = torch.aten.size.int %9433, %int1_7244 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7245 = torch.constant.int 4
%9575 = torch.aten.mul.int %int4_7245, %9574 : !torch.int, !torch.int -> !torch.int
%int1792_7246 = torch.constant.int 1792
%9576 = torch.prim.ListConstruct %9575, %int1792_7246 : (!torch.int, !torch.int) -> !torch.list<int>
%9577 = torch.aten.view %9538, %9576 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9577, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9578 = torch.aten.mm %9577, %9549 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9578, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7247 = torch.constant.int 4
%int4096_7248 = torch.constant.int 4096
%9579 = torch.prim.ListConstruct %int4_7247, %9574, %int4096_7248 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9580 = torch.aten.view %9578, %9579 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7249 = torch.constant.int 1
%9581 = torch.aten.size.int %9439, %int1_7249 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7250 = torch.constant.int 4
%9582 = torch.aten.mul.int %int4_7250, %9581 : !torch.int, !torch.int -> !torch.int
%int1792_7251 = torch.constant.int 1792
%9583 = torch.prim.ListConstruct %9582, %int1792_7251 : (!torch.int, !torch.int) -> !torch.list<int>
%9584 = torch.aten.view %9539, %9583 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9584, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9585 = torch.aten.mm %9584, %9551 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9585, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7252 = torch.constant.int 4
%int4096_7253 = torch.constant.int 4096
%9586 = torch.prim.ListConstruct %int4_7252, %9581, %int4096_7253 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9587 = torch.aten.view %9585, %9586 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7254 = torch.constant.int 1
%9588 = torch.aten.size.int %9445, %int1_7254 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7255 = torch.constant.int 4
%9589 = torch.aten.mul.int %int4_7255, %9588 : !torch.int, !torch.int -> !torch.int
%int1792_7256 = torch.constant.int 1792
%9590 = torch.prim.ListConstruct %9589, %int1792_7256 : (!torch.int, !torch.int) -> !torch.list<int>
%9591 = torch.aten.view %9540, %9590 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9591, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9592 = torch.aten.mm %9591, %9553 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9592, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7257 = torch.constant.int 4
%int4096_7258 = torch.constant.int 4096
%9593 = torch.prim.ListConstruct %int4_7257, %9588, %int4096_7258 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9594 = torch.aten.view %9592, %9593 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7259 = torch.constant.int 1
%9595 = torch.aten.size.int %9451, %int1_7259 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7260 = torch.constant.int 4
%9596 = torch.aten.mul.int %int4_7260, %9595 : !torch.int, !torch.int -> !torch.int
%int1792_7261 = torch.constant.int 1792
%9597 = torch.prim.ListConstruct %9596, %int1792_7261 : (!torch.int, !torch.int) -> !torch.list<int>
%9598 = torch.aten.view %9541, %9597 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9598, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9599 = torch.aten.mm %9598, %9555 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9599, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7262 = torch.constant.int 4
%int4096_7263 = torch.constant.int 4096
%9600 = torch.prim.ListConstruct %int4_7262, %9595, %int4096_7263 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9601 = torch.aten.view %9599, %9600 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7264 = torch.constant.int 1
%9602 = torch.aten.size.int %9457, %int1_7264 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7265 = torch.constant.int 4
%9603 = torch.aten.mul.int %int4_7265, %9602 : !torch.int, !torch.int -> !torch.int
%int1792_7266 = torch.constant.int 1792
%9604 = torch.prim.ListConstruct %9603, %int1792_7266 : (!torch.int, !torch.int) -> !torch.list<int>
%9605 = torch.aten.view %9542, %9604 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9605, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9606 = torch.aten.mm %9605, %9557 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9606, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7267 = torch.constant.int 4
%int4096_7268 = torch.constant.int 4096
%9607 = torch.prim.ListConstruct %int4_7267, %9602, %int4096_7268 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9608 = torch.aten.view %9606, %9607 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7269 = torch.constant.int 1
%9609 = torch.aten.size.int %9463, %int1_7269 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_7270 = torch.constant.int 4
%9610 = torch.aten.mul.int %int4_7270, %9609 : !torch.int, !torch.int -> !torch.int
%int1792_7271 = torch.constant.int 1792
%9611 = torch.prim.ListConstruct %9610, %int1792_7271 : (!torch.int, !torch.int) -> !torch.list<int>
%9612 = torch.aten.view %9543, %9611 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %9612, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%9613 = torch.aten.mm %9612, %9559 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9613, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_7272 = torch.constant.int 4
%int4096_7273 = torch.constant.int 4096
%9614 = torch.prim.ListConstruct %int4_7272, %9609, %int4096_7273 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9615 = torch.aten.view %9613, %9614 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9616 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7274 = arith.constant 1 : index
%dim_7275 = tensor.dim %9616, %c1_7274 : tensor<4x?x4096xf16>
%9617 = flow.tensor.transfer %9616 : tensor<4x?x4096xf16>{%dim_7275} to #hal.device.promise<@__device_0>
%9618 = torch_c.from_builtin_tensor %9617 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9619 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7276 = arith.constant 1 : index
%dim_7277 = tensor.dim %9619, %c1_7276 : tensor<4x?x4096xf16>
%9620 = flow.tensor.transfer %9619 : tensor<4x?x4096xf16>{%dim_7277} to #hal.device.promise<@__device_0>
%9621 = torch_c.from_builtin_tensor %9620 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9622 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7278 = arith.constant 1 : index
%dim_7279 = tensor.dim %9622, %c1_7278 : tensor<4x?x4096xf16>
%9623 = flow.tensor.transfer %9622 : tensor<4x?x4096xf16>{%dim_7279} to #hal.device.promise<@__device_0>
%9624 = torch_c.from_builtin_tensor %9623 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9625 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7280 = arith.constant 1 : index
%dim_7281 = tensor.dim %9625, %c1_7280 : tensor<4x?x4096xf16>
%9626 = flow.tensor.transfer %9625 : tensor<4x?x4096xf16>{%dim_7281} to #hal.device.promise<@__device_0>
%9627 = torch_c.from_builtin_tensor %9626 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9628 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7282 = arith.constant 1 : index
%dim_7283 = tensor.dim %9628, %c1_7282 : tensor<4x?x4096xf16>
%9629 = flow.tensor.transfer %9628 : tensor<4x?x4096xf16>{%dim_7283} to #hal.device.promise<@__device_0>
%9630 = torch_c.from_builtin_tensor %9629 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9631 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7284 = arith.constant 1 : index
%dim_7285 = tensor.dim %9631, %c1_7284 : tensor<4x?x4096xf16>
%9632 = flow.tensor.transfer %9631 : tensor<4x?x4096xf16>{%dim_7285} to #hal.device.promise<@__device_0>
%9633 = torch_c.from_builtin_tensor %9632 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9634 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7286 = arith.constant 1 : index
%dim_7287 = tensor.dim %9634, %c1_7286 : tensor<4x?x4096xf16>
%9635 = flow.tensor.transfer %9634 : tensor<4x?x4096xf16>{%dim_7287} to #hal.device.promise<@__device_0>
%9636 = torch_c.from_builtin_tensor %9635 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7288 = torch.constant.int 1
%9637 = torch.aten.add.Tensor %9566, %9618, %int1_7288 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7289 = torch.constant.int 1
%9638 = torch.aten.add.Tensor %9637, %9621, %int1_7289 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7290 = torch.constant.int 1
%9639 = torch.aten.add.Tensor %9638, %9624, %int1_7290 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7291 = torch.constant.int 1
%9640 = torch.aten.add.Tensor %9639, %9627, %int1_7291 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7292 = torch.constant.int 1
%9641 = torch.aten.add.Tensor %9640, %9630, %int1_7292 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7293 = torch.constant.int 1
%9642 = torch.aten.add.Tensor %9641, %9633, %int1_7293 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7294 = torch.constant.int 1
%9643 = torch.aten.add.Tensor %9642, %9636, %int1_7294 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9644 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7295 = arith.constant 1 : index
%dim_7296 = tensor.dim %9644, %c1_7295 : tensor<4x?x4096xf16>
%9645 = flow.tensor.transfer %9644 : tensor<4x?x4096xf16>{%dim_7296} to #hal.device.promise<@__device_1>
%9646 = torch_c.from_builtin_tensor %9645 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9647 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7297 = arith.constant 1 : index
%dim_7298 = tensor.dim %9647, %c1_7297 : tensor<4x?x4096xf16>
%9648 = flow.tensor.transfer %9647 : tensor<4x?x4096xf16>{%dim_7298} to #hal.device.promise<@__device_1>
%9649 = torch_c.from_builtin_tensor %9648 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9650 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7299 = arith.constant 1 : index
%dim_7300 = tensor.dim %9650, %c1_7299 : tensor<4x?x4096xf16>
%9651 = flow.tensor.transfer %9650 : tensor<4x?x4096xf16>{%dim_7300} to #hal.device.promise<@__device_1>
%9652 = torch_c.from_builtin_tensor %9651 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9653 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7301 = arith.constant 1 : index
%dim_7302 = tensor.dim %9653, %c1_7301 : tensor<4x?x4096xf16>
%9654 = flow.tensor.transfer %9653 : tensor<4x?x4096xf16>{%dim_7302} to #hal.device.promise<@__device_1>
%9655 = torch_c.from_builtin_tensor %9654 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9656 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7303 = arith.constant 1 : index
%dim_7304 = tensor.dim %9656, %c1_7303 : tensor<4x?x4096xf16>
%9657 = flow.tensor.transfer %9656 : tensor<4x?x4096xf16>{%dim_7304} to #hal.device.promise<@__device_1>
%9658 = torch_c.from_builtin_tensor %9657 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9659 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7305 = arith.constant 1 : index
%dim_7306 = tensor.dim %9659, %c1_7305 : tensor<4x?x4096xf16>
%9660 = flow.tensor.transfer %9659 : tensor<4x?x4096xf16>{%dim_7306} to #hal.device.promise<@__device_1>
%9661 = torch_c.from_builtin_tensor %9660 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9662 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7307 = arith.constant 1 : index
%dim_7308 = tensor.dim %9662, %c1_7307 : tensor<4x?x4096xf16>
%9663 = flow.tensor.transfer %9662 : tensor<4x?x4096xf16>{%dim_7308} to #hal.device.promise<@__device_1>
%9664 = torch_c.from_builtin_tensor %9663 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7309 = torch.constant.int 1
%9665 = torch.aten.add.Tensor %9646, %9573, %int1_7309 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7310 = torch.constant.int 1
%9666 = torch.aten.add.Tensor %9665, %9649, %int1_7310 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7311 = torch.constant.int 1
%9667 = torch.aten.add.Tensor %9666, %9652, %int1_7311 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7312 = torch.constant.int 1
%9668 = torch.aten.add.Tensor %9667, %9655, %int1_7312 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7313 = torch.constant.int 1
%9669 = torch.aten.add.Tensor %9668, %9658, %int1_7313 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7314 = torch.constant.int 1
%9670 = torch.aten.add.Tensor %9669, %9661, %int1_7314 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7315 = torch.constant.int 1
%9671 = torch.aten.add.Tensor %9670, %9664, %int1_7315 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9672 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7316 = arith.constant 1 : index
%dim_7317 = tensor.dim %9672, %c1_7316 : tensor<4x?x4096xf16>
%9673 = flow.tensor.transfer %9672 : tensor<4x?x4096xf16>{%dim_7317} to #hal.device.promise<@__device_2>
%9674 = torch_c.from_builtin_tensor %9673 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9675 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7318 = arith.constant 1 : index
%dim_7319 = tensor.dim %9675, %c1_7318 : tensor<4x?x4096xf16>
%9676 = flow.tensor.transfer %9675 : tensor<4x?x4096xf16>{%dim_7319} to #hal.device.promise<@__device_2>
%9677 = torch_c.from_builtin_tensor %9676 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9678 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7320 = arith.constant 1 : index
%dim_7321 = tensor.dim %9678, %c1_7320 : tensor<4x?x4096xf16>
%9679 = flow.tensor.transfer %9678 : tensor<4x?x4096xf16>{%dim_7321} to #hal.device.promise<@__device_2>
%9680 = torch_c.from_builtin_tensor %9679 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9681 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7322 = arith.constant 1 : index
%dim_7323 = tensor.dim %9681, %c1_7322 : tensor<4x?x4096xf16>
%9682 = flow.tensor.transfer %9681 : tensor<4x?x4096xf16>{%dim_7323} to #hal.device.promise<@__device_2>
%9683 = torch_c.from_builtin_tensor %9682 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9684 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7324 = arith.constant 1 : index
%dim_7325 = tensor.dim %9684, %c1_7324 : tensor<4x?x4096xf16>
%9685 = flow.tensor.transfer %9684 : tensor<4x?x4096xf16>{%dim_7325} to #hal.device.promise<@__device_2>
%9686 = torch_c.from_builtin_tensor %9685 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9687 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7326 = arith.constant 1 : index
%dim_7327 = tensor.dim %9687, %c1_7326 : tensor<4x?x4096xf16>
%9688 = flow.tensor.transfer %9687 : tensor<4x?x4096xf16>{%dim_7327} to #hal.device.promise<@__device_2>
%9689 = torch_c.from_builtin_tensor %9688 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9690 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7328 = arith.constant 1 : index
%dim_7329 = tensor.dim %9690, %c1_7328 : tensor<4x?x4096xf16>
%9691 = flow.tensor.transfer %9690 : tensor<4x?x4096xf16>{%dim_7329} to #hal.device.promise<@__device_2>
%9692 = torch_c.from_builtin_tensor %9691 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7330 = torch.constant.int 1
%9693 = torch.aten.add.Tensor %9674, %9677, %int1_7330 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7331 = torch.constant.int 1
%9694 = torch.aten.add.Tensor %9693, %9580, %int1_7331 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7332 = torch.constant.int 1
%9695 = torch.aten.add.Tensor %9694, %9680, %int1_7332 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7333 = torch.constant.int 1
%9696 = torch.aten.add.Tensor %9695, %9683, %int1_7333 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7334 = torch.constant.int 1
%9697 = torch.aten.add.Tensor %9696, %9686, %int1_7334 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7335 = torch.constant.int 1
%9698 = torch.aten.add.Tensor %9697, %9689, %int1_7335 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7336 = torch.constant.int 1
%9699 = torch.aten.add.Tensor %9698, %9692, %int1_7336 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9700 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7337 = arith.constant 1 : index
%dim_7338 = tensor.dim %9700, %c1_7337 : tensor<4x?x4096xf16>
%9701 = flow.tensor.transfer %9700 : tensor<4x?x4096xf16>{%dim_7338} to #hal.device.promise<@__device_3>
%9702 = torch_c.from_builtin_tensor %9701 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9703 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7339 = arith.constant 1 : index
%dim_7340 = tensor.dim %9703, %c1_7339 : tensor<4x?x4096xf16>
%9704 = flow.tensor.transfer %9703 : tensor<4x?x4096xf16>{%dim_7340} to #hal.device.promise<@__device_3>
%9705 = torch_c.from_builtin_tensor %9704 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9706 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7341 = arith.constant 1 : index
%dim_7342 = tensor.dim %9706, %c1_7341 : tensor<4x?x4096xf16>
%9707 = flow.tensor.transfer %9706 : tensor<4x?x4096xf16>{%dim_7342} to #hal.device.promise<@__device_3>
%9708 = torch_c.from_builtin_tensor %9707 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9709 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7343 = arith.constant 1 : index
%dim_7344 = tensor.dim %9709, %c1_7343 : tensor<4x?x4096xf16>
%9710 = flow.tensor.transfer %9709 : tensor<4x?x4096xf16>{%dim_7344} to #hal.device.promise<@__device_3>
%9711 = torch_c.from_builtin_tensor %9710 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9712 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7345 = arith.constant 1 : index
%dim_7346 = tensor.dim %9712, %c1_7345 : tensor<4x?x4096xf16>
%9713 = flow.tensor.transfer %9712 : tensor<4x?x4096xf16>{%dim_7346} to #hal.device.promise<@__device_3>
%9714 = torch_c.from_builtin_tensor %9713 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9715 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7347 = arith.constant 1 : index
%dim_7348 = tensor.dim %9715, %c1_7347 : tensor<4x?x4096xf16>
%9716 = flow.tensor.transfer %9715 : tensor<4x?x4096xf16>{%dim_7348} to #hal.device.promise<@__device_3>
%9717 = torch_c.from_builtin_tensor %9716 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9718 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7349 = arith.constant 1 : index
%dim_7350 = tensor.dim %9718, %c1_7349 : tensor<4x?x4096xf16>
%9719 = flow.tensor.transfer %9718 : tensor<4x?x4096xf16>{%dim_7350} to #hal.device.promise<@__device_3>
%9720 = torch_c.from_builtin_tensor %9719 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7351 = torch.constant.int 1
%9721 = torch.aten.add.Tensor %9702, %9705, %int1_7351 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7352 = torch.constant.int 1
%9722 = torch.aten.add.Tensor %9721, %9708, %int1_7352 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7353 = torch.constant.int 1
%9723 = torch.aten.add.Tensor %9722, %9587, %int1_7353 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7354 = torch.constant.int 1
%9724 = torch.aten.add.Tensor %9723, %9711, %int1_7354 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7355 = torch.constant.int 1
%9725 = torch.aten.add.Tensor %9724, %9714, %int1_7355 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7356 = torch.constant.int 1
%9726 = torch.aten.add.Tensor %9725, %9717, %int1_7356 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7357 = torch.constant.int 1
%9727 = torch.aten.add.Tensor %9726, %9720, %int1_7357 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9728 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7358 = arith.constant 1 : index
%dim_7359 = tensor.dim %9728, %c1_7358 : tensor<4x?x4096xf16>
%9729 = flow.tensor.transfer %9728 : tensor<4x?x4096xf16>{%dim_7359} to #hal.device.promise<@__device_4>
%9730 = torch_c.from_builtin_tensor %9729 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9731 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7360 = arith.constant 1 : index
%dim_7361 = tensor.dim %9731, %c1_7360 : tensor<4x?x4096xf16>
%9732 = flow.tensor.transfer %9731 : tensor<4x?x4096xf16>{%dim_7361} to #hal.device.promise<@__device_4>
%9733 = torch_c.from_builtin_tensor %9732 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9734 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7362 = arith.constant 1 : index
%dim_7363 = tensor.dim %9734, %c1_7362 : tensor<4x?x4096xf16>
%9735 = flow.tensor.transfer %9734 : tensor<4x?x4096xf16>{%dim_7363} to #hal.device.promise<@__device_4>
%9736 = torch_c.from_builtin_tensor %9735 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9737 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7364 = arith.constant 1 : index
%dim_7365 = tensor.dim %9737, %c1_7364 : tensor<4x?x4096xf16>
%9738 = flow.tensor.transfer %9737 : tensor<4x?x4096xf16>{%dim_7365} to #hal.device.promise<@__device_4>
%9739 = torch_c.from_builtin_tensor %9738 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9740 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7366 = arith.constant 1 : index
%dim_7367 = tensor.dim %9740, %c1_7366 : tensor<4x?x4096xf16>
%9741 = flow.tensor.transfer %9740 : tensor<4x?x4096xf16>{%dim_7367} to #hal.device.promise<@__device_4>
%9742 = torch_c.from_builtin_tensor %9741 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9743 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7368 = arith.constant 1 : index
%dim_7369 = tensor.dim %9743, %c1_7368 : tensor<4x?x4096xf16>
%9744 = flow.tensor.transfer %9743 : tensor<4x?x4096xf16>{%dim_7369} to #hal.device.promise<@__device_4>
%9745 = torch_c.from_builtin_tensor %9744 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9746 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7370 = arith.constant 1 : index
%dim_7371 = tensor.dim %9746, %c1_7370 : tensor<4x?x4096xf16>
%9747 = flow.tensor.transfer %9746 : tensor<4x?x4096xf16>{%dim_7371} to #hal.device.promise<@__device_4>
%9748 = torch_c.from_builtin_tensor %9747 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7372 = torch.constant.int 1
%9749 = torch.aten.add.Tensor %9730, %9733, %int1_7372 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7373 = torch.constant.int 1
%9750 = torch.aten.add.Tensor %9749, %9736, %int1_7373 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7374 = torch.constant.int 1
%9751 = torch.aten.add.Tensor %9750, %9739, %int1_7374 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7375 = torch.constant.int 1
%9752 = torch.aten.add.Tensor %9751, %9594, %int1_7375 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7376 = torch.constant.int 1
%9753 = torch.aten.add.Tensor %9752, %9742, %int1_7376 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7377 = torch.constant.int 1
%9754 = torch.aten.add.Tensor %9753, %9745, %int1_7377 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7378 = torch.constant.int 1
%9755 = torch.aten.add.Tensor %9754, %9748, %int1_7378 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9756 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7379 = arith.constant 1 : index
%dim_7380 = tensor.dim %9756, %c1_7379 : tensor<4x?x4096xf16>
%9757 = flow.tensor.transfer %9756 : tensor<4x?x4096xf16>{%dim_7380} to #hal.device.promise<@__device_5>
%9758 = torch_c.from_builtin_tensor %9757 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9759 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7381 = arith.constant 1 : index
%dim_7382 = tensor.dim %9759, %c1_7381 : tensor<4x?x4096xf16>
%9760 = flow.tensor.transfer %9759 : tensor<4x?x4096xf16>{%dim_7382} to #hal.device.promise<@__device_5>
%9761 = torch_c.from_builtin_tensor %9760 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9762 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7383 = arith.constant 1 : index
%dim_7384 = tensor.dim %9762, %c1_7383 : tensor<4x?x4096xf16>
%9763 = flow.tensor.transfer %9762 : tensor<4x?x4096xf16>{%dim_7384} to #hal.device.promise<@__device_5>
%9764 = torch_c.from_builtin_tensor %9763 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9765 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7385 = arith.constant 1 : index
%dim_7386 = tensor.dim %9765, %c1_7385 : tensor<4x?x4096xf16>
%9766 = flow.tensor.transfer %9765 : tensor<4x?x4096xf16>{%dim_7386} to #hal.device.promise<@__device_5>
%9767 = torch_c.from_builtin_tensor %9766 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9768 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7387 = arith.constant 1 : index
%dim_7388 = tensor.dim %9768, %c1_7387 : tensor<4x?x4096xf16>
%9769 = flow.tensor.transfer %9768 : tensor<4x?x4096xf16>{%dim_7388} to #hal.device.promise<@__device_5>
%9770 = torch_c.from_builtin_tensor %9769 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9771 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7389 = arith.constant 1 : index
%dim_7390 = tensor.dim %9771, %c1_7389 : tensor<4x?x4096xf16>
%9772 = flow.tensor.transfer %9771 : tensor<4x?x4096xf16>{%dim_7390} to #hal.device.promise<@__device_5>
%9773 = torch_c.from_builtin_tensor %9772 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9774 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7391 = arith.constant 1 : index
%dim_7392 = tensor.dim %9774, %c1_7391 : tensor<4x?x4096xf16>
%9775 = flow.tensor.transfer %9774 : tensor<4x?x4096xf16>{%dim_7392} to #hal.device.promise<@__device_5>
%9776 = torch_c.from_builtin_tensor %9775 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7393 = torch.constant.int 1
%9777 = torch.aten.add.Tensor %9758, %9761, %int1_7393 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7394 = torch.constant.int 1
%9778 = torch.aten.add.Tensor %9777, %9764, %int1_7394 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7395 = torch.constant.int 1
%9779 = torch.aten.add.Tensor %9778, %9767, %int1_7395 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7396 = torch.constant.int 1
%9780 = torch.aten.add.Tensor %9779, %9770, %int1_7396 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7397 = torch.constant.int 1
%9781 = torch.aten.add.Tensor %9780, %9601, %int1_7397 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7398 = torch.constant.int 1
%9782 = torch.aten.add.Tensor %9781, %9773, %int1_7398 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7399 = torch.constant.int 1
%9783 = torch.aten.add.Tensor %9782, %9776, %int1_7399 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9784 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7400 = arith.constant 1 : index
%dim_7401 = tensor.dim %9784, %c1_7400 : tensor<4x?x4096xf16>
%9785 = flow.tensor.transfer %9784 : tensor<4x?x4096xf16>{%dim_7401} to #hal.device.promise<@__device_6>
%9786 = torch_c.from_builtin_tensor %9785 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9787 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7402 = arith.constant 1 : index
%dim_7403 = tensor.dim %9787, %c1_7402 : tensor<4x?x4096xf16>
%9788 = flow.tensor.transfer %9787 : tensor<4x?x4096xf16>{%dim_7403} to #hal.device.promise<@__device_6>
%9789 = torch_c.from_builtin_tensor %9788 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9790 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7404 = arith.constant 1 : index
%dim_7405 = tensor.dim %9790, %c1_7404 : tensor<4x?x4096xf16>
%9791 = flow.tensor.transfer %9790 : tensor<4x?x4096xf16>{%dim_7405} to #hal.device.promise<@__device_6>
%9792 = torch_c.from_builtin_tensor %9791 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9793 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7406 = arith.constant 1 : index
%dim_7407 = tensor.dim %9793, %c1_7406 : tensor<4x?x4096xf16>
%9794 = flow.tensor.transfer %9793 : tensor<4x?x4096xf16>{%dim_7407} to #hal.device.promise<@__device_6>
%9795 = torch_c.from_builtin_tensor %9794 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9796 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7408 = arith.constant 1 : index
%dim_7409 = tensor.dim %9796, %c1_7408 : tensor<4x?x4096xf16>
%9797 = flow.tensor.transfer %9796 : tensor<4x?x4096xf16>{%dim_7409} to #hal.device.promise<@__device_6>
%9798 = torch_c.from_builtin_tensor %9797 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9799 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7410 = arith.constant 1 : index
%dim_7411 = tensor.dim %9799, %c1_7410 : tensor<4x?x4096xf16>
%9800 = flow.tensor.transfer %9799 : tensor<4x?x4096xf16>{%dim_7411} to #hal.device.promise<@__device_6>
%9801 = torch_c.from_builtin_tensor %9800 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9802 = torch_c.to_builtin_tensor %9615 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7412 = arith.constant 1 : index
%dim_7413 = tensor.dim %9802, %c1_7412 : tensor<4x?x4096xf16>
%9803 = flow.tensor.transfer %9802 : tensor<4x?x4096xf16>{%dim_7413} to #hal.device.promise<@__device_6>
%9804 = torch_c.from_builtin_tensor %9803 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7414 = torch.constant.int 1
%9805 = torch.aten.add.Tensor %9786, %9789, %int1_7414 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7415 = torch.constant.int 1
%9806 = torch.aten.add.Tensor %9805, %9792, %int1_7415 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7416 = torch.constant.int 1
%9807 = torch.aten.add.Tensor %9806, %9795, %int1_7416 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7417 = torch.constant.int 1
%9808 = torch.aten.add.Tensor %9807, %9798, %int1_7417 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7418 = torch.constant.int 1
%9809 = torch.aten.add.Tensor %9808, %9801, %int1_7418 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7419 = torch.constant.int 1
%9810 = torch.aten.add.Tensor %9809, %9608, %int1_7419 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7420 = torch.constant.int 1
%9811 = torch.aten.add.Tensor %9810, %9804, %int1_7420 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9812 = torch_c.to_builtin_tensor %9566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7421 = arith.constant 1 : index
%dim_7422 = tensor.dim %9812, %c1_7421 : tensor<4x?x4096xf16>
%9813 = flow.tensor.transfer %9812 : tensor<4x?x4096xf16>{%dim_7422} to #hal.device.promise<@__device_7>
%9814 = torch_c.from_builtin_tensor %9813 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9815 = torch_c.to_builtin_tensor %9573 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7423 = arith.constant 1 : index
%dim_7424 = tensor.dim %9815, %c1_7423 : tensor<4x?x4096xf16>
%9816 = flow.tensor.transfer %9815 : tensor<4x?x4096xf16>{%dim_7424} to #hal.device.promise<@__device_7>
%9817 = torch_c.from_builtin_tensor %9816 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9818 = torch_c.to_builtin_tensor %9580 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7425 = arith.constant 1 : index
%dim_7426 = tensor.dim %9818, %c1_7425 : tensor<4x?x4096xf16>
%9819 = flow.tensor.transfer %9818 : tensor<4x?x4096xf16>{%dim_7426} to #hal.device.promise<@__device_7>
%9820 = torch_c.from_builtin_tensor %9819 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9821 = torch_c.to_builtin_tensor %9587 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7427 = arith.constant 1 : index
%dim_7428 = tensor.dim %9821, %c1_7427 : tensor<4x?x4096xf16>
%9822 = flow.tensor.transfer %9821 : tensor<4x?x4096xf16>{%dim_7428} to #hal.device.promise<@__device_7>
%9823 = torch_c.from_builtin_tensor %9822 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9824 = torch_c.to_builtin_tensor %9594 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7429 = arith.constant 1 : index
%dim_7430 = tensor.dim %9824, %c1_7429 : tensor<4x?x4096xf16>
%9825 = flow.tensor.transfer %9824 : tensor<4x?x4096xf16>{%dim_7430} to #hal.device.promise<@__device_7>
%9826 = torch_c.from_builtin_tensor %9825 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9827 = torch_c.to_builtin_tensor %9601 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7431 = arith.constant 1 : index
%dim_7432 = tensor.dim %9827, %c1_7431 : tensor<4x?x4096xf16>
%9828 = flow.tensor.transfer %9827 : tensor<4x?x4096xf16>{%dim_7432} to #hal.device.promise<@__device_7>
%9829 = torch_c.from_builtin_tensor %9828 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%9830 = torch_c.to_builtin_tensor %9608 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_7433 = arith.constant 1 : index
%dim_7434 = tensor.dim %9830, %c1_7433 : tensor<4x?x4096xf16>
%9831 = flow.tensor.transfer %9830 : tensor<4x?x4096xf16>{%dim_7434} to #hal.device.promise<@__device_7>
%9832 = torch_c.from_builtin_tensor %9831 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7435 = torch.constant.int 1
%9833 = torch.aten.add.Tensor %9814, %9817, %int1_7435 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7436 = torch.constant.int 1
%9834 = torch.aten.add.Tensor %9833, %9820, %int1_7436 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7437 = torch.constant.int 1
%9835 = torch.aten.add.Tensor %9834, %9823, %int1_7437 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7438 = torch.constant.int 1
%9836 = torch.aten.add.Tensor %9835, %9826, %int1_7438 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7439 = torch.constant.int 1
%9837 = torch.aten.add.Tensor %9836, %9829, %int1_7439 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7440 = torch.constant.int 1
%9838 = torch.aten.add.Tensor %9837, %9832, %int1_7440 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7441 = torch.constant.int 1
%9839 = torch.aten.add.Tensor %9838, %9615, %int1_7441 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7442 = torch.constant.int 1
%9840 = torch.aten.add.Tensor %9320, %9643, %int1_7442 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7443 = torch.constant.int 1
%9841 = torch.aten.add.Tensor %9321, %9671, %int1_7443 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7444 = torch.constant.int 1
%9842 = torch.aten.add.Tensor %9322, %9699, %int1_7444 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7445 = torch.constant.int 1
%9843 = torch.aten.add.Tensor %9323, %9727, %int1_7445 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7446 = torch.constant.int 1
%9844 = torch.aten.add.Tensor %9324, %9755, %int1_7446 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7447 = torch.constant.int 1
%9845 = torch.aten.add.Tensor %9325, %9783, %int1_7447 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7448 = torch.constant.int 1
%9846 = torch.aten.add.Tensor %9326, %9811, %int1_7448 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7449 = torch.constant.int 1
%9847 = torch.aten.add.Tensor %9327, %9839, %int1_7449 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_7450 = torch.constant.int 6
%9848 = torch.prims.convert_element_type %9840, %int6_7450 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7451 = torch.constant.int 6
%9849 = torch.prims.convert_element_type %9841, %int6_7451 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7452 = torch.constant.int 6
%9850 = torch.prims.convert_element_type %9842, %int6_7452 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7453 = torch.constant.int 6
%9851 = torch.prims.convert_element_type %9843, %int6_7453 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7454 = torch.constant.int 6
%9852 = torch.prims.convert_element_type %9844, %int6_7454 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7455 = torch.constant.int 6
%9853 = torch.prims.convert_element_type %9845, %int6_7455 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7456 = torch.constant.int 6
%9854 = torch.prims.convert_element_type %9846, %int6_7456 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_7457 = torch.constant.int 6
%9855 = torch.prims.convert_element_type %9847, %int6_7457 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7458 = torch.constant.int 2
%9856 = torch.aten.pow.Tensor_Scalar %9848, %int2_7458 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7459 = torch.constant.int 2
%9857 = torch.aten.pow.Tensor_Scalar %9849, %int2_7459 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7460 = torch.constant.int 2
%9858 = torch.aten.pow.Tensor_Scalar %9850, %int2_7460 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7461 = torch.constant.int 2
%9859 = torch.aten.pow.Tensor_Scalar %9851, %int2_7461 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7462 = torch.constant.int 2
%9860 = torch.aten.pow.Tensor_Scalar %9852, %int2_7462 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7463 = torch.constant.int 2
%9861 = torch.aten.pow.Tensor_Scalar %9853, %int2_7463 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7464 = torch.constant.int 2
%9862 = torch.aten.pow.Tensor_Scalar %9854, %int2_7464 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_7465 = torch.constant.int 2
%9863 = torch.aten.pow.Tensor_Scalar %9855, %int2_7465 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_7466 = torch.constant.int -1
%9864 = torch.prim.ListConstruct %int-1_7466 : (!torch.int) -> !torch.list<int>
%true_7467 = torch.constant.bool true
%none_7468 = torch.constant.none
%9865 = torch.aten.mean.dim %9856, %9864, %true_7467, %none_7468 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7469 = torch.constant.int -1
%9866 = torch.prim.ListConstruct %int-1_7469 : (!torch.int) -> !torch.list<int>
%true_7470 = torch.constant.bool true
%none_7471 = torch.constant.none
%9867 = torch.aten.mean.dim %9857, %9866, %true_7470, %none_7471 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7472 = torch.constant.int -1
%9868 = torch.prim.ListConstruct %int-1_7472 : (!torch.int) -> !torch.list<int>
%true_7473 = torch.constant.bool true
%none_7474 = torch.constant.none
%9869 = torch.aten.mean.dim %9858, %9868, %true_7473, %none_7474 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7475 = torch.constant.int -1
%9870 = torch.prim.ListConstruct %int-1_7475 : (!torch.int) -> !torch.list<int>
%true_7476 = torch.constant.bool true
%none_7477 = torch.constant.none
%9871 = torch.aten.mean.dim %9859, %9870, %true_7476, %none_7477 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7478 = torch.constant.int -1
%9872 = torch.prim.ListConstruct %int-1_7478 : (!torch.int) -> !torch.list<int>
%true_7479 = torch.constant.bool true
%none_7480 = torch.constant.none
%9873 = torch.aten.mean.dim %9860, %9872, %true_7479, %none_7480 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7481 = torch.constant.int -1
%9874 = torch.prim.ListConstruct %int-1_7481 : (!torch.int) -> !torch.list<int>
%true_7482 = torch.constant.bool true
%none_7483 = torch.constant.none
%9875 = torch.aten.mean.dim %9861, %9874, %true_7482, %none_7483 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7484 = torch.constant.int -1
%9876 = torch.prim.ListConstruct %int-1_7484 : (!torch.int) -> !torch.list<int>
%true_7485 = torch.constant.bool true
%none_7486 = torch.constant.none
%9877 = torch.aten.mean.dim %9862, %9876, %true_7485, %none_7486 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_7487 = torch.constant.int -1
%9878 = torch.prim.ListConstruct %int-1_7487 : (!torch.int) -> !torch.list<int>
%true_7488 = torch.constant.bool true
%none_7489 = torch.constant.none
%9879 = torch.aten.mean.dim %9863, %9878, %true_7488, %none_7489 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7490 = torch.constant.float 9.9999997473787516E-6
%int1_7491 = torch.constant.int 1
%9880 = torch.aten.add.Scalar %9865, %float9.999990e-06_7490, %int1_7491 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7492 = torch.constant.float 9.9999997473787516E-6
%int1_7493 = torch.constant.int 1
%9881 = torch.aten.add.Scalar %9867, %float9.999990e-06_7492, %int1_7493 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7494 = torch.constant.float 9.9999997473787516E-6
%int1_7495 = torch.constant.int 1
%9882 = torch.aten.add.Scalar %9869, %float9.999990e-06_7494, %int1_7495 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7496 = torch.constant.float 9.9999997473787516E-6
%int1_7497 = torch.constant.int 1
%9883 = torch.aten.add.Scalar %9871, %float9.999990e-06_7496, %int1_7497 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7498 = torch.constant.float 9.9999997473787516E-6
%int1_7499 = torch.constant.int 1
%9884 = torch.aten.add.Scalar %9873, %float9.999990e-06_7498, %int1_7499 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7500 = torch.constant.float 9.9999997473787516E-6
%int1_7501 = torch.constant.int 1
%9885 = torch.aten.add.Scalar %9875, %float9.999990e-06_7500, %int1_7501 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7502 = torch.constant.float 9.9999997473787516E-6
%int1_7503 = torch.constant.int 1
%9886 = torch.aten.add.Scalar %9877, %float9.999990e-06_7502, %int1_7503 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_7504 = torch.constant.float 9.9999997473787516E-6
%int1_7505 = torch.constant.int 1
%9887 = torch.aten.add.Scalar %9879, %float9.999990e-06_7504, %int1_7505 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9888 = torch.aten.rsqrt %9880 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9889 = torch.aten.rsqrt %9881 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9890 = torch.aten.rsqrt %9882 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9891 = torch.aten.rsqrt %9883 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9892 = torch.aten.rsqrt %9884 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9893 = torch.aten.rsqrt %9885 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9894 = torch.aten.rsqrt %9886 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9895 = torch.aten.rsqrt %9887 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %9895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%9896 = torch.aten.mul.Tensor %9848, %9888 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9897 = torch.aten.mul.Tensor %9849, %9889 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9898 = torch.aten.mul.Tensor %9850, %9890 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9899 = torch.aten.mul.Tensor %9851, %9891 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9900 = torch.aten.mul.Tensor %9852, %9892 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9901 = torch.aten.mul.Tensor %9853, %9893 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9902 = torch.aten.mul.Tensor %9854, %9894 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9903 = torch.aten.mul.Tensor %9855, %9895 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9904 = torch.aten.mul.Tensor %296, %9896 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9905 = torch.aten.mul.Tensor %297, %9897 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9906 = torch.aten.mul.Tensor %298, %9898 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9907 = torch.aten.mul.Tensor %299, %9899 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9908 = torch.aten.mul.Tensor %300, %9900 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9909 = torch.aten.mul.Tensor %301, %9901 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9910 = torch.aten.mul.Tensor %302, %9902 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%9911 = torch.aten.mul.Tensor %303, %9903 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %9911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_7506 = torch.constant.int 5
%9912 = torch.prims.convert_element_type %9904, %int5_7506 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7507 = torch.constant.int 5
%9913 = torch.prims.convert_element_type %9905, %int5_7507 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7508 = torch.constant.int 5
%9914 = torch.prims.convert_element_type %9906, %int5_7508 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7509 = torch.constant.int 5
%9915 = torch.prims.convert_element_type %9907, %int5_7509 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7510 = torch.constant.int 5
%9916 = torch.prims.convert_element_type %9908, %int5_7510 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7511 = torch.constant.int 5
%9917 = torch.prims.convert_element_type %9909, %int5_7511 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7512 = torch.constant.int 5
%9918 = torch.prims.convert_element_type %9910, %int5_7512 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_7513 = torch.constant.int 5
%9919 = torch.prims.convert_element_type %9911, %int5_7513 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %9919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_7514 = torch.constant.int 1
%int0_7515 = torch.constant.int 0
%9920 = torch.prim.ListConstruct %int1_7514, %int0_7515 : (!torch.int, !torch.int) -> !torch.list<int>
%9921 = torch.aten.permute %304, %9920 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7516 = torch.constant.int 1
%int0_7517 = torch.constant.int 0
%9922 = torch.prim.ListConstruct %int1_7516, %int0_7517 : (!torch.int, !torch.int) -> !torch.list<int>
%9923 = torch.aten.permute %305, %9922 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7518 = torch.constant.int 1
%int0_7519 = torch.constant.int 0
%9924 = torch.prim.ListConstruct %int1_7518, %int0_7519 : (!torch.int, !torch.int) -> !torch.list<int>
%9925 = torch.aten.permute %306, %9924 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7520 = torch.constant.int 1
%int0_7521 = torch.constant.int 0
%9926 = torch.prim.ListConstruct %int1_7520, %int0_7521 : (!torch.int, !torch.int) -> !torch.list<int>
%9927 = torch.aten.permute %307, %9926 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7522 = torch.constant.int 1
%int0_7523 = torch.constant.int 0
%9928 = torch.prim.ListConstruct %int1_7522, %int0_7523 : (!torch.int, !torch.int) -> !torch.list<int>
%9929 = torch.aten.permute %308, %9928 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7524 = torch.constant.int 1
%int0_7525 = torch.constant.int 0
%9930 = torch.prim.ListConstruct %int1_7524, %int0_7525 : (!torch.int, !torch.int) -> !torch.list<int>
%9931 = torch.aten.permute %309, %9930 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7526 = torch.constant.int 1
%int0_7527 = torch.constant.int 0
%9932 = torch.prim.ListConstruct %int1_7526, %int0_7527 : (!torch.int, !torch.int) -> !torch.list<int>
%9933 = torch.aten.permute %310, %9932 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_7528 = torch.constant.int 1
%int0_7529 = torch.constant.int 0
%9934 = torch.prim.ListConstruct %int1_7528, %int0_7529 : (!torch.int, !torch.int) -> !torch.list<int>
%9935 = torch.aten.permute %311, %9934 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_7530 = torch.constant.int 4
%9936 = torch.aten.mul.int %int4_7530, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7531 = torch.constant.int 4096
%9937 = torch.prim.ListConstruct %9936, %int4096_7531 : (!torch.int, !torch.int) -> !torch.list<int>
%9938 = torch.aten.view %9912, %9937 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9938, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9939 = torch.aten.mm %9938, %9921 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9939, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7532 = torch.constant.int 4
%int512_7533 = torch.constant.int 512
%9940 = torch.prim.ListConstruct %int4_7532, %2482, %int512_7533 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9941 = torch.aten.view %9939, %9940 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7534 = torch.constant.int 4
%9942 = torch.aten.mul.int %int4_7534, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7535 = torch.constant.int 4096
%9943 = torch.prim.ListConstruct %9942, %int4096_7535 : (!torch.int, !torch.int) -> !torch.list<int>
%9944 = torch.aten.view %9913, %9943 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9944, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9945 = torch.aten.mm %9944, %9923 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9945, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7536 = torch.constant.int 4
%int512_7537 = torch.constant.int 512
%9946 = torch.prim.ListConstruct %int4_7536, %2482, %int512_7537 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9947 = torch.aten.view %9945, %9946 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7538 = torch.constant.int 4
%9948 = torch.aten.mul.int %int4_7538, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7539 = torch.constant.int 4096
%9949 = torch.prim.ListConstruct %9948, %int4096_7539 : (!torch.int, !torch.int) -> !torch.list<int>
%9950 = torch.aten.view %9914, %9949 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9950, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9951 = torch.aten.mm %9950, %9925 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9951, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7540 = torch.constant.int 4
%int512_7541 = torch.constant.int 512
%9952 = torch.prim.ListConstruct %int4_7540, %2482, %int512_7541 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9953 = torch.aten.view %9951, %9952 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7542 = torch.constant.int 4
%9954 = torch.aten.mul.int %int4_7542, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7543 = torch.constant.int 4096
%9955 = torch.prim.ListConstruct %9954, %int4096_7543 : (!torch.int, !torch.int) -> !torch.list<int>
%9956 = torch.aten.view %9915, %9955 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9956, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9957 = torch.aten.mm %9956, %9927 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9957, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7544 = torch.constant.int 4
%int512_7545 = torch.constant.int 512
%9958 = torch.prim.ListConstruct %int4_7544, %2482, %int512_7545 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9959 = torch.aten.view %9957, %9958 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7546 = torch.constant.int 4
%9960 = torch.aten.mul.int %int4_7546, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7547 = torch.constant.int 4096
%9961 = torch.prim.ListConstruct %9960, %int4096_7547 : (!torch.int, !torch.int) -> !torch.list<int>
%9962 = torch.aten.view %9916, %9961 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9962, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9963 = torch.aten.mm %9962, %9929 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9963, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7548 = torch.constant.int 4
%int512_7549 = torch.constant.int 512
%9964 = torch.prim.ListConstruct %int4_7548, %2482, %int512_7549 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9965 = torch.aten.view %9963, %9964 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7550 = torch.constant.int 4
%9966 = torch.aten.mul.int %int4_7550, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7551 = torch.constant.int 4096
%9967 = torch.prim.ListConstruct %9966, %int4096_7551 : (!torch.int, !torch.int) -> !torch.list<int>
%9968 = torch.aten.view %9917, %9967 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9968, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9969 = torch.aten.mm %9968, %9931 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9969, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7552 = torch.constant.int 4
%int512_7553 = torch.constant.int 512
%9970 = torch.prim.ListConstruct %int4_7552, %2482, %int512_7553 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9971 = torch.aten.view %9969, %9970 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7554 = torch.constant.int 4
%9972 = torch.aten.mul.int %int4_7554, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7555 = torch.constant.int 4096
%9973 = torch.prim.ListConstruct %9972, %int4096_7555 : (!torch.int, !torch.int) -> !torch.list<int>
%9974 = torch.aten.view %9918, %9973 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9974, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9975 = torch.aten.mm %9974, %9933 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9975, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7556 = torch.constant.int 4
%int512_7557 = torch.constant.int 512
%9976 = torch.prim.ListConstruct %int4_7556, %2482, %int512_7557 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9977 = torch.aten.view %9975, %9976 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_7558 = torch.constant.int 4
%9978 = torch.aten.mul.int %int4_7558, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7559 = torch.constant.int 4096
%9979 = torch.prim.ListConstruct %9978, %int4096_7559 : (!torch.int, !torch.int) -> !torch.list<int>
%9980 = torch.aten.view %9919, %9979 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %9980, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%9981 = torch.aten.mm %9980, %9935 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %9981, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_7560 = torch.constant.int 4
%int512_7561 = torch.constant.int 512
%9982 = torch.prim.ListConstruct %int4_7560, %2482, %int512_7561 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%9983 = torch.aten.view %9981, %9982 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %9983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_7562 = torch.constant.int 1
%int0_7563 = torch.constant.int 0
%9984 = torch.prim.ListConstruct %int1_7562, %int0_7563 : (!torch.int, !torch.int) -> !torch.list<int>
%9985 = torch.aten.permute %312, %9984 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7564 = torch.constant.int 1
%int0_7565 = torch.constant.int 0
%9986 = torch.prim.ListConstruct %int1_7564, %int0_7565 : (!torch.int, !torch.int) -> !torch.list<int>
%9987 = torch.aten.permute %313, %9986 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7566 = torch.constant.int 1
%int0_7567 = torch.constant.int 0
%9988 = torch.prim.ListConstruct %int1_7566, %int0_7567 : (!torch.int, !torch.int) -> !torch.list<int>
%9989 = torch.aten.permute %314, %9988 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7568 = torch.constant.int 1
%int0_7569 = torch.constant.int 0
%9990 = torch.prim.ListConstruct %int1_7568, %int0_7569 : (!torch.int, !torch.int) -> !torch.list<int>
%9991 = torch.aten.permute %315, %9990 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7570 = torch.constant.int 1
%int0_7571 = torch.constant.int 0
%9992 = torch.prim.ListConstruct %int1_7570, %int0_7571 : (!torch.int, !torch.int) -> !torch.list<int>
%9993 = torch.aten.permute %316, %9992 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7572 = torch.constant.int 1
%int0_7573 = torch.constant.int 0
%9994 = torch.prim.ListConstruct %int1_7572, %int0_7573 : (!torch.int, !torch.int) -> !torch.list<int>
%9995 = torch.aten.permute %317, %9994 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7574 = torch.constant.int 1
%int0_7575 = torch.constant.int 0
%9996 = torch.prim.ListConstruct %int1_7574, %int0_7575 : (!torch.int, !torch.int) -> !torch.list<int>
%9997 = torch.aten.permute %318, %9996 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7576 = torch.constant.int 1
%int0_7577 = torch.constant.int 0
%9998 = torch.prim.ListConstruct %int1_7576, %int0_7577 : (!torch.int, !torch.int) -> !torch.list<int>
%9999 = torch.aten.permute %319, %9998 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_7578 = torch.constant.int 4
%10000 = torch.aten.mul.int %int4_7578, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7579 = torch.constant.int 4096
%10001 = torch.prim.ListConstruct %10000, %int4096_7579 : (!torch.int, !torch.int) -> !torch.list<int>
%10002 = torch.aten.view %9912, %10001 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10002, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10003 = torch.aten.mm %10002, %9985 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10003, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7580 = torch.constant.int 4
%int128_7581 = torch.constant.int 128
%10004 = torch.prim.ListConstruct %int4_7580, %2482, %int128_7581 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10005 = torch.aten.view %10003, %10004 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7582 = torch.constant.int 4
%10006 = torch.aten.mul.int %int4_7582, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7583 = torch.constant.int 4096
%10007 = torch.prim.ListConstruct %10006, %int4096_7583 : (!torch.int, !torch.int) -> !torch.list<int>
%10008 = torch.aten.view %9913, %10007 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10008, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10009 = torch.aten.mm %10008, %9987 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10009, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7584 = torch.constant.int 4
%int128_7585 = torch.constant.int 128
%10010 = torch.prim.ListConstruct %int4_7584, %2482, %int128_7585 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10011 = torch.aten.view %10009, %10010 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7586 = torch.constant.int 4
%10012 = torch.aten.mul.int %int4_7586, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7587 = torch.constant.int 4096
%10013 = torch.prim.ListConstruct %10012, %int4096_7587 : (!torch.int, !torch.int) -> !torch.list<int>
%10014 = torch.aten.view %9914, %10013 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10014, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10015 = torch.aten.mm %10014, %9989 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10015, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7588 = torch.constant.int 4
%int128_7589 = torch.constant.int 128
%10016 = torch.prim.ListConstruct %int4_7588, %2482, %int128_7589 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10017 = torch.aten.view %10015, %10016 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7590 = torch.constant.int 4
%10018 = torch.aten.mul.int %int4_7590, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7591 = torch.constant.int 4096
%10019 = torch.prim.ListConstruct %10018, %int4096_7591 : (!torch.int, !torch.int) -> !torch.list<int>
%10020 = torch.aten.view %9915, %10019 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10020, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10021 = torch.aten.mm %10020, %9991 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10021, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7592 = torch.constant.int 4
%int128_7593 = torch.constant.int 128
%10022 = torch.prim.ListConstruct %int4_7592, %2482, %int128_7593 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10023 = torch.aten.view %10021, %10022 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7594 = torch.constant.int 4
%10024 = torch.aten.mul.int %int4_7594, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7595 = torch.constant.int 4096
%10025 = torch.prim.ListConstruct %10024, %int4096_7595 : (!torch.int, !torch.int) -> !torch.list<int>
%10026 = torch.aten.view %9916, %10025 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10026, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10027 = torch.aten.mm %10026, %9993 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10027, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7596 = torch.constant.int 4
%int128_7597 = torch.constant.int 128
%10028 = torch.prim.ListConstruct %int4_7596, %2482, %int128_7597 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10029 = torch.aten.view %10027, %10028 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7598 = torch.constant.int 4
%10030 = torch.aten.mul.int %int4_7598, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7599 = torch.constant.int 4096
%10031 = torch.prim.ListConstruct %10030, %int4096_7599 : (!torch.int, !torch.int) -> !torch.list<int>
%10032 = torch.aten.view %9917, %10031 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10032, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10033 = torch.aten.mm %10032, %9995 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10033, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7600 = torch.constant.int 4
%int128_7601 = torch.constant.int 128
%10034 = torch.prim.ListConstruct %int4_7600, %2482, %int128_7601 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10035 = torch.aten.view %10033, %10034 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7602 = torch.constant.int 4
%10036 = torch.aten.mul.int %int4_7602, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7603 = torch.constant.int 4096
%10037 = torch.prim.ListConstruct %10036, %int4096_7603 : (!torch.int, !torch.int) -> !torch.list<int>
%10038 = torch.aten.view %9918, %10037 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10038, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10039 = torch.aten.mm %10038, %9997 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10039, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7604 = torch.constant.int 4
%int128_7605 = torch.constant.int 128
%10040 = torch.prim.ListConstruct %int4_7604, %2482, %int128_7605 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10041 = torch.aten.view %10039, %10040 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7606 = torch.constant.int 4
%10042 = torch.aten.mul.int %int4_7606, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7607 = torch.constant.int 4096
%10043 = torch.prim.ListConstruct %10042, %int4096_7607 : (!torch.int, !torch.int) -> !torch.list<int>
%10044 = torch.aten.view %9919, %10043 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10044, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10045 = torch.aten.mm %10044, %9999 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10045, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7608 = torch.constant.int 4
%int128_7609 = torch.constant.int 128
%10046 = torch.prim.ListConstruct %int4_7608, %2482, %int128_7609 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10047 = torch.aten.view %10045, %10046 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_7610 = torch.constant.int 1
%int0_7611 = torch.constant.int 0
%10048 = torch.prim.ListConstruct %int1_7610, %int0_7611 : (!torch.int, !torch.int) -> !torch.list<int>
%10049 = torch.aten.permute %320, %10048 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7612 = torch.constant.int 1
%int0_7613 = torch.constant.int 0
%10050 = torch.prim.ListConstruct %int1_7612, %int0_7613 : (!torch.int, !torch.int) -> !torch.list<int>
%10051 = torch.aten.permute %321, %10050 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7614 = torch.constant.int 1
%int0_7615 = torch.constant.int 0
%10052 = torch.prim.ListConstruct %int1_7614, %int0_7615 : (!torch.int, !torch.int) -> !torch.list<int>
%10053 = torch.aten.permute %322, %10052 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7616 = torch.constant.int 1
%int0_7617 = torch.constant.int 0
%10054 = torch.prim.ListConstruct %int1_7616, %int0_7617 : (!torch.int, !torch.int) -> !torch.list<int>
%10055 = torch.aten.permute %323, %10054 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7618 = torch.constant.int 1
%int0_7619 = torch.constant.int 0
%10056 = torch.prim.ListConstruct %int1_7618, %int0_7619 : (!torch.int, !torch.int) -> !torch.list<int>
%10057 = torch.aten.permute %324, %10056 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7620 = torch.constant.int 1
%int0_7621 = torch.constant.int 0
%10058 = torch.prim.ListConstruct %int1_7620, %int0_7621 : (!torch.int, !torch.int) -> !torch.list<int>
%10059 = torch.aten.permute %325, %10058 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7622 = torch.constant.int 1
%int0_7623 = torch.constant.int 0
%10060 = torch.prim.ListConstruct %int1_7622, %int0_7623 : (!torch.int, !torch.int) -> !torch.list<int>
%10061 = torch.aten.permute %326, %10060 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_7624 = torch.constant.int 1
%int0_7625 = torch.constant.int 0
%10062 = torch.prim.ListConstruct %int1_7624, %int0_7625 : (!torch.int, !torch.int) -> !torch.list<int>
%10063 = torch.aten.permute %327, %10062 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_7626 = torch.constant.int 4
%10064 = torch.aten.mul.int %int4_7626, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7627 = torch.constant.int 4096
%10065 = torch.prim.ListConstruct %10064, %int4096_7627 : (!torch.int, !torch.int) -> !torch.list<int>
%10066 = torch.aten.view %9912, %10065 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10066, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10067 = torch.aten.mm %10066, %10049 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10067, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7628 = torch.constant.int 4
%int128_7629 = torch.constant.int 128
%10068 = torch.prim.ListConstruct %int4_7628, %2482, %int128_7629 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10069 = torch.aten.view %10067, %10068 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7630 = torch.constant.int 4
%10070 = torch.aten.mul.int %int4_7630, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7631 = torch.constant.int 4096
%10071 = torch.prim.ListConstruct %10070, %int4096_7631 : (!torch.int, !torch.int) -> !torch.list<int>
%10072 = torch.aten.view %9913, %10071 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10072, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10073 = torch.aten.mm %10072, %10051 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10073, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7632 = torch.constant.int 4
%int128_7633 = torch.constant.int 128
%10074 = torch.prim.ListConstruct %int4_7632, %2482, %int128_7633 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10075 = torch.aten.view %10073, %10074 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7634 = torch.constant.int 4
%10076 = torch.aten.mul.int %int4_7634, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7635 = torch.constant.int 4096
%10077 = torch.prim.ListConstruct %10076, %int4096_7635 : (!torch.int, !torch.int) -> !torch.list<int>
%10078 = torch.aten.view %9914, %10077 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10078, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10079 = torch.aten.mm %10078, %10053 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10079, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7636 = torch.constant.int 4
%int128_7637 = torch.constant.int 128
%10080 = torch.prim.ListConstruct %int4_7636, %2482, %int128_7637 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10081 = torch.aten.view %10079, %10080 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7638 = torch.constant.int 4
%10082 = torch.aten.mul.int %int4_7638, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7639 = torch.constant.int 4096
%10083 = torch.prim.ListConstruct %10082, %int4096_7639 : (!torch.int, !torch.int) -> !torch.list<int>
%10084 = torch.aten.view %9915, %10083 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10084, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10085 = torch.aten.mm %10084, %10055 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10085, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7640 = torch.constant.int 4
%int128_7641 = torch.constant.int 128
%10086 = torch.prim.ListConstruct %int4_7640, %2482, %int128_7641 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10087 = torch.aten.view %10085, %10086 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7642 = torch.constant.int 4
%10088 = torch.aten.mul.int %int4_7642, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7643 = torch.constant.int 4096
%10089 = torch.prim.ListConstruct %10088, %int4096_7643 : (!torch.int, !torch.int) -> !torch.list<int>
%10090 = torch.aten.view %9916, %10089 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10090, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10091 = torch.aten.mm %10090, %10057 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10091, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7644 = torch.constant.int 4
%int128_7645 = torch.constant.int 128
%10092 = torch.prim.ListConstruct %int4_7644, %2482, %int128_7645 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10093 = torch.aten.view %10091, %10092 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7646 = torch.constant.int 4
%10094 = torch.aten.mul.int %int4_7646, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7647 = torch.constant.int 4096
%10095 = torch.prim.ListConstruct %10094, %int4096_7647 : (!torch.int, !torch.int) -> !torch.list<int>
%10096 = torch.aten.view %9917, %10095 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10096, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10097 = torch.aten.mm %10096, %10059 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10097, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7648 = torch.constant.int 4
%int128_7649 = torch.constant.int 128
%10098 = torch.prim.ListConstruct %int4_7648, %2482, %int128_7649 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10099 = torch.aten.view %10097, %10098 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7650 = torch.constant.int 4
%10100 = torch.aten.mul.int %int4_7650, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7651 = torch.constant.int 4096
%10101 = torch.prim.ListConstruct %10100, %int4096_7651 : (!torch.int, !torch.int) -> !torch.list<int>
%10102 = torch.aten.view %9918, %10101 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10102, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10103 = torch.aten.mm %10102, %10061 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10103, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7652 = torch.constant.int 4
%int128_7653 = torch.constant.int 128
%10104 = torch.prim.ListConstruct %int4_7652, %2482, %int128_7653 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10105 = torch.aten.view %10103, %10104 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7654 = torch.constant.int 4
%10106 = torch.aten.mul.int %int4_7654, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_7655 = torch.constant.int 4096
%10107 = torch.prim.ListConstruct %10106, %int4096_7655 : (!torch.int, !torch.int) -> !torch.list<int>
%10108 = torch.aten.view %9919, %10107 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10108, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%10109 = torch.aten.mm %10108, %10063 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %10109, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_7656 = torch.constant.int 4
%int128_7657 = torch.constant.int 128
%10110 = torch.prim.ListConstruct %int4_7656, %2482, %int128_7657 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10111 = torch.aten.view %10109, %10110 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %10111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_7658 = torch.constant.int 4
%int4_7659 = torch.constant.int 4
%int128_7660 = torch.constant.int 128
%10112 = torch.prim.ListConstruct %int4_7658, %2482, %int4_7659, %int128_7660 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10113 = torch.aten.view %9941, %10112 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7661 = torch.constant.int 4
%int4_7662 = torch.constant.int 4
%int128_7663 = torch.constant.int 128
%10114 = torch.prim.ListConstruct %int4_7661, %2482, %int4_7662, %int128_7663 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10115 = torch.aten.view %9947, %10114 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7664 = torch.constant.int 4
%int4_7665 = torch.constant.int 4
%int128_7666 = torch.constant.int 128
%10116 = torch.prim.ListConstruct %int4_7664, %2482, %int4_7665, %int128_7666 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10117 = torch.aten.view %9953, %10116 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7667 = torch.constant.int 4
%int4_7668 = torch.constant.int 4
%int128_7669 = torch.constant.int 128
%10118 = torch.prim.ListConstruct %int4_7667, %2482, %int4_7668, %int128_7669 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10119 = torch.aten.view %9959, %10118 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7670 = torch.constant.int 4
%int4_7671 = torch.constant.int 4
%int128_7672 = torch.constant.int 128
%10120 = torch.prim.ListConstruct %int4_7670, %2482, %int4_7671, %int128_7672 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10121 = torch.aten.view %9965, %10120 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7673 = torch.constant.int 4
%int4_7674 = torch.constant.int 4
%int128_7675 = torch.constant.int 128
%10122 = torch.prim.ListConstruct %int4_7673, %2482, %int4_7674, %int128_7675 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10123 = torch.aten.view %9971, %10122 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7676 = torch.constant.int 4
%int4_7677 = torch.constant.int 4
%int128_7678 = torch.constant.int 128
%10124 = torch.prim.ListConstruct %int4_7676, %2482, %int4_7677, %int128_7678 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10125 = torch.aten.view %9977, %10124 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7679 = torch.constant.int 4
%int4_7680 = torch.constant.int 4
%int128_7681 = torch.constant.int 128
%10126 = torch.prim.ListConstruct %int4_7679, %2482, %int4_7680, %int128_7681 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10127 = torch.aten.view %9983, %10126 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_7682 = torch.constant.int 4
%int1_7683 = torch.constant.int 1
%int128_7684 = torch.constant.int 128
%10128 = torch.prim.ListConstruct %int4_7682, %2482, %int1_7683, %int128_7684 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10129 = torch.aten.view %10005, %10128 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7685 = torch.constant.int 4
%int1_7686 = torch.constant.int 1
%int128_7687 = torch.constant.int 128
%10130 = torch.prim.ListConstruct %int4_7685, %2482, %int1_7686, %int128_7687 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10131 = torch.aten.view %10011, %10130 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7688 = torch.constant.int 4
%int1_7689 = torch.constant.int 1
%int128_7690 = torch.constant.int 128
%10132 = torch.prim.ListConstruct %int4_7688, %2482, %int1_7689, %int128_7690 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10133 = torch.aten.view %10017, %10132 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7691 = torch.constant.int 4
%int1_7692 = torch.constant.int 1
%int128_7693 = torch.constant.int 128
%10134 = torch.prim.ListConstruct %int4_7691, %2482, %int1_7692, %int128_7693 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10135 = torch.aten.view %10023, %10134 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7694 = torch.constant.int 4
%int1_7695 = torch.constant.int 1
%int128_7696 = torch.constant.int 128
%10136 = torch.prim.ListConstruct %int4_7694, %2482, %int1_7695, %int128_7696 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10137 = torch.aten.view %10029, %10136 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7697 = torch.constant.int 4
%int1_7698 = torch.constant.int 1
%int128_7699 = torch.constant.int 128
%10138 = torch.prim.ListConstruct %int4_7697, %2482, %int1_7698, %int128_7699 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10139 = torch.aten.view %10035, %10138 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7700 = torch.constant.int 4
%int1_7701 = torch.constant.int 1
%int128_7702 = torch.constant.int 128
%10140 = torch.prim.ListConstruct %int4_7700, %2482, %int1_7701, %int128_7702 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10141 = torch.aten.view %10041, %10140 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7703 = torch.constant.int 4
%int1_7704 = torch.constant.int 1
%int128_7705 = torch.constant.int 128
%10142 = torch.prim.ListConstruct %int4_7703, %2482, %int1_7704, %int128_7705 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10143 = torch.aten.view %10047, %10142 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7706 = torch.constant.int 4
%int1_7707 = torch.constant.int 1
%int128_7708 = torch.constant.int 128
%10144 = torch.prim.ListConstruct %int4_7706, %2482, %int1_7707, %int128_7708 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10145 = torch.aten.view %10069, %10144 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7709 = torch.constant.int 4
%int1_7710 = torch.constant.int 1
%int128_7711 = torch.constant.int 128
%10146 = torch.prim.ListConstruct %int4_7709, %2482, %int1_7710, %int128_7711 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10147 = torch.aten.view %10075, %10146 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7712 = torch.constant.int 4
%int1_7713 = torch.constant.int 1
%int128_7714 = torch.constant.int 128
%10148 = torch.prim.ListConstruct %int4_7712, %2482, %int1_7713, %int128_7714 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10149 = torch.aten.view %10081, %10148 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7715 = torch.constant.int 4
%int1_7716 = torch.constant.int 1
%int128_7717 = torch.constant.int 128
%10150 = torch.prim.ListConstruct %int4_7715, %2482, %int1_7716, %int128_7717 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10151 = torch.aten.view %10087, %10150 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7718 = torch.constant.int 4
%int1_7719 = torch.constant.int 1
%int128_7720 = torch.constant.int 128
%10152 = torch.prim.ListConstruct %int4_7718, %2482, %int1_7719, %int128_7720 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10153 = torch.aten.view %10093, %10152 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7721 = torch.constant.int 4
%int1_7722 = torch.constant.int 1
%int128_7723 = torch.constant.int 128
%10154 = torch.prim.ListConstruct %int4_7721, %2482, %int1_7722, %int128_7723 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10155 = torch.aten.view %10099, %10154 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7724 = torch.constant.int 4
%int1_7725 = torch.constant.int 1
%int128_7726 = torch.constant.int 128
%10156 = torch.prim.ListConstruct %int4_7724, %2482, %int1_7725, %int128_7726 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10157 = torch.aten.view %10105, %10156 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_7727 = torch.constant.int 4
%int1_7728 = torch.constant.int 1
%int128_7729 = torch.constant.int 128
%10158 = torch.prim.ListConstruct %int4_7727, %2482, %int1_7728, %int128_7729 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10159 = torch.aten.view %10111, %10158 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_7730 = torch.constant.int 131072
%none_7731 = torch.constant.none
%none_7732 = torch.constant.none
%cpu_7733 = torch.constant.device "cpu"
%false_7734 = torch.constant.bool false
%10160 = torch.aten.arange %int131072_7730, %none_7731, %none_7732, %cpu_7733, %false_7734 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_7735 = torch.constant.int 0
%int128_7736 = torch.constant.int 128
%int2_7737 = torch.constant.int 2
%none_7738 = torch.constant.none
%none_7739 = torch.constant.none
%cpu_7740 = torch.constant.device "cpu"
%false_7741 = torch.constant.bool false
%10161 = torch.aten.arange.start_step %int0_7735, %int128_7736, %int2_7737, %none_7738, %none_7739, %cpu_7740, %false_7741 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_7742 = torch.constant.int 0
%int0_7743 = torch.constant.int 0
%int64_7744 = torch.constant.int 64
%int1_7745 = torch.constant.int 1
%10162 = torch.aten.slice.Tensor %10161, %int0_7742, %int0_7743, %int64_7744, %int1_7745 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_7746 = torch.constant.int 6
%10163 = torch.prims.convert_element_type %10162, %int6_7746 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_7747 = torch.constant.int 128
%10164 = torch.aten.div.Scalar %10163, %int128_7747 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_7748 = torch.constant.float 5.000000e+05
%10165 = torch.aten.pow.Scalar %float5.000000e05_7748, %10164 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%10166 = torch.aten.reciprocal %10165 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_7749 = torch.constant.float 1.000000e+00
%10167 = torch.aten.mul.Scalar %10166, %float1.000000e00_7749 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_7750 = torch.constant.int 131072
%int1_7751 = torch.constant.int 1
%10168 = torch.prim.ListConstruct %int131072_7750, %int1_7751 : (!torch.int, !torch.int) -> !torch.list<int>
%10169 = torch.aten.view %10160, %10168 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%10170 = torch.aten.mul.Tensor %10169, %10167 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%10171 = torch.aten.cos %10170 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%10172 = torch.aten.sin %10170 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%10173 = torch.aten.complex %10171, %10172 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%10174 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10175 = flow.tensor.transfer %10174 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%10176 = torch_c.from_builtin_tensor %10175 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10177 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10178 = flow.tensor.transfer %10177 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%10179 = torch_c.from_builtin_tensor %10178 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10180 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10181 = flow.tensor.transfer %10180 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%10182 = torch_c.from_builtin_tensor %10181 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10183 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10184 = flow.tensor.transfer %10183 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%10185 = torch_c.from_builtin_tensor %10184 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10186 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10187 = flow.tensor.transfer %10186 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%10188 = torch_c.from_builtin_tensor %10187 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10189 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10190 = flow.tensor.transfer %10189 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%10191 = torch_c.from_builtin_tensor %10190 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10192 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10193 = flow.tensor.transfer %10192 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%10194 = torch_c.from_builtin_tensor %10193 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10195 = torch_c.to_builtin_tensor %10173 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10196 = flow.tensor.transfer %10195 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%10197 = torch_c.from_builtin_tensor %10196 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_7752 = torch.constant.int 1
%10198 = torch.aten.size.int %9941, %int1_7752 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7753 = torch.constant.int 0
%10199 = torch.aten.add.int %int0_7753, %10198 : !torch.int, !torch.int -> !torch.int
%int0_7754 = torch.constant.int 0
%int0_7755 = torch.constant.int 0
%int1_7756 = torch.constant.int 1
%10200 = torch.aten.slice.Tensor %10176, %int0_7754, %int0_7755, %10199, %int1_7756 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10200, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7757 = torch.constant.int 1
%int0_7758 = torch.constant.int 0
%int9223372036854775807_7759 = torch.constant.int 9223372036854775807
%int1_7760 = torch.constant.int 1
%10201 = torch.aten.slice.Tensor %10200, %int1_7757, %int0_7758, %int9223372036854775807_7759, %int1_7760 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10201, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7761 = torch.constant.int 0
%10202 = torch.aten.unsqueeze %10201, %int0_7761 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10202, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7762 = torch.constant.int 2
%10203 = torch.aten.unsqueeze %10202, %int2_7762 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10203, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7763 = torch.constant.int 3
%int0_7764 = torch.constant.int 0
%int9223372036854775807_7765 = torch.constant.int 9223372036854775807
%int1_7766 = torch.constant.int 1
%10204 = torch.aten.slice.Tensor %10203, %int3_7763, %int0_7764, %int9223372036854775807_7765, %int1_7766 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10204, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10205 = torch_c.to_builtin_tensor %10113 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7767 = arith.constant 1 : index
%dim_7768 = tensor.dim %10205, %c1_7767 : tensor<4x?x4x128xf16>
%10206 = flow.tensor.bitcast %10205 : tensor<4x?x4x128xf16>{%dim_7768} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7768}
%10207 = torch_c.from_builtin_tensor %10206 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10208 = torch.aten.mul.Tensor %10207, %10204 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10209 = torch_c.to_builtin_tensor %10208 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7769 = arith.constant 1 : index
%dim_7770 = tensor.dim %10209, %c1_7769 : tensor<4x?x4x64xcomplex<f32>>
%10210 = flow.tensor.bitcast %10209 : tensor<4x?x4x64xcomplex<f32>>{%dim_7770} -> tensor<4x?x4x128xf32>{%dim_7770}
%10211 = torch_c.from_builtin_tensor %10210 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7771 = torch.constant.int 5
%10212 = torch.prims.convert_element_type %10211, %int5_7771 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7772 = torch.constant.int 1
%10213 = torch.aten.size.int %9947, %int1_7772 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7773 = torch.constant.int 0
%10214 = torch.aten.add.int %int0_7773, %10213 : !torch.int, !torch.int -> !torch.int
%int0_7774 = torch.constant.int 0
%int0_7775 = torch.constant.int 0
%int1_7776 = torch.constant.int 1
%10215 = torch.aten.slice.Tensor %10179, %int0_7774, %int0_7775, %10214, %int1_7776 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10215, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7777 = torch.constant.int 1
%int0_7778 = torch.constant.int 0
%int9223372036854775807_7779 = torch.constant.int 9223372036854775807
%int1_7780 = torch.constant.int 1
%10216 = torch.aten.slice.Tensor %10215, %int1_7777, %int0_7778, %int9223372036854775807_7779, %int1_7780 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10216, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7781 = torch.constant.int 0
%10217 = torch.aten.unsqueeze %10216, %int0_7781 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10217, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7782 = torch.constant.int 2
%10218 = torch.aten.unsqueeze %10217, %int2_7782 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10218, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7783 = torch.constant.int 3
%int0_7784 = torch.constant.int 0
%int9223372036854775807_7785 = torch.constant.int 9223372036854775807
%int1_7786 = torch.constant.int 1
%10219 = torch.aten.slice.Tensor %10218, %int3_7783, %int0_7784, %int9223372036854775807_7785, %int1_7786 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10219, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10220 = torch_c.to_builtin_tensor %10115 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7787 = arith.constant 1 : index
%dim_7788 = tensor.dim %10220, %c1_7787 : tensor<4x?x4x128xf16>
%10221 = flow.tensor.bitcast %10220 : tensor<4x?x4x128xf16>{%dim_7788} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7788}
%10222 = torch_c.from_builtin_tensor %10221 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10223 = torch.aten.mul.Tensor %10222, %10219 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10224 = torch_c.to_builtin_tensor %10223 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7789 = arith.constant 1 : index
%dim_7790 = tensor.dim %10224, %c1_7789 : tensor<4x?x4x64xcomplex<f32>>
%10225 = flow.tensor.bitcast %10224 : tensor<4x?x4x64xcomplex<f32>>{%dim_7790} -> tensor<4x?x4x128xf32>{%dim_7790}
%10226 = torch_c.from_builtin_tensor %10225 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7791 = torch.constant.int 5
%10227 = torch.prims.convert_element_type %10226, %int5_7791 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7792 = torch.constant.int 1
%10228 = torch.aten.size.int %9953, %int1_7792 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7793 = torch.constant.int 0
%10229 = torch.aten.add.int %int0_7793, %10228 : !torch.int, !torch.int -> !torch.int
%int0_7794 = torch.constant.int 0
%int0_7795 = torch.constant.int 0
%int1_7796 = torch.constant.int 1
%10230 = torch.aten.slice.Tensor %10182, %int0_7794, %int0_7795, %10229, %int1_7796 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10230, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7797 = torch.constant.int 1
%int0_7798 = torch.constant.int 0
%int9223372036854775807_7799 = torch.constant.int 9223372036854775807
%int1_7800 = torch.constant.int 1
%10231 = torch.aten.slice.Tensor %10230, %int1_7797, %int0_7798, %int9223372036854775807_7799, %int1_7800 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10231, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7801 = torch.constant.int 0
%10232 = torch.aten.unsqueeze %10231, %int0_7801 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10232, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7802 = torch.constant.int 2
%10233 = torch.aten.unsqueeze %10232, %int2_7802 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10233, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7803 = torch.constant.int 3
%int0_7804 = torch.constant.int 0
%int9223372036854775807_7805 = torch.constant.int 9223372036854775807
%int1_7806 = torch.constant.int 1
%10234 = torch.aten.slice.Tensor %10233, %int3_7803, %int0_7804, %int9223372036854775807_7805, %int1_7806 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10234, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10235 = torch_c.to_builtin_tensor %10117 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7807 = arith.constant 1 : index
%dim_7808 = tensor.dim %10235, %c1_7807 : tensor<4x?x4x128xf16>
%10236 = flow.tensor.bitcast %10235 : tensor<4x?x4x128xf16>{%dim_7808} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7808}
%10237 = torch_c.from_builtin_tensor %10236 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10238 = torch.aten.mul.Tensor %10237, %10234 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10239 = torch_c.to_builtin_tensor %10238 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7809 = arith.constant 1 : index
%dim_7810 = tensor.dim %10239, %c1_7809 : tensor<4x?x4x64xcomplex<f32>>
%10240 = flow.tensor.bitcast %10239 : tensor<4x?x4x64xcomplex<f32>>{%dim_7810} -> tensor<4x?x4x128xf32>{%dim_7810}
%10241 = torch_c.from_builtin_tensor %10240 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7811 = torch.constant.int 5
%10242 = torch.prims.convert_element_type %10241, %int5_7811 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7812 = torch.constant.int 1
%10243 = torch.aten.size.int %9959, %int1_7812 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7813 = torch.constant.int 0
%10244 = torch.aten.add.int %int0_7813, %10243 : !torch.int, !torch.int -> !torch.int
%int0_7814 = torch.constant.int 0
%int0_7815 = torch.constant.int 0
%int1_7816 = torch.constant.int 1
%10245 = torch.aten.slice.Tensor %10185, %int0_7814, %int0_7815, %10244, %int1_7816 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10245, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7817 = torch.constant.int 1
%int0_7818 = torch.constant.int 0
%int9223372036854775807_7819 = torch.constant.int 9223372036854775807
%int1_7820 = torch.constant.int 1
%10246 = torch.aten.slice.Tensor %10245, %int1_7817, %int0_7818, %int9223372036854775807_7819, %int1_7820 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10246, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7821 = torch.constant.int 0
%10247 = torch.aten.unsqueeze %10246, %int0_7821 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10247, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7822 = torch.constant.int 2
%10248 = torch.aten.unsqueeze %10247, %int2_7822 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10248, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7823 = torch.constant.int 3
%int0_7824 = torch.constant.int 0
%int9223372036854775807_7825 = torch.constant.int 9223372036854775807
%int1_7826 = torch.constant.int 1
%10249 = torch.aten.slice.Tensor %10248, %int3_7823, %int0_7824, %int9223372036854775807_7825, %int1_7826 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10249, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10250 = torch_c.to_builtin_tensor %10119 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7827 = arith.constant 1 : index
%dim_7828 = tensor.dim %10250, %c1_7827 : tensor<4x?x4x128xf16>
%10251 = flow.tensor.bitcast %10250 : tensor<4x?x4x128xf16>{%dim_7828} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7828}
%10252 = torch_c.from_builtin_tensor %10251 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10253 = torch.aten.mul.Tensor %10252, %10249 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10254 = torch_c.to_builtin_tensor %10253 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7829 = arith.constant 1 : index
%dim_7830 = tensor.dim %10254, %c1_7829 : tensor<4x?x4x64xcomplex<f32>>
%10255 = flow.tensor.bitcast %10254 : tensor<4x?x4x64xcomplex<f32>>{%dim_7830} -> tensor<4x?x4x128xf32>{%dim_7830}
%10256 = torch_c.from_builtin_tensor %10255 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7831 = torch.constant.int 5
%10257 = torch.prims.convert_element_type %10256, %int5_7831 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7832 = torch.constant.int 1
%10258 = torch.aten.size.int %9965, %int1_7832 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7833 = torch.constant.int 0
%10259 = torch.aten.add.int %int0_7833, %10258 : !torch.int, !torch.int -> !torch.int
%int0_7834 = torch.constant.int 0
%int0_7835 = torch.constant.int 0
%int1_7836 = torch.constant.int 1
%10260 = torch.aten.slice.Tensor %10188, %int0_7834, %int0_7835, %10259, %int1_7836 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10260, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7837 = torch.constant.int 1
%int0_7838 = torch.constant.int 0
%int9223372036854775807_7839 = torch.constant.int 9223372036854775807
%int1_7840 = torch.constant.int 1
%10261 = torch.aten.slice.Tensor %10260, %int1_7837, %int0_7838, %int9223372036854775807_7839, %int1_7840 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10261, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7841 = torch.constant.int 0
%10262 = torch.aten.unsqueeze %10261, %int0_7841 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10262, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7842 = torch.constant.int 2
%10263 = torch.aten.unsqueeze %10262, %int2_7842 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10263, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7843 = torch.constant.int 3
%int0_7844 = torch.constant.int 0
%int9223372036854775807_7845 = torch.constant.int 9223372036854775807
%int1_7846 = torch.constant.int 1
%10264 = torch.aten.slice.Tensor %10263, %int3_7843, %int0_7844, %int9223372036854775807_7845, %int1_7846 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10264, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10265 = torch_c.to_builtin_tensor %10121 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7847 = arith.constant 1 : index
%dim_7848 = tensor.dim %10265, %c1_7847 : tensor<4x?x4x128xf16>
%10266 = flow.tensor.bitcast %10265 : tensor<4x?x4x128xf16>{%dim_7848} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7848}
%10267 = torch_c.from_builtin_tensor %10266 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10268 = torch.aten.mul.Tensor %10267, %10264 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10269 = torch_c.to_builtin_tensor %10268 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7849 = arith.constant 1 : index
%dim_7850 = tensor.dim %10269, %c1_7849 : tensor<4x?x4x64xcomplex<f32>>
%10270 = flow.tensor.bitcast %10269 : tensor<4x?x4x64xcomplex<f32>>{%dim_7850} -> tensor<4x?x4x128xf32>{%dim_7850}
%10271 = torch_c.from_builtin_tensor %10270 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7851 = torch.constant.int 5
%10272 = torch.prims.convert_element_type %10271, %int5_7851 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7852 = torch.constant.int 1
%10273 = torch.aten.size.int %9971, %int1_7852 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7853 = torch.constant.int 0
%10274 = torch.aten.add.int %int0_7853, %10273 : !torch.int, !torch.int -> !torch.int
%int0_7854 = torch.constant.int 0
%int0_7855 = torch.constant.int 0
%int1_7856 = torch.constant.int 1
%10275 = torch.aten.slice.Tensor %10191, %int0_7854, %int0_7855, %10274, %int1_7856 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10275, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7857 = torch.constant.int 1
%int0_7858 = torch.constant.int 0
%int9223372036854775807_7859 = torch.constant.int 9223372036854775807
%int1_7860 = torch.constant.int 1
%10276 = torch.aten.slice.Tensor %10275, %int1_7857, %int0_7858, %int9223372036854775807_7859, %int1_7860 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10276, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7861 = torch.constant.int 0
%10277 = torch.aten.unsqueeze %10276, %int0_7861 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10277, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7862 = torch.constant.int 2
%10278 = torch.aten.unsqueeze %10277, %int2_7862 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10278, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7863 = torch.constant.int 3
%int0_7864 = torch.constant.int 0
%int9223372036854775807_7865 = torch.constant.int 9223372036854775807
%int1_7866 = torch.constant.int 1
%10279 = torch.aten.slice.Tensor %10278, %int3_7863, %int0_7864, %int9223372036854775807_7865, %int1_7866 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10279, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10280 = torch_c.to_builtin_tensor %10123 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7867 = arith.constant 1 : index
%dim_7868 = tensor.dim %10280, %c1_7867 : tensor<4x?x4x128xf16>
%10281 = flow.tensor.bitcast %10280 : tensor<4x?x4x128xf16>{%dim_7868} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7868}
%10282 = torch_c.from_builtin_tensor %10281 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10283 = torch.aten.mul.Tensor %10282, %10279 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10284 = torch_c.to_builtin_tensor %10283 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7869 = arith.constant 1 : index
%dim_7870 = tensor.dim %10284, %c1_7869 : tensor<4x?x4x64xcomplex<f32>>
%10285 = flow.tensor.bitcast %10284 : tensor<4x?x4x64xcomplex<f32>>{%dim_7870} -> tensor<4x?x4x128xf32>{%dim_7870}
%10286 = torch_c.from_builtin_tensor %10285 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7871 = torch.constant.int 5
%10287 = torch.prims.convert_element_type %10286, %int5_7871 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7872 = torch.constant.int 1
%10288 = torch.aten.size.int %9977, %int1_7872 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7873 = torch.constant.int 0
%10289 = torch.aten.add.int %int0_7873, %10288 : !torch.int, !torch.int -> !torch.int
%int0_7874 = torch.constant.int 0
%int0_7875 = torch.constant.int 0
%int1_7876 = torch.constant.int 1
%10290 = torch.aten.slice.Tensor %10194, %int0_7874, %int0_7875, %10289, %int1_7876 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10290, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7877 = torch.constant.int 1
%int0_7878 = torch.constant.int 0
%int9223372036854775807_7879 = torch.constant.int 9223372036854775807
%int1_7880 = torch.constant.int 1
%10291 = torch.aten.slice.Tensor %10290, %int1_7877, %int0_7878, %int9223372036854775807_7879, %int1_7880 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10291, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7881 = torch.constant.int 0
%10292 = torch.aten.unsqueeze %10291, %int0_7881 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10292, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7882 = torch.constant.int 2
%10293 = torch.aten.unsqueeze %10292, %int2_7882 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10293, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7883 = torch.constant.int 3
%int0_7884 = torch.constant.int 0
%int9223372036854775807_7885 = torch.constant.int 9223372036854775807
%int1_7886 = torch.constant.int 1
%10294 = torch.aten.slice.Tensor %10293, %int3_7883, %int0_7884, %int9223372036854775807_7885, %int1_7886 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10294, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10295 = torch_c.to_builtin_tensor %10125 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7887 = arith.constant 1 : index
%dim_7888 = tensor.dim %10295, %c1_7887 : tensor<4x?x4x128xf16>
%10296 = flow.tensor.bitcast %10295 : tensor<4x?x4x128xf16>{%dim_7888} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7888}
%10297 = torch_c.from_builtin_tensor %10296 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10298 = torch.aten.mul.Tensor %10297, %10294 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10299 = torch_c.to_builtin_tensor %10298 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7889 = arith.constant 1 : index
%dim_7890 = tensor.dim %10299, %c1_7889 : tensor<4x?x4x64xcomplex<f32>>
%10300 = flow.tensor.bitcast %10299 : tensor<4x?x4x64xcomplex<f32>>{%dim_7890} -> tensor<4x?x4x128xf32>{%dim_7890}
%10301 = torch_c.from_builtin_tensor %10300 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7891 = torch.constant.int 5
%10302 = torch.prims.convert_element_type %10301, %int5_7891 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_7892 = torch.constant.int 1
%10303 = torch.aten.size.int %9983, %int1_7892 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_7893 = torch.constant.int 0
%10304 = torch.aten.add.int %int0_7893, %10303 : !torch.int, !torch.int -> !torch.int
%int0_7894 = torch.constant.int 0
%int0_7895 = torch.constant.int 0
%int1_7896 = torch.constant.int 1
%10305 = torch.aten.slice.Tensor %10197, %int0_7894, %int0_7895, %10304, %int1_7896 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10305, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7897 = torch.constant.int 1
%int0_7898 = torch.constant.int 0
%int9223372036854775807_7899 = torch.constant.int 9223372036854775807
%int1_7900 = torch.constant.int 1
%10306 = torch.aten.slice.Tensor %10305, %int1_7897, %int0_7898, %int9223372036854775807_7899, %int1_7900 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10306, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7901 = torch.constant.int 0
%10307 = torch.aten.unsqueeze %10306, %int0_7901 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10307, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7902 = torch.constant.int 2
%10308 = torch.aten.unsqueeze %10307, %int2_7902 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10308, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7903 = torch.constant.int 3
%int0_7904 = torch.constant.int 0
%int9223372036854775807_7905 = torch.constant.int 9223372036854775807
%int1_7906 = torch.constant.int 1
%10309 = torch.aten.slice.Tensor %10308, %int3_7903, %int0_7904, %int9223372036854775807_7905, %int1_7906 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10309, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10310 = torch_c.to_builtin_tensor %10127 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_7907 = arith.constant 1 : index
%dim_7908 = tensor.dim %10310, %c1_7907 : tensor<4x?x4x128xf16>
%10311 = flow.tensor.bitcast %10310 : tensor<4x?x4x128xf16>{%dim_7908} -> tensor<4x?x4x64xcomplex<f16>>{%dim_7908}
%10312 = torch_c.from_builtin_tensor %10311 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %10312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%10313 = torch.aten.mul.Tensor %10312, %10309 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %10313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%10314 = torch_c.to_builtin_tensor %10313 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_7909 = arith.constant 1 : index
%dim_7910 = tensor.dim %10314, %c1_7909 : tensor<4x?x4x64xcomplex<f32>>
%10315 = flow.tensor.bitcast %10314 : tensor<4x?x4x64xcomplex<f32>>{%dim_7910} -> tensor<4x?x4x128xf32>{%dim_7910}
%10316 = torch_c.from_builtin_tensor %10315 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %10316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_7911 = torch.constant.int 5
%10317 = torch.prims.convert_element_type %10316, %int5_7911 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_7912 = torch.constant.int 131072
%none_7913 = torch.constant.none
%none_7914 = torch.constant.none
%cpu_7915 = torch.constant.device "cpu"
%false_7916 = torch.constant.bool false
%10318 = torch.aten.arange %int131072_7912, %none_7913, %none_7914, %cpu_7915, %false_7916 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_7917 = torch.constant.int 0
%int128_7918 = torch.constant.int 128
%int2_7919 = torch.constant.int 2
%none_7920 = torch.constant.none
%none_7921 = torch.constant.none
%cpu_7922 = torch.constant.device "cpu"
%false_7923 = torch.constant.bool false
%10319 = torch.aten.arange.start_step %int0_7917, %int128_7918, %int2_7919, %none_7920, %none_7921, %cpu_7922, %false_7923 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_7924 = torch.constant.int 0
%int0_7925 = torch.constant.int 0
%int64_7926 = torch.constant.int 64
%int1_7927 = torch.constant.int 1
%10320 = torch.aten.slice.Tensor %10319, %int0_7924, %int0_7925, %int64_7926, %int1_7927 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_7928 = torch.constant.int 6
%10321 = torch.prims.convert_element_type %10320, %int6_7928 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_7929 = torch.constant.int 128
%10322 = torch.aten.div.Scalar %10321, %int128_7929 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_7930 = torch.constant.float 5.000000e+05
%10323 = torch.aten.pow.Scalar %float5.000000e05_7930, %10322 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%10324 = torch.aten.reciprocal %10323 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_7931 = torch.constant.float 1.000000e+00
%10325 = torch.aten.mul.Scalar %10324, %float1.000000e00_7931 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_7932 = torch.constant.int 131072
%int1_7933 = torch.constant.int 1
%10326 = torch.prim.ListConstruct %int131072_7932, %int1_7933 : (!torch.int, !torch.int) -> !torch.list<int>
%10327 = torch.aten.view %10318, %10326 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%10328 = torch.aten.mul.Tensor %10327, %10325 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%10329 = torch.aten.cos %10328 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%10330 = torch.aten.sin %10328 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%10331 = torch.aten.complex %10329, %10330 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%10332 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10333 = flow.tensor.transfer %10332 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%10334 = torch_c.from_builtin_tensor %10333 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10335 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10336 = flow.tensor.transfer %10335 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%10337 = torch_c.from_builtin_tensor %10336 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10338 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10339 = flow.tensor.transfer %10338 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%10340 = torch_c.from_builtin_tensor %10339 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10341 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10342 = flow.tensor.transfer %10341 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%10343 = torch_c.from_builtin_tensor %10342 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10344 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10345 = flow.tensor.transfer %10344 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%10346 = torch_c.from_builtin_tensor %10345 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10347 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10348 = flow.tensor.transfer %10347 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%10349 = torch_c.from_builtin_tensor %10348 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10350 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10351 = flow.tensor.transfer %10350 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%10352 = torch_c.from_builtin_tensor %10351 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%10353 = torch_c.to_builtin_tensor %10331 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%10354 = flow.tensor.transfer %10353 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%10355 = torch_c.from_builtin_tensor %10354 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_7934 = torch.constant.int 1
%10356 = torch.aten.size.int %10005, %int1_7934 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_7935 = torch.constant.int 0
%10357 = torch.aten.add.int %int0_7935, %10356 : !torch.int, !torch.int -> !torch.int
%int0_7936 = torch.constant.int 0
%int0_7937 = torch.constant.int 0
%int1_7938 = torch.constant.int 1
%10358 = torch.aten.slice.Tensor %10334, %int0_7936, %int0_7937, %10357, %int1_7938 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10358, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7939 = torch.constant.int 1
%int0_7940 = torch.constant.int 0
%int9223372036854775807_7941 = torch.constant.int 9223372036854775807
%int1_7942 = torch.constant.int 1
%10359 = torch.aten.slice.Tensor %10358, %int1_7939, %int0_7940, %int9223372036854775807_7941, %int1_7942 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10359, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7943 = torch.constant.int 0
%10360 = torch.aten.unsqueeze %10359, %int0_7943 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10360, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7944 = torch.constant.int 2
%10361 = torch.aten.unsqueeze %10360, %int2_7944 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10361, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7945 = torch.constant.int 3
%int0_7946 = torch.constant.int 0
%int9223372036854775807_7947 = torch.constant.int 9223372036854775807
%int1_7948 = torch.constant.int 1
%10362 = torch.aten.slice.Tensor %10361, %int3_7945, %int0_7946, %int9223372036854775807_7947, %int1_7948 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10362, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10363 = torch_c.to_builtin_tensor %10129 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_7949 = arith.constant 1 : index
%dim_7950 = tensor.dim %10363, %c1_7949 : tensor<4x?x1x128xf16>
%10364 = flow.tensor.bitcast %10363 : tensor<4x?x1x128xf16>{%dim_7950} -> tensor<4x?x1x64xcomplex<f16>>{%dim_7950}
%10365 = torch_c.from_builtin_tensor %10364 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10366 = torch.aten.mul.Tensor %10365, %10362 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10367 = torch_c.to_builtin_tensor %10366 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_7951 = arith.constant 1 : index
%dim_7952 = tensor.dim %10367, %c1_7951 : tensor<4x?x1x64xcomplex<f32>>
%10368 = flow.tensor.bitcast %10367 : tensor<4x?x1x64xcomplex<f32>>{%dim_7952} -> tensor<4x?x1x128xf32>{%dim_7952}
%10369 = torch_c.from_builtin_tensor %10368 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_7953 = torch.constant.int 5
%10370 = torch.prims.convert_element_type %10369, %int5_7953 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_7954 = torch.constant.int 1
%10371 = torch.aten.size.int %10011, %int1_7954 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_7955 = torch.constant.int 0
%10372 = torch.aten.add.int %int0_7955, %10371 : !torch.int, !torch.int -> !torch.int
%int0_7956 = torch.constant.int 0
%int0_7957 = torch.constant.int 0
%int1_7958 = torch.constant.int 1
%10373 = torch.aten.slice.Tensor %10337, %int0_7956, %int0_7957, %10372, %int1_7958 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10373, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7959 = torch.constant.int 1
%int0_7960 = torch.constant.int 0
%int9223372036854775807_7961 = torch.constant.int 9223372036854775807
%int1_7962 = torch.constant.int 1
%10374 = torch.aten.slice.Tensor %10373, %int1_7959, %int0_7960, %int9223372036854775807_7961, %int1_7962 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10374, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7963 = torch.constant.int 0
%10375 = torch.aten.unsqueeze %10374, %int0_7963 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10375, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7964 = torch.constant.int 2
%10376 = torch.aten.unsqueeze %10375, %int2_7964 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10376, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7965 = torch.constant.int 3
%int0_7966 = torch.constant.int 0
%int9223372036854775807_7967 = torch.constant.int 9223372036854775807
%int1_7968 = torch.constant.int 1
%10377 = torch.aten.slice.Tensor %10376, %int3_7965, %int0_7966, %int9223372036854775807_7967, %int1_7968 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10377, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10378 = torch_c.to_builtin_tensor %10131 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_7969 = arith.constant 1 : index
%dim_7970 = tensor.dim %10378, %c1_7969 : tensor<4x?x1x128xf16>
%10379 = flow.tensor.bitcast %10378 : tensor<4x?x1x128xf16>{%dim_7970} -> tensor<4x?x1x64xcomplex<f16>>{%dim_7970}
%10380 = torch_c.from_builtin_tensor %10379 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10381 = torch.aten.mul.Tensor %10380, %10377 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10382 = torch_c.to_builtin_tensor %10381 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_7971 = arith.constant 1 : index
%dim_7972 = tensor.dim %10382, %c1_7971 : tensor<4x?x1x64xcomplex<f32>>
%10383 = flow.tensor.bitcast %10382 : tensor<4x?x1x64xcomplex<f32>>{%dim_7972} -> tensor<4x?x1x128xf32>{%dim_7972}
%10384 = torch_c.from_builtin_tensor %10383 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_7973 = torch.constant.int 5
%10385 = torch.prims.convert_element_type %10384, %int5_7973 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_7974 = torch.constant.int 1
%10386 = torch.aten.size.int %10017, %int1_7974 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_7975 = torch.constant.int 0
%10387 = torch.aten.add.int %int0_7975, %10386 : !torch.int, !torch.int -> !torch.int
%int0_7976 = torch.constant.int 0
%int0_7977 = torch.constant.int 0
%int1_7978 = torch.constant.int 1
%10388 = torch.aten.slice.Tensor %10340, %int0_7976, %int0_7977, %10387, %int1_7978 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10388, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7979 = torch.constant.int 1
%int0_7980 = torch.constant.int 0
%int9223372036854775807_7981 = torch.constant.int 9223372036854775807
%int1_7982 = torch.constant.int 1
%10389 = torch.aten.slice.Tensor %10388, %int1_7979, %int0_7980, %int9223372036854775807_7981, %int1_7982 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10389, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_7983 = torch.constant.int 0
%10390 = torch.aten.unsqueeze %10389, %int0_7983 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10390, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_7984 = torch.constant.int 2
%10391 = torch.aten.unsqueeze %10390, %int2_7984 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10391, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_7985 = torch.constant.int 3
%int0_7986 = torch.constant.int 0
%int9223372036854775807_7987 = torch.constant.int 9223372036854775807
%int1_7988 = torch.constant.int 1
%10392 = torch.aten.slice.Tensor %10391, %int3_7985, %int0_7986, %int9223372036854775807_7987, %int1_7988 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10392, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10393 = torch_c.to_builtin_tensor %10133 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_7989 = arith.constant 1 : index
%dim_7990 = tensor.dim %10393, %c1_7989 : tensor<4x?x1x128xf16>
%10394 = flow.tensor.bitcast %10393 : tensor<4x?x1x128xf16>{%dim_7990} -> tensor<4x?x1x64xcomplex<f16>>{%dim_7990}
%10395 = torch_c.from_builtin_tensor %10394 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10396 = torch.aten.mul.Tensor %10395, %10392 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10397 = torch_c.to_builtin_tensor %10396 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_7991 = arith.constant 1 : index
%dim_7992 = tensor.dim %10397, %c1_7991 : tensor<4x?x1x64xcomplex<f32>>
%10398 = flow.tensor.bitcast %10397 : tensor<4x?x1x64xcomplex<f32>>{%dim_7992} -> tensor<4x?x1x128xf32>{%dim_7992}
%10399 = torch_c.from_builtin_tensor %10398 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_7993 = torch.constant.int 5
%10400 = torch.prims.convert_element_type %10399, %int5_7993 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_7994 = torch.constant.int 1
%10401 = torch.aten.size.int %10023, %int1_7994 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_7995 = torch.constant.int 0
%10402 = torch.aten.add.int %int0_7995, %10401 : !torch.int, !torch.int -> !torch.int
%int0_7996 = torch.constant.int 0
%int0_7997 = torch.constant.int 0
%int1_7998 = torch.constant.int 1
%10403 = torch.aten.slice.Tensor %10343, %int0_7996, %int0_7997, %10402, %int1_7998 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10403, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_7999 = torch.constant.int 1
%int0_8000 = torch.constant.int 0
%int9223372036854775807_8001 = torch.constant.int 9223372036854775807
%int1_8002 = torch.constant.int 1
%10404 = torch.aten.slice.Tensor %10403, %int1_7999, %int0_8000, %int9223372036854775807_8001, %int1_8002 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10404, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_8003 = torch.constant.int 0
%10405 = torch.aten.unsqueeze %10404, %int0_8003 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10405, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_8004 = torch.constant.int 2
%10406 = torch.aten.unsqueeze %10405, %int2_8004 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10406, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_8005 = torch.constant.int 3
%int0_8006 = torch.constant.int 0
%int9223372036854775807_8007 = torch.constant.int 9223372036854775807
%int1_8008 = torch.constant.int 1
%10407 = torch.aten.slice.Tensor %10406, %int3_8005, %int0_8006, %int9223372036854775807_8007, %int1_8008 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10407, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10408 = torch_c.to_builtin_tensor %10135 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_8009 = arith.constant 1 : index
%dim_8010 = tensor.dim %10408, %c1_8009 : tensor<4x?x1x128xf16>
%10409 = flow.tensor.bitcast %10408 : tensor<4x?x1x128xf16>{%dim_8010} -> tensor<4x?x1x64xcomplex<f16>>{%dim_8010}
%10410 = torch_c.from_builtin_tensor %10409 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10411 = torch.aten.mul.Tensor %10410, %10407 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10412 = torch_c.to_builtin_tensor %10411 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_8011 = arith.constant 1 : index
%dim_8012 = tensor.dim %10412, %c1_8011 : tensor<4x?x1x64xcomplex<f32>>
%10413 = flow.tensor.bitcast %10412 : tensor<4x?x1x64xcomplex<f32>>{%dim_8012} -> tensor<4x?x1x128xf32>{%dim_8012}
%10414 = torch_c.from_builtin_tensor %10413 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_8013 = torch.constant.int 5
%10415 = torch.prims.convert_element_type %10414, %int5_8013 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_8014 = torch.constant.int 1
%10416 = torch.aten.size.int %10029, %int1_8014 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_8015 = torch.constant.int 0
%10417 = torch.aten.add.int %int0_8015, %10416 : !torch.int, !torch.int -> !torch.int
%int0_8016 = torch.constant.int 0
%int0_8017 = torch.constant.int 0
%int1_8018 = torch.constant.int 1
%10418 = torch.aten.slice.Tensor %10346, %int0_8016, %int0_8017, %10417, %int1_8018 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10418, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_8019 = torch.constant.int 1
%int0_8020 = torch.constant.int 0
%int9223372036854775807_8021 = torch.constant.int 9223372036854775807
%int1_8022 = torch.constant.int 1
%10419 = torch.aten.slice.Tensor %10418, %int1_8019, %int0_8020, %int9223372036854775807_8021, %int1_8022 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10419, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_8023 = torch.constant.int 0
%10420 = torch.aten.unsqueeze %10419, %int0_8023 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10420, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_8024 = torch.constant.int 2
%10421 = torch.aten.unsqueeze %10420, %int2_8024 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10421, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_8025 = torch.constant.int 3
%int0_8026 = torch.constant.int 0
%int9223372036854775807_8027 = torch.constant.int 9223372036854775807
%int1_8028 = torch.constant.int 1
%10422 = torch.aten.slice.Tensor %10421, %int3_8025, %int0_8026, %int9223372036854775807_8027, %int1_8028 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10422, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10423 = torch_c.to_builtin_tensor %10137 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_8029 = arith.constant 1 : index
%dim_8030 = tensor.dim %10423, %c1_8029 : tensor<4x?x1x128xf16>
%10424 = flow.tensor.bitcast %10423 : tensor<4x?x1x128xf16>{%dim_8030} -> tensor<4x?x1x64xcomplex<f16>>{%dim_8030}
%10425 = torch_c.from_builtin_tensor %10424 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10426 = torch.aten.mul.Tensor %10425, %10422 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10427 = torch_c.to_builtin_tensor %10426 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_8031 = arith.constant 1 : index
%dim_8032 = tensor.dim %10427, %c1_8031 : tensor<4x?x1x64xcomplex<f32>>
%10428 = flow.tensor.bitcast %10427 : tensor<4x?x1x64xcomplex<f32>>{%dim_8032} -> tensor<4x?x1x128xf32>{%dim_8032}
%10429 = torch_c.from_builtin_tensor %10428 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_8033 = torch.constant.int 5
%10430 = torch.prims.convert_element_type %10429, %int5_8033 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_8034 = torch.constant.int 1
%10431 = torch.aten.size.int %10035, %int1_8034 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_8035 = torch.constant.int 0
%10432 = torch.aten.add.int %int0_8035, %10431 : !torch.int, !torch.int -> !torch.int
%int0_8036 = torch.constant.int 0
%int0_8037 = torch.constant.int 0
%int1_8038 = torch.constant.int 1
%10433 = torch.aten.slice.Tensor %10349, %int0_8036, %int0_8037, %10432, %int1_8038 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10433, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_8039 = torch.constant.int 1
%int0_8040 = torch.constant.int 0
%int9223372036854775807_8041 = torch.constant.int 9223372036854775807
%int1_8042 = torch.constant.int 1
%10434 = torch.aten.slice.Tensor %10433, %int1_8039, %int0_8040, %int9223372036854775807_8041, %int1_8042 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10434, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_8043 = torch.constant.int 0
%10435 = torch.aten.unsqueeze %10434, %int0_8043 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10435, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_8044 = torch.constant.int 2
%10436 = torch.aten.unsqueeze %10435, %int2_8044 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10436, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_8045 = torch.constant.int 3
%int0_8046 = torch.constant.int 0
%int9223372036854775807_8047 = torch.constant.int 9223372036854775807
%int1_8048 = torch.constant.int 1
%10437 = torch.aten.slice.Tensor %10436, %int3_8045, %int0_8046, %int9223372036854775807_8047, %int1_8048 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10437, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10438 = torch_c.to_builtin_tensor %10139 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_8049 = arith.constant 1 : index
%dim_8050 = tensor.dim %10438, %c1_8049 : tensor<4x?x1x128xf16>
%10439 = flow.tensor.bitcast %10438 : tensor<4x?x1x128xf16>{%dim_8050} -> tensor<4x?x1x64xcomplex<f16>>{%dim_8050}
%10440 = torch_c.from_builtin_tensor %10439 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10441 = torch.aten.mul.Tensor %10440, %10437 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10442 = torch_c.to_builtin_tensor %10441 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_8051 = arith.constant 1 : index
%dim_8052 = tensor.dim %10442, %c1_8051 : tensor<4x?x1x64xcomplex<f32>>
%10443 = flow.tensor.bitcast %10442 : tensor<4x?x1x64xcomplex<f32>>{%dim_8052} -> tensor<4x?x1x128xf32>{%dim_8052}
%10444 = torch_c.from_builtin_tensor %10443 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_8053 = torch.constant.int 5
%10445 = torch.prims.convert_element_type %10444, %int5_8053 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_8054 = torch.constant.int 1
%10446 = torch.aten.size.int %10041, %int1_8054 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_8055 = torch.constant.int 0
%10447 = torch.aten.add.int %int0_8055, %10446 : !torch.int, !torch.int -> !torch.int
%int0_8056 = torch.constant.int 0
%int0_8057 = torch.constant.int 0
%int1_8058 = torch.constant.int 1
%10448 = torch.aten.slice.Tensor %10352, %int0_8056, %int0_8057, %10447, %int1_8058 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10448, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_8059 = torch.constant.int 1
%int0_8060 = torch.constant.int 0
%int9223372036854775807_8061 = torch.constant.int 9223372036854775807
%int1_8062 = torch.constant.int 1
%10449 = torch.aten.slice.Tensor %10448, %int1_8059, %int0_8060, %int9223372036854775807_8061, %int1_8062 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10449, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_8063 = torch.constant.int 0
%10450 = torch.aten.unsqueeze %10449, %int0_8063 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10450, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_8064 = torch.constant.int 2
%10451 = torch.aten.unsqueeze %10450, %int2_8064 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10451, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_8065 = torch.constant.int 3
%int0_8066 = torch.constant.int 0
%int9223372036854775807_8067 = torch.constant.int 9223372036854775807
%int1_8068 = torch.constant.int 1
%10452 = torch.aten.slice.Tensor %10451, %int3_8065, %int0_8066, %int9223372036854775807_8067, %int1_8068 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10452, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10453 = torch_c.to_builtin_tensor %10141 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_8069 = arith.constant 1 : index
%dim_8070 = tensor.dim %10453, %c1_8069 : tensor<4x?x1x128xf16>
%10454 = flow.tensor.bitcast %10453 : tensor<4x?x1x128xf16>{%dim_8070} -> tensor<4x?x1x64xcomplex<f16>>{%dim_8070}
%10455 = torch_c.from_builtin_tensor %10454 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10456 = torch.aten.mul.Tensor %10455, %10452 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10457 = torch_c.to_builtin_tensor %10456 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_8071 = arith.constant 1 : index
%dim_8072 = tensor.dim %10457, %c1_8071 : tensor<4x?x1x64xcomplex<f32>>
%10458 = flow.tensor.bitcast %10457 : tensor<4x?x1x64xcomplex<f32>>{%dim_8072} -> tensor<4x?x1x128xf32>{%dim_8072}
%10459 = torch_c.from_builtin_tensor %10458 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_8073 = torch.constant.int 5
%10460 = torch.prims.convert_element_type %10459, %int5_8073 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_8074 = torch.constant.int 1
%10461 = torch.aten.size.int %10047, %int1_8074 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_8075 = torch.constant.int 0
%10462 = torch.aten.add.int %int0_8075, %10461 : !torch.int, !torch.int -> !torch.int
%int0_8076 = torch.constant.int 0
%int0_8077 = torch.constant.int 0
%int1_8078 = torch.constant.int 1
%10463 = torch.aten.slice.Tensor %10355, %int0_8076, %int0_8077, %10462, %int1_8078 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10463, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_8079 = torch.constant.int 1
%int0_8080 = torch.constant.int 0
%int9223372036854775807_8081 = torch.constant.int 9223372036854775807
%int1_8082 = torch.constant.int 1
%10464 = torch.aten.slice.Tensor %10463, %int1_8079, %int0_8080, %int9223372036854775807_8081, %int1_8082 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %10464, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_8083 = torch.constant.int 0
%10465 = torch.aten.unsqueeze %10464, %int0_8083 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %10465, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_8084 = torch.constant.int 2
%10466 = torch.aten.unsqueeze %10465, %int2_8084 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10466, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_8085 = torch.constant.int 3
%int0_8086 = torch.constant.int 0
%int9223372036854775807_8087 = torch.constant.int 9223372036854775807
%int1_8088 = torch.constant.int 1
%10467 = torch.aten.slice.Tensor %10466, %int3_8085, %int0_8086, %int9223372036854775807_8087, %int1_8088 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10467, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%10468 = torch_c.to_builtin_tensor %10143 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_8089 = arith.constant 1 : index
%dim_8090 = tensor.dim %10468, %c1_8089 : tensor<4x?x1x128xf16>
%10469 = flow.tensor.bitcast %10468 : tensor<4x?x1x128xf16>{%dim_8090} -> tensor<4x?x1x64xcomplex<f16>>{%dim_8090}
%10470 = torch_c.from_builtin_tensor %10469 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %10470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%10471 = torch.aten.mul.Tensor %10470, %10467 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %10471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%10472 = torch_c.to_builtin_tensor %10471 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_8091 = arith.constant 1 : index
%dim_8092 = tensor.dim %10472, %c1_8091 : tensor<4x?x1x64xcomplex<f32>>
%10473 = flow.tensor.bitcast %10472 : tensor<4x?x1x64xcomplex<f32>>{%dim_8092} -> tensor<4x?x1x128xf32>{%dim_8092}
%10474 = torch_c.from_builtin_tensor %10473 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %10474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_8093 = torch.constant.int 5
%10475 = torch.prims.convert_element_type %10474, %int5_8093 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %10475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_8094 = torch.constant.int 64
%10476 = torch.aten.mul.Scalar %2364, %int64_8094 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10476, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8095 = torch.constant.int 64
%10477 = torch.aten.mul.Scalar %2367, %int64_8095 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10477, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8096 = torch.constant.int 64
%10478 = torch.aten.mul.Scalar %2370, %int64_8096 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10478, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8097 = torch.constant.int 64
%10479 = torch.aten.mul.Scalar %2373, %int64_8097 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10479, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8098 = torch.constant.int 64
%10480 = torch.aten.mul.Scalar %2376, %int64_8098 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10480, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8099 = torch.constant.int 64
%10481 = torch.aten.mul.Scalar %2379, %int64_8099 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10481, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8100 = torch.constant.int 64
%10482 = torch.aten.mul.Scalar %2382, %int64_8100 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10482, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_8101 = torch.constant.int 64
%10483 = torch.aten.mul.Scalar %2385, %int64_8101 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10483, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8 = torch.constant.int 8
%int1_8102 = torch.constant.int 1
%10484 = torch.aten.add.Scalar %10476, %int8, %int1_8102 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10484, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8103 = torch.constant.int 8
%int1_8104 = torch.constant.int 1
%10485 = torch.aten.add.Scalar %10477, %int8_8103, %int1_8104 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10485, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8105 = torch.constant.int 8
%int1_8106 = torch.constant.int 1
%10486 = torch.aten.add.Scalar %10478, %int8_8105, %int1_8106 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10486, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8107 = torch.constant.int 8
%int1_8108 = torch.constant.int 1
%10487 = torch.aten.add.Scalar %10479, %int8_8107, %int1_8108 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10487, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8109 = torch.constant.int 8
%int1_8110 = torch.constant.int 1
%10488 = torch.aten.add.Scalar %10480, %int8_8109, %int1_8110 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10488, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8111 = torch.constant.int 8
%int1_8112 = torch.constant.int 1
%10489 = torch.aten.add.Scalar %10481, %int8_8111, %int1_8112 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10489, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8113 = torch.constant.int 8
%int1_8114 = torch.constant.int 1
%10490 = torch.aten.add.Scalar %10482, %int8_8113, %int1_8114 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10490, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int8_8115 = torch.constant.int 8
%int1_8116 = torch.constant.int 1
%10491 = torch.aten.add.Scalar %10483, %int8_8115, %int1_8116 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10491, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_8117 = torch.constant.int 4
%int16_8118 = torch.constant.int 16
%int1_8119 = torch.constant.int 1
%int128_8120 = torch.constant.int 128
%10492 = torch.prim.ListConstruct %int4_8117, %3095, %int16_8118, %int1_8119, %int128_8120 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10493 = torch.aten.view %10370, %10492 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10493, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8121 = torch.constant.int 4
%int16_8122 = torch.constant.int 16
%int1_8123 = torch.constant.int 1
%int128_8124 = torch.constant.int 128
%10494 = torch.prim.ListConstruct %int4_8121, %3095, %int16_8122, %int1_8123, %int128_8124 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10495 = torch.aten.view %10385, %10494 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10495, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8125 = torch.constant.int 4
%int16_8126 = torch.constant.int 16
%int1_8127 = torch.constant.int 1
%int128_8128 = torch.constant.int 128
%10496 = torch.prim.ListConstruct %int4_8125, %3095, %int16_8126, %int1_8127, %int128_8128 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10497 = torch.aten.view %10400, %10496 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10497, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8129 = torch.constant.int 4
%int16_8130 = torch.constant.int 16
%int1_8131 = torch.constant.int 1
%int128_8132 = torch.constant.int 128
%10498 = torch.prim.ListConstruct %int4_8129, %3095, %int16_8130, %int1_8131, %int128_8132 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10499 = torch.aten.view %10415, %10498 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10499, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8133 = torch.constant.int 4
%int16_8134 = torch.constant.int 16
%int1_8135 = torch.constant.int 1
%int128_8136 = torch.constant.int 128
%10500 = torch.prim.ListConstruct %int4_8133, %3095, %int16_8134, %int1_8135, %int128_8136 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10501 = torch.aten.view %10430, %10500 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10501, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8137 = torch.constant.int 4
%int16_8138 = torch.constant.int 16
%int1_8139 = torch.constant.int 1
%int128_8140 = torch.constant.int 128
%10502 = torch.prim.ListConstruct %int4_8137, %3095, %int16_8138, %int1_8139, %int128_8140 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10503 = torch.aten.view %10445, %10502 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10503, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8141 = torch.constant.int 4
%int16_8142 = torch.constant.int 16
%int1_8143 = torch.constant.int 1
%int128_8144 = torch.constant.int 128
%10504 = torch.prim.ListConstruct %int4_8141, %3095, %int16_8142, %int1_8143, %int128_8144 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10505 = torch.aten.view %10460, %10504 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10505, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8145 = torch.constant.int 4
%int16_8146 = torch.constant.int 16
%int1_8147 = torch.constant.int 1
%int128_8148 = torch.constant.int 128
%10506 = torch.prim.ListConstruct %int4_8145, %3095, %int16_8146, %int1_8147, %int128_8148 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10507 = torch.aten.view %10475, %10506 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10507, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8149 = torch.constant.int 4
%10508 = torch.aten.mul.int %int4_8149, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8150 = torch.constant.int 16
%int1_8151 = torch.constant.int 1
%int128_8152 = torch.constant.int 128
%10509 = torch.prim.ListConstruct %10508, %int16_8150, %int1_8151, %int128_8152 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10510 = torch.aten.view %10493, %10509 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10510, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8153 = torch.constant.int 4
%10511 = torch.aten.mul.int %int4_8153, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8154 = torch.constant.int 16
%int1_8155 = torch.constant.int 1
%int128_8156 = torch.constant.int 128
%10512 = torch.prim.ListConstruct %10511, %int16_8154, %int1_8155, %int128_8156 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10513 = torch.aten.view %10495, %10512 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10513, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8157 = torch.constant.int 4
%10514 = torch.aten.mul.int %int4_8157, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8158 = torch.constant.int 16
%int1_8159 = torch.constant.int 1
%int128_8160 = torch.constant.int 128
%10515 = torch.prim.ListConstruct %10514, %int16_8158, %int1_8159, %int128_8160 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10516 = torch.aten.view %10497, %10515 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10516, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8161 = torch.constant.int 4
%10517 = torch.aten.mul.int %int4_8161, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8162 = torch.constant.int 16
%int1_8163 = torch.constant.int 1
%int128_8164 = torch.constant.int 128
%10518 = torch.prim.ListConstruct %10517, %int16_8162, %int1_8163, %int128_8164 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10519 = torch.aten.view %10499, %10518 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10519, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8165 = torch.constant.int 4
%10520 = torch.aten.mul.int %int4_8165, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8166 = torch.constant.int 16
%int1_8167 = torch.constant.int 1
%int128_8168 = torch.constant.int 128
%10521 = torch.prim.ListConstruct %10520, %int16_8166, %int1_8167, %int128_8168 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10522 = torch.aten.view %10501, %10521 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10522, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8169 = torch.constant.int 4
%10523 = torch.aten.mul.int %int4_8169, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8170 = torch.constant.int 16
%int1_8171 = torch.constant.int 1
%int128_8172 = torch.constant.int 128
%10524 = torch.prim.ListConstruct %10523, %int16_8170, %int1_8171, %int128_8172 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10525 = torch.aten.view %10503, %10524 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10525, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8173 = torch.constant.int 4
%10526 = torch.aten.mul.int %int4_8173, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8174 = torch.constant.int 16
%int1_8175 = torch.constant.int 1
%int128_8176 = torch.constant.int 128
%10527 = torch.prim.ListConstruct %10526, %int16_8174, %int1_8175, %int128_8176 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10528 = torch.aten.view %10505, %10527 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10528, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8177 = torch.constant.int 4
%10529 = torch.aten.mul.int %int4_8177, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8178 = torch.constant.int 16
%int1_8179 = torch.constant.int 1
%int128_8180 = torch.constant.int 128
%10530 = torch.prim.ListConstruct %10529, %int16_8178, %int1_8179, %int128_8180 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10531 = torch.aten.view %10507, %10530 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10531, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8181 = torch.constant.int 4
%10532 = torch.aten.mul.int %int4_8181, %3095 : !torch.int, !torch.int -> !torch.int
%10533 = torch.prim.ListConstruct %10532 : (!torch.int) -> !torch.list<int>
%10534 = torch.aten.view %10484, %10533 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10534, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8182 = torch.constant.int 4
%10535 = torch.aten.mul.int %int4_8182, %3095 : !torch.int, !torch.int -> !torch.int
%10536 = torch.prim.ListConstruct %10535 : (!torch.int) -> !torch.list<int>
%10537 = torch.aten.view %10485, %10536 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10537, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8183 = torch.constant.int 4
%10538 = torch.aten.mul.int %int4_8183, %3095 : !torch.int, !torch.int -> !torch.int
%10539 = torch.prim.ListConstruct %10538 : (!torch.int) -> !torch.list<int>
%10540 = torch.aten.view %10486, %10539 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10540, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8184 = torch.constant.int 4
%10541 = torch.aten.mul.int %int4_8184, %3095 : !torch.int, !torch.int -> !torch.int
%10542 = torch.prim.ListConstruct %10541 : (!torch.int) -> !torch.list<int>
%10543 = torch.aten.view %10487, %10542 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10543, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8185 = torch.constant.int 4
%10544 = torch.aten.mul.int %int4_8185, %3095 : !torch.int, !torch.int -> !torch.int
%10545 = torch.prim.ListConstruct %10544 : (!torch.int) -> !torch.list<int>
%10546 = torch.aten.view %10488, %10545 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10546, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8186 = torch.constant.int 4
%10547 = torch.aten.mul.int %int4_8186, %3095 : !torch.int, !torch.int -> !torch.int
%10548 = torch.prim.ListConstruct %10547 : (!torch.int) -> !torch.list<int>
%10549 = torch.aten.view %10489, %10548 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10549, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8187 = torch.constant.int 4
%10550 = torch.aten.mul.int %int4_8187, %3095 : !torch.int, !torch.int -> !torch.int
%10551 = torch.prim.ListConstruct %10550 : (!torch.int) -> !torch.list<int>
%10552 = torch.aten.view %10490, %10551 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10552, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8188 = torch.constant.int 4
%10553 = torch.aten.mul.int %int4_8188, %3095 : !torch.int, !torch.int -> !torch.int
%10554 = torch.prim.ListConstruct %10553 : (!torch.int) -> !torch.list<int>
%10555 = torch.aten.view %10491, %10554 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10555, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8189 = torch.constant.int 4
%int16_8190 = torch.constant.int 16
%int1_8191 = torch.constant.int 1
%int128_8192 = torch.constant.int 128
%10556 = torch.prim.ListConstruct %int4_8189, %3095, %int16_8190, %int1_8191, %int128_8192 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10557 = torch.aten.view %10145, %10556 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10557, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8193 = torch.constant.int 4
%int16_8194 = torch.constant.int 16
%int1_8195 = torch.constant.int 1
%int128_8196 = torch.constant.int 128
%10558 = torch.prim.ListConstruct %int4_8193, %3095, %int16_8194, %int1_8195, %int128_8196 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10559 = torch.aten.view %10147, %10558 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10559, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8197 = torch.constant.int 4
%int16_8198 = torch.constant.int 16
%int1_8199 = torch.constant.int 1
%int128_8200 = torch.constant.int 128
%10560 = torch.prim.ListConstruct %int4_8197, %3095, %int16_8198, %int1_8199, %int128_8200 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10561 = torch.aten.view %10149, %10560 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10561, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8201 = torch.constant.int 4
%int16_8202 = torch.constant.int 16
%int1_8203 = torch.constant.int 1
%int128_8204 = torch.constant.int 128
%10562 = torch.prim.ListConstruct %int4_8201, %3095, %int16_8202, %int1_8203, %int128_8204 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10563 = torch.aten.view %10151, %10562 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10563, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8205 = torch.constant.int 4
%int16_8206 = torch.constant.int 16
%int1_8207 = torch.constant.int 1
%int128_8208 = torch.constant.int 128
%10564 = torch.prim.ListConstruct %int4_8205, %3095, %int16_8206, %int1_8207, %int128_8208 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10565 = torch.aten.view %10153, %10564 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10565, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8209 = torch.constant.int 4
%int16_8210 = torch.constant.int 16
%int1_8211 = torch.constant.int 1
%int128_8212 = torch.constant.int 128
%10566 = torch.prim.ListConstruct %int4_8209, %3095, %int16_8210, %int1_8211, %int128_8212 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10567 = torch.aten.view %10155, %10566 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10567, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8213 = torch.constant.int 4
%int16_8214 = torch.constant.int 16
%int1_8215 = torch.constant.int 1
%int128_8216 = torch.constant.int 128
%10568 = torch.prim.ListConstruct %int4_8213, %3095, %int16_8214, %int1_8215, %int128_8216 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10569 = torch.aten.view %10157, %10568 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10569, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8217 = torch.constant.int 4
%int16_8218 = torch.constant.int 16
%int1_8219 = torch.constant.int 1
%int128_8220 = torch.constant.int 128
%10570 = torch.prim.ListConstruct %int4_8217, %3095, %int16_8218, %int1_8219, %int128_8220 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10571 = torch.aten.view %10159, %10570 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %10571, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_8221 = torch.constant.int 4
%10572 = torch.aten.mul.int %int4_8221, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8222 = torch.constant.int 16
%int1_8223 = torch.constant.int 1
%int128_8224 = torch.constant.int 128
%10573 = torch.prim.ListConstruct %10572, %int16_8222, %int1_8223, %int128_8224 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10574 = torch.aten.view %10557, %10573 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10574, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8225 = torch.constant.int 4
%10575 = torch.aten.mul.int %int4_8225, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8226 = torch.constant.int 16
%int1_8227 = torch.constant.int 1
%int128_8228 = torch.constant.int 128
%10576 = torch.prim.ListConstruct %10575, %int16_8226, %int1_8227, %int128_8228 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10577 = torch.aten.view %10559, %10576 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10577, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8229 = torch.constant.int 4
%10578 = torch.aten.mul.int %int4_8229, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8230 = torch.constant.int 16
%int1_8231 = torch.constant.int 1
%int128_8232 = torch.constant.int 128
%10579 = torch.prim.ListConstruct %10578, %int16_8230, %int1_8231, %int128_8232 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10580 = torch.aten.view %10561, %10579 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10580, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8233 = torch.constant.int 4
%10581 = torch.aten.mul.int %int4_8233, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8234 = torch.constant.int 16
%int1_8235 = torch.constant.int 1
%int128_8236 = torch.constant.int 128
%10582 = torch.prim.ListConstruct %10581, %int16_8234, %int1_8235, %int128_8236 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10583 = torch.aten.view %10563, %10582 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10583, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8237 = torch.constant.int 4
%10584 = torch.aten.mul.int %int4_8237, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8238 = torch.constant.int 16
%int1_8239 = torch.constant.int 1
%int128_8240 = torch.constant.int 128
%10585 = torch.prim.ListConstruct %10584, %int16_8238, %int1_8239, %int128_8240 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10586 = torch.aten.view %10565, %10585 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10586, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8241 = torch.constant.int 4
%10587 = torch.aten.mul.int %int4_8241, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8242 = torch.constant.int 16
%int1_8243 = torch.constant.int 1
%int128_8244 = torch.constant.int 128
%10588 = torch.prim.ListConstruct %10587, %int16_8242, %int1_8243, %int128_8244 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10589 = torch.aten.view %10567, %10588 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10589, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8245 = torch.constant.int 4
%10590 = torch.aten.mul.int %int4_8245, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8246 = torch.constant.int 16
%int1_8247 = torch.constant.int 1
%int128_8248 = torch.constant.int 128
%10591 = torch.prim.ListConstruct %10590, %int16_8246, %int1_8247, %int128_8248 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10592 = torch.aten.view %10569, %10591 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10592, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_8249 = torch.constant.int 4
%10593 = torch.aten.mul.int %int4_8249, %3095 : !torch.int, !torch.int -> !torch.int
%int16_8250 = torch.constant.int 16
%int1_8251 = torch.constant.int 1
%int128_8252 = torch.constant.int 128
%10594 = torch.prim.ListConstruct %10593, %int16_8250, %int1_8251, %int128_8252 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10595 = torch.aten.view %10571, %10594 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10595, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_8253 = torch.constant.int 1
%int1_8254 = torch.constant.int 1
%10596 = torch.aten.add.Scalar %10484, %int1_8253, %int1_8254 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10596, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8255 = torch.constant.int 1
%int1_8256 = torch.constant.int 1
%10597 = torch.aten.add.Scalar %10485, %int1_8255, %int1_8256 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10597, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8257 = torch.constant.int 1
%int1_8258 = torch.constant.int 1
%10598 = torch.aten.add.Scalar %10486, %int1_8257, %int1_8258 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10598, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8259 = torch.constant.int 1
%int1_8260 = torch.constant.int 1
%10599 = torch.aten.add.Scalar %10487, %int1_8259, %int1_8260 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10599, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8261 = torch.constant.int 1
%int1_8262 = torch.constant.int 1
%10600 = torch.aten.add.Scalar %10488, %int1_8261, %int1_8262 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10600, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8263 = torch.constant.int 1
%int1_8264 = torch.constant.int 1
%10601 = torch.aten.add.Scalar %10489, %int1_8263, %int1_8264 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10601, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8265 = torch.constant.int 1
%int1_8266 = torch.constant.int 1
%10602 = torch.aten.add.Scalar %10490, %int1_8265, %int1_8266 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10602, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_8267 = torch.constant.int 1
%int1_8268 = torch.constant.int 1
%10603 = torch.aten.add.Scalar %10491, %int1_8267, %int1_8268 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %10603, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_8269 = torch.constant.int 4
%10604 = torch.aten.mul.int %int4_8269, %3095 : !torch.int, !torch.int -> !torch.int
%10605 = torch.prim.ListConstruct %10604 : (!torch.int) -> !torch.list<int>
%10606 = torch.aten.view %10596, %10605 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10606, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8270 = torch.constant.int 4
%10607 = torch.aten.mul.int %int4_8270, %3095 : !torch.int, !torch.int -> !torch.int
%10608 = torch.prim.ListConstruct %10607 : (!torch.int) -> !torch.list<int>
%10609 = torch.aten.view %10597, %10608 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10609, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8271 = torch.constant.int 4
%10610 = torch.aten.mul.int %int4_8271, %3095 : !torch.int, !torch.int -> !torch.int
%10611 = torch.prim.ListConstruct %10610 : (!torch.int) -> !torch.list<int>
%10612 = torch.aten.view %10598, %10611 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10612, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8272 = torch.constant.int 4
%10613 = torch.aten.mul.int %int4_8272, %3095 : !torch.int, !torch.int -> !torch.int
%10614 = torch.prim.ListConstruct %10613 : (!torch.int) -> !torch.list<int>
%10615 = torch.aten.view %10599, %10614 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10615, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8273 = torch.constant.int 4
%10616 = torch.aten.mul.int %int4_8273, %3095 : !torch.int, !torch.int -> !torch.int
%10617 = torch.prim.ListConstruct %10616 : (!torch.int) -> !torch.list<int>
%10618 = torch.aten.view %10600, %10617 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10618, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8274 = torch.constant.int 4
%10619 = torch.aten.mul.int %int4_8274, %3095 : !torch.int, !torch.int -> !torch.int
%10620 = torch.prim.ListConstruct %10619 : (!torch.int) -> !torch.list<int>
%10621 = torch.aten.view %10601, %10620 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10621, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8275 = torch.constant.int 4
%10622 = torch.aten.mul.int %int4_8275, %3095 : !torch.int, !torch.int -> !torch.int
%10623 = torch.prim.ListConstruct %10622 : (!torch.int) -> !torch.list<int>
%10624 = torch.aten.view %10602, %10623 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10624, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_8276 = torch.constant.int 4
%10625 = torch.aten.mul.int %int4_8276, %3095 : !torch.int, !torch.int -> !torch.int
%10626 = torch.prim.ListConstruct %10625 : (!torch.int) -> !torch.list<int>
%10627 = torch.aten.view %10603, %10626 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10627, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%10628 = torch.prim.ListConstruct %10534, %10606 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8277 = torch.constant.int 0
%10629 = torch.aten.cat %10628, %int0_8277 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10629, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10630 = torch.prim.ListConstruct %10537, %10609 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8278 = torch.constant.int 0
%10631 = torch.aten.cat %10630, %int0_8278 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10631, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10632 = torch.prim.ListConstruct %10540, %10612 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8279 = torch.constant.int 0
%10633 = torch.aten.cat %10632, %int0_8279 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10633, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10634 = torch.prim.ListConstruct %10543, %10615 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8280 = torch.constant.int 0
%10635 = torch.aten.cat %10634, %int0_8280 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10635, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10636 = torch.prim.ListConstruct %10546, %10618 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8281 = torch.constant.int 0
%10637 = torch.aten.cat %10636, %int0_8281 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10637, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10638 = torch.prim.ListConstruct %10549, %10621 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8282 = torch.constant.int 0
%10639 = torch.aten.cat %10638, %int0_8282 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10639, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10640 = torch.prim.ListConstruct %10552, %10624 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8283 = torch.constant.int 0
%10641 = torch.aten.cat %10640, %int0_8283 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10641, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10642 = torch.prim.ListConstruct %10555, %10627 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_8284 = torch.constant.int 0
%10643 = torch.aten.cat %10642, %int0_8284 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %10643, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%10644 = torch.prim.ListConstruct %10510, %10574 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8285 = torch.constant.int 0
%10645 = torch.aten.cat %10644, %int0_8285 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10645, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10646 = torch.prim.ListConstruct %10513, %10577 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8286 = torch.constant.int 0
%10647 = torch.aten.cat %10646, %int0_8286 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10647, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10648 = torch.prim.ListConstruct %10516, %10580 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8287 = torch.constant.int 0
%10649 = torch.aten.cat %10648, %int0_8287 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10649, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10650 = torch.prim.ListConstruct %10519, %10583 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8288 = torch.constant.int 0
%10651 = torch.aten.cat %10650, %int0_8288 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10651, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10652 = torch.prim.ListConstruct %10522, %10586 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8289 = torch.constant.int 0
%10653 = torch.aten.cat %10652, %int0_8289 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10653, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10654 = torch.prim.ListConstruct %10525, %10589 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8290 = torch.constant.int 0
%10655 = torch.aten.cat %10654, %int0_8290 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10655, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10656 = torch.prim.ListConstruct %10528, %10592 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8291 = torch.constant.int 0
%10657 = torch.aten.cat %10656, %int0_8291 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10657, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10658 = torch.prim.ListConstruct %10531, %10595 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_8292 = torch.constant.int 0
%10659 = torch.aten.cat %10658, %int0_8292 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10659, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8293 = torch.constant.int 32
%int2_8294 = torch.constant.int 2
%int16_8295 = torch.constant.int 16
%int1_8296 = torch.constant.int 1
%int128_8297 = torch.constant.int 128
%10660 = torch.prim.ListConstruct %3023, %int32_8293, %int2_8294, %int16_8295, %int1_8296, %int128_8297 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10661 = torch.aten.view %8810, %10660 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10661, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8298 = torch.constant.int 32
%10662 = torch.aten.mul.int %3023, %int32_8298 : !torch.int, !torch.int -> !torch.int
%int2_8299 = torch.constant.int 2
%10663 = torch.aten.mul.int %10662, %int2_8299 : !torch.int, !torch.int -> !torch.int
%int16_8300 = torch.constant.int 16
%int1_8301 = torch.constant.int 1
%int128_8302 = torch.constant.int 128
%10664 = torch.prim.ListConstruct %10663, %int16_8300, %int1_8301, %int128_8302 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10665 = torch.aten.view %10661, %10664 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10665, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10666 = torch.prim.ListConstruct %10629 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8303 = torch.constant.bool false
%10667 = torch.aten.index_put %10665, %10666, %10645, %false_8303 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10667, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8304 = torch.constant.int 32
%int2_8305 = torch.constant.int 2
%int16_8306 = torch.constant.int 16
%int1_8307 = torch.constant.int 1
%int128_8308 = torch.constant.int 128
%10668 = torch.prim.ListConstruct %3023, %int32_8304, %int2_8305, %int16_8306, %int1_8307, %int128_8308 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10669 = torch.aten.view %10667, %10668 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10669, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8309 = torch.constant.int 131072
%10670 = torch.prim.ListConstruct %3023, %int131072_8309 : (!torch.int, !torch.int) -> !torch.list<int>
%10671 = torch.aten.view %10669, %10670 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10671, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8310 = torch.constant.int 32
%int2_8311 = torch.constant.int 2
%int16_8312 = torch.constant.int 16
%int1_8313 = torch.constant.int 1
%int128_8314 = torch.constant.int 128
%10672 = torch.prim.ListConstruct %3026, %int32_8310, %int2_8311, %int16_8312, %int1_8313, %int128_8314 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10673 = torch.aten.view %8822, %10672 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10673, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8315 = torch.constant.int 32
%10674 = torch.aten.mul.int %3026, %int32_8315 : !torch.int, !torch.int -> !torch.int
%int2_8316 = torch.constant.int 2
%10675 = torch.aten.mul.int %10674, %int2_8316 : !torch.int, !torch.int -> !torch.int
%int16_8317 = torch.constant.int 16
%int1_8318 = torch.constant.int 1
%int128_8319 = torch.constant.int 128
%10676 = torch.prim.ListConstruct %10675, %int16_8317, %int1_8318, %int128_8319 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10677 = torch.aten.view %10673, %10676 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10677, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10678 = torch.prim.ListConstruct %10631 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8320 = torch.constant.bool false
%10679 = torch.aten.index_put %10677, %10678, %10647, %false_8320 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10679, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8321 = torch.constant.int 32
%int2_8322 = torch.constant.int 2
%int16_8323 = torch.constant.int 16
%int1_8324 = torch.constant.int 1
%int128_8325 = torch.constant.int 128
%10680 = torch.prim.ListConstruct %3026, %int32_8321, %int2_8322, %int16_8323, %int1_8324, %int128_8325 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10681 = torch.aten.view %10679, %10680 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10681, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8326 = torch.constant.int 131072
%10682 = torch.prim.ListConstruct %3026, %int131072_8326 : (!torch.int, !torch.int) -> !torch.list<int>
%10683 = torch.aten.view %10681, %10682 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10683, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8327 = torch.constant.int 32
%int2_8328 = torch.constant.int 2
%int16_8329 = torch.constant.int 16
%int1_8330 = torch.constant.int 1
%int128_8331 = torch.constant.int 128
%10684 = torch.prim.ListConstruct %3029, %int32_8327, %int2_8328, %int16_8329, %int1_8330, %int128_8331 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10685 = torch.aten.view %8834, %10684 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10685, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8332 = torch.constant.int 32
%10686 = torch.aten.mul.int %3029, %int32_8332 : !torch.int, !torch.int -> !torch.int
%int2_8333 = torch.constant.int 2
%10687 = torch.aten.mul.int %10686, %int2_8333 : !torch.int, !torch.int -> !torch.int
%int16_8334 = torch.constant.int 16
%int1_8335 = torch.constant.int 1
%int128_8336 = torch.constant.int 128
%10688 = torch.prim.ListConstruct %10687, %int16_8334, %int1_8335, %int128_8336 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10689 = torch.aten.view %10685, %10688 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10689, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10690 = torch.prim.ListConstruct %10633 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8337 = torch.constant.bool false
%10691 = torch.aten.index_put %10689, %10690, %10649, %false_8337 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10691, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8338 = torch.constant.int 32
%int2_8339 = torch.constant.int 2
%int16_8340 = torch.constant.int 16
%int1_8341 = torch.constant.int 1
%int128_8342 = torch.constant.int 128
%10692 = torch.prim.ListConstruct %3029, %int32_8338, %int2_8339, %int16_8340, %int1_8341, %int128_8342 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10693 = torch.aten.view %10691, %10692 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10693, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8343 = torch.constant.int 131072
%10694 = torch.prim.ListConstruct %3029, %int131072_8343 : (!torch.int, !torch.int) -> !torch.list<int>
%10695 = torch.aten.view %10693, %10694 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10695, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8344 = torch.constant.int 32
%int2_8345 = torch.constant.int 2
%int16_8346 = torch.constant.int 16
%int1_8347 = torch.constant.int 1
%int128_8348 = torch.constant.int 128
%10696 = torch.prim.ListConstruct %3032, %int32_8344, %int2_8345, %int16_8346, %int1_8347, %int128_8348 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10697 = torch.aten.view %8846, %10696 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10697, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8349 = torch.constant.int 32
%10698 = torch.aten.mul.int %3032, %int32_8349 : !torch.int, !torch.int -> !torch.int
%int2_8350 = torch.constant.int 2
%10699 = torch.aten.mul.int %10698, %int2_8350 : !torch.int, !torch.int -> !torch.int
%int16_8351 = torch.constant.int 16
%int1_8352 = torch.constant.int 1
%int128_8353 = torch.constant.int 128
%10700 = torch.prim.ListConstruct %10699, %int16_8351, %int1_8352, %int128_8353 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10701 = torch.aten.view %10697, %10700 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10701, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10702 = torch.prim.ListConstruct %10635 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8354 = torch.constant.bool false
%10703 = torch.aten.index_put %10701, %10702, %10651, %false_8354 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10703, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8355 = torch.constant.int 32
%int2_8356 = torch.constant.int 2
%int16_8357 = torch.constant.int 16
%int1_8358 = torch.constant.int 1
%int128_8359 = torch.constant.int 128
%10704 = torch.prim.ListConstruct %3032, %int32_8355, %int2_8356, %int16_8357, %int1_8358, %int128_8359 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10705 = torch.aten.view %10703, %10704 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10705, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8360 = torch.constant.int 131072
%10706 = torch.prim.ListConstruct %3032, %int131072_8360 : (!torch.int, !torch.int) -> !torch.list<int>
%10707 = torch.aten.view %10705, %10706 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10707, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8361 = torch.constant.int 32
%int2_8362 = torch.constant.int 2
%int16_8363 = torch.constant.int 16
%int1_8364 = torch.constant.int 1
%int128_8365 = torch.constant.int 128
%10708 = torch.prim.ListConstruct %3035, %int32_8361, %int2_8362, %int16_8363, %int1_8364, %int128_8365 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10709 = torch.aten.view %8858, %10708 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10709, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8366 = torch.constant.int 32
%10710 = torch.aten.mul.int %3035, %int32_8366 : !torch.int, !torch.int -> !torch.int
%int2_8367 = torch.constant.int 2
%10711 = torch.aten.mul.int %10710, %int2_8367 : !torch.int, !torch.int -> !torch.int
%int16_8368 = torch.constant.int 16
%int1_8369 = torch.constant.int 1
%int128_8370 = torch.constant.int 128
%10712 = torch.prim.ListConstruct %10711, %int16_8368, %int1_8369, %int128_8370 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10713 = torch.aten.view %10709, %10712 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10713, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10714 = torch.prim.ListConstruct %10637 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8371 = torch.constant.bool false
%10715 = torch.aten.index_put %10713, %10714, %10653, %false_8371 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10715, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8372 = torch.constant.int 32
%int2_8373 = torch.constant.int 2
%int16_8374 = torch.constant.int 16
%int1_8375 = torch.constant.int 1
%int128_8376 = torch.constant.int 128
%10716 = torch.prim.ListConstruct %3035, %int32_8372, %int2_8373, %int16_8374, %int1_8375, %int128_8376 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10717 = torch.aten.view %10715, %10716 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10717, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8377 = torch.constant.int 131072
%10718 = torch.prim.ListConstruct %3035, %int131072_8377 : (!torch.int, !torch.int) -> !torch.list<int>
%10719 = torch.aten.view %10717, %10718 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10719, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8378 = torch.constant.int 32
%int2_8379 = torch.constant.int 2
%int16_8380 = torch.constant.int 16
%int1_8381 = torch.constant.int 1
%int128_8382 = torch.constant.int 128
%10720 = torch.prim.ListConstruct %3038, %int32_8378, %int2_8379, %int16_8380, %int1_8381, %int128_8382 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10721 = torch.aten.view %8870, %10720 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10721, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8383 = torch.constant.int 32
%10722 = torch.aten.mul.int %3038, %int32_8383 : !torch.int, !torch.int -> !torch.int
%int2_8384 = torch.constant.int 2
%10723 = torch.aten.mul.int %10722, %int2_8384 : !torch.int, !torch.int -> !torch.int
%int16_8385 = torch.constant.int 16
%int1_8386 = torch.constant.int 1
%int128_8387 = torch.constant.int 128
%10724 = torch.prim.ListConstruct %10723, %int16_8385, %int1_8386, %int128_8387 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10725 = torch.aten.view %10721, %10724 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10725, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10726 = torch.prim.ListConstruct %10639 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8388 = torch.constant.bool false
%10727 = torch.aten.index_put %10725, %10726, %10655, %false_8388 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10727, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8389 = torch.constant.int 32
%int2_8390 = torch.constant.int 2
%int16_8391 = torch.constant.int 16
%int1_8392 = torch.constant.int 1
%int128_8393 = torch.constant.int 128
%10728 = torch.prim.ListConstruct %3038, %int32_8389, %int2_8390, %int16_8391, %int1_8392, %int128_8393 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10729 = torch.aten.view %10727, %10728 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10729, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8394 = torch.constant.int 131072
%10730 = torch.prim.ListConstruct %3038, %int131072_8394 : (!torch.int, !torch.int) -> !torch.list<int>
%10731 = torch.aten.view %10729, %10730 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10731, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8395 = torch.constant.int 32
%int2_8396 = torch.constant.int 2
%int16_8397 = torch.constant.int 16
%int1_8398 = torch.constant.int 1
%int128_8399 = torch.constant.int 128
%10732 = torch.prim.ListConstruct %3041, %int32_8395, %int2_8396, %int16_8397, %int1_8398, %int128_8399 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10733 = torch.aten.view %8882, %10732 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10733, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8400 = torch.constant.int 32
%10734 = torch.aten.mul.int %3041, %int32_8400 : !torch.int, !torch.int -> !torch.int
%int2_8401 = torch.constant.int 2
%10735 = torch.aten.mul.int %10734, %int2_8401 : !torch.int, !torch.int -> !torch.int
%int16_8402 = torch.constant.int 16
%int1_8403 = torch.constant.int 1
%int128_8404 = torch.constant.int 128
%10736 = torch.prim.ListConstruct %10735, %int16_8402, %int1_8403, %int128_8404 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10737 = torch.aten.view %10733, %10736 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10737, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10738 = torch.prim.ListConstruct %10641 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8405 = torch.constant.bool false
%10739 = torch.aten.index_put %10737, %10738, %10657, %false_8405 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10739, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8406 = torch.constant.int 32
%int2_8407 = torch.constant.int 2
%int16_8408 = torch.constant.int 16
%int1_8409 = torch.constant.int 1
%int128_8410 = torch.constant.int 128
%10740 = torch.prim.ListConstruct %3041, %int32_8406, %int2_8407, %int16_8408, %int1_8409, %int128_8410 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10741 = torch.aten.view %10739, %10740 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10741, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8411 = torch.constant.int 131072
%10742 = torch.prim.ListConstruct %3041, %int131072_8411 : (!torch.int, !torch.int) -> !torch.list<int>
%10743 = torch.aten.view %10741, %10742 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10743, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_8412 = torch.constant.int 32
%int2_8413 = torch.constant.int 2
%int16_8414 = torch.constant.int 16
%int1_8415 = torch.constant.int 1
%int128_8416 = torch.constant.int 128
%10744 = torch.prim.ListConstruct %3044, %int32_8412, %int2_8413, %int16_8414, %int1_8415, %int128_8416 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10745 = torch.aten.view %8894, %10744 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10745, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_8417 = torch.constant.int 32
%10746 = torch.aten.mul.int %3044, %int32_8417 : !torch.int, !torch.int -> !torch.int
%int2_8418 = torch.constant.int 2
%10747 = torch.aten.mul.int %10746, %int2_8418 : !torch.int, !torch.int -> !torch.int
%int16_8419 = torch.constant.int 16
%int1_8420 = torch.constant.int 1
%int128_8421 = torch.constant.int 128
%10748 = torch.prim.ListConstruct %10747, %int16_8419, %int1_8420, %int128_8421 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10749 = torch.aten.view %10745, %10748 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10749, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%10750 = torch.prim.ListConstruct %10643 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_8422 = torch.constant.bool false
%10751 = torch.aten.index_put %10749, %10750, %10659, %false_8422 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %10751, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_8423 = torch.constant.int 32
%int2_8424 = torch.constant.int 2
%int16_8425 = torch.constant.int 16
%int1_8426 = torch.constant.int 1
%int128_8427 = torch.constant.int 128
%10752 = torch.prim.ListConstruct %3044, %int32_8423, %int2_8424, %int16_8425, %int1_8426, %int128_8427 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10753 = torch.aten.view %10751, %10752 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %10753, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_8428 = torch.constant.int 131072
%10754 = torch.prim.ListConstruct %3044, %int131072_8428 : (!torch.int, !torch.int) -> !torch.list<int>
%10755 = torch.aten.view %10753, %10754 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %10755, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_8429 = torch.constant.int -2
%10756 = torch.aten.unsqueeze %10370, %int-2_8429 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8430 = torch.constant.int -2
%10757 = torch.aten.unsqueeze %10385, %int-2_8430 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8431 = torch.constant.int -2
%10758 = torch.aten.unsqueeze %10400, %int-2_8431 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8432 = torch.constant.int -2
%10759 = torch.aten.unsqueeze %10415, %int-2_8432 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8433 = torch.constant.int -2
%10760 = torch.aten.unsqueeze %10430, %int-2_8433 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8434 = torch.constant.int -2
%10761 = torch.aten.unsqueeze %10445, %int-2_8434 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8435 = torch.constant.int -2
%10762 = torch.aten.unsqueeze %10460, %int-2_8435 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8436 = torch.constant.int -2
%10763 = torch.aten.unsqueeze %10475, %int-2_8436 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_8437 = torch.constant.int 4
%int1_8438 = torch.constant.int 1
%int4_8439 = torch.constant.int 4
%int128_8440 = torch.constant.int 128
%10764 = torch.prim.ListConstruct %int4_8437, %10356, %int1_8438, %int4_8439, %int128_8440 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8441 = torch.constant.bool false
%10765 = torch.aten.expand %10756, %10764, %false_8441 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8442 = torch.constant.int 4
%int1_8443 = torch.constant.int 1
%int4_8444 = torch.constant.int 4
%int128_8445 = torch.constant.int 128
%10766 = torch.prim.ListConstruct %int4_8442, %10356, %int1_8443, %int4_8444, %int128_8445 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8446 = torch.constant.bool false
%10767 = torch.aten.expand %10757, %10766, %false_8446 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8447 = torch.constant.int 4
%int1_8448 = torch.constant.int 1
%int4_8449 = torch.constant.int 4
%int128_8450 = torch.constant.int 128
%10768 = torch.prim.ListConstruct %int4_8447, %10356, %int1_8448, %int4_8449, %int128_8450 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8451 = torch.constant.bool false
%10769 = torch.aten.expand %10758, %10768, %false_8451 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8452 = torch.constant.int 4
%int1_8453 = torch.constant.int 1
%int4_8454 = torch.constant.int 4
%int128_8455 = torch.constant.int 128
%10770 = torch.prim.ListConstruct %int4_8452, %10356, %int1_8453, %int4_8454, %int128_8455 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8456 = torch.constant.bool false
%10771 = torch.aten.expand %10759, %10770, %false_8456 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8457 = torch.constant.int 4
%int1_8458 = torch.constant.int 1
%int4_8459 = torch.constant.int 4
%int128_8460 = torch.constant.int 128
%10772 = torch.prim.ListConstruct %int4_8457, %10356, %int1_8458, %int4_8459, %int128_8460 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8461 = torch.constant.bool false
%10773 = torch.aten.expand %10760, %10772, %false_8461 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8462 = torch.constant.int 4
%int1_8463 = torch.constant.int 1
%int4_8464 = torch.constant.int 4
%int128_8465 = torch.constant.int 128
%10774 = torch.prim.ListConstruct %int4_8462, %10356, %int1_8463, %int4_8464, %int128_8465 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8466 = torch.constant.bool false
%10775 = torch.aten.expand %10761, %10774, %false_8466 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8467 = torch.constant.int 4
%int1_8468 = torch.constant.int 1
%int4_8469 = torch.constant.int 4
%int128_8470 = torch.constant.int 128
%10776 = torch.prim.ListConstruct %int4_8467, %10356, %int1_8468, %int4_8469, %int128_8470 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8471 = torch.constant.bool false
%10777 = torch.aten.expand %10762, %10776, %false_8471 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8472 = torch.constant.int 4
%int1_8473 = torch.constant.int 1
%int4_8474 = torch.constant.int 4
%int128_8475 = torch.constant.int 128
%10778 = torch.prim.ListConstruct %int4_8472, %10356, %int1_8473, %int4_8474, %int128_8475 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8476 = torch.constant.bool false
%10779 = torch.aten.expand %10763, %10778, %false_8476 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8477 = torch.constant.int 4
%int4_8478 = torch.constant.int 4
%int128_8479 = torch.constant.int 128
%10780 = torch.prim.ListConstruct %int4_8477, %10356, %int4_8478, %int128_8479 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10781 = torch.aten.view %10765, %10780 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8480 = torch.constant.int 4
%int4_8481 = torch.constant.int 4
%int128_8482 = torch.constant.int 128
%10782 = torch.prim.ListConstruct %int4_8480, %10356, %int4_8481, %int128_8482 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10783 = torch.aten.view %10767, %10782 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8483 = torch.constant.int 4
%int4_8484 = torch.constant.int 4
%int128_8485 = torch.constant.int 128
%10784 = torch.prim.ListConstruct %int4_8483, %10356, %int4_8484, %int128_8485 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10785 = torch.aten.view %10769, %10784 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8486 = torch.constant.int 4
%int4_8487 = torch.constant.int 4
%int128_8488 = torch.constant.int 128
%10786 = torch.prim.ListConstruct %int4_8486, %10356, %int4_8487, %int128_8488 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10787 = torch.aten.view %10771, %10786 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8489 = torch.constant.int 4
%int4_8490 = torch.constant.int 4
%int128_8491 = torch.constant.int 128
%10788 = torch.prim.ListConstruct %int4_8489, %10356, %int4_8490, %int128_8491 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10789 = torch.aten.view %10773, %10788 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8492 = torch.constant.int 4
%int4_8493 = torch.constant.int 4
%int128_8494 = torch.constant.int 128
%10790 = torch.prim.ListConstruct %int4_8492, %10356, %int4_8493, %int128_8494 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10791 = torch.aten.view %10775, %10790 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8495 = torch.constant.int 4
%int4_8496 = torch.constant.int 4
%int128_8497 = torch.constant.int 128
%10792 = torch.prim.ListConstruct %int4_8495, %10356, %int4_8496, %int128_8497 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10793 = torch.aten.view %10777, %10792 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8498 = torch.constant.int 4
%int4_8499 = torch.constant.int 4
%int128_8500 = torch.constant.int 128
%10794 = torch.prim.ListConstruct %int4_8498, %10356, %int4_8499, %int128_8500 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10795 = torch.aten.view %10779, %10794 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_8501 = torch.constant.int -2
%10796 = torch.aten.unsqueeze %10145, %int-2_8501 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8502 = torch.constant.int -2
%10797 = torch.aten.unsqueeze %10147, %int-2_8502 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8503 = torch.constant.int -2
%10798 = torch.aten.unsqueeze %10149, %int-2_8503 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8504 = torch.constant.int -2
%10799 = torch.aten.unsqueeze %10151, %int-2_8504 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8505 = torch.constant.int -2
%10800 = torch.aten.unsqueeze %10153, %int-2_8505 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8506 = torch.constant.int -2
%10801 = torch.aten.unsqueeze %10155, %int-2_8506 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8507 = torch.constant.int -2
%10802 = torch.aten.unsqueeze %10157, %int-2_8507 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_8508 = torch.constant.int -2
%10803 = torch.aten.unsqueeze %10159, %int-2_8508 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %10803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_8509 = torch.constant.int 1
%10804 = torch.aten.size.int %10069, %int1_8509 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_8510 = torch.constant.int 4
%int1_8511 = torch.constant.int 1
%int4_8512 = torch.constant.int 4
%int128_8513 = torch.constant.int 128
%10805 = torch.prim.ListConstruct %int4_8510, %10804, %int1_8511, %int4_8512, %int128_8513 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8514 = torch.constant.bool false
%10806 = torch.aten.expand %10796, %10805, %false_8514 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8515 = torch.constant.int 4
%int1_8516 = torch.constant.int 1
%int4_8517 = torch.constant.int 4
%int128_8518 = torch.constant.int 128
%10807 = torch.prim.ListConstruct %int4_8515, %10804, %int1_8516, %int4_8517, %int128_8518 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8519 = torch.constant.bool false
%10808 = torch.aten.expand %10797, %10807, %false_8519 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8520 = torch.constant.int 4
%int1_8521 = torch.constant.int 1
%int4_8522 = torch.constant.int 4
%int128_8523 = torch.constant.int 128
%10809 = torch.prim.ListConstruct %int4_8520, %10804, %int1_8521, %int4_8522, %int128_8523 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8524 = torch.constant.bool false
%10810 = torch.aten.expand %10798, %10809, %false_8524 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8525 = torch.constant.int 4
%int1_8526 = torch.constant.int 1
%int4_8527 = torch.constant.int 4
%int128_8528 = torch.constant.int 128
%10811 = torch.prim.ListConstruct %int4_8525, %10804, %int1_8526, %int4_8527, %int128_8528 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8529 = torch.constant.bool false
%10812 = torch.aten.expand %10799, %10811, %false_8529 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8530 = torch.constant.int 4
%int1_8531 = torch.constant.int 1
%int4_8532 = torch.constant.int 4
%int128_8533 = torch.constant.int 128
%10813 = torch.prim.ListConstruct %int4_8530, %10804, %int1_8531, %int4_8532, %int128_8533 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8534 = torch.constant.bool false
%10814 = torch.aten.expand %10800, %10813, %false_8534 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8535 = torch.constant.int 4
%int1_8536 = torch.constant.int 1
%int4_8537 = torch.constant.int 4
%int128_8538 = torch.constant.int 128
%10815 = torch.prim.ListConstruct %int4_8535, %10804, %int1_8536, %int4_8537, %int128_8538 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8539 = torch.constant.bool false
%10816 = torch.aten.expand %10801, %10815, %false_8539 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8540 = torch.constant.int 4
%int1_8541 = torch.constant.int 1
%int4_8542 = torch.constant.int 4
%int128_8543 = torch.constant.int 128
%10817 = torch.prim.ListConstruct %int4_8540, %10804, %int1_8541, %int4_8542, %int128_8543 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8544 = torch.constant.bool false
%10818 = torch.aten.expand %10802, %10817, %false_8544 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8545 = torch.constant.int 4
%int1_8546 = torch.constant.int 1
%int4_8547 = torch.constant.int 4
%int128_8548 = torch.constant.int 128
%10819 = torch.prim.ListConstruct %int4_8545, %10804, %int1_8546, %int4_8547, %int128_8548 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_8549 = torch.constant.bool false
%10820 = torch.aten.expand %10803, %10819, %false_8549 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %10820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_8550 = torch.constant.int 4
%int4_8551 = torch.constant.int 4
%int128_8552 = torch.constant.int 128
%10821 = torch.prim.ListConstruct %int4_8550, %10804, %int4_8551, %int128_8552 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10822 = torch.aten.view %10806, %10821 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8553 = torch.constant.int 4
%int4_8554 = torch.constant.int 4
%int128_8555 = torch.constant.int 128
%10823 = torch.prim.ListConstruct %int4_8553, %10804, %int4_8554, %int128_8555 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10824 = torch.aten.view %10808, %10823 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8556 = torch.constant.int 4
%int4_8557 = torch.constant.int 4
%int128_8558 = torch.constant.int 128
%10825 = torch.prim.ListConstruct %int4_8556, %10804, %int4_8557, %int128_8558 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10826 = torch.aten.view %10810, %10825 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8559 = torch.constant.int 4
%int4_8560 = torch.constant.int 4
%int128_8561 = torch.constant.int 128
%10827 = torch.prim.ListConstruct %int4_8559, %10804, %int4_8560, %int128_8561 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10828 = torch.aten.view %10812, %10827 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8562 = torch.constant.int 4
%int4_8563 = torch.constant.int 4
%int128_8564 = torch.constant.int 128
%10829 = torch.prim.ListConstruct %int4_8562, %10804, %int4_8563, %int128_8564 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10830 = torch.aten.view %10814, %10829 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8565 = torch.constant.int 4
%int4_8566 = torch.constant.int 4
%int128_8567 = torch.constant.int 128
%10831 = torch.prim.ListConstruct %int4_8565, %10804, %int4_8566, %int128_8567 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10832 = torch.aten.view %10816, %10831 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8568 = torch.constant.int 4
%int4_8569 = torch.constant.int 4
%int128_8570 = torch.constant.int 128
%10833 = torch.prim.ListConstruct %int4_8568, %10804, %int4_8569, %int128_8570 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10834 = torch.aten.view %10818, %10833 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8571 = torch.constant.int 4
%int4_8572 = torch.constant.int 4
%int128_8573 = torch.constant.int 128
%10835 = torch.prim.ListConstruct %int4_8571, %10804, %int4_8572, %int128_8573 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10836 = torch.aten.view %10820, %10835 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8574 = torch.constant.int 1
%int2_8575 = torch.constant.int 2
%10837 = torch.aten.transpose.int %10212, %int1_8574, %int2_8575 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10837, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8576 = torch.constant.int 1
%int2_8577 = torch.constant.int 2
%10838 = torch.aten.transpose.int %10227, %int1_8576, %int2_8577 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10838, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8578 = torch.constant.int 1
%int2_8579 = torch.constant.int 2
%10839 = torch.aten.transpose.int %10242, %int1_8578, %int2_8579 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10839, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8580 = torch.constant.int 1
%int2_8581 = torch.constant.int 2
%10840 = torch.aten.transpose.int %10257, %int1_8580, %int2_8581 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10840, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8582 = torch.constant.int 1
%int2_8583 = torch.constant.int 2
%10841 = torch.aten.transpose.int %10272, %int1_8582, %int2_8583 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10841, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8584 = torch.constant.int 1
%int2_8585 = torch.constant.int 2
%10842 = torch.aten.transpose.int %10287, %int1_8584, %int2_8585 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10842, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8586 = torch.constant.int 1
%int2_8587 = torch.constant.int 2
%10843 = torch.aten.transpose.int %10302, %int1_8586, %int2_8587 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10843, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8588 = torch.constant.int 1
%int2_8589 = torch.constant.int 2
%10844 = torch.aten.transpose.int %10317, %int1_8588, %int2_8589 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10844, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8590 = torch.constant.int 1
%int2_8591 = torch.constant.int 2
%10845 = torch.aten.transpose.int %10781, %int1_8590, %int2_8591 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10845, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8592 = torch.constant.int 1
%int2_8593 = torch.constant.int 2
%10846 = torch.aten.transpose.int %10783, %int1_8592, %int2_8593 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10846, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8594 = torch.constant.int 1
%int2_8595 = torch.constant.int 2
%10847 = torch.aten.transpose.int %10785, %int1_8594, %int2_8595 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10847, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8596 = torch.constant.int 1
%int2_8597 = torch.constant.int 2
%10848 = torch.aten.transpose.int %10787, %int1_8596, %int2_8597 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10848, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8598 = torch.constant.int 1
%int2_8599 = torch.constant.int 2
%10849 = torch.aten.transpose.int %10789, %int1_8598, %int2_8599 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10849, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8600 = torch.constant.int 1
%int2_8601 = torch.constant.int 2
%10850 = torch.aten.transpose.int %10791, %int1_8600, %int2_8601 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10850, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8602 = torch.constant.int 1
%int2_8603 = torch.constant.int 2
%10851 = torch.aten.transpose.int %10793, %int1_8602, %int2_8603 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10851, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8604 = torch.constant.int 1
%int2_8605 = torch.constant.int 2
%10852 = torch.aten.transpose.int %10795, %int1_8604, %int2_8605 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10852, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8606 = torch.constant.int 1
%int2_8607 = torch.constant.int 2
%10853 = torch.aten.transpose.int %10822, %int1_8606, %int2_8607 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10853, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8608 = torch.constant.int 1
%int2_8609 = torch.constant.int 2
%10854 = torch.aten.transpose.int %10824, %int1_8608, %int2_8609 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10854, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8610 = torch.constant.int 1
%int2_8611 = torch.constant.int 2
%10855 = torch.aten.transpose.int %10826, %int1_8610, %int2_8611 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10855, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8612 = torch.constant.int 1
%int2_8613 = torch.constant.int 2
%10856 = torch.aten.transpose.int %10828, %int1_8612, %int2_8613 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10856, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8614 = torch.constant.int 1
%int2_8615 = torch.constant.int 2
%10857 = torch.aten.transpose.int %10830, %int1_8614, %int2_8615 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10857, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8616 = torch.constant.int 1
%int2_8617 = torch.constant.int 2
%10858 = torch.aten.transpose.int %10832, %int1_8616, %int2_8617 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10858, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8618 = torch.constant.int 1
%int2_8619 = torch.constant.int 2
%10859 = torch.aten.transpose.int %10834, %int1_8618, %int2_8619 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10859, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8620 = torch.constant.int 1
%int2_8621 = torch.constant.int 2
%10860 = torch.aten.transpose.int %10836, %int1_8620, %int2_8621 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %10860, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8622 = torch.constant.float 0.000000e+00
%true_8623 = torch.constant.bool true
%none_8624 = torch.constant.none
%none_8625 = torch.constant.none
%10861:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10837, %10845, %10853, %float0.000000e00_8622, %true_8623, %none_8624, %none_8625) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10861#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8626 = torch.constant.float 0.000000e+00
%true_8627 = torch.constant.bool true
%none_8628 = torch.constant.none
%none_8629 = torch.constant.none
%10862:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10838, %10846, %10854, %float0.000000e00_8626, %true_8627, %none_8628, %none_8629) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10862#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8630 = torch.constant.float 0.000000e+00
%true_8631 = torch.constant.bool true
%none_8632 = torch.constant.none
%none_8633 = torch.constant.none
%10863:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10839, %10847, %10855, %float0.000000e00_8630, %true_8631, %none_8632, %none_8633) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10863#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8634 = torch.constant.float 0.000000e+00
%true_8635 = torch.constant.bool true
%none_8636 = torch.constant.none
%none_8637 = torch.constant.none
%10864:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10840, %10848, %10856, %float0.000000e00_8634, %true_8635, %none_8636, %none_8637) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10864#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8638 = torch.constant.float 0.000000e+00
%true_8639 = torch.constant.bool true
%none_8640 = torch.constant.none
%none_8641 = torch.constant.none
%10865:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10841, %10849, %10857, %float0.000000e00_8638, %true_8639, %none_8640, %none_8641) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10865#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8642 = torch.constant.float 0.000000e+00
%true_8643 = torch.constant.bool true
%none_8644 = torch.constant.none
%none_8645 = torch.constant.none
%10866:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10842, %10850, %10858, %float0.000000e00_8642, %true_8643, %none_8644, %none_8645) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10866#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8646 = torch.constant.float 0.000000e+00
%true_8647 = torch.constant.bool true
%none_8648 = torch.constant.none
%none_8649 = torch.constant.none
%10867:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10843, %10851, %10859, %float0.000000e00_8646, %true_8647, %none_8648, %none_8649) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10867#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_8650 = torch.constant.float 0.000000e+00
%true_8651 = torch.constant.bool true
%none_8652 = torch.constant.none
%none_8653 = torch.constant.none
%10868:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%10844, %10852, %10860, %float0.000000e00_8650, %true_8651, %none_8652, %none_8653) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %10868#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_8654 = torch.constant.int 1
%int2_8655 = torch.constant.int 2
%10869 = torch.aten.transpose.int %10861#0, %int1_8654, %int2_8655 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8656 = torch.constant.int 1
%int2_8657 = torch.constant.int 2
%10870 = torch.aten.transpose.int %10862#0, %int1_8656, %int2_8657 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8658 = torch.constant.int 1
%int2_8659 = torch.constant.int 2
%10871 = torch.aten.transpose.int %10863#0, %int1_8658, %int2_8659 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8660 = torch.constant.int 1
%int2_8661 = torch.constant.int 2
%10872 = torch.aten.transpose.int %10864#0, %int1_8660, %int2_8661 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8662 = torch.constant.int 1
%int2_8663 = torch.constant.int 2
%10873 = torch.aten.transpose.int %10865#0, %int1_8662, %int2_8663 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8664 = torch.constant.int 1
%int2_8665 = torch.constant.int 2
%10874 = torch.aten.transpose.int %10866#0, %int1_8664, %int2_8665 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8666 = torch.constant.int 1
%int2_8667 = torch.constant.int 2
%10875 = torch.aten.transpose.int %10867#0, %int1_8666, %int2_8667 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_8668 = torch.constant.int 1
%int2_8669 = torch.constant.int 2
%10876 = torch.aten.transpose.int %10868#0, %int1_8668, %int2_8669 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %10876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_8670 = torch.constant.int 4
%int512_8671 = torch.constant.int 512
%10877 = torch.prim.ListConstruct %int4_8670, %10198, %int512_8671 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10878 = torch.aten.view %10869, %10877 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8672 = torch.constant.int 4
%int512_8673 = torch.constant.int 512
%10879 = torch.prim.ListConstruct %int4_8672, %10213, %int512_8673 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10880 = torch.aten.view %10870, %10879 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8674 = torch.constant.int 4
%int512_8675 = torch.constant.int 512
%10881 = torch.prim.ListConstruct %int4_8674, %10228, %int512_8675 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10882 = torch.aten.view %10871, %10881 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8676 = torch.constant.int 4
%int512_8677 = torch.constant.int 512
%10883 = torch.prim.ListConstruct %int4_8676, %10243, %int512_8677 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10884 = torch.aten.view %10872, %10883 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8678 = torch.constant.int 4
%int512_8679 = torch.constant.int 512
%10885 = torch.prim.ListConstruct %int4_8678, %10258, %int512_8679 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10886 = torch.aten.view %10873, %10885 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8680 = torch.constant.int 4
%int512_8681 = torch.constant.int 512
%10887 = torch.prim.ListConstruct %int4_8680, %10273, %int512_8681 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10888 = torch.aten.view %10874, %10887 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8682 = torch.constant.int 4
%int512_8683 = torch.constant.int 512
%10889 = torch.prim.ListConstruct %int4_8682, %10288, %int512_8683 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10890 = torch.aten.view %10875, %10889 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_8684 = torch.constant.int 4
%int512_8685 = torch.constant.int 512
%10891 = torch.prim.ListConstruct %int4_8684, %10303, %int512_8685 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10892 = torch.aten.view %10876, %10891 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %10892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_8686 = torch.constant.int 1
%int0_8687 = torch.constant.int 0
%10893 = torch.prim.ListConstruct %int1_8686, %int0_8687 : (!torch.int, !torch.int) -> !torch.list<int>
%10894 = torch.aten.permute %328, %10893 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8688 = torch.constant.int 1
%int0_8689 = torch.constant.int 0
%10895 = torch.prim.ListConstruct %int1_8688, %int0_8689 : (!torch.int, !torch.int) -> !torch.list<int>
%10896 = torch.aten.permute %329, %10895 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8690 = torch.constant.int 1
%int0_8691 = torch.constant.int 0
%10897 = torch.prim.ListConstruct %int1_8690, %int0_8691 : (!torch.int, !torch.int) -> !torch.list<int>
%10898 = torch.aten.permute %330, %10897 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8692 = torch.constant.int 1
%int0_8693 = torch.constant.int 0
%10899 = torch.prim.ListConstruct %int1_8692, %int0_8693 : (!torch.int, !torch.int) -> !torch.list<int>
%10900 = torch.aten.permute %331, %10899 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8694 = torch.constant.int 1
%int0_8695 = torch.constant.int 0
%10901 = torch.prim.ListConstruct %int1_8694, %int0_8695 : (!torch.int, !torch.int) -> !torch.list<int>
%10902 = torch.aten.permute %332, %10901 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8696 = torch.constant.int 1
%int0_8697 = torch.constant.int 0
%10903 = torch.prim.ListConstruct %int1_8696, %int0_8697 : (!torch.int, !torch.int) -> !torch.list<int>
%10904 = torch.aten.permute %333, %10903 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8698 = torch.constant.int 1
%int0_8699 = torch.constant.int 0
%10905 = torch.prim.ListConstruct %int1_8698, %int0_8699 : (!torch.int, !torch.int) -> !torch.list<int>
%10906 = torch.aten.permute %334, %10905 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_8700 = torch.constant.int 1
%int0_8701 = torch.constant.int 0
%10907 = torch.prim.ListConstruct %int1_8700, %int0_8701 : (!torch.int, !torch.int) -> !torch.list<int>
%10908 = torch.aten.permute %335, %10907 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_8702 = torch.constant.int 4
%10909 = torch.aten.mul.int %int4_8702, %10198 : !torch.int, !torch.int -> !torch.int
%int512_8703 = torch.constant.int 512
%10910 = torch.prim.ListConstruct %10909, %int512_8703 : (!torch.int, !torch.int) -> !torch.list<int>
%10911 = torch.aten.view %10878, %10910 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10911, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10912 = torch.aten.mm %10911, %10894 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10912, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8704 = torch.constant.int 4
%int4096_8705 = torch.constant.int 4096
%10913 = torch.prim.ListConstruct %int4_8704, %10198, %int4096_8705 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10914 = torch.aten.view %10912, %10913 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8706 = torch.constant.int 4
%10915 = torch.aten.mul.int %int4_8706, %10213 : !torch.int, !torch.int -> !torch.int
%int512_8707 = torch.constant.int 512
%10916 = torch.prim.ListConstruct %10915, %int512_8707 : (!torch.int, !torch.int) -> !torch.list<int>
%10917 = torch.aten.view %10880, %10916 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10917, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10918 = torch.aten.mm %10917, %10896 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10918, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8708 = torch.constant.int 4
%int4096_8709 = torch.constant.int 4096
%10919 = torch.prim.ListConstruct %int4_8708, %10213, %int4096_8709 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10920 = torch.aten.view %10918, %10919 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8710 = torch.constant.int 4
%10921 = torch.aten.mul.int %int4_8710, %10228 : !torch.int, !torch.int -> !torch.int
%int512_8711 = torch.constant.int 512
%10922 = torch.prim.ListConstruct %10921, %int512_8711 : (!torch.int, !torch.int) -> !torch.list<int>
%10923 = torch.aten.view %10882, %10922 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10923, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10924 = torch.aten.mm %10923, %10898 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10924, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8712 = torch.constant.int 4
%int4096_8713 = torch.constant.int 4096
%10925 = torch.prim.ListConstruct %int4_8712, %10228, %int4096_8713 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10926 = torch.aten.view %10924, %10925 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8714 = torch.constant.int 4
%10927 = torch.aten.mul.int %int4_8714, %10243 : !torch.int, !torch.int -> !torch.int
%int512_8715 = torch.constant.int 512
%10928 = torch.prim.ListConstruct %10927, %int512_8715 : (!torch.int, !torch.int) -> !torch.list<int>
%10929 = torch.aten.view %10884, %10928 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10929, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10930 = torch.aten.mm %10929, %10900 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10930, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8716 = torch.constant.int 4
%int4096_8717 = torch.constant.int 4096
%10931 = torch.prim.ListConstruct %int4_8716, %10243, %int4096_8717 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10932 = torch.aten.view %10930, %10931 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8718 = torch.constant.int 4
%10933 = torch.aten.mul.int %int4_8718, %10258 : !torch.int, !torch.int -> !torch.int
%int512_8719 = torch.constant.int 512
%10934 = torch.prim.ListConstruct %10933, %int512_8719 : (!torch.int, !torch.int) -> !torch.list<int>
%10935 = torch.aten.view %10886, %10934 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10935, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10936 = torch.aten.mm %10935, %10902 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10936, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8720 = torch.constant.int 4
%int4096_8721 = torch.constant.int 4096
%10937 = torch.prim.ListConstruct %int4_8720, %10258, %int4096_8721 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10938 = torch.aten.view %10936, %10937 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8722 = torch.constant.int 4
%10939 = torch.aten.mul.int %int4_8722, %10273 : !torch.int, !torch.int -> !torch.int
%int512_8723 = torch.constant.int 512
%10940 = torch.prim.ListConstruct %10939, %int512_8723 : (!torch.int, !torch.int) -> !torch.list<int>
%10941 = torch.aten.view %10888, %10940 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10941, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10942 = torch.aten.mm %10941, %10904 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10942, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8724 = torch.constant.int 4
%int4096_8725 = torch.constant.int 4096
%10943 = torch.prim.ListConstruct %int4_8724, %10273, %int4096_8725 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10944 = torch.aten.view %10942, %10943 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8726 = torch.constant.int 4
%10945 = torch.aten.mul.int %int4_8726, %10288 : !torch.int, !torch.int -> !torch.int
%int512_8727 = torch.constant.int 512
%10946 = torch.prim.ListConstruct %10945, %int512_8727 : (!torch.int, !torch.int) -> !torch.list<int>
%10947 = torch.aten.view %10890, %10946 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10947, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10948 = torch.aten.mm %10947, %10906 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10948, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8728 = torch.constant.int 4
%int4096_8729 = torch.constant.int 4096
%10949 = torch.prim.ListConstruct %int4_8728, %10288, %int4096_8729 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10950 = torch.aten.view %10948, %10949 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_8730 = torch.constant.int 4
%10951 = torch.aten.mul.int %int4_8730, %10303 : !torch.int, !torch.int -> !torch.int
%int512_8731 = torch.constant.int 512
%10952 = torch.prim.ListConstruct %10951, %int512_8731 : (!torch.int, !torch.int) -> !torch.list<int>
%10953 = torch.aten.view %10892, %10952 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %10953, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%10954 = torch.aten.mm %10953, %10908 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %10954, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_8732 = torch.constant.int 4
%int4096_8733 = torch.constant.int 4096
%10955 = torch.prim.ListConstruct %int4_8732, %10303, %int4096_8733 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%10956 = torch.aten.view %10954, %10955 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10957 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8734 = arith.constant 1 : index
%dim_8735 = tensor.dim %10957, %c1_8734 : tensor<4x?x4096xf16>
%10958 = flow.tensor.transfer %10957 : tensor<4x?x4096xf16>{%dim_8735} to #hal.device.promise<@__device_0>
%10959 = torch_c.from_builtin_tensor %10958 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10960 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8736 = arith.constant 1 : index
%dim_8737 = tensor.dim %10960, %c1_8736 : tensor<4x?x4096xf16>
%10961 = flow.tensor.transfer %10960 : tensor<4x?x4096xf16>{%dim_8737} to #hal.device.promise<@__device_0>
%10962 = torch_c.from_builtin_tensor %10961 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10963 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8738 = arith.constant 1 : index
%dim_8739 = tensor.dim %10963, %c1_8738 : tensor<4x?x4096xf16>
%10964 = flow.tensor.transfer %10963 : tensor<4x?x4096xf16>{%dim_8739} to #hal.device.promise<@__device_0>
%10965 = torch_c.from_builtin_tensor %10964 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10966 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8740 = arith.constant 1 : index
%dim_8741 = tensor.dim %10966, %c1_8740 : tensor<4x?x4096xf16>
%10967 = flow.tensor.transfer %10966 : tensor<4x?x4096xf16>{%dim_8741} to #hal.device.promise<@__device_0>
%10968 = torch_c.from_builtin_tensor %10967 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10969 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8742 = arith.constant 1 : index
%dim_8743 = tensor.dim %10969, %c1_8742 : tensor<4x?x4096xf16>
%10970 = flow.tensor.transfer %10969 : tensor<4x?x4096xf16>{%dim_8743} to #hal.device.promise<@__device_0>
%10971 = torch_c.from_builtin_tensor %10970 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10972 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8744 = arith.constant 1 : index
%dim_8745 = tensor.dim %10972, %c1_8744 : tensor<4x?x4096xf16>
%10973 = flow.tensor.transfer %10972 : tensor<4x?x4096xf16>{%dim_8745} to #hal.device.promise<@__device_0>
%10974 = torch_c.from_builtin_tensor %10973 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10975 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8746 = arith.constant 1 : index
%dim_8747 = tensor.dim %10975, %c1_8746 : tensor<4x?x4096xf16>
%10976 = flow.tensor.transfer %10975 : tensor<4x?x4096xf16>{%dim_8747} to #hal.device.promise<@__device_0>
%10977 = torch_c.from_builtin_tensor %10976 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8748 = torch.constant.int 1
%10978 = torch.aten.add.Tensor %10914, %10959, %int1_8748 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8749 = torch.constant.int 1
%10979 = torch.aten.add.Tensor %10978, %10962, %int1_8749 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8750 = torch.constant.int 1
%10980 = torch.aten.add.Tensor %10979, %10965, %int1_8750 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8751 = torch.constant.int 1
%10981 = torch.aten.add.Tensor %10980, %10968, %int1_8751 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8752 = torch.constant.int 1
%10982 = torch.aten.add.Tensor %10981, %10971, %int1_8752 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8753 = torch.constant.int 1
%10983 = torch.aten.add.Tensor %10982, %10974, %int1_8753 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8754 = torch.constant.int 1
%10984 = torch.aten.add.Tensor %10983, %10977, %int1_8754 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10985 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8755 = arith.constant 1 : index
%dim_8756 = tensor.dim %10985, %c1_8755 : tensor<4x?x4096xf16>
%10986 = flow.tensor.transfer %10985 : tensor<4x?x4096xf16>{%dim_8756} to #hal.device.promise<@__device_1>
%10987 = torch_c.from_builtin_tensor %10986 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10988 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8757 = arith.constant 1 : index
%dim_8758 = tensor.dim %10988, %c1_8757 : tensor<4x?x4096xf16>
%10989 = flow.tensor.transfer %10988 : tensor<4x?x4096xf16>{%dim_8758} to #hal.device.promise<@__device_1>
%10990 = torch_c.from_builtin_tensor %10989 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10991 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8759 = arith.constant 1 : index
%dim_8760 = tensor.dim %10991, %c1_8759 : tensor<4x?x4096xf16>
%10992 = flow.tensor.transfer %10991 : tensor<4x?x4096xf16>{%dim_8760} to #hal.device.promise<@__device_1>
%10993 = torch_c.from_builtin_tensor %10992 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10994 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8761 = arith.constant 1 : index
%dim_8762 = tensor.dim %10994, %c1_8761 : tensor<4x?x4096xf16>
%10995 = flow.tensor.transfer %10994 : tensor<4x?x4096xf16>{%dim_8762} to #hal.device.promise<@__device_1>
%10996 = torch_c.from_builtin_tensor %10995 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%10997 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8763 = arith.constant 1 : index
%dim_8764 = tensor.dim %10997, %c1_8763 : tensor<4x?x4096xf16>
%10998 = flow.tensor.transfer %10997 : tensor<4x?x4096xf16>{%dim_8764} to #hal.device.promise<@__device_1>
%10999 = torch_c.from_builtin_tensor %10998 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %10999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11000 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8765 = arith.constant 1 : index
%dim_8766 = tensor.dim %11000, %c1_8765 : tensor<4x?x4096xf16>
%11001 = flow.tensor.transfer %11000 : tensor<4x?x4096xf16>{%dim_8766} to #hal.device.promise<@__device_1>
%11002 = torch_c.from_builtin_tensor %11001 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11003 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8767 = arith.constant 1 : index
%dim_8768 = tensor.dim %11003, %c1_8767 : tensor<4x?x4096xf16>
%11004 = flow.tensor.transfer %11003 : tensor<4x?x4096xf16>{%dim_8768} to #hal.device.promise<@__device_1>
%11005 = torch_c.from_builtin_tensor %11004 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8769 = torch.constant.int 1
%11006 = torch.aten.add.Tensor %10987, %10920, %int1_8769 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8770 = torch.constant.int 1
%11007 = torch.aten.add.Tensor %11006, %10990, %int1_8770 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8771 = torch.constant.int 1
%11008 = torch.aten.add.Tensor %11007, %10993, %int1_8771 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8772 = torch.constant.int 1
%11009 = torch.aten.add.Tensor %11008, %10996, %int1_8772 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8773 = torch.constant.int 1
%11010 = torch.aten.add.Tensor %11009, %10999, %int1_8773 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8774 = torch.constant.int 1
%11011 = torch.aten.add.Tensor %11010, %11002, %int1_8774 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8775 = torch.constant.int 1
%11012 = torch.aten.add.Tensor %11011, %11005, %int1_8775 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11013 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8776 = arith.constant 1 : index
%dim_8777 = tensor.dim %11013, %c1_8776 : tensor<4x?x4096xf16>
%11014 = flow.tensor.transfer %11013 : tensor<4x?x4096xf16>{%dim_8777} to #hal.device.promise<@__device_2>
%11015 = torch_c.from_builtin_tensor %11014 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11016 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8778 = arith.constant 1 : index
%dim_8779 = tensor.dim %11016, %c1_8778 : tensor<4x?x4096xf16>
%11017 = flow.tensor.transfer %11016 : tensor<4x?x4096xf16>{%dim_8779} to #hal.device.promise<@__device_2>
%11018 = torch_c.from_builtin_tensor %11017 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11019 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8780 = arith.constant 1 : index
%dim_8781 = tensor.dim %11019, %c1_8780 : tensor<4x?x4096xf16>
%11020 = flow.tensor.transfer %11019 : tensor<4x?x4096xf16>{%dim_8781} to #hal.device.promise<@__device_2>
%11021 = torch_c.from_builtin_tensor %11020 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11022 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8782 = arith.constant 1 : index
%dim_8783 = tensor.dim %11022, %c1_8782 : tensor<4x?x4096xf16>
%11023 = flow.tensor.transfer %11022 : tensor<4x?x4096xf16>{%dim_8783} to #hal.device.promise<@__device_2>
%11024 = torch_c.from_builtin_tensor %11023 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11025 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8784 = arith.constant 1 : index
%dim_8785 = tensor.dim %11025, %c1_8784 : tensor<4x?x4096xf16>
%11026 = flow.tensor.transfer %11025 : tensor<4x?x4096xf16>{%dim_8785} to #hal.device.promise<@__device_2>
%11027 = torch_c.from_builtin_tensor %11026 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11028 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8786 = arith.constant 1 : index
%dim_8787 = tensor.dim %11028, %c1_8786 : tensor<4x?x4096xf16>
%11029 = flow.tensor.transfer %11028 : tensor<4x?x4096xf16>{%dim_8787} to #hal.device.promise<@__device_2>
%11030 = torch_c.from_builtin_tensor %11029 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11031 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8788 = arith.constant 1 : index
%dim_8789 = tensor.dim %11031, %c1_8788 : tensor<4x?x4096xf16>
%11032 = flow.tensor.transfer %11031 : tensor<4x?x4096xf16>{%dim_8789} to #hal.device.promise<@__device_2>
%11033 = torch_c.from_builtin_tensor %11032 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8790 = torch.constant.int 1
%11034 = torch.aten.add.Tensor %11015, %11018, %int1_8790 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8791 = torch.constant.int 1
%11035 = torch.aten.add.Tensor %11034, %10926, %int1_8791 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8792 = torch.constant.int 1
%11036 = torch.aten.add.Tensor %11035, %11021, %int1_8792 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8793 = torch.constant.int 1
%11037 = torch.aten.add.Tensor %11036, %11024, %int1_8793 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8794 = torch.constant.int 1
%11038 = torch.aten.add.Tensor %11037, %11027, %int1_8794 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8795 = torch.constant.int 1
%11039 = torch.aten.add.Tensor %11038, %11030, %int1_8795 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8796 = torch.constant.int 1
%11040 = torch.aten.add.Tensor %11039, %11033, %int1_8796 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11041 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8797 = arith.constant 1 : index
%dim_8798 = tensor.dim %11041, %c1_8797 : tensor<4x?x4096xf16>
%11042 = flow.tensor.transfer %11041 : tensor<4x?x4096xf16>{%dim_8798} to #hal.device.promise<@__device_3>
%11043 = torch_c.from_builtin_tensor %11042 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11044 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8799 = arith.constant 1 : index
%dim_8800 = tensor.dim %11044, %c1_8799 : tensor<4x?x4096xf16>
%11045 = flow.tensor.transfer %11044 : tensor<4x?x4096xf16>{%dim_8800} to #hal.device.promise<@__device_3>
%11046 = torch_c.from_builtin_tensor %11045 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11047 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8801 = arith.constant 1 : index
%dim_8802 = tensor.dim %11047, %c1_8801 : tensor<4x?x4096xf16>
%11048 = flow.tensor.transfer %11047 : tensor<4x?x4096xf16>{%dim_8802} to #hal.device.promise<@__device_3>
%11049 = torch_c.from_builtin_tensor %11048 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11050 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8803 = arith.constant 1 : index
%dim_8804 = tensor.dim %11050, %c1_8803 : tensor<4x?x4096xf16>
%11051 = flow.tensor.transfer %11050 : tensor<4x?x4096xf16>{%dim_8804} to #hal.device.promise<@__device_3>
%11052 = torch_c.from_builtin_tensor %11051 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11053 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8805 = arith.constant 1 : index
%dim_8806 = tensor.dim %11053, %c1_8805 : tensor<4x?x4096xf16>
%11054 = flow.tensor.transfer %11053 : tensor<4x?x4096xf16>{%dim_8806} to #hal.device.promise<@__device_3>
%11055 = torch_c.from_builtin_tensor %11054 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11056 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8807 = arith.constant 1 : index
%dim_8808 = tensor.dim %11056, %c1_8807 : tensor<4x?x4096xf16>
%11057 = flow.tensor.transfer %11056 : tensor<4x?x4096xf16>{%dim_8808} to #hal.device.promise<@__device_3>
%11058 = torch_c.from_builtin_tensor %11057 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11059 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8809 = arith.constant 1 : index
%dim_8810 = tensor.dim %11059, %c1_8809 : tensor<4x?x4096xf16>
%11060 = flow.tensor.transfer %11059 : tensor<4x?x4096xf16>{%dim_8810} to #hal.device.promise<@__device_3>
%11061 = torch_c.from_builtin_tensor %11060 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8811 = torch.constant.int 1
%11062 = torch.aten.add.Tensor %11043, %11046, %int1_8811 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8812 = torch.constant.int 1
%11063 = torch.aten.add.Tensor %11062, %11049, %int1_8812 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8813 = torch.constant.int 1
%11064 = torch.aten.add.Tensor %11063, %10932, %int1_8813 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8814 = torch.constant.int 1
%11065 = torch.aten.add.Tensor %11064, %11052, %int1_8814 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8815 = torch.constant.int 1
%11066 = torch.aten.add.Tensor %11065, %11055, %int1_8815 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8816 = torch.constant.int 1
%11067 = torch.aten.add.Tensor %11066, %11058, %int1_8816 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8817 = torch.constant.int 1
%11068 = torch.aten.add.Tensor %11067, %11061, %int1_8817 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11069 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8818 = arith.constant 1 : index
%dim_8819 = tensor.dim %11069, %c1_8818 : tensor<4x?x4096xf16>
%11070 = flow.tensor.transfer %11069 : tensor<4x?x4096xf16>{%dim_8819} to #hal.device.promise<@__device_4>
%11071 = torch_c.from_builtin_tensor %11070 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11072 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8820 = arith.constant 1 : index
%dim_8821 = tensor.dim %11072, %c1_8820 : tensor<4x?x4096xf16>
%11073 = flow.tensor.transfer %11072 : tensor<4x?x4096xf16>{%dim_8821} to #hal.device.promise<@__device_4>
%11074 = torch_c.from_builtin_tensor %11073 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11075 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8822 = arith.constant 1 : index
%dim_8823 = tensor.dim %11075, %c1_8822 : tensor<4x?x4096xf16>
%11076 = flow.tensor.transfer %11075 : tensor<4x?x4096xf16>{%dim_8823} to #hal.device.promise<@__device_4>
%11077 = torch_c.from_builtin_tensor %11076 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11078 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8824 = arith.constant 1 : index
%dim_8825 = tensor.dim %11078, %c1_8824 : tensor<4x?x4096xf16>
%11079 = flow.tensor.transfer %11078 : tensor<4x?x4096xf16>{%dim_8825} to #hal.device.promise<@__device_4>
%11080 = torch_c.from_builtin_tensor %11079 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11081 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8826 = arith.constant 1 : index
%dim_8827 = tensor.dim %11081, %c1_8826 : tensor<4x?x4096xf16>
%11082 = flow.tensor.transfer %11081 : tensor<4x?x4096xf16>{%dim_8827} to #hal.device.promise<@__device_4>
%11083 = torch_c.from_builtin_tensor %11082 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11084 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8828 = arith.constant 1 : index
%dim_8829 = tensor.dim %11084, %c1_8828 : tensor<4x?x4096xf16>
%11085 = flow.tensor.transfer %11084 : tensor<4x?x4096xf16>{%dim_8829} to #hal.device.promise<@__device_4>
%11086 = torch_c.from_builtin_tensor %11085 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11087 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8830 = arith.constant 1 : index
%dim_8831 = tensor.dim %11087, %c1_8830 : tensor<4x?x4096xf16>
%11088 = flow.tensor.transfer %11087 : tensor<4x?x4096xf16>{%dim_8831} to #hal.device.promise<@__device_4>
%11089 = torch_c.from_builtin_tensor %11088 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8832 = torch.constant.int 1
%11090 = torch.aten.add.Tensor %11071, %11074, %int1_8832 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8833 = torch.constant.int 1
%11091 = torch.aten.add.Tensor %11090, %11077, %int1_8833 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8834 = torch.constant.int 1
%11092 = torch.aten.add.Tensor %11091, %11080, %int1_8834 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8835 = torch.constant.int 1
%11093 = torch.aten.add.Tensor %11092, %10938, %int1_8835 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8836 = torch.constant.int 1
%11094 = torch.aten.add.Tensor %11093, %11083, %int1_8836 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8837 = torch.constant.int 1
%11095 = torch.aten.add.Tensor %11094, %11086, %int1_8837 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8838 = torch.constant.int 1
%11096 = torch.aten.add.Tensor %11095, %11089, %int1_8838 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11097 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8839 = arith.constant 1 : index
%dim_8840 = tensor.dim %11097, %c1_8839 : tensor<4x?x4096xf16>
%11098 = flow.tensor.transfer %11097 : tensor<4x?x4096xf16>{%dim_8840} to #hal.device.promise<@__device_5>
%11099 = torch_c.from_builtin_tensor %11098 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11100 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8841 = arith.constant 1 : index
%dim_8842 = tensor.dim %11100, %c1_8841 : tensor<4x?x4096xf16>
%11101 = flow.tensor.transfer %11100 : tensor<4x?x4096xf16>{%dim_8842} to #hal.device.promise<@__device_5>
%11102 = torch_c.from_builtin_tensor %11101 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11103 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8843 = arith.constant 1 : index
%dim_8844 = tensor.dim %11103, %c1_8843 : tensor<4x?x4096xf16>
%11104 = flow.tensor.transfer %11103 : tensor<4x?x4096xf16>{%dim_8844} to #hal.device.promise<@__device_5>
%11105 = torch_c.from_builtin_tensor %11104 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11106 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8845 = arith.constant 1 : index
%dim_8846 = tensor.dim %11106, %c1_8845 : tensor<4x?x4096xf16>
%11107 = flow.tensor.transfer %11106 : tensor<4x?x4096xf16>{%dim_8846} to #hal.device.promise<@__device_5>
%11108 = torch_c.from_builtin_tensor %11107 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11109 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8847 = arith.constant 1 : index
%dim_8848 = tensor.dim %11109, %c1_8847 : tensor<4x?x4096xf16>
%11110 = flow.tensor.transfer %11109 : tensor<4x?x4096xf16>{%dim_8848} to #hal.device.promise<@__device_5>
%11111 = torch_c.from_builtin_tensor %11110 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11112 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8849 = arith.constant 1 : index
%dim_8850 = tensor.dim %11112, %c1_8849 : tensor<4x?x4096xf16>
%11113 = flow.tensor.transfer %11112 : tensor<4x?x4096xf16>{%dim_8850} to #hal.device.promise<@__device_5>
%11114 = torch_c.from_builtin_tensor %11113 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11115 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8851 = arith.constant 1 : index
%dim_8852 = tensor.dim %11115, %c1_8851 : tensor<4x?x4096xf16>
%11116 = flow.tensor.transfer %11115 : tensor<4x?x4096xf16>{%dim_8852} to #hal.device.promise<@__device_5>
%11117 = torch_c.from_builtin_tensor %11116 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8853 = torch.constant.int 1
%11118 = torch.aten.add.Tensor %11099, %11102, %int1_8853 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8854 = torch.constant.int 1
%11119 = torch.aten.add.Tensor %11118, %11105, %int1_8854 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8855 = torch.constant.int 1
%11120 = torch.aten.add.Tensor %11119, %11108, %int1_8855 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8856 = torch.constant.int 1
%11121 = torch.aten.add.Tensor %11120, %11111, %int1_8856 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8857 = torch.constant.int 1
%11122 = torch.aten.add.Tensor %11121, %10944, %int1_8857 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8858 = torch.constant.int 1
%11123 = torch.aten.add.Tensor %11122, %11114, %int1_8858 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8859 = torch.constant.int 1
%11124 = torch.aten.add.Tensor %11123, %11117, %int1_8859 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11124, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11125 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8860 = arith.constant 1 : index
%dim_8861 = tensor.dim %11125, %c1_8860 : tensor<4x?x4096xf16>
%11126 = flow.tensor.transfer %11125 : tensor<4x?x4096xf16>{%dim_8861} to #hal.device.promise<@__device_6>
%11127 = torch_c.from_builtin_tensor %11126 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11128 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8862 = arith.constant 1 : index
%dim_8863 = tensor.dim %11128, %c1_8862 : tensor<4x?x4096xf16>
%11129 = flow.tensor.transfer %11128 : tensor<4x?x4096xf16>{%dim_8863} to #hal.device.promise<@__device_6>
%11130 = torch_c.from_builtin_tensor %11129 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11131 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8864 = arith.constant 1 : index
%dim_8865 = tensor.dim %11131, %c1_8864 : tensor<4x?x4096xf16>
%11132 = flow.tensor.transfer %11131 : tensor<4x?x4096xf16>{%dim_8865} to #hal.device.promise<@__device_6>
%11133 = torch_c.from_builtin_tensor %11132 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11134 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8866 = arith.constant 1 : index
%dim_8867 = tensor.dim %11134, %c1_8866 : tensor<4x?x4096xf16>
%11135 = flow.tensor.transfer %11134 : tensor<4x?x4096xf16>{%dim_8867} to #hal.device.promise<@__device_6>
%11136 = torch_c.from_builtin_tensor %11135 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11137 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8868 = arith.constant 1 : index
%dim_8869 = tensor.dim %11137, %c1_8868 : tensor<4x?x4096xf16>
%11138 = flow.tensor.transfer %11137 : tensor<4x?x4096xf16>{%dim_8869} to #hal.device.promise<@__device_6>
%11139 = torch_c.from_builtin_tensor %11138 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11140 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8870 = arith.constant 1 : index
%dim_8871 = tensor.dim %11140, %c1_8870 : tensor<4x?x4096xf16>
%11141 = flow.tensor.transfer %11140 : tensor<4x?x4096xf16>{%dim_8871} to #hal.device.promise<@__device_6>
%11142 = torch_c.from_builtin_tensor %11141 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11143 = torch_c.to_builtin_tensor %10956 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8872 = arith.constant 1 : index
%dim_8873 = tensor.dim %11143, %c1_8872 : tensor<4x?x4096xf16>
%11144 = flow.tensor.transfer %11143 : tensor<4x?x4096xf16>{%dim_8873} to #hal.device.promise<@__device_6>
%11145 = torch_c.from_builtin_tensor %11144 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8874 = torch.constant.int 1
%11146 = torch.aten.add.Tensor %11127, %11130, %int1_8874 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8875 = torch.constant.int 1
%11147 = torch.aten.add.Tensor %11146, %11133, %int1_8875 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8876 = torch.constant.int 1
%11148 = torch.aten.add.Tensor %11147, %11136, %int1_8876 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8877 = torch.constant.int 1
%11149 = torch.aten.add.Tensor %11148, %11139, %int1_8877 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8878 = torch.constant.int 1
%11150 = torch.aten.add.Tensor %11149, %11142, %int1_8878 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8879 = torch.constant.int 1
%11151 = torch.aten.add.Tensor %11150, %10950, %int1_8879 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8880 = torch.constant.int 1
%11152 = torch.aten.add.Tensor %11151, %11145, %int1_8880 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11153 = torch_c.to_builtin_tensor %10914 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8881 = arith.constant 1 : index
%dim_8882 = tensor.dim %11153, %c1_8881 : tensor<4x?x4096xf16>
%11154 = flow.tensor.transfer %11153 : tensor<4x?x4096xf16>{%dim_8882} to #hal.device.promise<@__device_7>
%11155 = torch_c.from_builtin_tensor %11154 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11156 = torch_c.to_builtin_tensor %10920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8883 = arith.constant 1 : index
%dim_8884 = tensor.dim %11156, %c1_8883 : tensor<4x?x4096xf16>
%11157 = flow.tensor.transfer %11156 : tensor<4x?x4096xf16>{%dim_8884} to #hal.device.promise<@__device_7>
%11158 = torch_c.from_builtin_tensor %11157 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11159 = torch_c.to_builtin_tensor %10926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8885 = arith.constant 1 : index
%dim_8886 = tensor.dim %11159, %c1_8885 : tensor<4x?x4096xf16>
%11160 = flow.tensor.transfer %11159 : tensor<4x?x4096xf16>{%dim_8886} to #hal.device.promise<@__device_7>
%11161 = torch_c.from_builtin_tensor %11160 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11162 = torch_c.to_builtin_tensor %10932 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8887 = arith.constant 1 : index
%dim_8888 = tensor.dim %11162, %c1_8887 : tensor<4x?x4096xf16>
%11163 = flow.tensor.transfer %11162 : tensor<4x?x4096xf16>{%dim_8888} to #hal.device.promise<@__device_7>
%11164 = torch_c.from_builtin_tensor %11163 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11165 = torch_c.to_builtin_tensor %10938 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8889 = arith.constant 1 : index
%dim_8890 = tensor.dim %11165, %c1_8889 : tensor<4x?x4096xf16>
%11166 = flow.tensor.transfer %11165 : tensor<4x?x4096xf16>{%dim_8890} to #hal.device.promise<@__device_7>
%11167 = torch_c.from_builtin_tensor %11166 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11168 = torch_c.to_builtin_tensor %10944 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8891 = arith.constant 1 : index
%dim_8892 = tensor.dim %11168, %c1_8891 : tensor<4x?x4096xf16>
%11169 = flow.tensor.transfer %11168 : tensor<4x?x4096xf16>{%dim_8892} to #hal.device.promise<@__device_7>
%11170 = torch_c.from_builtin_tensor %11169 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11171 = torch_c.to_builtin_tensor %10950 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_8893 = arith.constant 1 : index
%dim_8894 = tensor.dim %11171, %c1_8893 : tensor<4x?x4096xf16>
%11172 = flow.tensor.transfer %11171 : tensor<4x?x4096xf16>{%dim_8894} to #hal.device.promise<@__device_7>
%11173 = torch_c.from_builtin_tensor %11172 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8895 = torch.constant.int 1
%11174 = torch.aten.add.Tensor %11155, %11158, %int1_8895 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8896 = torch.constant.int 1
%11175 = torch.aten.add.Tensor %11174, %11161, %int1_8896 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8897 = torch.constant.int 1
%11176 = torch.aten.add.Tensor %11175, %11164, %int1_8897 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8898 = torch.constant.int 1
%11177 = torch.aten.add.Tensor %11176, %11167, %int1_8898 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8899 = torch.constant.int 1
%11178 = torch.aten.add.Tensor %11177, %11170, %int1_8899 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8900 = torch.constant.int 1
%11179 = torch.aten.add.Tensor %11178, %11173, %int1_8900 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8901 = torch.constant.int 1
%11180 = torch.aten.add.Tensor %11179, %10956, %int1_8901 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8902 = torch.constant.int 1
%11181 = torch.aten.add.Tensor %9840, %10984, %int1_8902 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8903 = torch.constant.int 1
%11182 = torch.aten.add.Tensor %9841, %11012, %int1_8903 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8904 = torch.constant.int 1
%11183 = torch.aten.add.Tensor %9842, %11040, %int1_8904 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8905 = torch.constant.int 1
%11184 = torch.aten.add.Tensor %9843, %11068, %int1_8905 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8906 = torch.constant.int 1
%11185 = torch.aten.add.Tensor %9844, %11096, %int1_8906 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8907 = torch.constant.int 1
%11186 = torch.aten.add.Tensor %9845, %11124, %int1_8907 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8908 = torch.constant.int 1
%11187 = torch.aten.add.Tensor %9846, %11152, %int1_8908 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8909 = torch.constant.int 1
%11188 = torch.aten.add.Tensor %9847, %11180, %int1_8909 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_8910 = torch.constant.int 6
%11189 = torch.prims.convert_element_type %11181, %int6_8910 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8911 = torch.constant.int 6
%11190 = torch.prims.convert_element_type %11182, %int6_8911 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8912 = torch.constant.int 6
%11191 = torch.prims.convert_element_type %11183, %int6_8912 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8913 = torch.constant.int 6
%11192 = torch.prims.convert_element_type %11184, %int6_8913 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8914 = torch.constant.int 6
%11193 = torch.prims.convert_element_type %11185, %int6_8914 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8915 = torch.constant.int 6
%11194 = torch.prims.convert_element_type %11186, %int6_8915 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8916 = torch.constant.int 6
%11195 = torch.prims.convert_element_type %11187, %int6_8916 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_8917 = torch.constant.int 6
%11196 = torch.prims.convert_element_type %11188, %int6_8917 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8918 = torch.constant.int 2
%11197 = torch.aten.pow.Tensor_Scalar %11189, %int2_8918 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8919 = torch.constant.int 2
%11198 = torch.aten.pow.Tensor_Scalar %11190, %int2_8919 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8920 = torch.constant.int 2
%11199 = torch.aten.pow.Tensor_Scalar %11191, %int2_8920 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8921 = torch.constant.int 2
%11200 = torch.aten.pow.Tensor_Scalar %11192, %int2_8921 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8922 = torch.constant.int 2
%11201 = torch.aten.pow.Tensor_Scalar %11193, %int2_8922 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8923 = torch.constant.int 2
%11202 = torch.aten.pow.Tensor_Scalar %11194, %int2_8923 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8924 = torch.constant.int 2
%11203 = torch.aten.pow.Tensor_Scalar %11195, %int2_8924 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_8925 = torch.constant.int 2
%11204 = torch.aten.pow.Tensor_Scalar %11196, %int2_8925 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_8926 = torch.constant.int -1
%11205 = torch.prim.ListConstruct %int-1_8926 : (!torch.int) -> !torch.list<int>
%true_8927 = torch.constant.bool true
%none_8928 = torch.constant.none
%11206 = torch.aten.mean.dim %11197, %11205, %true_8927, %none_8928 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8929 = torch.constant.int -1
%11207 = torch.prim.ListConstruct %int-1_8929 : (!torch.int) -> !torch.list<int>
%true_8930 = torch.constant.bool true
%none_8931 = torch.constant.none
%11208 = torch.aten.mean.dim %11198, %11207, %true_8930, %none_8931 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8932 = torch.constant.int -1
%11209 = torch.prim.ListConstruct %int-1_8932 : (!torch.int) -> !torch.list<int>
%true_8933 = torch.constant.bool true
%none_8934 = torch.constant.none
%11210 = torch.aten.mean.dim %11199, %11209, %true_8933, %none_8934 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8935 = torch.constant.int -1
%11211 = torch.prim.ListConstruct %int-1_8935 : (!torch.int) -> !torch.list<int>
%true_8936 = torch.constant.bool true
%none_8937 = torch.constant.none
%11212 = torch.aten.mean.dim %11200, %11211, %true_8936, %none_8937 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8938 = torch.constant.int -1
%11213 = torch.prim.ListConstruct %int-1_8938 : (!torch.int) -> !torch.list<int>
%true_8939 = torch.constant.bool true
%none_8940 = torch.constant.none
%11214 = torch.aten.mean.dim %11201, %11213, %true_8939, %none_8940 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8941 = torch.constant.int -1
%11215 = torch.prim.ListConstruct %int-1_8941 : (!torch.int) -> !torch.list<int>
%true_8942 = torch.constant.bool true
%none_8943 = torch.constant.none
%11216 = torch.aten.mean.dim %11202, %11215, %true_8942, %none_8943 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8944 = torch.constant.int -1
%11217 = torch.prim.ListConstruct %int-1_8944 : (!torch.int) -> !torch.list<int>
%true_8945 = torch.constant.bool true
%none_8946 = torch.constant.none
%11218 = torch.aten.mean.dim %11203, %11217, %true_8945, %none_8946 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_8947 = torch.constant.int -1
%11219 = torch.prim.ListConstruct %int-1_8947 : (!torch.int) -> !torch.list<int>
%true_8948 = torch.constant.bool true
%none_8949 = torch.constant.none
%11220 = torch.aten.mean.dim %11204, %11219, %true_8948, %none_8949 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8950 = torch.constant.float 9.9999997473787516E-6
%int1_8951 = torch.constant.int 1
%11221 = torch.aten.add.Scalar %11206, %float9.999990e-06_8950, %int1_8951 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8952 = torch.constant.float 9.9999997473787516E-6
%int1_8953 = torch.constant.int 1
%11222 = torch.aten.add.Scalar %11208, %float9.999990e-06_8952, %int1_8953 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8954 = torch.constant.float 9.9999997473787516E-6
%int1_8955 = torch.constant.int 1
%11223 = torch.aten.add.Scalar %11210, %float9.999990e-06_8954, %int1_8955 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8956 = torch.constant.float 9.9999997473787516E-6
%int1_8957 = torch.constant.int 1
%11224 = torch.aten.add.Scalar %11212, %float9.999990e-06_8956, %int1_8957 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8958 = torch.constant.float 9.9999997473787516E-6
%int1_8959 = torch.constant.int 1
%11225 = torch.aten.add.Scalar %11214, %float9.999990e-06_8958, %int1_8959 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8960 = torch.constant.float 9.9999997473787516E-6
%int1_8961 = torch.constant.int 1
%11226 = torch.aten.add.Scalar %11216, %float9.999990e-06_8960, %int1_8961 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8962 = torch.constant.float 9.9999997473787516E-6
%int1_8963 = torch.constant.int 1
%11227 = torch.aten.add.Scalar %11218, %float9.999990e-06_8962, %int1_8963 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_8964 = torch.constant.float 9.9999997473787516E-6
%int1_8965 = torch.constant.int 1
%11228 = torch.aten.add.Scalar %11220, %float9.999990e-06_8964, %int1_8965 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11229 = torch.aten.rsqrt %11221 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11230 = torch.aten.rsqrt %11222 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11231 = torch.aten.rsqrt %11223 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11232 = torch.aten.rsqrt %11224 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11233 = torch.aten.rsqrt %11225 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11234 = torch.aten.rsqrt %11226 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11235 = torch.aten.rsqrt %11227 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11236 = torch.aten.rsqrt %11228 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11236, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11237 = torch.aten.mul.Tensor %11189, %11229 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11238 = torch.aten.mul.Tensor %11190, %11230 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11239 = torch.aten.mul.Tensor %11191, %11231 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11240 = torch.aten.mul.Tensor %11192, %11232 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11241 = torch.aten.mul.Tensor %11193, %11233 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11242 = torch.aten.mul.Tensor %11194, %11234 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11243 = torch.aten.mul.Tensor %11195, %11235 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11244 = torch.aten.mul.Tensor %11196, %11236 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11245 = torch.aten.mul.Tensor %336, %11237 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11246 = torch.aten.mul.Tensor %337, %11238 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11247 = torch.aten.mul.Tensor %338, %11239 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11248 = torch.aten.mul.Tensor %339, %11240 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11249 = torch.aten.mul.Tensor %340, %11241 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11250 = torch.aten.mul.Tensor %341, %11242 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11251 = torch.aten.mul.Tensor %342, %11243 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11252 = torch.aten.mul.Tensor %343, %11244 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_8966 = torch.constant.int 5
%11253 = torch.prims.convert_element_type %11245, %int5_8966 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8967 = torch.constant.int 5
%11254 = torch.prims.convert_element_type %11246, %int5_8967 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8968 = torch.constant.int 5
%11255 = torch.prims.convert_element_type %11247, %int5_8968 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8969 = torch.constant.int 5
%11256 = torch.prims.convert_element_type %11248, %int5_8969 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8970 = torch.constant.int 5
%11257 = torch.prims.convert_element_type %11249, %int5_8970 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8971 = torch.constant.int 5
%11258 = torch.prims.convert_element_type %11250, %int5_8971 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8972 = torch.constant.int 5
%11259 = torch.prims.convert_element_type %11251, %int5_8972 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_8973 = torch.constant.int 5
%11260 = torch.prims.convert_element_type %11252, %int5_8973 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_8974 = torch.constant.int 1
%int0_8975 = torch.constant.int 0
%11261 = torch.prim.ListConstruct %int1_8974, %int0_8975 : (!torch.int, !torch.int) -> !torch.list<int>
%11262 = torch.aten.permute %344, %11261 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8976 = torch.constant.int 1
%int0_8977 = torch.constant.int 0
%11263 = torch.prim.ListConstruct %int1_8976, %int0_8977 : (!torch.int, !torch.int) -> !torch.list<int>
%11264 = torch.aten.permute %345, %11263 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8978 = torch.constant.int 1
%int0_8979 = torch.constant.int 0
%11265 = torch.prim.ListConstruct %int1_8978, %int0_8979 : (!torch.int, !torch.int) -> !torch.list<int>
%11266 = torch.aten.permute %346, %11265 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8980 = torch.constant.int 1
%int0_8981 = torch.constant.int 0
%11267 = torch.prim.ListConstruct %int1_8980, %int0_8981 : (!torch.int, !torch.int) -> !torch.list<int>
%11268 = torch.aten.permute %347, %11267 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8982 = torch.constant.int 1
%int0_8983 = torch.constant.int 0
%11269 = torch.prim.ListConstruct %int1_8982, %int0_8983 : (!torch.int, !torch.int) -> !torch.list<int>
%11270 = torch.aten.permute %348, %11269 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8984 = torch.constant.int 1
%int0_8985 = torch.constant.int 0
%11271 = torch.prim.ListConstruct %int1_8984, %int0_8985 : (!torch.int, !torch.int) -> !torch.list<int>
%11272 = torch.aten.permute %349, %11271 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8986 = torch.constant.int 1
%int0_8987 = torch.constant.int 0
%11273 = torch.prim.ListConstruct %int1_8986, %int0_8987 : (!torch.int, !torch.int) -> !torch.list<int>
%11274 = torch.aten.permute %350, %11273 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_8988 = torch.constant.int 1
%int0_8989 = torch.constant.int 0
%11275 = torch.prim.ListConstruct %int1_8988, %int0_8989 : (!torch.int, !torch.int) -> !torch.list<int>
%11276 = torch.aten.permute %351, %11275 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_8990 = torch.constant.int 4
%11277 = torch.aten.mul.int %int4_8990, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_8991 = torch.constant.int 4096
%11278 = torch.prim.ListConstruct %11277, %int4096_8991 : (!torch.int, !torch.int) -> !torch.list<int>
%11279 = torch.aten.view %11253, %11278 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11279, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11280 = torch.aten.mm %11279, %11262 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11280, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_8992 = torch.constant.int 4
%int1792_8993 = torch.constant.int 1792
%11281 = torch.prim.ListConstruct %int4_8992, %2482, %int1792_8993 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11282 = torch.aten.view %11280, %11281 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_8994 = torch.constant.int 4
%11283 = torch.aten.mul.int %int4_8994, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_8995 = torch.constant.int 4096
%11284 = torch.prim.ListConstruct %11283, %int4096_8995 : (!torch.int, !torch.int) -> !torch.list<int>
%11285 = torch.aten.view %11254, %11284 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11285, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11286 = torch.aten.mm %11285, %11264 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11286, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_8996 = torch.constant.int 4
%int1792_8997 = torch.constant.int 1792
%11287 = torch.prim.ListConstruct %int4_8996, %2482, %int1792_8997 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11288 = torch.aten.view %11286, %11287 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_8998 = torch.constant.int 4
%11289 = torch.aten.mul.int %int4_8998, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_8999 = torch.constant.int 4096
%11290 = torch.prim.ListConstruct %11289, %int4096_8999 : (!torch.int, !torch.int) -> !torch.list<int>
%11291 = torch.aten.view %11255, %11290 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11291, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11292 = torch.aten.mm %11291, %11266 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11292, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9000 = torch.constant.int 4
%int1792_9001 = torch.constant.int 1792
%11293 = torch.prim.ListConstruct %int4_9000, %2482, %int1792_9001 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11294 = torch.aten.view %11292, %11293 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9002 = torch.constant.int 4
%11295 = torch.aten.mul.int %int4_9002, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9003 = torch.constant.int 4096
%11296 = torch.prim.ListConstruct %11295, %int4096_9003 : (!torch.int, !torch.int) -> !torch.list<int>
%11297 = torch.aten.view %11256, %11296 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11297, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11298 = torch.aten.mm %11297, %11268 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11298, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9004 = torch.constant.int 4
%int1792_9005 = torch.constant.int 1792
%11299 = torch.prim.ListConstruct %int4_9004, %2482, %int1792_9005 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11300 = torch.aten.view %11298, %11299 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9006 = torch.constant.int 4
%11301 = torch.aten.mul.int %int4_9006, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9007 = torch.constant.int 4096
%11302 = torch.prim.ListConstruct %11301, %int4096_9007 : (!torch.int, !torch.int) -> !torch.list<int>
%11303 = torch.aten.view %11257, %11302 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11303, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11304 = torch.aten.mm %11303, %11270 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11304, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9008 = torch.constant.int 4
%int1792_9009 = torch.constant.int 1792
%11305 = torch.prim.ListConstruct %int4_9008, %2482, %int1792_9009 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11306 = torch.aten.view %11304, %11305 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9010 = torch.constant.int 4
%11307 = torch.aten.mul.int %int4_9010, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9011 = torch.constant.int 4096
%11308 = torch.prim.ListConstruct %11307, %int4096_9011 : (!torch.int, !torch.int) -> !torch.list<int>
%11309 = torch.aten.view %11258, %11308 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11309, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11310 = torch.aten.mm %11309, %11272 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11310, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9012 = torch.constant.int 4
%int1792_9013 = torch.constant.int 1792
%11311 = torch.prim.ListConstruct %int4_9012, %2482, %int1792_9013 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11312 = torch.aten.view %11310, %11311 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9014 = torch.constant.int 4
%11313 = torch.aten.mul.int %int4_9014, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9015 = torch.constant.int 4096
%11314 = torch.prim.ListConstruct %11313, %int4096_9015 : (!torch.int, !torch.int) -> !torch.list<int>
%11315 = torch.aten.view %11259, %11314 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11315, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11316 = torch.aten.mm %11315, %11274 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11316, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9016 = torch.constant.int 4
%int1792_9017 = torch.constant.int 1792
%11317 = torch.prim.ListConstruct %int4_9016, %2482, %int1792_9017 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11318 = torch.aten.view %11316, %11317 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9018 = torch.constant.int 4
%11319 = torch.aten.mul.int %int4_9018, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9019 = torch.constant.int 4096
%11320 = torch.prim.ListConstruct %11319, %int4096_9019 : (!torch.int, !torch.int) -> !torch.list<int>
%11321 = torch.aten.view %11260, %11320 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11321, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11322 = torch.aten.mm %11321, %11276 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11322, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9020 = torch.constant.int 4
%int1792_9021 = torch.constant.int 1792
%11323 = torch.prim.ListConstruct %int4_9020, %2482, %int1792_9021 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11324 = torch.aten.view %11322, %11323 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11325 = torch.aten.silu %11282 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11326 = torch.aten.silu %11288 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11327 = torch.aten.silu %11294 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11328 = torch.aten.silu %11300 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11329 = torch.aten.silu %11306 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11330 = torch.aten.silu %11312 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11331 = torch.aten.silu %11318 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11332 = torch.aten.silu %11324 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_9022 = torch.constant.int 1
%int0_9023 = torch.constant.int 0
%11333 = torch.prim.ListConstruct %int1_9022, %int0_9023 : (!torch.int, !torch.int) -> !torch.list<int>
%11334 = torch.aten.permute %352, %11333 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9024 = torch.constant.int 1
%int0_9025 = torch.constant.int 0
%11335 = torch.prim.ListConstruct %int1_9024, %int0_9025 : (!torch.int, !torch.int) -> !torch.list<int>
%11336 = torch.aten.permute %353, %11335 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9026 = torch.constant.int 1
%int0_9027 = torch.constant.int 0
%11337 = torch.prim.ListConstruct %int1_9026, %int0_9027 : (!torch.int, !torch.int) -> !torch.list<int>
%11338 = torch.aten.permute %354, %11337 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9028 = torch.constant.int 1
%int0_9029 = torch.constant.int 0
%11339 = torch.prim.ListConstruct %int1_9028, %int0_9029 : (!torch.int, !torch.int) -> !torch.list<int>
%11340 = torch.aten.permute %355, %11339 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9030 = torch.constant.int 1
%int0_9031 = torch.constant.int 0
%11341 = torch.prim.ListConstruct %int1_9030, %int0_9031 : (!torch.int, !torch.int) -> !torch.list<int>
%11342 = torch.aten.permute %356, %11341 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9032 = torch.constant.int 1
%int0_9033 = torch.constant.int 0
%11343 = torch.prim.ListConstruct %int1_9032, %int0_9033 : (!torch.int, !torch.int) -> !torch.list<int>
%11344 = torch.aten.permute %357, %11343 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9034 = torch.constant.int 1
%int0_9035 = torch.constant.int 0
%11345 = torch.prim.ListConstruct %int1_9034, %int0_9035 : (!torch.int, !torch.int) -> !torch.list<int>
%11346 = torch.aten.permute %358, %11345 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_9036 = torch.constant.int 1
%int0_9037 = torch.constant.int 0
%11347 = torch.prim.ListConstruct %int1_9036, %int0_9037 : (!torch.int, !torch.int) -> !torch.list<int>
%11348 = torch.aten.permute %359, %11347 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_9038 = torch.constant.int 4
%11349 = torch.aten.mul.int %int4_9038, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9039 = torch.constant.int 4096
%11350 = torch.prim.ListConstruct %11349, %int4096_9039 : (!torch.int, !torch.int) -> !torch.list<int>
%11351 = torch.aten.view %11253, %11350 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11351, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11352 = torch.aten.mm %11351, %11334 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11352, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9040 = torch.constant.int 4
%int1792_9041 = torch.constant.int 1792
%11353 = torch.prim.ListConstruct %int4_9040, %2482, %int1792_9041 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11354 = torch.aten.view %11352, %11353 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9042 = torch.constant.int 4
%11355 = torch.aten.mul.int %int4_9042, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9043 = torch.constant.int 4096
%11356 = torch.prim.ListConstruct %11355, %int4096_9043 : (!torch.int, !torch.int) -> !torch.list<int>
%11357 = torch.aten.view %11254, %11356 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11357, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11358 = torch.aten.mm %11357, %11336 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11358, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9044 = torch.constant.int 4
%int1792_9045 = torch.constant.int 1792
%11359 = torch.prim.ListConstruct %int4_9044, %2482, %int1792_9045 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11360 = torch.aten.view %11358, %11359 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9046 = torch.constant.int 4
%11361 = torch.aten.mul.int %int4_9046, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9047 = torch.constant.int 4096
%11362 = torch.prim.ListConstruct %11361, %int4096_9047 : (!torch.int, !torch.int) -> !torch.list<int>
%11363 = torch.aten.view %11255, %11362 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11363, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11364 = torch.aten.mm %11363, %11338 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11364, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9048 = torch.constant.int 4
%int1792_9049 = torch.constant.int 1792
%11365 = torch.prim.ListConstruct %int4_9048, %2482, %int1792_9049 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11366 = torch.aten.view %11364, %11365 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9050 = torch.constant.int 4
%11367 = torch.aten.mul.int %int4_9050, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9051 = torch.constant.int 4096
%11368 = torch.prim.ListConstruct %11367, %int4096_9051 : (!torch.int, !torch.int) -> !torch.list<int>
%11369 = torch.aten.view %11256, %11368 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11369, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11370 = torch.aten.mm %11369, %11340 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11370, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9052 = torch.constant.int 4
%int1792_9053 = torch.constant.int 1792
%11371 = torch.prim.ListConstruct %int4_9052, %2482, %int1792_9053 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11372 = torch.aten.view %11370, %11371 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9054 = torch.constant.int 4
%11373 = torch.aten.mul.int %int4_9054, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9055 = torch.constant.int 4096
%11374 = torch.prim.ListConstruct %11373, %int4096_9055 : (!torch.int, !torch.int) -> !torch.list<int>
%11375 = torch.aten.view %11257, %11374 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11375, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11376 = torch.aten.mm %11375, %11342 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11376, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9056 = torch.constant.int 4
%int1792_9057 = torch.constant.int 1792
%11377 = torch.prim.ListConstruct %int4_9056, %2482, %int1792_9057 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11378 = torch.aten.view %11376, %11377 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9058 = torch.constant.int 4
%11379 = torch.aten.mul.int %int4_9058, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9059 = torch.constant.int 4096
%11380 = torch.prim.ListConstruct %11379, %int4096_9059 : (!torch.int, !torch.int) -> !torch.list<int>
%11381 = torch.aten.view %11258, %11380 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11381, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11382 = torch.aten.mm %11381, %11344 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11382, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9060 = torch.constant.int 4
%int1792_9061 = torch.constant.int 1792
%11383 = torch.prim.ListConstruct %int4_9060, %2482, %int1792_9061 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11384 = torch.aten.view %11382, %11383 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9062 = torch.constant.int 4
%11385 = torch.aten.mul.int %int4_9062, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9063 = torch.constant.int 4096
%11386 = torch.prim.ListConstruct %11385, %int4096_9063 : (!torch.int, !torch.int) -> !torch.list<int>
%11387 = torch.aten.view %11259, %11386 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11387, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11388 = torch.aten.mm %11387, %11346 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11388, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9064 = torch.constant.int 4
%int1792_9065 = torch.constant.int 1792
%11389 = torch.prim.ListConstruct %int4_9064, %2482, %int1792_9065 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11390 = torch.aten.view %11388, %11389 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_9066 = torch.constant.int 4
%11391 = torch.aten.mul.int %int4_9066, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9067 = torch.constant.int 4096
%11392 = torch.prim.ListConstruct %11391, %int4096_9067 : (!torch.int, !torch.int) -> !torch.list<int>
%11393 = torch.aten.view %11260, %11392 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11393, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11394 = torch.aten.mm %11393, %11348 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11394, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_9068 = torch.constant.int 4
%int1792_9069 = torch.constant.int 1792
%11395 = torch.prim.ListConstruct %int4_9068, %2482, %int1792_9069 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11396 = torch.aten.view %11394, %11395 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11397 = torch.aten.mul.Tensor %11325, %11354 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11398 = torch.aten.mul.Tensor %11326, %11360 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11399 = torch.aten.mul.Tensor %11327, %11366 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11400 = torch.aten.mul.Tensor %11328, %11372 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11401 = torch.aten.mul.Tensor %11329, %11378 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11402 = torch.aten.mul.Tensor %11330, %11384 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11402, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11403 = torch.aten.mul.Tensor %11331, %11390 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%11404 = torch.aten.mul.Tensor %11332, %11396 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %11404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_9070 = torch.constant.int 1
%int0_9071 = torch.constant.int 0
%11405 = torch.prim.ListConstruct %int1_9070, %int0_9071 : (!torch.int, !torch.int) -> !torch.list<int>
%11406 = torch.aten.permute %360, %11405 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9072 = torch.constant.int 1
%int0_9073 = torch.constant.int 0
%11407 = torch.prim.ListConstruct %int1_9072, %int0_9073 : (!torch.int, !torch.int) -> !torch.list<int>
%11408 = torch.aten.permute %361, %11407 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9074 = torch.constant.int 1
%int0_9075 = torch.constant.int 0
%11409 = torch.prim.ListConstruct %int1_9074, %int0_9075 : (!torch.int, !torch.int) -> !torch.list<int>
%11410 = torch.aten.permute %362, %11409 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9076 = torch.constant.int 1
%int0_9077 = torch.constant.int 0
%11411 = torch.prim.ListConstruct %int1_9076, %int0_9077 : (!torch.int, !torch.int) -> !torch.list<int>
%11412 = torch.aten.permute %363, %11411 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9078 = torch.constant.int 1
%int0_9079 = torch.constant.int 0
%11413 = torch.prim.ListConstruct %int1_9078, %int0_9079 : (!torch.int, !torch.int) -> !torch.list<int>
%11414 = torch.aten.permute %364, %11413 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9080 = torch.constant.int 1
%int0_9081 = torch.constant.int 0
%11415 = torch.prim.ListConstruct %int1_9080, %int0_9081 : (!torch.int, !torch.int) -> !torch.list<int>
%11416 = torch.aten.permute %365, %11415 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9082 = torch.constant.int 1
%int0_9083 = torch.constant.int 0
%11417 = torch.prim.ListConstruct %int1_9082, %int0_9083 : (!torch.int, !torch.int) -> !torch.list<int>
%11418 = torch.aten.permute %366, %11417 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9084 = torch.constant.int 1
%int0_9085 = torch.constant.int 0
%11419 = torch.prim.ListConstruct %int1_9084, %int0_9085 : (!torch.int, !torch.int) -> !torch.list<int>
%11420 = torch.aten.permute %367, %11419 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_9086 = torch.constant.int 1
%11421 = torch.aten.size.int %11282, %int1_9086 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9087 = torch.constant.int 4
%11422 = torch.aten.mul.int %int4_9087, %11421 : !torch.int, !torch.int -> !torch.int
%int1792_9088 = torch.constant.int 1792
%11423 = torch.prim.ListConstruct %11422, %int1792_9088 : (!torch.int, !torch.int) -> !torch.list<int>
%11424 = torch.aten.view %11397, %11423 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11424, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11425 = torch.aten.mm %11424, %11406 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11425, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9089 = torch.constant.int 4
%int4096_9090 = torch.constant.int 4096
%11426 = torch.prim.ListConstruct %int4_9089, %11421, %int4096_9090 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11427 = torch.aten.view %11425, %11426 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9091 = torch.constant.int 1
%11428 = torch.aten.size.int %11288, %int1_9091 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9092 = torch.constant.int 4
%11429 = torch.aten.mul.int %int4_9092, %11428 : !torch.int, !torch.int -> !torch.int
%int1792_9093 = torch.constant.int 1792
%11430 = torch.prim.ListConstruct %11429, %int1792_9093 : (!torch.int, !torch.int) -> !torch.list<int>
%11431 = torch.aten.view %11398, %11430 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11431, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11432 = torch.aten.mm %11431, %11408 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11432, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9094 = torch.constant.int 4
%int4096_9095 = torch.constant.int 4096
%11433 = torch.prim.ListConstruct %int4_9094, %11428, %int4096_9095 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11434 = torch.aten.view %11432, %11433 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9096 = torch.constant.int 1
%11435 = torch.aten.size.int %11294, %int1_9096 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9097 = torch.constant.int 4
%11436 = torch.aten.mul.int %int4_9097, %11435 : !torch.int, !torch.int -> !torch.int
%int1792_9098 = torch.constant.int 1792
%11437 = torch.prim.ListConstruct %11436, %int1792_9098 : (!torch.int, !torch.int) -> !torch.list<int>
%11438 = torch.aten.view %11399, %11437 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11438, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11439 = torch.aten.mm %11438, %11410 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11439, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9099 = torch.constant.int 4
%int4096_9100 = torch.constant.int 4096
%11440 = torch.prim.ListConstruct %int4_9099, %11435, %int4096_9100 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11441 = torch.aten.view %11439, %11440 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9101 = torch.constant.int 1
%11442 = torch.aten.size.int %11300, %int1_9101 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9102 = torch.constant.int 4
%11443 = torch.aten.mul.int %int4_9102, %11442 : !torch.int, !torch.int -> !torch.int
%int1792_9103 = torch.constant.int 1792
%11444 = torch.prim.ListConstruct %11443, %int1792_9103 : (!torch.int, !torch.int) -> !torch.list<int>
%11445 = torch.aten.view %11400, %11444 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11445, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11446 = torch.aten.mm %11445, %11412 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11446, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9104 = torch.constant.int 4
%int4096_9105 = torch.constant.int 4096
%11447 = torch.prim.ListConstruct %int4_9104, %11442, %int4096_9105 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11448 = torch.aten.view %11446, %11447 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9106 = torch.constant.int 1
%11449 = torch.aten.size.int %11306, %int1_9106 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9107 = torch.constant.int 4
%11450 = torch.aten.mul.int %int4_9107, %11449 : !torch.int, !torch.int -> !torch.int
%int1792_9108 = torch.constant.int 1792
%11451 = torch.prim.ListConstruct %11450, %int1792_9108 : (!torch.int, !torch.int) -> !torch.list<int>
%11452 = torch.aten.view %11401, %11451 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11452, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11453 = torch.aten.mm %11452, %11414 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11453, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9109 = torch.constant.int 4
%int4096_9110 = torch.constant.int 4096
%11454 = torch.prim.ListConstruct %int4_9109, %11449, %int4096_9110 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11455 = torch.aten.view %11453, %11454 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9111 = torch.constant.int 1
%11456 = torch.aten.size.int %11312, %int1_9111 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9112 = torch.constant.int 4
%11457 = torch.aten.mul.int %int4_9112, %11456 : !torch.int, !torch.int -> !torch.int
%int1792_9113 = torch.constant.int 1792
%11458 = torch.prim.ListConstruct %11457, %int1792_9113 : (!torch.int, !torch.int) -> !torch.list<int>
%11459 = torch.aten.view %11402, %11458 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11459, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11460 = torch.aten.mm %11459, %11416 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11460, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9114 = torch.constant.int 4
%int4096_9115 = torch.constant.int 4096
%11461 = torch.prim.ListConstruct %int4_9114, %11456, %int4096_9115 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11462 = torch.aten.view %11460, %11461 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9116 = torch.constant.int 1
%11463 = torch.aten.size.int %11318, %int1_9116 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9117 = torch.constant.int 4
%11464 = torch.aten.mul.int %int4_9117, %11463 : !torch.int, !torch.int -> !torch.int
%int1792_9118 = torch.constant.int 1792
%11465 = torch.prim.ListConstruct %11464, %int1792_9118 : (!torch.int, !torch.int) -> !torch.list<int>
%11466 = torch.aten.view %11403, %11465 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11466, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11467 = torch.aten.mm %11466, %11418 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11467, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9119 = torch.constant.int 4
%int4096_9120 = torch.constant.int 4096
%11468 = torch.prim.ListConstruct %int4_9119, %11463, %int4096_9120 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11469 = torch.aten.view %11467, %11468 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9121 = torch.constant.int 1
%11470 = torch.aten.size.int %11324, %int1_9121 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_9122 = torch.constant.int 4
%11471 = torch.aten.mul.int %int4_9122, %11470 : !torch.int, !torch.int -> !torch.int
%int1792_9123 = torch.constant.int 1792
%11472 = torch.prim.ListConstruct %11471, %int1792_9123 : (!torch.int, !torch.int) -> !torch.list<int>
%11473 = torch.aten.view %11404, %11472 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %11473, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%11474 = torch.aten.mm %11473, %11420 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11474, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_9124 = torch.constant.int 4
%int4096_9125 = torch.constant.int 4096
%11475 = torch.prim.ListConstruct %int4_9124, %11470, %int4096_9125 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11476 = torch.aten.view %11474, %11475 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11477 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9126 = arith.constant 1 : index
%dim_9127 = tensor.dim %11477, %c1_9126 : tensor<4x?x4096xf16>
%11478 = flow.tensor.transfer %11477 : tensor<4x?x4096xf16>{%dim_9127} to #hal.device.promise<@__device_0>
%11479 = torch_c.from_builtin_tensor %11478 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11480 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9128 = arith.constant 1 : index
%dim_9129 = tensor.dim %11480, %c1_9128 : tensor<4x?x4096xf16>
%11481 = flow.tensor.transfer %11480 : tensor<4x?x4096xf16>{%dim_9129} to #hal.device.promise<@__device_0>
%11482 = torch_c.from_builtin_tensor %11481 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11483 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9130 = arith.constant 1 : index
%dim_9131 = tensor.dim %11483, %c1_9130 : tensor<4x?x4096xf16>
%11484 = flow.tensor.transfer %11483 : tensor<4x?x4096xf16>{%dim_9131} to #hal.device.promise<@__device_0>
%11485 = torch_c.from_builtin_tensor %11484 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11486 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9132 = arith.constant 1 : index
%dim_9133 = tensor.dim %11486, %c1_9132 : tensor<4x?x4096xf16>
%11487 = flow.tensor.transfer %11486 : tensor<4x?x4096xf16>{%dim_9133} to #hal.device.promise<@__device_0>
%11488 = torch_c.from_builtin_tensor %11487 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11489 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9134 = arith.constant 1 : index
%dim_9135 = tensor.dim %11489, %c1_9134 : tensor<4x?x4096xf16>
%11490 = flow.tensor.transfer %11489 : tensor<4x?x4096xf16>{%dim_9135} to #hal.device.promise<@__device_0>
%11491 = torch_c.from_builtin_tensor %11490 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11492 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9136 = arith.constant 1 : index
%dim_9137 = tensor.dim %11492, %c1_9136 : tensor<4x?x4096xf16>
%11493 = flow.tensor.transfer %11492 : tensor<4x?x4096xf16>{%dim_9137} to #hal.device.promise<@__device_0>
%11494 = torch_c.from_builtin_tensor %11493 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11495 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9138 = arith.constant 1 : index
%dim_9139 = tensor.dim %11495, %c1_9138 : tensor<4x?x4096xf16>
%11496 = flow.tensor.transfer %11495 : tensor<4x?x4096xf16>{%dim_9139} to #hal.device.promise<@__device_0>
%11497 = torch_c.from_builtin_tensor %11496 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9140 = torch.constant.int 1
%11498 = torch.aten.add.Tensor %11427, %11479, %int1_9140 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9141 = torch.constant.int 1
%11499 = torch.aten.add.Tensor %11498, %11482, %int1_9141 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9142 = torch.constant.int 1
%11500 = torch.aten.add.Tensor %11499, %11485, %int1_9142 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9143 = torch.constant.int 1
%11501 = torch.aten.add.Tensor %11500, %11488, %int1_9143 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9144 = torch.constant.int 1
%11502 = torch.aten.add.Tensor %11501, %11491, %int1_9144 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9145 = torch.constant.int 1
%11503 = torch.aten.add.Tensor %11502, %11494, %int1_9145 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9146 = torch.constant.int 1
%11504 = torch.aten.add.Tensor %11503, %11497, %int1_9146 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11505 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9147 = arith.constant 1 : index
%dim_9148 = tensor.dim %11505, %c1_9147 : tensor<4x?x4096xf16>
%11506 = flow.tensor.transfer %11505 : tensor<4x?x4096xf16>{%dim_9148} to #hal.device.promise<@__device_1>
%11507 = torch_c.from_builtin_tensor %11506 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11508 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9149 = arith.constant 1 : index
%dim_9150 = tensor.dim %11508, %c1_9149 : tensor<4x?x4096xf16>
%11509 = flow.tensor.transfer %11508 : tensor<4x?x4096xf16>{%dim_9150} to #hal.device.promise<@__device_1>
%11510 = torch_c.from_builtin_tensor %11509 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11511 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9151 = arith.constant 1 : index
%dim_9152 = tensor.dim %11511, %c1_9151 : tensor<4x?x4096xf16>
%11512 = flow.tensor.transfer %11511 : tensor<4x?x4096xf16>{%dim_9152} to #hal.device.promise<@__device_1>
%11513 = torch_c.from_builtin_tensor %11512 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11514 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9153 = arith.constant 1 : index
%dim_9154 = tensor.dim %11514, %c1_9153 : tensor<4x?x4096xf16>
%11515 = flow.tensor.transfer %11514 : tensor<4x?x4096xf16>{%dim_9154} to #hal.device.promise<@__device_1>
%11516 = torch_c.from_builtin_tensor %11515 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11517 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9155 = arith.constant 1 : index
%dim_9156 = tensor.dim %11517, %c1_9155 : tensor<4x?x4096xf16>
%11518 = flow.tensor.transfer %11517 : tensor<4x?x4096xf16>{%dim_9156} to #hal.device.promise<@__device_1>
%11519 = torch_c.from_builtin_tensor %11518 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11520 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9157 = arith.constant 1 : index
%dim_9158 = tensor.dim %11520, %c1_9157 : tensor<4x?x4096xf16>
%11521 = flow.tensor.transfer %11520 : tensor<4x?x4096xf16>{%dim_9158} to #hal.device.promise<@__device_1>
%11522 = torch_c.from_builtin_tensor %11521 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11523 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9159 = arith.constant 1 : index
%dim_9160 = tensor.dim %11523, %c1_9159 : tensor<4x?x4096xf16>
%11524 = flow.tensor.transfer %11523 : tensor<4x?x4096xf16>{%dim_9160} to #hal.device.promise<@__device_1>
%11525 = torch_c.from_builtin_tensor %11524 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9161 = torch.constant.int 1
%11526 = torch.aten.add.Tensor %11507, %11434, %int1_9161 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9162 = torch.constant.int 1
%11527 = torch.aten.add.Tensor %11526, %11510, %int1_9162 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9163 = torch.constant.int 1
%11528 = torch.aten.add.Tensor %11527, %11513, %int1_9163 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9164 = torch.constant.int 1
%11529 = torch.aten.add.Tensor %11528, %11516, %int1_9164 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9165 = torch.constant.int 1
%11530 = torch.aten.add.Tensor %11529, %11519, %int1_9165 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9166 = torch.constant.int 1
%11531 = torch.aten.add.Tensor %11530, %11522, %int1_9166 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9167 = torch.constant.int 1
%11532 = torch.aten.add.Tensor %11531, %11525, %int1_9167 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11533 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9168 = arith.constant 1 : index
%dim_9169 = tensor.dim %11533, %c1_9168 : tensor<4x?x4096xf16>
%11534 = flow.tensor.transfer %11533 : tensor<4x?x4096xf16>{%dim_9169} to #hal.device.promise<@__device_2>
%11535 = torch_c.from_builtin_tensor %11534 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11536 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9170 = arith.constant 1 : index
%dim_9171 = tensor.dim %11536, %c1_9170 : tensor<4x?x4096xf16>
%11537 = flow.tensor.transfer %11536 : tensor<4x?x4096xf16>{%dim_9171} to #hal.device.promise<@__device_2>
%11538 = torch_c.from_builtin_tensor %11537 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11539 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9172 = arith.constant 1 : index
%dim_9173 = tensor.dim %11539, %c1_9172 : tensor<4x?x4096xf16>
%11540 = flow.tensor.transfer %11539 : tensor<4x?x4096xf16>{%dim_9173} to #hal.device.promise<@__device_2>
%11541 = torch_c.from_builtin_tensor %11540 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11541, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11542 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9174 = arith.constant 1 : index
%dim_9175 = tensor.dim %11542, %c1_9174 : tensor<4x?x4096xf16>
%11543 = flow.tensor.transfer %11542 : tensor<4x?x4096xf16>{%dim_9175} to #hal.device.promise<@__device_2>
%11544 = torch_c.from_builtin_tensor %11543 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11545 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9176 = arith.constant 1 : index
%dim_9177 = tensor.dim %11545, %c1_9176 : tensor<4x?x4096xf16>
%11546 = flow.tensor.transfer %11545 : tensor<4x?x4096xf16>{%dim_9177} to #hal.device.promise<@__device_2>
%11547 = torch_c.from_builtin_tensor %11546 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11548 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9178 = arith.constant 1 : index
%dim_9179 = tensor.dim %11548, %c1_9178 : tensor<4x?x4096xf16>
%11549 = flow.tensor.transfer %11548 : tensor<4x?x4096xf16>{%dim_9179} to #hal.device.promise<@__device_2>
%11550 = torch_c.from_builtin_tensor %11549 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11551 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9180 = arith.constant 1 : index
%dim_9181 = tensor.dim %11551, %c1_9180 : tensor<4x?x4096xf16>
%11552 = flow.tensor.transfer %11551 : tensor<4x?x4096xf16>{%dim_9181} to #hal.device.promise<@__device_2>
%11553 = torch_c.from_builtin_tensor %11552 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9182 = torch.constant.int 1
%11554 = torch.aten.add.Tensor %11535, %11538, %int1_9182 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9183 = torch.constant.int 1
%11555 = torch.aten.add.Tensor %11554, %11441, %int1_9183 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9184 = torch.constant.int 1
%11556 = torch.aten.add.Tensor %11555, %11541, %int1_9184 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9185 = torch.constant.int 1
%11557 = torch.aten.add.Tensor %11556, %11544, %int1_9185 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9186 = torch.constant.int 1
%11558 = torch.aten.add.Tensor %11557, %11547, %int1_9186 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9187 = torch.constant.int 1
%11559 = torch.aten.add.Tensor %11558, %11550, %int1_9187 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9188 = torch.constant.int 1
%11560 = torch.aten.add.Tensor %11559, %11553, %int1_9188 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11561 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9189 = arith.constant 1 : index
%dim_9190 = tensor.dim %11561, %c1_9189 : tensor<4x?x4096xf16>
%11562 = flow.tensor.transfer %11561 : tensor<4x?x4096xf16>{%dim_9190} to #hal.device.promise<@__device_3>
%11563 = torch_c.from_builtin_tensor %11562 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11564 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9191 = arith.constant 1 : index
%dim_9192 = tensor.dim %11564, %c1_9191 : tensor<4x?x4096xf16>
%11565 = flow.tensor.transfer %11564 : tensor<4x?x4096xf16>{%dim_9192} to #hal.device.promise<@__device_3>
%11566 = torch_c.from_builtin_tensor %11565 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11567 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9193 = arith.constant 1 : index
%dim_9194 = tensor.dim %11567, %c1_9193 : tensor<4x?x4096xf16>
%11568 = flow.tensor.transfer %11567 : tensor<4x?x4096xf16>{%dim_9194} to #hal.device.promise<@__device_3>
%11569 = torch_c.from_builtin_tensor %11568 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11570 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9195 = arith.constant 1 : index
%dim_9196 = tensor.dim %11570, %c1_9195 : tensor<4x?x4096xf16>
%11571 = flow.tensor.transfer %11570 : tensor<4x?x4096xf16>{%dim_9196} to #hal.device.promise<@__device_3>
%11572 = torch_c.from_builtin_tensor %11571 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11573 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9197 = arith.constant 1 : index
%dim_9198 = tensor.dim %11573, %c1_9197 : tensor<4x?x4096xf16>
%11574 = flow.tensor.transfer %11573 : tensor<4x?x4096xf16>{%dim_9198} to #hal.device.promise<@__device_3>
%11575 = torch_c.from_builtin_tensor %11574 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11576 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9199 = arith.constant 1 : index
%dim_9200 = tensor.dim %11576, %c1_9199 : tensor<4x?x4096xf16>
%11577 = flow.tensor.transfer %11576 : tensor<4x?x4096xf16>{%dim_9200} to #hal.device.promise<@__device_3>
%11578 = torch_c.from_builtin_tensor %11577 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11579 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9201 = arith.constant 1 : index
%dim_9202 = tensor.dim %11579, %c1_9201 : tensor<4x?x4096xf16>
%11580 = flow.tensor.transfer %11579 : tensor<4x?x4096xf16>{%dim_9202} to #hal.device.promise<@__device_3>
%11581 = torch_c.from_builtin_tensor %11580 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9203 = torch.constant.int 1
%11582 = torch.aten.add.Tensor %11563, %11566, %int1_9203 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9204 = torch.constant.int 1
%11583 = torch.aten.add.Tensor %11582, %11569, %int1_9204 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9205 = torch.constant.int 1
%11584 = torch.aten.add.Tensor %11583, %11448, %int1_9205 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9206 = torch.constant.int 1
%11585 = torch.aten.add.Tensor %11584, %11572, %int1_9206 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9207 = torch.constant.int 1
%11586 = torch.aten.add.Tensor %11585, %11575, %int1_9207 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9208 = torch.constant.int 1
%11587 = torch.aten.add.Tensor %11586, %11578, %int1_9208 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9209 = torch.constant.int 1
%11588 = torch.aten.add.Tensor %11587, %11581, %int1_9209 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11589 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9210 = arith.constant 1 : index
%dim_9211 = tensor.dim %11589, %c1_9210 : tensor<4x?x4096xf16>
%11590 = flow.tensor.transfer %11589 : tensor<4x?x4096xf16>{%dim_9211} to #hal.device.promise<@__device_4>
%11591 = torch_c.from_builtin_tensor %11590 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11592 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9212 = arith.constant 1 : index
%dim_9213 = tensor.dim %11592, %c1_9212 : tensor<4x?x4096xf16>
%11593 = flow.tensor.transfer %11592 : tensor<4x?x4096xf16>{%dim_9213} to #hal.device.promise<@__device_4>
%11594 = torch_c.from_builtin_tensor %11593 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11595 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9214 = arith.constant 1 : index
%dim_9215 = tensor.dim %11595, %c1_9214 : tensor<4x?x4096xf16>
%11596 = flow.tensor.transfer %11595 : tensor<4x?x4096xf16>{%dim_9215} to #hal.device.promise<@__device_4>
%11597 = torch_c.from_builtin_tensor %11596 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11598 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9216 = arith.constant 1 : index
%dim_9217 = tensor.dim %11598, %c1_9216 : tensor<4x?x4096xf16>
%11599 = flow.tensor.transfer %11598 : tensor<4x?x4096xf16>{%dim_9217} to #hal.device.promise<@__device_4>
%11600 = torch_c.from_builtin_tensor %11599 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11601 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9218 = arith.constant 1 : index
%dim_9219 = tensor.dim %11601, %c1_9218 : tensor<4x?x4096xf16>
%11602 = flow.tensor.transfer %11601 : tensor<4x?x4096xf16>{%dim_9219} to #hal.device.promise<@__device_4>
%11603 = torch_c.from_builtin_tensor %11602 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11604 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9220 = arith.constant 1 : index
%dim_9221 = tensor.dim %11604, %c1_9220 : tensor<4x?x4096xf16>
%11605 = flow.tensor.transfer %11604 : tensor<4x?x4096xf16>{%dim_9221} to #hal.device.promise<@__device_4>
%11606 = torch_c.from_builtin_tensor %11605 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11607 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9222 = arith.constant 1 : index
%dim_9223 = tensor.dim %11607, %c1_9222 : tensor<4x?x4096xf16>
%11608 = flow.tensor.transfer %11607 : tensor<4x?x4096xf16>{%dim_9223} to #hal.device.promise<@__device_4>
%11609 = torch_c.from_builtin_tensor %11608 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9224 = torch.constant.int 1
%11610 = torch.aten.add.Tensor %11591, %11594, %int1_9224 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9225 = torch.constant.int 1
%11611 = torch.aten.add.Tensor %11610, %11597, %int1_9225 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9226 = torch.constant.int 1
%11612 = torch.aten.add.Tensor %11611, %11600, %int1_9226 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9227 = torch.constant.int 1
%11613 = torch.aten.add.Tensor %11612, %11455, %int1_9227 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9228 = torch.constant.int 1
%11614 = torch.aten.add.Tensor %11613, %11603, %int1_9228 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9229 = torch.constant.int 1
%11615 = torch.aten.add.Tensor %11614, %11606, %int1_9229 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9230 = torch.constant.int 1
%11616 = torch.aten.add.Tensor %11615, %11609, %int1_9230 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11617 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9231 = arith.constant 1 : index
%dim_9232 = tensor.dim %11617, %c1_9231 : tensor<4x?x4096xf16>
%11618 = flow.tensor.transfer %11617 : tensor<4x?x4096xf16>{%dim_9232} to #hal.device.promise<@__device_5>
%11619 = torch_c.from_builtin_tensor %11618 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11620 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9233 = arith.constant 1 : index
%dim_9234 = tensor.dim %11620, %c1_9233 : tensor<4x?x4096xf16>
%11621 = flow.tensor.transfer %11620 : tensor<4x?x4096xf16>{%dim_9234} to #hal.device.promise<@__device_5>
%11622 = torch_c.from_builtin_tensor %11621 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11623 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9235 = arith.constant 1 : index
%dim_9236 = tensor.dim %11623, %c1_9235 : tensor<4x?x4096xf16>
%11624 = flow.tensor.transfer %11623 : tensor<4x?x4096xf16>{%dim_9236} to #hal.device.promise<@__device_5>
%11625 = torch_c.from_builtin_tensor %11624 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11626 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9237 = arith.constant 1 : index
%dim_9238 = tensor.dim %11626, %c1_9237 : tensor<4x?x4096xf16>
%11627 = flow.tensor.transfer %11626 : tensor<4x?x4096xf16>{%dim_9238} to #hal.device.promise<@__device_5>
%11628 = torch_c.from_builtin_tensor %11627 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11629 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9239 = arith.constant 1 : index
%dim_9240 = tensor.dim %11629, %c1_9239 : tensor<4x?x4096xf16>
%11630 = flow.tensor.transfer %11629 : tensor<4x?x4096xf16>{%dim_9240} to #hal.device.promise<@__device_5>
%11631 = torch_c.from_builtin_tensor %11630 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11632 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9241 = arith.constant 1 : index
%dim_9242 = tensor.dim %11632, %c1_9241 : tensor<4x?x4096xf16>
%11633 = flow.tensor.transfer %11632 : tensor<4x?x4096xf16>{%dim_9242} to #hal.device.promise<@__device_5>
%11634 = torch_c.from_builtin_tensor %11633 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11635 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9243 = arith.constant 1 : index
%dim_9244 = tensor.dim %11635, %c1_9243 : tensor<4x?x4096xf16>
%11636 = flow.tensor.transfer %11635 : tensor<4x?x4096xf16>{%dim_9244} to #hal.device.promise<@__device_5>
%11637 = torch_c.from_builtin_tensor %11636 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9245 = torch.constant.int 1
%11638 = torch.aten.add.Tensor %11619, %11622, %int1_9245 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9246 = torch.constant.int 1
%11639 = torch.aten.add.Tensor %11638, %11625, %int1_9246 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9247 = torch.constant.int 1
%11640 = torch.aten.add.Tensor %11639, %11628, %int1_9247 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9248 = torch.constant.int 1
%11641 = torch.aten.add.Tensor %11640, %11631, %int1_9248 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9249 = torch.constant.int 1
%11642 = torch.aten.add.Tensor %11641, %11462, %int1_9249 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9250 = torch.constant.int 1
%11643 = torch.aten.add.Tensor %11642, %11634, %int1_9250 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9251 = torch.constant.int 1
%11644 = torch.aten.add.Tensor %11643, %11637, %int1_9251 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11645 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9252 = arith.constant 1 : index
%dim_9253 = tensor.dim %11645, %c1_9252 : tensor<4x?x4096xf16>
%11646 = flow.tensor.transfer %11645 : tensor<4x?x4096xf16>{%dim_9253} to #hal.device.promise<@__device_6>
%11647 = torch_c.from_builtin_tensor %11646 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11648 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9254 = arith.constant 1 : index
%dim_9255 = tensor.dim %11648, %c1_9254 : tensor<4x?x4096xf16>
%11649 = flow.tensor.transfer %11648 : tensor<4x?x4096xf16>{%dim_9255} to #hal.device.promise<@__device_6>
%11650 = torch_c.from_builtin_tensor %11649 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11651 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9256 = arith.constant 1 : index
%dim_9257 = tensor.dim %11651, %c1_9256 : tensor<4x?x4096xf16>
%11652 = flow.tensor.transfer %11651 : tensor<4x?x4096xf16>{%dim_9257} to #hal.device.promise<@__device_6>
%11653 = torch_c.from_builtin_tensor %11652 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11654 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9258 = arith.constant 1 : index
%dim_9259 = tensor.dim %11654, %c1_9258 : tensor<4x?x4096xf16>
%11655 = flow.tensor.transfer %11654 : tensor<4x?x4096xf16>{%dim_9259} to #hal.device.promise<@__device_6>
%11656 = torch_c.from_builtin_tensor %11655 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11657 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9260 = arith.constant 1 : index
%dim_9261 = tensor.dim %11657, %c1_9260 : tensor<4x?x4096xf16>
%11658 = flow.tensor.transfer %11657 : tensor<4x?x4096xf16>{%dim_9261} to #hal.device.promise<@__device_6>
%11659 = torch_c.from_builtin_tensor %11658 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11660 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9262 = arith.constant 1 : index
%dim_9263 = tensor.dim %11660, %c1_9262 : tensor<4x?x4096xf16>
%11661 = flow.tensor.transfer %11660 : tensor<4x?x4096xf16>{%dim_9263} to #hal.device.promise<@__device_6>
%11662 = torch_c.from_builtin_tensor %11661 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11663 = torch_c.to_builtin_tensor %11476 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9264 = arith.constant 1 : index
%dim_9265 = tensor.dim %11663, %c1_9264 : tensor<4x?x4096xf16>
%11664 = flow.tensor.transfer %11663 : tensor<4x?x4096xf16>{%dim_9265} to #hal.device.promise<@__device_6>
%11665 = torch_c.from_builtin_tensor %11664 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9266 = torch.constant.int 1
%11666 = torch.aten.add.Tensor %11647, %11650, %int1_9266 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9267 = torch.constant.int 1
%11667 = torch.aten.add.Tensor %11666, %11653, %int1_9267 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9268 = torch.constant.int 1
%11668 = torch.aten.add.Tensor %11667, %11656, %int1_9268 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9269 = torch.constant.int 1
%11669 = torch.aten.add.Tensor %11668, %11659, %int1_9269 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9270 = torch.constant.int 1
%11670 = torch.aten.add.Tensor %11669, %11662, %int1_9270 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9271 = torch.constant.int 1
%11671 = torch.aten.add.Tensor %11670, %11469, %int1_9271 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9272 = torch.constant.int 1
%11672 = torch.aten.add.Tensor %11671, %11665, %int1_9272 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11673 = torch_c.to_builtin_tensor %11427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9273 = arith.constant 1 : index
%dim_9274 = tensor.dim %11673, %c1_9273 : tensor<4x?x4096xf16>
%11674 = flow.tensor.transfer %11673 : tensor<4x?x4096xf16>{%dim_9274} to #hal.device.promise<@__device_7>
%11675 = torch_c.from_builtin_tensor %11674 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11676 = torch_c.to_builtin_tensor %11434 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9275 = arith.constant 1 : index
%dim_9276 = tensor.dim %11676, %c1_9275 : tensor<4x?x4096xf16>
%11677 = flow.tensor.transfer %11676 : tensor<4x?x4096xf16>{%dim_9276} to #hal.device.promise<@__device_7>
%11678 = torch_c.from_builtin_tensor %11677 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11679 = torch_c.to_builtin_tensor %11441 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9277 = arith.constant 1 : index
%dim_9278 = tensor.dim %11679, %c1_9277 : tensor<4x?x4096xf16>
%11680 = flow.tensor.transfer %11679 : tensor<4x?x4096xf16>{%dim_9278} to #hal.device.promise<@__device_7>
%11681 = torch_c.from_builtin_tensor %11680 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11682 = torch_c.to_builtin_tensor %11448 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9279 = arith.constant 1 : index
%dim_9280 = tensor.dim %11682, %c1_9279 : tensor<4x?x4096xf16>
%11683 = flow.tensor.transfer %11682 : tensor<4x?x4096xf16>{%dim_9280} to #hal.device.promise<@__device_7>
%11684 = torch_c.from_builtin_tensor %11683 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11685 = torch_c.to_builtin_tensor %11455 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9281 = arith.constant 1 : index
%dim_9282 = tensor.dim %11685, %c1_9281 : tensor<4x?x4096xf16>
%11686 = flow.tensor.transfer %11685 : tensor<4x?x4096xf16>{%dim_9282} to #hal.device.promise<@__device_7>
%11687 = torch_c.from_builtin_tensor %11686 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11688 = torch_c.to_builtin_tensor %11462 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9283 = arith.constant 1 : index
%dim_9284 = tensor.dim %11688, %c1_9283 : tensor<4x?x4096xf16>
%11689 = flow.tensor.transfer %11688 : tensor<4x?x4096xf16>{%dim_9284} to #hal.device.promise<@__device_7>
%11690 = torch_c.from_builtin_tensor %11689 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%11691 = torch_c.to_builtin_tensor %11469 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_9285 = arith.constant 1 : index
%dim_9286 = tensor.dim %11691, %c1_9285 : tensor<4x?x4096xf16>
%11692 = flow.tensor.transfer %11691 : tensor<4x?x4096xf16>{%dim_9286} to #hal.device.promise<@__device_7>
%11693 = torch_c.from_builtin_tensor %11692 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9287 = torch.constant.int 1
%11694 = torch.aten.add.Tensor %11675, %11678, %int1_9287 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9288 = torch.constant.int 1
%11695 = torch.aten.add.Tensor %11694, %11681, %int1_9288 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9289 = torch.constant.int 1
%11696 = torch.aten.add.Tensor %11695, %11684, %int1_9289 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9290 = torch.constant.int 1
%11697 = torch.aten.add.Tensor %11696, %11687, %int1_9290 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9291 = torch.constant.int 1
%11698 = torch.aten.add.Tensor %11697, %11690, %int1_9291 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9292 = torch.constant.int 1
%11699 = torch.aten.add.Tensor %11698, %11693, %int1_9292 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9293 = torch.constant.int 1
%11700 = torch.aten.add.Tensor %11699, %11476, %int1_9293 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9294 = torch.constant.int 1
%11701 = torch.aten.add.Tensor %11181, %11504, %int1_9294 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9295 = torch.constant.int 1
%11702 = torch.aten.add.Tensor %11182, %11532, %int1_9295 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9296 = torch.constant.int 1
%11703 = torch.aten.add.Tensor %11183, %11560, %int1_9296 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9297 = torch.constant.int 1
%11704 = torch.aten.add.Tensor %11184, %11588, %int1_9297 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9298 = torch.constant.int 1
%11705 = torch.aten.add.Tensor %11185, %11616, %int1_9298 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9299 = torch.constant.int 1
%11706 = torch.aten.add.Tensor %11186, %11644, %int1_9299 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9300 = torch.constant.int 1
%11707 = torch.aten.add.Tensor %11187, %11672, %int1_9300 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9301 = torch.constant.int 1
%11708 = torch.aten.add.Tensor %11188, %11700, %int1_9301 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_9302 = torch.constant.int 6
%11709 = torch.prims.convert_element_type %11701, %int6_9302 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9303 = torch.constant.int 6
%11710 = torch.prims.convert_element_type %11702, %int6_9303 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9304 = torch.constant.int 6
%11711 = torch.prims.convert_element_type %11703, %int6_9304 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9305 = torch.constant.int 6
%11712 = torch.prims.convert_element_type %11704, %int6_9305 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9306 = torch.constant.int 6
%11713 = torch.prims.convert_element_type %11705, %int6_9306 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9307 = torch.constant.int 6
%11714 = torch.prims.convert_element_type %11706, %int6_9307 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9308 = torch.constant.int 6
%11715 = torch.prims.convert_element_type %11707, %int6_9308 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_9309 = torch.constant.int 6
%11716 = torch.prims.convert_element_type %11708, %int6_9309 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9310 = torch.constant.int 2
%11717 = torch.aten.pow.Tensor_Scalar %11709, %int2_9310 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9311 = torch.constant.int 2
%11718 = torch.aten.pow.Tensor_Scalar %11710, %int2_9311 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9312 = torch.constant.int 2
%11719 = torch.aten.pow.Tensor_Scalar %11711, %int2_9312 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9313 = torch.constant.int 2
%11720 = torch.aten.pow.Tensor_Scalar %11712, %int2_9313 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9314 = torch.constant.int 2
%11721 = torch.aten.pow.Tensor_Scalar %11713, %int2_9314 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9315 = torch.constant.int 2
%11722 = torch.aten.pow.Tensor_Scalar %11714, %int2_9315 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9316 = torch.constant.int 2
%11723 = torch.aten.pow.Tensor_Scalar %11715, %int2_9316 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_9317 = torch.constant.int 2
%11724 = torch.aten.pow.Tensor_Scalar %11716, %int2_9317 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_9318 = torch.constant.int -1
%11725 = torch.prim.ListConstruct %int-1_9318 : (!torch.int) -> !torch.list<int>
%true_9319 = torch.constant.bool true
%none_9320 = torch.constant.none
%11726 = torch.aten.mean.dim %11717, %11725, %true_9319, %none_9320 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9321 = torch.constant.int -1
%11727 = torch.prim.ListConstruct %int-1_9321 : (!torch.int) -> !torch.list<int>
%true_9322 = torch.constant.bool true
%none_9323 = torch.constant.none
%11728 = torch.aten.mean.dim %11718, %11727, %true_9322, %none_9323 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9324 = torch.constant.int -1
%11729 = torch.prim.ListConstruct %int-1_9324 : (!torch.int) -> !torch.list<int>
%true_9325 = torch.constant.bool true
%none_9326 = torch.constant.none
%11730 = torch.aten.mean.dim %11719, %11729, %true_9325, %none_9326 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9327 = torch.constant.int -1
%11731 = torch.prim.ListConstruct %int-1_9327 : (!torch.int) -> !torch.list<int>
%true_9328 = torch.constant.bool true
%none_9329 = torch.constant.none
%11732 = torch.aten.mean.dim %11720, %11731, %true_9328, %none_9329 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9330 = torch.constant.int -1
%11733 = torch.prim.ListConstruct %int-1_9330 : (!torch.int) -> !torch.list<int>
%true_9331 = torch.constant.bool true
%none_9332 = torch.constant.none
%11734 = torch.aten.mean.dim %11721, %11733, %true_9331, %none_9332 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9333 = torch.constant.int -1
%11735 = torch.prim.ListConstruct %int-1_9333 : (!torch.int) -> !torch.list<int>
%true_9334 = torch.constant.bool true
%none_9335 = torch.constant.none
%11736 = torch.aten.mean.dim %11722, %11735, %true_9334, %none_9335 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9336 = torch.constant.int -1
%11737 = torch.prim.ListConstruct %int-1_9336 : (!torch.int) -> !torch.list<int>
%true_9337 = torch.constant.bool true
%none_9338 = torch.constant.none
%11738 = torch.aten.mean.dim %11723, %11737, %true_9337, %none_9338 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_9339 = torch.constant.int -1
%11739 = torch.prim.ListConstruct %int-1_9339 : (!torch.int) -> !torch.list<int>
%true_9340 = torch.constant.bool true
%none_9341 = torch.constant.none
%11740 = torch.aten.mean.dim %11724, %11739, %true_9340, %none_9341 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9342 = torch.constant.float 9.9999997473787516E-6
%int1_9343 = torch.constant.int 1
%11741 = torch.aten.add.Scalar %11726, %float9.999990e-06_9342, %int1_9343 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9344 = torch.constant.float 9.9999997473787516E-6
%int1_9345 = torch.constant.int 1
%11742 = torch.aten.add.Scalar %11728, %float9.999990e-06_9344, %int1_9345 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9346 = torch.constant.float 9.9999997473787516E-6
%int1_9347 = torch.constant.int 1
%11743 = torch.aten.add.Scalar %11730, %float9.999990e-06_9346, %int1_9347 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9348 = torch.constant.float 9.9999997473787516E-6
%int1_9349 = torch.constant.int 1
%11744 = torch.aten.add.Scalar %11732, %float9.999990e-06_9348, %int1_9349 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9350 = torch.constant.float 9.9999997473787516E-6
%int1_9351 = torch.constant.int 1
%11745 = torch.aten.add.Scalar %11734, %float9.999990e-06_9350, %int1_9351 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9352 = torch.constant.float 9.9999997473787516E-6
%int1_9353 = torch.constant.int 1
%11746 = torch.aten.add.Scalar %11736, %float9.999990e-06_9352, %int1_9353 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9354 = torch.constant.float 9.9999997473787516E-6
%int1_9355 = torch.constant.int 1
%11747 = torch.aten.add.Scalar %11738, %float9.999990e-06_9354, %int1_9355 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_9356 = torch.constant.float 9.9999997473787516E-6
%int1_9357 = torch.constant.int 1
%11748 = torch.aten.add.Scalar %11740, %float9.999990e-06_9356, %int1_9357 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11749 = torch.aten.rsqrt %11741 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11750 = torch.aten.rsqrt %11742 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11751 = torch.aten.rsqrt %11743 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11752 = torch.aten.rsqrt %11744 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11753 = torch.aten.rsqrt %11745 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11754 = torch.aten.rsqrt %11746 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11755 = torch.aten.rsqrt %11747 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11756 = torch.aten.rsqrt %11748 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %11756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%11757 = torch.aten.mul.Tensor %11709, %11749 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11758 = torch.aten.mul.Tensor %11710, %11750 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11759 = torch.aten.mul.Tensor %11711, %11751 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11760 = torch.aten.mul.Tensor %11712, %11752 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11761 = torch.aten.mul.Tensor %11713, %11753 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11762 = torch.aten.mul.Tensor %11714, %11754 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11763 = torch.aten.mul.Tensor %11715, %11755 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11764 = torch.aten.mul.Tensor %11716, %11756 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11765 = torch.aten.mul.Tensor %368, %11757 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11766 = torch.aten.mul.Tensor %369, %11758 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11767 = torch.aten.mul.Tensor %370, %11759 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11768 = torch.aten.mul.Tensor %371, %11760 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11769 = torch.aten.mul.Tensor %372, %11761 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11770 = torch.aten.mul.Tensor %373, %11762 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11771 = torch.aten.mul.Tensor %374, %11763 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%11772 = torch.aten.mul.Tensor %375, %11764 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %11772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_9358 = torch.constant.int 5
%11773 = torch.prims.convert_element_type %11765, %int5_9358 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9359 = torch.constant.int 5
%11774 = torch.prims.convert_element_type %11766, %int5_9359 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9360 = torch.constant.int 5
%11775 = torch.prims.convert_element_type %11767, %int5_9360 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9361 = torch.constant.int 5
%11776 = torch.prims.convert_element_type %11768, %int5_9361 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9362 = torch.constant.int 5
%11777 = torch.prims.convert_element_type %11769, %int5_9362 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9363 = torch.constant.int 5
%11778 = torch.prims.convert_element_type %11770, %int5_9363 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9364 = torch.constant.int 5
%11779 = torch.prims.convert_element_type %11771, %int5_9364 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_9365 = torch.constant.int 5
%11780 = torch.prims.convert_element_type %11772, %int5_9365 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %11780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_9366 = torch.constant.int 1
%int0_9367 = torch.constant.int 0
%11781 = torch.prim.ListConstruct %int1_9366, %int0_9367 : (!torch.int, !torch.int) -> !torch.list<int>
%11782 = torch.aten.permute %376, %11781 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9368 = torch.constant.int 1
%int0_9369 = torch.constant.int 0
%11783 = torch.prim.ListConstruct %int1_9368, %int0_9369 : (!torch.int, !torch.int) -> !torch.list<int>
%11784 = torch.aten.permute %377, %11783 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9370 = torch.constant.int 1
%int0_9371 = torch.constant.int 0
%11785 = torch.prim.ListConstruct %int1_9370, %int0_9371 : (!torch.int, !torch.int) -> !torch.list<int>
%11786 = torch.aten.permute %378, %11785 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9372 = torch.constant.int 1
%int0_9373 = torch.constant.int 0
%11787 = torch.prim.ListConstruct %int1_9372, %int0_9373 : (!torch.int, !torch.int) -> !torch.list<int>
%11788 = torch.aten.permute %379, %11787 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9374 = torch.constant.int 1
%int0_9375 = torch.constant.int 0
%11789 = torch.prim.ListConstruct %int1_9374, %int0_9375 : (!torch.int, !torch.int) -> !torch.list<int>
%11790 = torch.aten.permute %380, %11789 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9376 = torch.constant.int 1
%int0_9377 = torch.constant.int 0
%11791 = torch.prim.ListConstruct %int1_9376, %int0_9377 : (!torch.int, !torch.int) -> !torch.list<int>
%11792 = torch.aten.permute %381, %11791 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9378 = torch.constant.int 1
%int0_9379 = torch.constant.int 0
%11793 = torch.prim.ListConstruct %int1_9378, %int0_9379 : (!torch.int, !torch.int) -> !torch.list<int>
%11794 = torch.aten.permute %382, %11793 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_9380 = torch.constant.int 1
%int0_9381 = torch.constant.int 0
%11795 = torch.prim.ListConstruct %int1_9380, %int0_9381 : (!torch.int, !torch.int) -> !torch.list<int>
%11796 = torch.aten.permute %383, %11795 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_9382 = torch.constant.int 4
%11797 = torch.aten.mul.int %int4_9382, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9383 = torch.constant.int 4096
%11798 = torch.prim.ListConstruct %11797, %int4096_9383 : (!torch.int, !torch.int) -> !torch.list<int>
%11799 = torch.aten.view %11773, %11798 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11799, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11800 = torch.aten.mm %11799, %11782 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11800, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9384 = torch.constant.int 4
%int512_9385 = torch.constant.int 512
%11801 = torch.prim.ListConstruct %int4_9384, %2482, %int512_9385 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11802 = torch.aten.view %11800, %11801 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9386 = torch.constant.int 4
%11803 = torch.aten.mul.int %int4_9386, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9387 = torch.constant.int 4096
%11804 = torch.prim.ListConstruct %11803, %int4096_9387 : (!torch.int, !torch.int) -> !torch.list<int>
%11805 = torch.aten.view %11774, %11804 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11805, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11806 = torch.aten.mm %11805, %11784 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11806, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9388 = torch.constant.int 4
%int512_9389 = torch.constant.int 512
%11807 = torch.prim.ListConstruct %int4_9388, %2482, %int512_9389 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11808 = torch.aten.view %11806, %11807 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9390 = torch.constant.int 4
%11809 = torch.aten.mul.int %int4_9390, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9391 = torch.constant.int 4096
%11810 = torch.prim.ListConstruct %11809, %int4096_9391 : (!torch.int, !torch.int) -> !torch.list<int>
%11811 = torch.aten.view %11775, %11810 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11811, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11812 = torch.aten.mm %11811, %11786 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11812, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9392 = torch.constant.int 4
%int512_9393 = torch.constant.int 512
%11813 = torch.prim.ListConstruct %int4_9392, %2482, %int512_9393 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11814 = torch.aten.view %11812, %11813 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9394 = torch.constant.int 4
%11815 = torch.aten.mul.int %int4_9394, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9395 = torch.constant.int 4096
%11816 = torch.prim.ListConstruct %11815, %int4096_9395 : (!torch.int, !torch.int) -> !torch.list<int>
%11817 = torch.aten.view %11776, %11816 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11817, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11818 = torch.aten.mm %11817, %11788 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11818, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9396 = torch.constant.int 4
%int512_9397 = torch.constant.int 512
%11819 = torch.prim.ListConstruct %int4_9396, %2482, %int512_9397 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11820 = torch.aten.view %11818, %11819 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9398 = torch.constant.int 4
%11821 = torch.aten.mul.int %int4_9398, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9399 = torch.constant.int 4096
%11822 = torch.prim.ListConstruct %11821, %int4096_9399 : (!torch.int, !torch.int) -> !torch.list<int>
%11823 = torch.aten.view %11777, %11822 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11823, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11824 = torch.aten.mm %11823, %11790 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11824, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9400 = torch.constant.int 4
%int512_9401 = torch.constant.int 512
%11825 = torch.prim.ListConstruct %int4_9400, %2482, %int512_9401 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11826 = torch.aten.view %11824, %11825 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9402 = torch.constant.int 4
%11827 = torch.aten.mul.int %int4_9402, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9403 = torch.constant.int 4096
%11828 = torch.prim.ListConstruct %11827, %int4096_9403 : (!torch.int, !torch.int) -> !torch.list<int>
%11829 = torch.aten.view %11778, %11828 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11829, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11830 = torch.aten.mm %11829, %11792 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11830, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9404 = torch.constant.int 4
%int512_9405 = torch.constant.int 512
%11831 = torch.prim.ListConstruct %int4_9404, %2482, %int512_9405 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11832 = torch.aten.view %11830, %11831 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9406 = torch.constant.int 4
%11833 = torch.aten.mul.int %int4_9406, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9407 = torch.constant.int 4096
%11834 = torch.prim.ListConstruct %11833, %int4096_9407 : (!torch.int, !torch.int) -> !torch.list<int>
%11835 = torch.aten.view %11779, %11834 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11835, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11836 = torch.aten.mm %11835, %11794 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11836, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9408 = torch.constant.int 4
%int512_9409 = torch.constant.int 512
%11837 = torch.prim.ListConstruct %int4_9408, %2482, %int512_9409 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11838 = torch.aten.view %11836, %11837 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_9410 = torch.constant.int 4
%11839 = torch.aten.mul.int %int4_9410, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9411 = torch.constant.int 4096
%11840 = torch.prim.ListConstruct %11839, %int4096_9411 : (!torch.int, !torch.int) -> !torch.list<int>
%11841 = torch.aten.view %11780, %11840 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11841, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11842 = torch.aten.mm %11841, %11796 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %11842, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_9412 = torch.constant.int 4
%int512_9413 = torch.constant.int 512
%11843 = torch.prim.ListConstruct %int4_9412, %2482, %int512_9413 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11844 = torch.aten.view %11842, %11843 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %11844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_9414 = torch.constant.int 1
%int0_9415 = torch.constant.int 0
%11845 = torch.prim.ListConstruct %int1_9414, %int0_9415 : (!torch.int, !torch.int) -> !torch.list<int>
%11846 = torch.aten.permute %384, %11845 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9416 = torch.constant.int 1
%int0_9417 = torch.constant.int 0
%11847 = torch.prim.ListConstruct %int1_9416, %int0_9417 : (!torch.int, !torch.int) -> !torch.list<int>
%11848 = torch.aten.permute %385, %11847 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9418 = torch.constant.int 1
%int0_9419 = torch.constant.int 0
%11849 = torch.prim.ListConstruct %int1_9418, %int0_9419 : (!torch.int, !torch.int) -> !torch.list<int>
%11850 = torch.aten.permute %386, %11849 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9420 = torch.constant.int 1
%int0_9421 = torch.constant.int 0
%11851 = torch.prim.ListConstruct %int1_9420, %int0_9421 : (!torch.int, !torch.int) -> !torch.list<int>
%11852 = torch.aten.permute %387, %11851 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9422 = torch.constant.int 1
%int0_9423 = torch.constant.int 0
%11853 = torch.prim.ListConstruct %int1_9422, %int0_9423 : (!torch.int, !torch.int) -> !torch.list<int>
%11854 = torch.aten.permute %388, %11853 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9424 = torch.constant.int 1
%int0_9425 = torch.constant.int 0
%11855 = torch.prim.ListConstruct %int1_9424, %int0_9425 : (!torch.int, !torch.int) -> !torch.list<int>
%11856 = torch.aten.permute %389, %11855 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9426 = torch.constant.int 1
%int0_9427 = torch.constant.int 0
%11857 = torch.prim.ListConstruct %int1_9426, %int0_9427 : (!torch.int, !torch.int) -> !torch.list<int>
%11858 = torch.aten.permute %390, %11857 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9428 = torch.constant.int 1
%int0_9429 = torch.constant.int 0
%11859 = torch.prim.ListConstruct %int1_9428, %int0_9429 : (!torch.int, !torch.int) -> !torch.list<int>
%11860 = torch.aten.permute %391, %11859 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_9430 = torch.constant.int 4
%11861 = torch.aten.mul.int %int4_9430, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9431 = torch.constant.int 4096
%11862 = torch.prim.ListConstruct %11861, %int4096_9431 : (!torch.int, !torch.int) -> !torch.list<int>
%11863 = torch.aten.view %11773, %11862 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11863, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11864 = torch.aten.mm %11863, %11846 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11864, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9432 = torch.constant.int 4
%int128_9433 = torch.constant.int 128
%11865 = torch.prim.ListConstruct %int4_9432, %2482, %int128_9433 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11866 = torch.aten.view %11864, %11865 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9434 = torch.constant.int 4
%11867 = torch.aten.mul.int %int4_9434, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9435 = torch.constant.int 4096
%11868 = torch.prim.ListConstruct %11867, %int4096_9435 : (!torch.int, !torch.int) -> !torch.list<int>
%11869 = torch.aten.view %11774, %11868 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11869, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11870 = torch.aten.mm %11869, %11848 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11870, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9436 = torch.constant.int 4
%int128_9437 = torch.constant.int 128
%11871 = torch.prim.ListConstruct %int4_9436, %2482, %int128_9437 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11872 = torch.aten.view %11870, %11871 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9438 = torch.constant.int 4
%11873 = torch.aten.mul.int %int4_9438, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9439 = torch.constant.int 4096
%11874 = torch.prim.ListConstruct %11873, %int4096_9439 : (!torch.int, !torch.int) -> !torch.list<int>
%11875 = torch.aten.view %11775, %11874 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11875, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11876 = torch.aten.mm %11875, %11850 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11876, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9440 = torch.constant.int 4
%int128_9441 = torch.constant.int 128
%11877 = torch.prim.ListConstruct %int4_9440, %2482, %int128_9441 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11878 = torch.aten.view %11876, %11877 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9442 = torch.constant.int 4
%11879 = torch.aten.mul.int %int4_9442, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9443 = torch.constant.int 4096
%11880 = torch.prim.ListConstruct %11879, %int4096_9443 : (!torch.int, !torch.int) -> !torch.list<int>
%11881 = torch.aten.view %11776, %11880 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11881, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11882 = torch.aten.mm %11881, %11852 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11882, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9444 = torch.constant.int 4
%int128_9445 = torch.constant.int 128
%11883 = torch.prim.ListConstruct %int4_9444, %2482, %int128_9445 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11884 = torch.aten.view %11882, %11883 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9446 = torch.constant.int 4
%11885 = torch.aten.mul.int %int4_9446, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9447 = torch.constant.int 4096
%11886 = torch.prim.ListConstruct %11885, %int4096_9447 : (!torch.int, !torch.int) -> !torch.list<int>
%11887 = torch.aten.view %11777, %11886 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11887, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11888 = torch.aten.mm %11887, %11854 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11888, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9448 = torch.constant.int 4
%int128_9449 = torch.constant.int 128
%11889 = torch.prim.ListConstruct %int4_9448, %2482, %int128_9449 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11890 = torch.aten.view %11888, %11889 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9450 = torch.constant.int 4
%11891 = torch.aten.mul.int %int4_9450, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9451 = torch.constant.int 4096
%11892 = torch.prim.ListConstruct %11891, %int4096_9451 : (!torch.int, !torch.int) -> !torch.list<int>
%11893 = torch.aten.view %11778, %11892 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11893, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11894 = torch.aten.mm %11893, %11856 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11894, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9452 = torch.constant.int 4
%int128_9453 = torch.constant.int 128
%11895 = torch.prim.ListConstruct %int4_9452, %2482, %int128_9453 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11896 = torch.aten.view %11894, %11895 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9454 = torch.constant.int 4
%11897 = torch.aten.mul.int %int4_9454, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9455 = torch.constant.int 4096
%11898 = torch.prim.ListConstruct %11897, %int4096_9455 : (!torch.int, !torch.int) -> !torch.list<int>
%11899 = torch.aten.view %11779, %11898 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11899, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11900 = torch.aten.mm %11899, %11858 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11900, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9456 = torch.constant.int 4
%int128_9457 = torch.constant.int 128
%11901 = torch.prim.ListConstruct %int4_9456, %2482, %int128_9457 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11902 = torch.aten.view %11900, %11901 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9458 = torch.constant.int 4
%11903 = torch.aten.mul.int %int4_9458, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9459 = torch.constant.int 4096
%11904 = torch.prim.ListConstruct %11903, %int4096_9459 : (!torch.int, !torch.int) -> !torch.list<int>
%11905 = torch.aten.view %11780, %11904 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11905, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11906 = torch.aten.mm %11905, %11860 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11906, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9460 = torch.constant.int 4
%int128_9461 = torch.constant.int 128
%11907 = torch.prim.ListConstruct %int4_9460, %2482, %int128_9461 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11908 = torch.aten.view %11906, %11907 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_9462 = torch.constant.int 1
%int0_9463 = torch.constant.int 0
%11909 = torch.prim.ListConstruct %int1_9462, %int0_9463 : (!torch.int, !torch.int) -> !torch.list<int>
%11910 = torch.aten.permute %392, %11909 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9464 = torch.constant.int 1
%int0_9465 = torch.constant.int 0
%11911 = torch.prim.ListConstruct %int1_9464, %int0_9465 : (!torch.int, !torch.int) -> !torch.list<int>
%11912 = torch.aten.permute %393, %11911 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9466 = torch.constant.int 1
%int0_9467 = torch.constant.int 0
%11913 = torch.prim.ListConstruct %int1_9466, %int0_9467 : (!torch.int, !torch.int) -> !torch.list<int>
%11914 = torch.aten.permute %394, %11913 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9468 = torch.constant.int 1
%int0_9469 = torch.constant.int 0
%11915 = torch.prim.ListConstruct %int1_9468, %int0_9469 : (!torch.int, !torch.int) -> !torch.list<int>
%11916 = torch.aten.permute %395, %11915 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9470 = torch.constant.int 1
%int0_9471 = torch.constant.int 0
%11917 = torch.prim.ListConstruct %int1_9470, %int0_9471 : (!torch.int, !torch.int) -> !torch.list<int>
%11918 = torch.aten.permute %396, %11917 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9472 = torch.constant.int 1
%int0_9473 = torch.constant.int 0
%11919 = torch.prim.ListConstruct %int1_9472, %int0_9473 : (!torch.int, !torch.int) -> !torch.list<int>
%11920 = torch.aten.permute %397, %11919 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9474 = torch.constant.int 1
%int0_9475 = torch.constant.int 0
%11921 = torch.prim.ListConstruct %int1_9474, %int0_9475 : (!torch.int, !torch.int) -> !torch.list<int>
%11922 = torch.aten.permute %398, %11921 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_9476 = torch.constant.int 1
%int0_9477 = torch.constant.int 0
%11923 = torch.prim.ListConstruct %int1_9476, %int0_9477 : (!torch.int, !torch.int) -> !torch.list<int>
%11924 = torch.aten.permute %399, %11923 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_9478 = torch.constant.int 4
%11925 = torch.aten.mul.int %int4_9478, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9479 = torch.constant.int 4096
%11926 = torch.prim.ListConstruct %11925, %int4096_9479 : (!torch.int, !torch.int) -> !torch.list<int>
%11927 = torch.aten.view %11773, %11926 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11927, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11928 = torch.aten.mm %11927, %11910 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11928, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9480 = torch.constant.int 4
%int128_9481 = torch.constant.int 128
%11929 = torch.prim.ListConstruct %int4_9480, %2482, %int128_9481 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11930 = torch.aten.view %11928, %11929 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9482 = torch.constant.int 4
%11931 = torch.aten.mul.int %int4_9482, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9483 = torch.constant.int 4096
%11932 = torch.prim.ListConstruct %11931, %int4096_9483 : (!torch.int, !torch.int) -> !torch.list<int>
%11933 = torch.aten.view %11774, %11932 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11933, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11934 = torch.aten.mm %11933, %11912 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11934, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9484 = torch.constant.int 4
%int128_9485 = torch.constant.int 128
%11935 = torch.prim.ListConstruct %int4_9484, %2482, %int128_9485 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11936 = torch.aten.view %11934, %11935 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9486 = torch.constant.int 4
%11937 = torch.aten.mul.int %int4_9486, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9487 = torch.constant.int 4096
%11938 = torch.prim.ListConstruct %11937, %int4096_9487 : (!torch.int, !torch.int) -> !torch.list<int>
%11939 = torch.aten.view %11775, %11938 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11939, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11940 = torch.aten.mm %11939, %11914 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11940, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9488 = torch.constant.int 4
%int128_9489 = torch.constant.int 128
%11941 = torch.prim.ListConstruct %int4_9488, %2482, %int128_9489 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11942 = torch.aten.view %11940, %11941 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9490 = torch.constant.int 4
%11943 = torch.aten.mul.int %int4_9490, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9491 = torch.constant.int 4096
%11944 = torch.prim.ListConstruct %11943, %int4096_9491 : (!torch.int, !torch.int) -> !torch.list<int>
%11945 = torch.aten.view %11776, %11944 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11945, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11946 = torch.aten.mm %11945, %11916 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11946, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9492 = torch.constant.int 4
%int128_9493 = torch.constant.int 128
%11947 = torch.prim.ListConstruct %int4_9492, %2482, %int128_9493 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11948 = torch.aten.view %11946, %11947 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9494 = torch.constant.int 4
%11949 = torch.aten.mul.int %int4_9494, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9495 = torch.constant.int 4096
%11950 = torch.prim.ListConstruct %11949, %int4096_9495 : (!torch.int, !torch.int) -> !torch.list<int>
%11951 = torch.aten.view %11777, %11950 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11951, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11952 = torch.aten.mm %11951, %11918 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11952, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9496 = torch.constant.int 4
%int128_9497 = torch.constant.int 128
%11953 = torch.prim.ListConstruct %int4_9496, %2482, %int128_9497 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11954 = torch.aten.view %11952, %11953 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9498 = torch.constant.int 4
%11955 = torch.aten.mul.int %int4_9498, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9499 = torch.constant.int 4096
%11956 = torch.prim.ListConstruct %11955, %int4096_9499 : (!torch.int, !torch.int) -> !torch.list<int>
%11957 = torch.aten.view %11778, %11956 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11957, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11958 = torch.aten.mm %11957, %11920 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11958, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9500 = torch.constant.int 4
%int128_9501 = torch.constant.int 128
%11959 = torch.prim.ListConstruct %int4_9500, %2482, %int128_9501 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11960 = torch.aten.view %11958, %11959 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9502 = torch.constant.int 4
%11961 = torch.aten.mul.int %int4_9502, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9503 = torch.constant.int 4096
%11962 = torch.prim.ListConstruct %11961, %int4096_9503 : (!torch.int, !torch.int) -> !torch.list<int>
%11963 = torch.aten.view %11779, %11962 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11963, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11964 = torch.aten.mm %11963, %11922 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11964, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9504 = torch.constant.int 4
%int128_9505 = torch.constant.int 128
%11965 = torch.prim.ListConstruct %int4_9504, %2482, %int128_9505 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11966 = torch.aten.view %11964, %11965 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9506 = torch.constant.int 4
%11967 = torch.aten.mul.int %int4_9506, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_9507 = torch.constant.int 4096
%11968 = torch.prim.ListConstruct %11967, %int4096_9507 : (!torch.int, !torch.int) -> !torch.list<int>
%11969 = torch.aten.view %11780, %11968 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %11969, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%11970 = torch.aten.mm %11969, %11924 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %11970, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_9508 = torch.constant.int 4
%int128_9509 = torch.constant.int 128
%11971 = torch.prim.ListConstruct %int4_9508, %2482, %int128_9509 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11972 = torch.aten.view %11970, %11971 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %11972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_9510 = torch.constant.int 4
%int4_9511 = torch.constant.int 4
%int128_9512 = torch.constant.int 128
%11973 = torch.prim.ListConstruct %int4_9510, %2482, %int4_9511, %int128_9512 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11974 = torch.aten.view %11802, %11973 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9513 = torch.constant.int 4
%int4_9514 = torch.constant.int 4
%int128_9515 = torch.constant.int 128
%11975 = torch.prim.ListConstruct %int4_9513, %2482, %int4_9514, %int128_9515 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11976 = torch.aten.view %11808, %11975 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9516 = torch.constant.int 4
%int4_9517 = torch.constant.int 4
%int128_9518 = torch.constant.int 128
%11977 = torch.prim.ListConstruct %int4_9516, %2482, %int4_9517, %int128_9518 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11978 = torch.aten.view %11814, %11977 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9519 = torch.constant.int 4
%int4_9520 = torch.constant.int 4
%int128_9521 = torch.constant.int 128
%11979 = torch.prim.ListConstruct %int4_9519, %2482, %int4_9520, %int128_9521 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11980 = torch.aten.view %11820, %11979 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9522 = torch.constant.int 4
%int4_9523 = torch.constant.int 4
%int128_9524 = torch.constant.int 128
%11981 = torch.prim.ListConstruct %int4_9522, %2482, %int4_9523, %int128_9524 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11982 = torch.aten.view %11826, %11981 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9525 = torch.constant.int 4
%int4_9526 = torch.constant.int 4
%int128_9527 = torch.constant.int 128
%11983 = torch.prim.ListConstruct %int4_9525, %2482, %int4_9526, %int128_9527 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11984 = torch.aten.view %11832, %11983 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9528 = torch.constant.int 4
%int4_9529 = torch.constant.int 4
%int128_9530 = torch.constant.int 128
%11985 = torch.prim.ListConstruct %int4_9528, %2482, %int4_9529, %int128_9530 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11986 = torch.aten.view %11838, %11985 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9531 = torch.constant.int 4
%int4_9532 = torch.constant.int 4
%int128_9533 = torch.constant.int 128
%11987 = torch.prim.ListConstruct %int4_9531, %2482, %int4_9532, %int128_9533 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11988 = torch.aten.view %11844, %11987 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %11988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_9534 = torch.constant.int 4
%int1_9535 = torch.constant.int 1
%int128_9536 = torch.constant.int 128
%11989 = torch.prim.ListConstruct %int4_9534, %2482, %int1_9535, %int128_9536 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11990 = torch.aten.view %11866, %11989 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %11990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9537 = torch.constant.int 4
%int1_9538 = torch.constant.int 1
%int128_9539 = torch.constant.int 128
%11991 = torch.prim.ListConstruct %int4_9537, %2482, %int1_9538, %int128_9539 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11992 = torch.aten.view %11872, %11991 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %11992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9540 = torch.constant.int 4
%int1_9541 = torch.constant.int 1
%int128_9542 = torch.constant.int 128
%11993 = torch.prim.ListConstruct %int4_9540, %2482, %int1_9541, %int128_9542 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11994 = torch.aten.view %11878, %11993 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %11994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9543 = torch.constant.int 4
%int1_9544 = torch.constant.int 1
%int128_9545 = torch.constant.int 128
%11995 = torch.prim.ListConstruct %int4_9543, %2482, %int1_9544, %int128_9545 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11996 = torch.aten.view %11884, %11995 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %11996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9546 = torch.constant.int 4
%int1_9547 = torch.constant.int 1
%int128_9548 = torch.constant.int 128
%11997 = torch.prim.ListConstruct %int4_9546, %2482, %int1_9547, %int128_9548 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%11998 = torch.aten.view %11890, %11997 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %11998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9549 = torch.constant.int 4
%int1_9550 = torch.constant.int 1
%int128_9551 = torch.constant.int 128
%11999 = torch.prim.ListConstruct %int4_9549, %2482, %int1_9550, %int128_9551 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12000 = torch.aten.view %11896, %11999 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9552 = torch.constant.int 4
%int1_9553 = torch.constant.int 1
%int128_9554 = torch.constant.int 128
%12001 = torch.prim.ListConstruct %int4_9552, %2482, %int1_9553, %int128_9554 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12002 = torch.aten.view %11902, %12001 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9555 = torch.constant.int 4
%int1_9556 = torch.constant.int 1
%int128_9557 = torch.constant.int 128
%12003 = torch.prim.ListConstruct %int4_9555, %2482, %int1_9556, %int128_9557 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12004 = torch.aten.view %11908, %12003 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9558 = torch.constant.int 4
%int1_9559 = torch.constant.int 1
%int128_9560 = torch.constant.int 128
%12005 = torch.prim.ListConstruct %int4_9558, %2482, %int1_9559, %int128_9560 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12006 = torch.aten.view %11930, %12005 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9561 = torch.constant.int 4
%int1_9562 = torch.constant.int 1
%int128_9563 = torch.constant.int 128
%12007 = torch.prim.ListConstruct %int4_9561, %2482, %int1_9562, %int128_9563 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12008 = torch.aten.view %11936, %12007 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9564 = torch.constant.int 4
%int1_9565 = torch.constant.int 1
%int128_9566 = torch.constant.int 128
%12009 = torch.prim.ListConstruct %int4_9564, %2482, %int1_9565, %int128_9566 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12010 = torch.aten.view %11942, %12009 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9567 = torch.constant.int 4
%int1_9568 = torch.constant.int 1
%int128_9569 = torch.constant.int 128
%12011 = torch.prim.ListConstruct %int4_9567, %2482, %int1_9568, %int128_9569 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12012 = torch.aten.view %11948, %12011 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9570 = torch.constant.int 4
%int1_9571 = torch.constant.int 1
%int128_9572 = torch.constant.int 128
%12013 = torch.prim.ListConstruct %int4_9570, %2482, %int1_9571, %int128_9572 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12014 = torch.aten.view %11954, %12013 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9573 = torch.constant.int 4
%int1_9574 = torch.constant.int 1
%int128_9575 = torch.constant.int 128
%12015 = torch.prim.ListConstruct %int4_9573, %2482, %int1_9574, %int128_9575 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12016 = torch.aten.view %11960, %12015 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9576 = torch.constant.int 4
%int1_9577 = torch.constant.int 1
%int128_9578 = torch.constant.int 128
%12017 = torch.prim.ListConstruct %int4_9576, %2482, %int1_9577, %int128_9578 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12018 = torch.aten.view %11966, %12017 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_9579 = torch.constant.int 4
%int1_9580 = torch.constant.int 1
%int128_9581 = torch.constant.int 128
%12019 = torch.prim.ListConstruct %int4_9579, %2482, %int1_9580, %int128_9581 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12020 = torch.aten.view %11972, %12019 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_9582 = torch.constant.int 131072
%none_9583 = torch.constant.none
%none_9584 = torch.constant.none
%cpu_9585 = torch.constant.device "cpu"
%false_9586 = torch.constant.bool false
%12021 = torch.aten.arange %int131072_9582, %none_9583, %none_9584, %cpu_9585, %false_9586 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_9587 = torch.constant.int 0
%int128_9588 = torch.constant.int 128
%int2_9589 = torch.constant.int 2
%none_9590 = torch.constant.none
%none_9591 = torch.constant.none
%cpu_9592 = torch.constant.device "cpu"
%false_9593 = torch.constant.bool false
%12022 = torch.aten.arange.start_step %int0_9587, %int128_9588, %int2_9589, %none_9590, %none_9591, %cpu_9592, %false_9593 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_9594 = torch.constant.int 0
%int0_9595 = torch.constant.int 0
%int64_9596 = torch.constant.int 64
%int1_9597 = torch.constant.int 1
%12023 = torch.aten.slice.Tensor %12022, %int0_9594, %int0_9595, %int64_9596, %int1_9597 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_9598 = torch.constant.int 6
%12024 = torch.prims.convert_element_type %12023, %int6_9598 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_9599 = torch.constant.int 128
%12025 = torch.aten.div.Scalar %12024, %int128_9599 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_9600 = torch.constant.float 5.000000e+05
%12026 = torch.aten.pow.Scalar %float5.000000e05_9600, %12025 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%12027 = torch.aten.reciprocal %12026 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_9601 = torch.constant.float 1.000000e+00
%12028 = torch.aten.mul.Scalar %12027, %float1.000000e00_9601 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_9602 = torch.constant.int 131072
%int1_9603 = torch.constant.int 1
%12029 = torch.prim.ListConstruct %int131072_9602, %int1_9603 : (!torch.int, !torch.int) -> !torch.list<int>
%12030 = torch.aten.view %12021, %12029 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%12031 = torch.aten.mul.Tensor %12030, %12028 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%12032 = torch.aten.cos %12031 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%12033 = torch.aten.sin %12031 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%12034 = torch.aten.complex %12032, %12033 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%12035 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12036 = flow.tensor.transfer %12035 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%12037 = torch_c.from_builtin_tensor %12036 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12038 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12039 = flow.tensor.transfer %12038 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%12040 = torch_c.from_builtin_tensor %12039 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12041 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12042 = flow.tensor.transfer %12041 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%12043 = torch_c.from_builtin_tensor %12042 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12044 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12045 = flow.tensor.transfer %12044 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%12046 = torch_c.from_builtin_tensor %12045 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12047 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12048 = flow.tensor.transfer %12047 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%12049 = torch_c.from_builtin_tensor %12048 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12050 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12051 = flow.tensor.transfer %12050 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%12052 = torch_c.from_builtin_tensor %12051 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12053 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12054 = flow.tensor.transfer %12053 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%12055 = torch_c.from_builtin_tensor %12054 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12056 = torch_c.to_builtin_tensor %12034 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12057 = flow.tensor.transfer %12056 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%12058 = torch_c.from_builtin_tensor %12057 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_9604 = torch.constant.int 1
%12059 = torch.aten.size.int %11802, %int1_9604 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9605 = torch.constant.int 0
%12060 = torch.aten.add.int %int0_9605, %12059 : !torch.int, !torch.int -> !torch.int
%int0_9606 = torch.constant.int 0
%int0_9607 = torch.constant.int 0
%int1_9608 = torch.constant.int 1
%12061 = torch.aten.slice.Tensor %12037, %int0_9606, %int0_9607, %12060, %int1_9608 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12061, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9609 = torch.constant.int 1
%int0_9610 = torch.constant.int 0
%int9223372036854775807_9611 = torch.constant.int 9223372036854775807
%int1_9612 = torch.constant.int 1
%12062 = torch.aten.slice.Tensor %12061, %int1_9609, %int0_9610, %int9223372036854775807_9611, %int1_9612 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12062, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9613 = torch.constant.int 0
%12063 = torch.aten.unsqueeze %12062, %int0_9613 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12063, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9614 = torch.constant.int 2
%12064 = torch.aten.unsqueeze %12063, %int2_9614 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12064, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9615 = torch.constant.int 3
%int0_9616 = torch.constant.int 0
%int9223372036854775807_9617 = torch.constant.int 9223372036854775807
%int1_9618 = torch.constant.int 1
%12065 = torch.aten.slice.Tensor %12064, %int3_9615, %int0_9616, %int9223372036854775807_9617, %int1_9618 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12065, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12066 = torch_c.to_builtin_tensor %11974 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9619 = arith.constant 1 : index
%dim_9620 = tensor.dim %12066, %c1_9619 : tensor<4x?x4x128xf16>
%12067 = flow.tensor.bitcast %12066 : tensor<4x?x4x128xf16>{%dim_9620} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9620}
%12068 = torch_c.from_builtin_tensor %12067 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12069 = torch.aten.mul.Tensor %12068, %12065 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12070 = torch_c.to_builtin_tensor %12069 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9621 = arith.constant 1 : index
%dim_9622 = tensor.dim %12070, %c1_9621 : tensor<4x?x4x64xcomplex<f32>>
%12071 = flow.tensor.bitcast %12070 : tensor<4x?x4x64xcomplex<f32>>{%dim_9622} -> tensor<4x?x4x128xf32>{%dim_9622}
%12072 = torch_c.from_builtin_tensor %12071 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9623 = torch.constant.int 5
%12073 = torch.prims.convert_element_type %12072, %int5_9623 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9624 = torch.constant.int 1
%12074 = torch.aten.size.int %11808, %int1_9624 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9625 = torch.constant.int 0
%12075 = torch.aten.add.int %int0_9625, %12074 : !torch.int, !torch.int -> !torch.int
%int0_9626 = torch.constant.int 0
%int0_9627 = torch.constant.int 0
%int1_9628 = torch.constant.int 1
%12076 = torch.aten.slice.Tensor %12040, %int0_9626, %int0_9627, %12075, %int1_9628 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12076, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9629 = torch.constant.int 1
%int0_9630 = torch.constant.int 0
%int9223372036854775807_9631 = torch.constant.int 9223372036854775807
%int1_9632 = torch.constant.int 1
%12077 = torch.aten.slice.Tensor %12076, %int1_9629, %int0_9630, %int9223372036854775807_9631, %int1_9632 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12077, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9633 = torch.constant.int 0
%12078 = torch.aten.unsqueeze %12077, %int0_9633 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12078, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9634 = torch.constant.int 2
%12079 = torch.aten.unsqueeze %12078, %int2_9634 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12079, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9635 = torch.constant.int 3
%int0_9636 = torch.constant.int 0
%int9223372036854775807_9637 = torch.constant.int 9223372036854775807
%int1_9638 = torch.constant.int 1
%12080 = torch.aten.slice.Tensor %12079, %int3_9635, %int0_9636, %int9223372036854775807_9637, %int1_9638 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12080, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12081 = torch_c.to_builtin_tensor %11976 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9639 = arith.constant 1 : index
%dim_9640 = tensor.dim %12081, %c1_9639 : tensor<4x?x4x128xf16>
%12082 = flow.tensor.bitcast %12081 : tensor<4x?x4x128xf16>{%dim_9640} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9640}
%12083 = torch_c.from_builtin_tensor %12082 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12084 = torch.aten.mul.Tensor %12083, %12080 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12085 = torch_c.to_builtin_tensor %12084 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9641 = arith.constant 1 : index
%dim_9642 = tensor.dim %12085, %c1_9641 : tensor<4x?x4x64xcomplex<f32>>
%12086 = flow.tensor.bitcast %12085 : tensor<4x?x4x64xcomplex<f32>>{%dim_9642} -> tensor<4x?x4x128xf32>{%dim_9642}
%12087 = torch_c.from_builtin_tensor %12086 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9643 = torch.constant.int 5
%12088 = torch.prims.convert_element_type %12087, %int5_9643 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9644 = torch.constant.int 1
%12089 = torch.aten.size.int %11814, %int1_9644 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9645 = torch.constant.int 0
%12090 = torch.aten.add.int %int0_9645, %12089 : !torch.int, !torch.int -> !torch.int
%int0_9646 = torch.constant.int 0
%int0_9647 = torch.constant.int 0
%int1_9648 = torch.constant.int 1
%12091 = torch.aten.slice.Tensor %12043, %int0_9646, %int0_9647, %12090, %int1_9648 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12091, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9649 = torch.constant.int 1
%int0_9650 = torch.constant.int 0
%int9223372036854775807_9651 = torch.constant.int 9223372036854775807
%int1_9652 = torch.constant.int 1
%12092 = torch.aten.slice.Tensor %12091, %int1_9649, %int0_9650, %int9223372036854775807_9651, %int1_9652 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12092, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9653 = torch.constant.int 0
%12093 = torch.aten.unsqueeze %12092, %int0_9653 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12093, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9654 = torch.constant.int 2
%12094 = torch.aten.unsqueeze %12093, %int2_9654 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12094, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9655 = torch.constant.int 3
%int0_9656 = torch.constant.int 0
%int9223372036854775807_9657 = torch.constant.int 9223372036854775807
%int1_9658 = torch.constant.int 1
%12095 = torch.aten.slice.Tensor %12094, %int3_9655, %int0_9656, %int9223372036854775807_9657, %int1_9658 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12095, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12096 = torch_c.to_builtin_tensor %11978 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9659 = arith.constant 1 : index
%dim_9660 = tensor.dim %12096, %c1_9659 : tensor<4x?x4x128xf16>
%12097 = flow.tensor.bitcast %12096 : tensor<4x?x4x128xf16>{%dim_9660} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9660}
%12098 = torch_c.from_builtin_tensor %12097 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12099 = torch.aten.mul.Tensor %12098, %12095 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12100 = torch_c.to_builtin_tensor %12099 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9661 = arith.constant 1 : index
%dim_9662 = tensor.dim %12100, %c1_9661 : tensor<4x?x4x64xcomplex<f32>>
%12101 = flow.tensor.bitcast %12100 : tensor<4x?x4x64xcomplex<f32>>{%dim_9662} -> tensor<4x?x4x128xf32>{%dim_9662}
%12102 = torch_c.from_builtin_tensor %12101 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9663 = torch.constant.int 5
%12103 = torch.prims.convert_element_type %12102, %int5_9663 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9664 = torch.constant.int 1
%12104 = torch.aten.size.int %11820, %int1_9664 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9665 = torch.constant.int 0
%12105 = torch.aten.add.int %int0_9665, %12104 : !torch.int, !torch.int -> !torch.int
%int0_9666 = torch.constant.int 0
%int0_9667 = torch.constant.int 0
%int1_9668 = torch.constant.int 1
%12106 = torch.aten.slice.Tensor %12046, %int0_9666, %int0_9667, %12105, %int1_9668 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12106, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9669 = torch.constant.int 1
%int0_9670 = torch.constant.int 0
%int9223372036854775807_9671 = torch.constant.int 9223372036854775807
%int1_9672 = torch.constant.int 1
%12107 = torch.aten.slice.Tensor %12106, %int1_9669, %int0_9670, %int9223372036854775807_9671, %int1_9672 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12107, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9673 = torch.constant.int 0
%12108 = torch.aten.unsqueeze %12107, %int0_9673 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12108, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9674 = torch.constant.int 2
%12109 = torch.aten.unsqueeze %12108, %int2_9674 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12109, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9675 = torch.constant.int 3
%int0_9676 = torch.constant.int 0
%int9223372036854775807_9677 = torch.constant.int 9223372036854775807
%int1_9678 = torch.constant.int 1
%12110 = torch.aten.slice.Tensor %12109, %int3_9675, %int0_9676, %int9223372036854775807_9677, %int1_9678 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12110, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12111 = torch_c.to_builtin_tensor %11980 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9679 = arith.constant 1 : index
%dim_9680 = tensor.dim %12111, %c1_9679 : tensor<4x?x4x128xf16>
%12112 = flow.tensor.bitcast %12111 : tensor<4x?x4x128xf16>{%dim_9680} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9680}
%12113 = torch_c.from_builtin_tensor %12112 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12114 = torch.aten.mul.Tensor %12113, %12110 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12115 = torch_c.to_builtin_tensor %12114 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9681 = arith.constant 1 : index
%dim_9682 = tensor.dim %12115, %c1_9681 : tensor<4x?x4x64xcomplex<f32>>
%12116 = flow.tensor.bitcast %12115 : tensor<4x?x4x64xcomplex<f32>>{%dim_9682} -> tensor<4x?x4x128xf32>{%dim_9682}
%12117 = torch_c.from_builtin_tensor %12116 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9683 = torch.constant.int 5
%12118 = torch.prims.convert_element_type %12117, %int5_9683 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9684 = torch.constant.int 1
%12119 = torch.aten.size.int %11826, %int1_9684 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9685 = torch.constant.int 0
%12120 = torch.aten.add.int %int0_9685, %12119 : !torch.int, !torch.int -> !torch.int
%int0_9686 = torch.constant.int 0
%int0_9687 = torch.constant.int 0
%int1_9688 = torch.constant.int 1
%12121 = torch.aten.slice.Tensor %12049, %int0_9686, %int0_9687, %12120, %int1_9688 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12121, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9689 = torch.constant.int 1
%int0_9690 = torch.constant.int 0
%int9223372036854775807_9691 = torch.constant.int 9223372036854775807
%int1_9692 = torch.constant.int 1
%12122 = torch.aten.slice.Tensor %12121, %int1_9689, %int0_9690, %int9223372036854775807_9691, %int1_9692 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12122, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9693 = torch.constant.int 0
%12123 = torch.aten.unsqueeze %12122, %int0_9693 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12123, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9694 = torch.constant.int 2
%12124 = torch.aten.unsqueeze %12123, %int2_9694 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12124, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9695 = torch.constant.int 3
%int0_9696 = torch.constant.int 0
%int9223372036854775807_9697 = torch.constant.int 9223372036854775807
%int1_9698 = torch.constant.int 1
%12125 = torch.aten.slice.Tensor %12124, %int3_9695, %int0_9696, %int9223372036854775807_9697, %int1_9698 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12125, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12126 = torch_c.to_builtin_tensor %11982 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9699 = arith.constant 1 : index
%dim_9700 = tensor.dim %12126, %c1_9699 : tensor<4x?x4x128xf16>
%12127 = flow.tensor.bitcast %12126 : tensor<4x?x4x128xf16>{%dim_9700} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9700}
%12128 = torch_c.from_builtin_tensor %12127 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12129 = torch.aten.mul.Tensor %12128, %12125 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12130 = torch_c.to_builtin_tensor %12129 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9701 = arith.constant 1 : index
%dim_9702 = tensor.dim %12130, %c1_9701 : tensor<4x?x4x64xcomplex<f32>>
%12131 = flow.tensor.bitcast %12130 : tensor<4x?x4x64xcomplex<f32>>{%dim_9702} -> tensor<4x?x4x128xf32>{%dim_9702}
%12132 = torch_c.from_builtin_tensor %12131 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9703 = torch.constant.int 5
%12133 = torch.prims.convert_element_type %12132, %int5_9703 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9704 = torch.constant.int 1
%12134 = torch.aten.size.int %11832, %int1_9704 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9705 = torch.constant.int 0
%12135 = torch.aten.add.int %int0_9705, %12134 : !torch.int, !torch.int -> !torch.int
%int0_9706 = torch.constant.int 0
%int0_9707 = torch.constant.int 0
%int1_9708 = torch.constant.int 1
%12136 = torch.aten.slice.Tensor %12052, %int0_9706, %int0_9707, %12135, %int1_9708 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12136, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9709 = torch.constant.int 1
%int0_9710 = torch.constant.int 0
%int9223372036854775807_9711 = torch.constant.int 9223372036854775807
%int1_9712 = torch.constant.int 1
%12137 = torch.aten.slice.Tensor %12136, %int1_9709, %int0_9710, %int9223372036854775807_9711, %int1_9712 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12137, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9713 = torch.constant.int 0
%12138 = torch.aten.unsqueeze %12137, %int0_9713 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12138, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9714 = torch.constant.int 2
%12139 = torch.aten.unsqueeze %12138, %int2_9714 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12139, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9715 = torch.constant.int 3
%int0_9716 = torch.constant.int 0
%int9223372036854775807_9717 = torch.constant.int 9223372036854775807
%int1_9718 = torch.constant.int 1
%12140 = torch.aten.slice.Tensor %12139, %int3_9715, %int0_9716, %int9223372036854775807_9717, %int1_9718 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12140, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12141 = torch_c.to_builtin_tensor %11984 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9719 = arith.constant 1 : index
%dim_9720 = tensor.dim %12141, %c1_9719 : tensor<4x?x4x128xf16>
%12142 = flow.tensor.bitcast %12141 : tensor<4x?x4x128xf16>{%dim_9720} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9720}
%12143 = torch_c.from_builtin_tensor %12142 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12144 = torch.aten.mul.Tensor %12143, %12140 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12145 = torch_c.to_builtin_tensor %12144 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9721 = arith.constant 1 : index
%dim_9722 = tensor.dim %12145, %c1_9721 : tensor<4x?x4x64xcomplex<f32>>
%12146 = flow.tensor.bitcast %12145 : tensor<4x?x4x64xcomplex<f32>>{%dim_9722} -> tensor<4x?x4x128xf32>{%dim_9722}
%12147 = torch_c.from_builtin_tensor %12146 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9723 = torch.constant.int 5
%12148 = torch.prims.convert_element_type %12147, %int5_9723 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9724 = torch.constant.int 1
%12149 = torch.aten.size.int %11838, %int1_9724 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9725 = torch.constant.int 0
%12150 = torch.aten.add.int %int0_9725, %12149 : !torch.int, !torch.int -> !torch.int
%int0_9726 = torch.constant.int 0
%int0_9727 = torch.constant.int 0
%int1_9728 = torch.constant.int 1
%12151 = torch.aten.slice.Tensor %12055, %int0_9726, %int0_9727, %12150, %int1_9728 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12151, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9729 = torch.constant.int 1
%int0_9730 = torch.constant.int 0
%int9223372036854775807_9731 = torch.constant.int 9223372036854775807
%int1_9732 = torch.constant.int 1
%12152 = torch.aten.slice.Tensor %12151, %int1_9729, %int0_9730, %int9223372036854775807_9731, %int1_9732 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12152, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9733 = torch.constant.int 0
%12153 = torch.aten.unsqueeze %12152, %int0_9733 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12153, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9734 = torch.constant.int 2
%12154 = torch.aten.unsqueeze %12153, %int2_9734 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12154, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9735 = torch.constant.int 3
%int0_9736 = torch.constant.int 0
%int9223372036854775807_9737 = torch.constant.int 9223372036854775807
%int1_9738 = torch.constant.int 1
%12155 = torch.aten.slice.Tensor %12154, %int3_9735, %int0_9736, %int9223372036854775807_9737, %int1_9738 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12155, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12156 = torch_c.to_builtin_tensor %11986 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9739 = arith.constant 1 : index
%dim_9740 = tensor.dim %12156, %c1_9739 : tensor<4x?x4x128xf16>
%12157 = flow.tensor.bitcast %12156 : tensor<4x?x4x128xf16>{%dim_9740} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9740}
%12158 = torch_c.from_builtin_tensor %12157 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12159 = torch.aten.mul.Tensor %12158, %12155 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12160 = torch_c.to_builtin_tensor %12159 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9741 = arith.constant 1 : index
%dim_9742 = tensor.dim %12160, %c1_9741 : tensor<4x?x4x64xcomplex<f32>>
%12161 = flow.tensor.bitcast %12160 : tensor<4x?x4x64xcomplex<f32>>{%dim_9742} -> tensor<4x?x4x128xf32>{%dim_9742}
%12162 = torch_c.from_builtin_tensor %12161 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9743 = torch.constant.int 5
%12163 = torch.prims.convert_element_type %12162, %int5_9743 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_9744 = torch.constant.int 1
%12164 = torch.aten.size.int %11844, %int1_9744 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_9745 = torch.constant.int 0
%12165 = torch.aten.add.int %int0_9745, %12164 : !torch.int, !torch.int -> !torch.int
%int0_9746 = torch.constant.int 0
%int0_9747 = torch.constant.int 0
%int1_9748 = torch.constant.int 1
%12166 = torch.aten.slice.Tensor %12058, %int0_9746, %int0_9747, %12165, %int1_9748 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12166, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9749 = torch.constant.int 1
%int0_9750 = torch.constant.int 0
%int9223372036854775807_9751 = torch.constant.int 9223372036854775807
%int1_9752 = torch.constant.int 1
%12167 = torch.aten.slice.Tensor %12166, %int1_9749, %int0_9750, %int9223372036854775807_9751, %int1_9752 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12167, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9753 = torch.constant.int 0
%12168 = torch.aten.unsqueeze %12167, %int0_9753 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12168, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9754 = torch.constant.int 2
%12169 = torch.aten.unsqueeze %12168, %int2_9754 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12169, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9755 = torch.constant.int 3
%int0_9756 = torch.constant.int 0
%int9223372036854775807_9757 = torch.constant.int 9223372036854775807
%int1_9758 = torch.constant.int 1
%12170 = torch.aten.slice.Tensor %12169, %int3_9755, %int0_9756, %int9223372036854775807_9757, %int1_9758 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12170, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12171 = torch_c.to_builtin_tensor %11988 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_9759 = arith.constant 1 : index
%dim_9760 = tensor.dim %12171, %c1_9759 : tensor<4x?x4x128xf16>
%12172 = flow.tensor.bitcast %12171 : tensor<4x?x4x128xf16>{%dim_9760} -> tensor<4x?x4x64xcomplex<f16>>{%dim_9760}
%12173 = torch_c.from_builtin_tensor %12172 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %12173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%12174 = torch.aten.mul.Tensor %12173, %12170 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %12174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%12175 = torch_c.to_builtin_tensor %12174 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_9761 = arith.constant 1 : index
%dim_9762 = tensor.dim %12175, %c1_9761 : tensor<4x?x4x64xcomplex<f32>>
%12176 = flow.tensor.bitcast %12175 : tensor<4x?x4x64xcomplex<f32>>{%dim_9762} -> tensor<4x?x4x128xf32>{%dim_9762}
%12177 = torch_c.from_builtin_tensor %12176 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %12177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_9763 = torch.constant.int 5
%12178 = torch.prims.convert_element_type %12177, %int5_9763 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_9764 = torch.constant.int 131072
%none_9765 = torch.constant.none
%none_9766 = torch.constant.none
%cpu_9767 = torch.constant.device "cpu"
%false_9768 = torch.constant.bool false
%12179 = torch.aten.arange %int131072_9764, %none_9765, %none_9766, %cpu_9767, %false_9768 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_9769 = torch.constant.int 0
%int128_9770 = torch.constant.int 128
%int2_9771 = torch.constant.int 2
%none_9772 = torch.constant.none
%none_9773 = torch.constant.none
%cpu_9774 = torch.constant.device "cpu"
%false_9775 = torch.constant.bool false
%12180 = torch.aten.arange.start_step %int0_9769, %int128_9770, %int2_9771, %none_9772, %none_9773, %cpu_9774, %false_9775 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_9776 = torch.constant.int 0
%int0_9777 = torch.constant.int 0
%int64_9778 = torch.constant.int 64
%int1_9779 = torch.constant.int 1
%12181 = torch.aten.slice.Tensor %12180, %int0_9776, %int0_9777, %int64_9778, %int1_9779 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_9780 = torch.constant.int 6
%12182 = torch.prims.convert_element_type %12181, %int6_9780 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_9781 = torch.constant.int 128
%12183 = torch.aten.div.Scalar %12182, %int128_9781 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_9782 = torch.constant.float 5.000000e+05
%12184 = torch.aten.pow.Scalar %float5.000000e05_9782, %12183 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%12185 = torch.aten.reciprocal %12184 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_9783 = torch.constant.float 1.000000e+00
%12186 = torch.aten.mul.Scalar %12185, %float1.000000e00_9783 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_9784 = torch.constant.int 131072
%int1_9785 = torch.constant.int 1
%12187 = torch.prim.ListConstruct %int131072_9784, %int1_9785 : (!torch.int, !torch.int) -> !torch.list<int>
%12188 = torch.aten.view %12179, %12187 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%12189 = torch.aten.mul.Tensor %12188, %12186 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%12190 = torch.aten.cos %12189 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%12191 = torch.aten.sin %12189 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%12192 = torch.aten.complex %12190, %12191 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%12193 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12194 = flow.tensor.transfer %12193 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%12195 = torch_c.from_builtin_tensor %12194 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12196 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12197 = flow.tensor.transfer %12196 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%12198 = torch_c.from_builtin_tensor %12197 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12199 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12200 = flow.tensor.transfer %12199 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%12201 = torch_c.from_builtin_tensor %12200 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12202 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12203 = flow.tensor.transfer %12202 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%12204 = torch_c.from_builtin_tensor %12203 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12205 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12206 = flow.tensor.transfer %12205 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%12207 = torch_c.from_builtin_tensor %12206 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12208 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12209 = flow.tensor.transfer %12208 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%12210 = torch_c.from_builtin_tensor %12209 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12211 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12212 = flow.tensor.transfer %12211 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%12213 = torch_c.from_builtin_tensor %12212 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%12214 = torch_c.to_builtin_tensor %12192 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%12215 = flow.tensor.transfer %12214 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%12216 = torch_c.from_builtin_tensor %12215 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_9786 = torch.constant.int 1
%12217 = torch.aten.size.int %11866, %int1_9786 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9787 = torch.constant.int 0
%12218 = torch.aten.add.int %int0_9787, %12217 : !torch.int, !torch.int -> !torch.int
%int0_9788 = torch.constant.int 0
%int0_9789 = torch.constant.int 0
%int1_9790 = torch.constant.int 1
%12219 = torch.aten.slice.Tensor %12195, %int0_9788, %int0_9789, %12218, %int1_9790 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12219, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9791 = torch.constant.int 1
%int0_9792 = torch.constant.int 0
%int9223372036854775807_9793 = torch.constant.int 9223372036854775807
%int1_9794 = torch.constant.int 1
%12220 = torch.aten.slice.Tensor %12219, %int1_9791, %int0_9792, %int9223372036854775807_9793, %int1_9794 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12220, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9795 = torch.constant.int 0
%12221 = torch.aten.unsqueeze %12220, %int0_9795 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12221, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9796 = torch.constant.int 2
%12222 = torch.aten.unsqueeze %12221, %int2_9796 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12222, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9797 = torch.constant.int 3
%int0_9798 = torch.constant.int 0
%int9223372036854775807_9799 = torch.constant.int 9223372036854775807
%int1_9800 = torch.constant.int 1
%12223 = torch.aten.slice.Tensor %12222, %int3_9797, %int0_9798, %int9223372036854775807_9799, %int1_9800 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12223, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12224 = torch_c.to_builtin_tensor %11990 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9801 = arith.constant 1 : index
%dim_9802 = tensor.dim %12224, %c1_9801 : tensor<4x?x1x128xf16>
%12225 = flow.tensor.bitcast %12224 : tensor<4x?x1x128xf16>{%dim_9802} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9802}
%12226 = torch_c.from_builtin_tensor %12225 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12227 = torch.aten.mul.Tensor %12226, %12223 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12228 = torch_c.to_builtin_tensor %12227 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9803 = arith.constant 1 : index
%dim_9804 = tensor.dim %12228, %c1_9803 : tensor<4x?x1x64xcomplex<f32>>
%12229 = flow.tensor.bitcast %12228 : tensor<4x?x1x64xcomplex<f32>>{%dim_9804} -> tensor<4x?x1x128xf32>{%dim_9804}
%12230 = torch_c.from_builtin_tensor %12229 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9805 = torch.constant.int 5
%12231 = torch.prims.convert_element_type %12230, %int5_9805 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9806 = torch.constant.int 1
%12232 = torch.aten.size.int %11872, %int1_9806 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9807 = torch.constant.int 0
%12233 = torch.aten.add.int %int0_9807, %12232 : !torch.int, !torch.int -> !torch.int
%int0_9808 = torch.constant.int 0
%int0_9809 = torch.constant.int 0
%int1_9810 = torch.constant.int 1
%12234 = torch.aten.slice.Tensor %12198, %int0_9808, %int0_9809, %12233, %int1_9810 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12234, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9811 = torch.constant.int 1
%int0_9812 = torch.constant.int 0
%int9223372036854775807_9813 = torch.constant.int 9223372036854775807
%int1_9814 = torch.constant.int 1
%12235 = torch.aten.slice.Tensor %12234, %int1_9811, %int0_9812, %int9223372036854775807_9813, %int1_9814 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12235, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9815 = torch.constant.int 0
%12236 = torch.aten.unsqueeze %12235, %int0_9815 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12236, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9816 = torch.constant.int 2
%12237 = torch.aten.unsqueeze %12236, %int2_9816 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12237, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9817 = torch.constant.int 3
%int0_9818 = torch.constant.int 0
%int9223372036854775807_9819 = torch.constant.int 9223372036854775807
%int1_9820 = torch.constant.int 1
%12238 = torch.aten.slice.Tensor %12237, %int3_9817, %int0_9818, %int9223372036854775807_9819, %int1_9820 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12238, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12239 = torch_c.to_builtin_tensor %11992 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9821 = arith.constant 1 : index
%dim_9822 = tensor.dim %12239, %c1_9821 : tensor<4x?x1x128xf16>
%12240 = flow.tensor.bitcast %12239 : tensor<4x?x1x128xf16>{%dim_9822} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9822}
%12241 = torch_c.from_builtin_tensor %12240 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12242 = torch.aten.mul.Tensor %12241, %12238 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12243 = torch_c.to_builtin_tensor %12242 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9823 = arith.constant 1 : index
%dim_9824 = tensor.dim %12243, %c1_9823 : tensor<4x?x1x64xcomplex<f32>>
%12244 = flow.tensor.bitcast %12243 : tensor<4x?x1x64xcomplex<f32>>{%dim_9824} -> tensor<4x?x1x128xf32>{%dim_9824}
%12245 = torch_c.from_builtin_tensor %12244 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9825 = torch.constant.int 5
%12246 = torch.prims.convert_element_type %12245, %int5_9825 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9826 = torch.constant.int 1
%12247 = torch.aten.size.int %11878, %int1_9826 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9827 = torch.constant.int 0
%12248 = torch.aten.add.int %int0_9827, %12247 : !torch.int, !torch.int -> !torch.int
%int0_9828 = torch.constant.int 0
%int0_9829 = torch.constant.int 0
%int1_9830 = torch.constant.int 1
%12249 = torch.aten.slice.Tensor %12201, %int0_9828, %int0_9829, %12248, %int1_9830 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12249, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9831 = torch.constant.int 1
%int0_9832 = torch.constant.int 0
%int9223372036854775807_9833 = torch.constant.int 9223372036854775807
%int1_9834 = torch.constant.int 1
%12250 = torch.aten.slice.Tensor %12249, %int1_9831, %int0_9832, %int9223372036854775807_9833, %int1_9834 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12250, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9835 = torch.constant.int 0
%12251 = torch.aten.unsqueeze %12250, %int0_9835 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12251, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9836 = torch.constant.int 2
%12252 = torch.aten.unsqueeze %12251, %int2_9836 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12252, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9837 = torch.constant.int 3
%int0_9838 = torch.constant.int 0
%int9223372036854775807_9839 = torch.constant.int 9223372036854775807
%int1_9840 = torch.constant.int 1
%12253 = torch.aten.slice.Tensor %12252, %int3_9837, %int0_9838, %int9223372036854775807_9839, %int1_9840 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12253, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12254 = torch_c.to_builtin_tensor %11994 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9841 = arith.constant 1 : index
%dim_9842 = tensor.dim %12254, %c1_9841 : tensor<4x?x1x128xf16>
%12255 = flow.tensor.bitcast %12254 : tensor<4x?x1x128xf16>{%dim_9842} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9842}
%12256 = torch_c.from_builtin_tensor %12255 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12257 = torch.aten.mul.Tensor %12256, %12253 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12258 = torch_c.to_builtin_tensor %12257 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9843 = arith.constant 1 : index
%dim_9844 = tensor.dim %12258, %c1_9843 : tensor<4x?x1x64xcomplex<f32>>
%12259 = flow.tensor.bitcast %12258 : tensor<4x?x1x64xcomplex<f32>>{%dim_9844} -> tensor<4x?x1x128xf32>{%dim_9844}
%12260 = torch_c.from_builtin_tensor %12259 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9845 = torch.constant.int 5
%12261 = torch.prims.convert_element_type %12260, %int5_9845 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9846 = torch.constant.int 1
%12262 = torch.aten.size.int %11884, %int1_9846 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9847 = torch.constant.int 0
%12263 = torch.aten.add.int %int0_9847, %12262 : !torch.int, !torch.int -> !torch.int
%int0_9848 = torch.constant.int 0
%int0_9849 = torch.constant.int 0
%int1_9850 = torch.constant.int 1
%12264 = torch.aten.slice.Tensor %12204, %int0_9848, %int0_9849, %12263, %int1_9850 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12264, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9851 = torch.constant.int 1
%int0_9852 = torch.constant.int 0
%int9223372036854775807_9853 = torch.constant.int 9223372036854775807
%int1_9854 = torch.constant.int 1
%12265 = torch.aten.slice.Tensor %12264, %int1_9851, %int0_9852, %int9223372036854775807_9853, %int1_9854 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12265, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9855 = torch.constant.int 0
%12266 = torch.aten.unsqueeze %12265, %int0_9855 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12266, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9856 = torch.constant.int 2
%12267 = torch.aten.unsqueeze %12266, %int2_9856 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12267, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9857 = torch.constant.int 3
%int0_9858 = torch.constant.int 0
%int9223372036854775807_9859 = torch.constant.int 9223372036854775807
%int1_9860 = torch.constant.int 1
%12268 = torch.aten.slice.Tensor %12267, %int3_9857, %int0_9858, %int9223372036854775807_9859, %int1_9860 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12268, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12269 = torch_c.to_builtin_tensor %11996 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9861 = arith.constant 1 : index
%dim_9862 = tensor.dim %12269, %c1_9861 : tensor<4x?x1x128xf16>
%12270 = flow.tensor.bitcast %12269 : tensor<4x?x1x128xf16>{%dim_9862} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9862}
%12271 = torch_c.from_builtin_tensor %12270 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12272 = torch.aten.mul.Tensor %12271, %12268 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12273 = torch_c.to_builtin_tensor %12272 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9863 = arith.constant 1 : index
%dim_9864 = tensor.dim %12273, %c1_9863 : tensor<4x?x1x64xcomplex<f32>>
%12274 = flow.tensor.bitcast %12273 : tensor<4x?x1x64xcomplex<f32>>{%dim_9864} -> tensor<4x?x1x128xf32>{%dim_9864}
%12275 = torch_c.from_builtin_tensor %12274 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9865 = torch.constant.int 5
%12276 = torch.prims.convert_element_type %12275, %int5_9865 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9866 = torch.constant.int 1
%12277 = torch.aten.size.int %11890, %int1_9866 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9867 = torch.constant.int 0
%12278 = torch.aten.add.int %int0_9867, %12277 : !torch.int, !torch.int -> !torch.int
%int0_9868 = torch.constant.int 0
%int0_9869 = torch.constant.int 0
%int1_9870 = torch.constant.int 1
%12279 = torch.aten.slice.Tensor %12207, %int0_9868, %int0_9869, %12278, %int1_9870 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12279, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9871 = torch.constant.int 1
%int0_9872 = torch.constant.int 0
%int9223372036854775807_9873 = torch.constant.int 9223372036854775807
%int1_9874 = torch.constant.int 1
%12280 = torch.aten.slice.Tensor %12279, %int1_9871, %int0_9872, %int9223372036854775807_9873, %int1_9874 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12280, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9875 = torch.constant.int 0
%12281 = torch.aten.unsqueeze %12280, %int0_9875 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12281, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9876 = torch.constant.int 2
%12282 = torch.aten.unsqueeze %12281, %int2_9876 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12282, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9877 = torch.constant.int 3
%int0_9878 = torch.constant.int 0
%int9223372036854775807_9879 = torch.constant.int 9223372036854775807
%int1_9880 = torch.constant.int 1
%12283 = torch.aten.slice.Tensor %12282, %int3_9877, %int0_9878, %int9223372036854775807_9879, %int1_9880 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12283, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12284 = torch_c.to_builtin_tensor %11998 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9881 = arith.constant 1 : index
%dim_9882 = tensor.dim %12284, %c1_9881 : tensor<4x?x1x128xf16>
%12285 = flow.tensor.bitcast %12284 : tensor<4x?x1x128xf16>{%dim_9882} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9882}
%12286 = torch_c.from_builtin_tensor %12285 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12287 = torch.aten.mul.Tensor %12286, %12283 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12288 = torch_c.to_builtin_tensor %12287 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9883 = arith.constant 1 : index
%dim_9884 = tensor.dim %12288, %c1_9883 : tensor<4x?x1x64xcomplex<f32>>
%12289 = flow.tensor.bitcast %12288 : tensor<4x?x1x64xcomplex<f32>>{%dim_9884} -> tensor<4x?x1x128xf32>{%dim_9884}
%12290 = torch_c.from_builtin_tensor %12289 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9885 = torch.constant.int 5
%12291 = torch.prims.convert_element_type %12290, %int5_9885 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9886 = torch.constant.int 1
%12292 = torch.aten.size.int %11896, %int1_9886 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9887 = torch.constant.int 0
%12293 = torch.aten.add.int %int0_9887, %12292 : !torch.int, !torch.int -> !torch.int
%int0_9888 = torch.constant.int 0
%int0_9889 = torch.constant.int 0
%int1_9890 = torch.constant.int 1
%12294 = torch.aten.slice.Tensor %12210, %int0_9888, %int0_9889, %12293, %int1_9890 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12294, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9891 = torch.constant.int 1
%int0_9892 = torch.constant.int 0
%int9223372036854775807_9893 = torch.constant.int 9223372036854775807
%int1_9894 = torch.constant.int 1
%12295 = torch.aten.slice.Tensor %12294, %int1_9891, %int0_9892, %int9223372036854775807_9893, %int1_9894 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12295, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9895 = torch.constant.int 0
%12296 = torch.aten.unsqueeze %12295, %int0_9895 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12296, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9896 = torch.constant.int 2
%12297 = torch.aten.unsqueeze %12296, %int2_9896 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12297, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9897 = torch.constant.int 3
%int0_9898 = torch.constant.int 0
%int9223372036854775807_9899 = torch.constant.int 9223372036854775807
%int1_9900 = torch.constant.int 1
%12298 = torch.aten.slice.Tensor %12297, %int3_9897, %int0_9898, %int9223372036854775807_9899, %int1_9900 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12298, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12299 = torch_c.to_builtin_tensor %12000 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9901 = arith.constant 1 : index
%dim_9902 = tensor.dim %12299, %c1_9901 : tensor<4x?x1x128xf16>
%12300 = flow.tensor.bitcast %12299 : tensor<4x?x1x128xf16>{%dim_9902} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9902}
%12301 = torch_c.from_builtin_tensor %12300 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12302 = torch.aten.mul.Tensor %12301, %12298 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12303 = torch_c.to_builtin_tensor %12302 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9903 = arith.constant 1 : index
%dim_9904 = tensor.dim %12303, %c1_9903 : tensor<4x?x1x64xcomplex<f32>>
%12304 = flow.tensor.bitcast %12303 : tensor<4x?x1x64xcomplex<f32>>{%dim_9904} -> tensor<4x?x1x128xf32>{%dim_9904}
%12305 = torch_c.from_builtin_tensor %12304 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9905 = torch.constant.int 5
%12306 = torch.prims.convert_element_type %12305, %int5_9905 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9906 = torch.constant.int 1
%12307 = torch.aten.size.int %11902, %int1_9906 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9907 = torch.constant.int 0
%12308 = torch.aten.add.int %int0_9907, %12307 : !torch.int, !torch.int -> !torch.int
%int0_9908 = torch.constant.int 0
%int0_9909 = torch.constant.int 0
%int1_9910 = torch.constant.int 1
%12309 = torch.aten.slice.Tensor %12213, %int0_9908, %int0_9909, %12308, %int1_9910 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12309, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9911 = torch.constant.int 1
%int0_9912 = torch.constant.int 0
%int9223372036854775807_9913 = torch.constant.int 9223372036854775807
%int1_9914 = torch.constant.int 1
%12310 = torch.aten.slice.Tensor %12309, %int1_9911, %int0_9912, %int9223372036854775807_9913, %int1_9914 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12310, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9915 = torch.constant.int 0
%12311 = torch.aten.unsqueeze %12310, %int0_9915 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12311, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9916 = torch.constant.int 2
%12312 = torch.aten.unsqueeze %12311, %int2_9916 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12312, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9917 = torch.constant.int 3
%int0_9918 = torch.constant.int 0
%int9223372036854775807_9919 = torch.constant.int 9223372036854775807
%int1_9920 = torch.constant.int 1
%12313 = torch.aten.slice.Tensor %12312, %int3_9917, %int0_9918, %int9223372036854775807_9919, %int1_9920 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12313, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12314 = torch_c.to_builtin_tensor %12002 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9921 = arith.constant 1 : index
%dim_9922 = tensor.dim %12314, %c1_9921 : tensor<4x?x1x128xf16>
%12315 = flow.tensor.bitcast %12314 : tensor<4x?x1x128xf16>{%dim_9922} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9922}
%12316 = torch_c.from_builtin_tensor %12315 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12317 = torch.aten.mul.Tensor %12316, %12313 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12318 = torch_c.to_builtin_tensor %12317 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9923 = arith.constant 1 : index
%dim_9924 = tensor.dim %12318, %c1_9923 : tensor<4x?x1x64xcomplex<f32>>
%12319 = flow.tensor.bitcast %12318 : tensor<4x?x1x64xcomplex<f32>>{%dim_9924} -> tensor<4x?x1x128xf32>{%dim_9924}
%12320 = torch_c.from_builtin_tensor %12319 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9925 = torch.constant.int 5
%12321 = torch.prims.convert_element_type %12320, %int5_9925 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_9926 = torch.constant.int 1
%12322 = torch.aten.size.int %11908, %int1_9926 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_9927 = torch.constant.int 0
%12323 = torch.aten.add.int %int0_9927, %12322 : !torch.int, !torch.int -> !torch.int
%int0_9928 = torch.constant.int 0
%int0_9929 = torch.constant.int 0
%int1_9930 = torch.constant.int 1
%12324 = torch.aten.slice.Tensor %12216, %int0_9928, %int0_9929, %12323, %int1_9930 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12324, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_9931 = torch.constant.int 1
%int0_9932 = torch.constant.int 0
%int9223372036854775807_9933 = torch.constant.int 9223372036854775807
%int1_9934 = torch.constant.int 1
%12325 = torch.aten.slice.Tensor %12324, %int1_9931, %int0_9932, %int9223372036854775807_9933, %int1_9934 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %12325, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_9935 = torch.constant.int 0
%12326 = torch.aten.unsqueeze %12325, %int0_9935 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %12326, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_9936 = torch.constant.int 2
%12327 = torch.aten.unsqueeze %12326, %int2_9936 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12327, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_9937 = torch.constant.int 3
%int0_9938 = torch.constant.int 0
%int9223372036854775807_9939 = torch.constant.int 9223372036854775807
%int1_9940 = torch.constant.int 1
%12328 = torch.aten.slice.Tensor %12327, %int3_9937, %int0_9938, %int9223372036854775807_9939, %int1_9940 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12328, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%12329 = torch_c.to_builtin_tensor %12004 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_9941 = arith.constant 1 : index
%dim_9942 = tensor.dim %12329, %c1_9941 : tensor<4x?x1x128xf16>
%12330 = flow.tensor.bitcast %12329 : tensor<4x?x1x128xf16>{%dim_9942} -> tensor<4x?x1x64xcomplex<f16>>{%dim_9942}
%12331 = torch_c.from_builtin_tensor %12330 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %12331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%12332 = torch.aten.mul.Tensor %12331, %12328 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %12332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%12333 = torch_c.to_builtin_tensor %12332 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_9943 = arith.constant 1 : index
%dim_9944 = tensor.dim %12333, %c1_9943 : tensor<4x?x1x64xcomplex<f32>>
%12334 = flow.tensor.bitcast %12333 : tensor<4x?x1x64xcomplex<f32>>{%dim_9944} -> tensor<4x?x1x128xf32>{%dim_9944}
%12335 = torch_c.from_builtin_tensor %12334 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %12335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_9945 = torch.constant.int 5
%12336 = torch.prims.convert_element_type %12335, %int5_9945 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %12336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_9946 = torch.constant.int 64
%12337 = torch.aten.mul.Scalar %2364, %int64_9946 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12337, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9947 = torch.constant.int 64
%12338 = torch.aten.mul.Scalar %2367, %int64_9947 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12338, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9948 = torch.constant.int 64
%12339 = torch.aten.mul.Scalar %2370, %int64_9948 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12339, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9949 = torch.constant.int 64
%12340 = torch.aten.mul.Scalar %2373, %int64_9949 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12340, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9950 = torch.constant.int 64
%12341 = torch.aten.mul.Scalar %2376, %int64_9950 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12341, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9951 = torch.constant.int 64
%12342 = torch.aten.mul.Scalar %2379, %int64_9951 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12342, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9952 = torch.constant.int 64
%12343 = torch.aten.mul.Scalar %2382, %int64_9952 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12343, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_9953 = torch.constant.int 64
%12344 = torch.aten.mul.Scalar %2385, %int64_9953 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12344, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10 = torch.constant.int 10
%int1_9954 = torch.constant.int 1
%12345 = torch.aten.add.Scalar %12337, %int10, %int1_9954 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12345, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9955 = torch.constant.int 10
%int1_9956 = torch.constant.int 1
%12346 = torch.aten.add.Scalar %12338, %int10_9955, %int1_9956 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12346, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9957 = torch.constant.int 10
%int1_9958 = torch.constant.int 1
%12347 = torch.aten.add.Scalar %12339, %int10_9957, %int1_9958 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12347, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9959 = torch.constant.int 10
%int1_9960 = torch.constant.int 1
%12348 = torch.aten.add.Scalar %12340, %int10_9959, %int1_9960 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12348, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9961 = torch.constant.int 10
%int1_9962 = torch.constant.int 1
%12349 = torch.aten.add.Scalar %12341, %int10_9961, %int1_9962 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12349, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9963 = torch.constant.int 10
%int1_9964 = torch.constant.int 1
%12350 = torch.aten.add.Scalar %12342, %int10_9963, %int1_9964 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12350, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9965 = torch.constant.int 10
%int1_9966 = torch.constant.int 1
%12351 = torch.aten.add.Scalar %12343, %int10_9965, %int1_9966 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12351, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int10_9967 = torch.constant.int 10
%int1_9968 = torch.constant.int 1
%12352 = torch.aten.add.Scalar %12344, %int10_9967, %int1_9968 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12352, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_9969 = torch.constant.int 4
%int16_9970 = torch.constant.int 16
%int1_9971 = torch.constant.int 1
%int128_9972 = torch.constant.int 128
%12353 = torch.prim.ListConstruct %int4_9969, %3095, %int16_9970, %int1_9971, %int128_9972 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12354 = torch.aten.view %12231, %12353 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12354, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9973 = torch.constant.int 4
%int16_9974 = torch.constant.int 16
%int1_9975 = torch.constant.int 1
%int128_9976 = torch.constant.int 128
%12355 = torch.prim.ListConstruct %int4_9973, %3095, %int16_9974, %int1_9975, %int128_9976 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12356 = torch.aten.view %12246, %12355 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12356, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9977 = torch.constant.int 4
%int16_9978 = torch.constant.int 16
%int1_9979 = torch.constant.int 1
%int128_9980 = torch.constant.int 128
%12357 = torch.prim.ListConstruct %int4_9977, %3095, %int16_9978, %int1_9979, %int128_9980 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12358 = torch.aten.view %12261, %12357 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12358, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9981 = torch.constant.int 4
%int16_9982 = torch.constant.int 16
%int1_9983 = torch.constant.int 1
%int128_9984 = torch.constant.int 128
%12359 = torch.prim.ListConstruct %int4_9981, %3095, %int16_9982, %int1_9983, %int128_9984 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12360 = torch.aten.view %12276, %12359 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12360, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9985 = torch.constant.int 4
%int16_9986 = torch.constant.int 16
%int1_9987 = torch.constant.int 1
%int128_9988 = torch.constant.int 128
%12361 = torch.prim.ListConstruct %int4_9985, %3095, %int16_9986, %int1_9987, %int128_9988 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12362 = torch.aten.view %12291, %12361 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12362, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9989 = torch.constant.int 4
%int16_9990 = torch.constant.int 16
%int1_9991 = torch.constant.int 1
%int128_9992 = torch.constant.int 128
%12363 = torch.prim.ListConstruct %int4_9989, %3095, %int16_9990, %int1_9991, %int128_9992 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12364 = torch.aten.view %12306, %12363 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12364, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9993 = torch.constant.int 4
%int16_9994 = torch.constant.int 16
%int1_9995 = torch.constant.int 1
%int128_9996 = torch.constant.int 128
%12365 = torch.prim.ListConstruct %int4_9993, %3095, %int16_9994, %int1_9995, %int128_9996 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12366 = torch.aten.view %12321, %12365 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12366, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_9997 = torch.constant.int 4
%int16_9998 = torch.constant.int 16
%int1_9999 = torch.constant.int 1
%int128_10000 = torch.constant.int 128
%12367 = torch.prim.ListConstruct %int4_9997, %3095, %int16_9998, %int1_9999, %int128_10000 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12368 = torch.aten.view %12336, %12367 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12368, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10001 = torch.constant.int 4
%12369 = torch.aten.mul.int %int4_10001, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10002 = torch.constant.int 16
%int1_10003 = torch.constant.int 1
%int128_10004 = torch.constant.int 128
%12370 = torch.prim.ListConstruct %12369, %int16_10002, %int1_10003, %int128_10004 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12371 = torch.aten.view %12354, %12370 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12371, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10005 = torch.constant.int 4
%12372 = torch.aten.mul.int %int4_10005, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10006 = torch.constant.int 16
%int1_10007 = torch.constant.int 1
%int128_10008 = torch.constant.int 128
%12373 = torch.prim.ListConstruct %12372, %int16_10006, %int1_10007, %int128_10008 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12374 = torch.aten.view %12356, %12373 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12374, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10009 = torch.constant.int 4
%12375 = torch.aten.mul.int %int4_10009, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10010 = torch.constant.int 16
%int1_10011 = torch.constant.int 1
%int128_10012 = torch.constant.int 128
%12376 = torch.prim.ListConstruct %12375, %int16_10010, %int1_10011, %int128_10012 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12377 = torch.aten.view %12358, %12376 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12377, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10013 = torch.constant.int 4
%12378 = torch.aten.mul.int %int4_10013, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10014 = torch.constant.int 16
%int1_10015 = torch.constant.int 1
%int128_10016 = torch.constant.int 128
%12379 = torch.prim.ListConstruct %12378, %int16_10014, %int1_10015, %int128_10016 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12380 = torch.aten.view %12360, %12379 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12380, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10017 = torch.constant.int 4
%12381 = torch.aten.mul.int %int4_10017, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10018 = torch.constant.int 16
%int1_10019 = torch.constant.int 1
%int128_10020 = torch.constant.int 128
%12382 = torch.prim.ListConstruct %12381, %int16_10018, %int1_10019, %int128_10020 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12383 = torch.aten.view %12362, %12382 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12383, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10021 = torch.constant.int 4
%12384 = torch.aten.mul.int %int4_10021, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10022 = torch.constant.int 16
%int1_10023 = torch.constant.int 1
%int128_10024 = torch.constant.int 128
%12385 = torch.prim.ListConstruct %12384, %int16_10022, %int1_10023, %int128_10024 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12386 = torch.aten.view %12364, %12385 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12386, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10025 = torch.constant.int 4
%12387 = torch.aten.mul.int %int4_10025, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10026 = torch.constant.int 16
%int1_10027 = torch.constant.int 1
%int128_10028 = torch.constant.int 128
%12388 = torch.prim.ListConstruct %12387, %int16_10026, %int1_10027, %int128_10028 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12389 = torch.aten.view %12366, %12388 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12389, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10029 = torch.constant.int 4
%12390 = torch.aten.mul.int %int4_10029, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10030 = torch.constant.int 16
%int1_10031 = torch.constant.int 1
%int128_10032 = torch.constant.int 128
%12391 = torch.prim.ListConstruct %12390, %int16_10030, %int1_10031, %int128_10032 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12392 = torch.aten.view %12368, %12391 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12392, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10033 = torch.constant.int 4
%12393 = torch.aten.mul.int %int4_10033, %3095 : !torch.int, !torch.int -> !torch.int
%12394 = torch.prim.ListConstruct %12393 : (!torch.int) -> !torch.list<int>
%12395 = torch.aten.view %12345, %12394 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12395, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10034 = torch.constant.int 4
%12396 = torch.aten.mul.int %int4_10034, %3095 : !torch.int, !torch.int -> !torch.int
%12397 = torch.prim.ListConstruct %12396 : (!torch.int) -> !torch.list<int>
%12398 = torch.aten.view %12346, %12397 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12398, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10035 = torch.constant.int 4
%12399 = torch.aten.mul.int %int4_10035, %3095 : !torch.int, !torch.int -> !torch.int
%12400 = torch.prim.ListConstruct %12399 : (!torch.int) -> !torch.list<int>
%12401 = torch.aten.view %12347, %12400 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12401, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10036 = torch.constant.int 4
%12402 = torch.aten.mul.int %int4_10036, %3095 : !torch.int, !torch.int -> !torch.int
%12403 = torch.prim.ListConstruct %12402 : (!torch.int) -> !torch.list<int>
%12404 = torch.aten.view %12348, %12403 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12404, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10037 = torch.constant.int 4
%12405 = torch.aten.mul.int %int4_10037, %3095 : !torch.int, !torch.int -> !torch.int
%12406 = torch.prim.ListConstruct %12405 : (!torch.int) -> !torch.list<int>
%12407 = torch.aten.view %12349, %12406 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12407, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10038 = torch.constant.int 4
%12408 = torch.aten.mul.int %int4_10038, %3095 : !torch.int, !torch.int -> !torch.int
%12409 = torch.prim.ListConstruct %12408 : (!torch.int) -> !torch.list<int>
%12410 = torch.aten.view %12350, %12409 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12410, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10039 = torch.constant.int 4
%12411 = torch.aten.mul.int %int4_10039, %3095 : !torch.int, !torch.int -> !torch.int
%12412 = torch.prim.ListConstruct %12411 : (!torch.int) -> !torch.list<int>
%12413 = torch.aten.view %12351, %12412 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12413, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10040 = torch.constant.int 4
%12414 = torch.aten.mul.int %int4_10040, %3095 : !torch.int, !torch.int -> !torch.int
%12415 = torch.prim.ListConstruct %12414 : (!torch.int) -> !torch.list<int>
%12416 = torch.aten.view %12352, %12415 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12416, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10041 = torch.constant.int 4
%int16_10042 = torch.constant.int 16
%int1_10043 = torch.constant.int 1
%int128_10044 = torch.constant.int 128
%12417 = torch.prim.ListConstruct %int4_10041, %3095, %int16_10042, %int1_10043, %int128_10044 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12418 = torch.aten.view %12006, %12417 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12418, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10045 = torch.constant.int 4
%int16_10046 = torch.constant.int 16
%int1_10047 = torch.constant.int 1
%int128_10048 = torch.constant.int 128
%12419 = torch.prim.ListConstruct %int4_10045, %3095, %int16_10046, %int1_10047, %int128_10048 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12420 = torch.aten.view %12008, %12419 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12420, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10049 = torch.constant.int 4
%int16_10050 = torch.constant.int 16
%int1_10051 = torch.constant.int 1
%int128_10052 = torch.constant.int 128
%12421 = torch.prim.ListConstruct %int4_10049, %3095, %int16_10050, %int1_10051, %int128_10052 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12422 = torch.aten.view %12010, %12421 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12422, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10053 = torch.constant.int 4
%int16_10054 = torch.constant.int 16
%int1_10055 = torch.constant.int 1
%int128_10056 = torch.constant.int 128
%12423 = torch.prim.ListConstruct %int4_10053, %3095, %int16_10054, %int1_10055, %int128_10056 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12424 = torch.aten.view %12012, %12423 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12424, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10057 = torch.constant.int 4
%int16_10058 = torch.constant.int 16
%int1_10059 = torch.constant.int 1
%int128_10060 = torch.constant.int 128
%12425 = torch.prim.ListConstruct %int4_10057, %3095, %int16_10058, %int1_10059, %int128_10060 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12426 = torch.aten.view %12014, %12425 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12426, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10061 = torch.constant.int 4
%int16_10062 = torch.constant.int 16
%int1_10063 = torch.constant.int 1
%int128_10064 = torch.constant.int 128
%12427 = torch.prim.ListConstruct %int4_10061, %3095, %int16_10062, %int1_10063, %int128_10064 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12428 = torch.aten.view %12016, %12427 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12428, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10065 = torch.constant.int 4
%int16_10066 = torch.constant.int 16
%int1_10067 = torch.constant.int 1
%int128_10068 = torch.constant.int 128
%12429 = torch.prim.ListConstruct %int4_10065, %3095, %int16_10066, %int1_10067, %int128_10068 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12430 = torch.aten.view %12018, %12429 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12430, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10069 = torch.constant.int 4
%int16_10070 = torch.constant.int 16
%int1_10071 = torch.constant.int 1
%int128_10072 = torch.constant.int 128
%12431 = torch.prim.ListConstruct %int4_10069, %3095, %int16_10070, %int1_10071, %int128_10072 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12432 = torch.aten.view %12020, %12431 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %12432, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_10073 = torch.constant.int 4
%12433 = torch.aten.mul.int %int4_10073, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10074 = torch.constant.int 16
%int1_10075 = torch.constant.int 1
%int128_10076 = torch.constant.int 128
%12434 = torch.prim.ListConstruct %12433, %int16_10074, %int1_10075, %int128_10076 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12435 = torch.aten.view %12418, %12434 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12435, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10077 = torch.constant.int 4
%12436 = torch.aten.mul.int %int4_10077, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10078 = torch.constant.int 16
%int1_10079 = torch.constant.int 1
%int128_10080 = torch.constant.int 128
%12437 = torch.prim.ListConstruct %12436, %int16_10078, %int1_10079, %int128_10080 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12438 = torch.aten.view %12420, %12437 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12438, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10081 = torch.constant.int 4
%12439 = torch.aten.mul.int %int4_10081, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10082 = torch.constant.int 16
%int1_10083 = torch.constant.int 1
%int128_10084 = torch.constant.int 128
%12440 = torch.prim.ListConstruct %12439, %int16_10082, %int1_10083, %int128_10084 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12441 = torch.aten.view %12422, %12440 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12441, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10085 = torch.constant.int 4
%12442 = torch.aten.mul.int %int4_10085, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10086 = torch.constant.int 16
%int1_10087 = torch.constant.int 1
%int128_10088 = torch.constant.int 128
%12443 = torch.prim.ListConstruct %12442, %int16_10086, %int1_10087, %int128_10088 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12444 = torch.aten.view %12424, %12443 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12444, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10089 = torch.constant.int 4
%12445 = torch.aten.mul.int %int4_10089, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10090 = torch.constant.int 16
%int1_10091 = torch.constant.int 1
%int128_10092 = torch.constant.int 128
%12446 = torch.prim.ListConstruct %12445, %int16_10090, %int1_10091, %int128_10092 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12447 = torch.aten.view %12426, %12446 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12447, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10093 = torch.constant.int 4
%12448 = torch.aten.mul.int %int4_10093, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10094 = torch.constant.int 16
%int1_10095 = torch.constant.int 1
%int128_10096 = torch.constant.int 128
%12449 = torch.prim.ListConstruct %12448, %int16_10094, %int1_10095, %int128_10096 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12450 = torch.aten.view %12428, %12449 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12450, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10097 = torch.constant.int 4
%12451 = torch.aten.mul.int %int4_10097, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10098 = torch.constant.int 16
%int1_10099 = torch.constant.int 1
%int128_10100 = torch.constant.int 128
%12452 = torch.prim.ListConstruct %12451, %int16_10098, %int1_10099, %int128_10100 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12453 = torch.aten.view %12430, %12452 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12453, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_10101 = torch.constant.int 4
%12454 = torch.aten.mul.int %int4_10101, %3095 : !torch.int, !torch.int -> !torch.int
%int16_10102 = torch.constant.int 16
%int1_10103 = torch.constant.int 1
%int128_10104 = torch.constant.int 128
%12455 = torch.prim.ListConstruct %12454, %int16_10102, %int1_10103, %int128_10104 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12456 = torch.aten.view %12432, %12455 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12456, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_10105 = torch.constant.int 1
%int1_10106 = torch.constant.int 1
%12457 = torch.aten.add.Scalar %12345, %int1_10105, %int1_10106 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12457, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10107 = torch.constant.int 1
%int1_10108 = torch.constant.int 1
%12458 = torch.aten.add.Scalar %12346, %int1_10107, %int1_10108 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12458, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10109 = torch.constant.int 1
%int1_10110 = torch.constant.int 1
%12459 = torch.aten.add.Scalar %12347, %int1_10109, %int1_10110 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12459, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10111 = torch.constant.int 1
%int1_10112 = torch.constant.int 1
%12460 = torch.aten.add.Scalar %12348, %int1_10111, %int1_10112 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12460, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10113 = torch.constant.int 1
%int1_10114 = torch.constant.int 1
%12461 = torch.aten.add.Scalar %12349, %int1_10113, %int1_10114 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12461, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10115 = torch.constant.int 1
%int1_10116 = torch.constant.int 1
%12462 = torch.aten.add.Scalar %12350, %int1_10115, %int1_10116 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12462, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10117 = torch.constant.int 1
%int1_10118 = torch.constant.int 1
%12463 = torch.aten.add.Scalar %12351, %int1_10117, %int1_10118 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12463, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_10119 = torch.constant.int 1
%int1_10120 = torch.constant.int 1
%12464 = torch.aten.add.Scalar %12352, %int1_10119, %int1_10120 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %12464, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_10121 = torch.constant.int 4
%12465 = torch.aten.mul.int %int4_10121, %3095 : !torch.int, !torch.int -> !torch.int
%12466 = torch.prim.ListConstruct %12465 : (!torch.int) -> !torch.list<int>
%12467 = torch.aten.view %12457, %12466 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12467, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10122 = torch.constant.int 4
%12468 = torch.aten.mul.int %int4_10122, %3095 : !torch.int, !torch.int -> !torch.int
%12469 = torch.prim.ListConstruct %12468 : (!torch.int) -> !torch.list<int>
%12470 = torch.aten.view %12458, %12469 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12470, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10123 = torch.constant.int 4
%12471 = torch.aten.mul.int %int4_10123, %3095 : !torch.int, !torch.int -> !torch.int
%12472 = torch.prim.ListConstruct %12471 : (!torch.int) -> !torch.list<int>
%12473 = torch.aten.view %12459, %12472 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12473, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10124 = torch.constant.int 4
%12474 = torch.aten.mul.int %int4_10124, %3095 : !torch.int, !torch.int -> !torch.int
%12475 = torch.prim.ListConstruct %12474 : (!torch.int) -> !torch.list<int>
%12476 = torch.aten.view %12460, %12475 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12476, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10125 = torch.constant.int 4
%12477 = torch.aten.mul.int %int4_10125, %3095 : !torch.int, !torch.int -> !torch.int
%12478 = torch.prim.ListConstruct %12477 : (!torch.int) -> !torch.list<int>
%12479 = torch.aten.view %12461, %12478 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12479, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10126 = torch.constant.int 4
%12480 = torch.aten.mul.int %int4_10126, %3095 : !torch.int, !torch.int -> !torch.int
%12481 = torch.prim.ListConstruct %12480 : (!torch.int) -> !torch.list<int>
%12482 = torch.aten.view %12462, %12481 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12482, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10127 = torch.constant.int 4
%12483 = torch.aten.mul.int %int4_10127, %3095 : !torch.int, !torch.int -> !torch.int
%12484 = torch.prim.ListConstruct %12483 : (!torch.int) -> !torch.list<int>
%12485 = torch.aten.view %12463, %12484 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12485, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_10128 = torch.constant.int 4
%12486 = torch.aten.mul.int %int4_10128, %3095 : !torch.int, !torch.int -> !torch.int
%12487 = torch.prim.ListConstruct %12486 : (!torch.int) -> !torch.list<int>
%12488 = torch.aten.view %12464, %12487 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12488, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%12489 = torch.prim.ListConstruct %12395, %12467 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10129 = torch.constant.int 0
%12490 = torch.aten.cat %12489, %int0_10129 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12490, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12491 = torch.prim.ListConstruct %12398, %12470 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10130 = torch.constant.int 0
%12492 = torch.aten.cat %12491, %int0_10130 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12492, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12493 = torch.prim.ListConstruct %12401, %12473 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10131 = torch.constant.int 0
%12494 = torch.aten.cat %12493, %int0_10131 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12494, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12495 = torch.prim.ListConstruct %12404, %12476 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10132 = torch.constant.int 0
%12496 = torch.aten.cat %12495, %int0_10132 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12496, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12497 = torch.prim.ListConstruct %12407, %12479 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10133 = torch.constant.int 0
%12498 = torch.aten.cat %12497, %int0_10133 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12498, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12499 = torch.prim.ListConstruct %12410, %12482 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10134 = torch.constant.int 0
%12500 = torch.aten.cat %12499, %int0_10134 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12500, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12501 = torch.prim.ListConstruct %12413, %12485 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10135 = torch.constant.int 0
%12502 = torch.aten.cat %12501, %int0_10135 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12502, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12503 = torch.prim.ListConstruct %12416, %12488 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_10136 = torch.constant.int 0
%12504 = torch.aten.cat %12503, %int0_10136 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %12504, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%12505 = torch.prim.ListConstruct %12371, %12435 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10137 = torch.constant.int 0
%12506 = torch.aten.cat %12505, %int0_10137 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12506, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12507 = torch.prim.ListConstruct %12374, %12438 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10138 = torch.constant.int 0
%12508 = torch.aten.cat %12507, %int0_10138 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12508, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12509 = torch.prim.ListConstruct %12377, %12441 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10139 = torch.constant.int 0
%12510 = torch.aten.cat %12509, %int0_10139 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12510, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12511 = torch.prim.ListConstruct %12380, %12444 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10140 = torch.constant.int 0
%12512 = torch.aten.cat %12511, %int0_10140 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12512, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12513 = torch.prim.ListConstruct %12383, %12447 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10141 = torch.constant.int 0
%12514 = torch.aten.cat %12513, %int0_10141 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12514, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12515 = torch.prim.ListConstruct %12386, %12450 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10142 = torch.constant.int 0
%12516 = torch.aten.cat %12515, %int0_10142 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12516, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12517 = torch.prim.ListConstruct %12389, %12453 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10143 = torch.constant.int 0
%12518 = torch.aten.cat %12517, %int0_10143 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12518, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12519 = torch.prim.ListConstruct %12392, %12456 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_10144 = torch.constant.int 0
%12520 = torch.aten.cat %12519, %int0_10144 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12520, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10145 = torch.constant.int 32
%int2_10146 = torch.constant.int 2
%int16_10147 = torch.constant.int 16
%int1_10148 = torch.constant.int 1
%int128_10149 = torch.constant.int 128
%12521 = torch.prim.ListConstruct %3023, %int32_10145, %int2_10146, %int16_10147, %int1_10148, %int128_10149 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12522 = torch.aten.view %10671, %12521 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12522, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10150 = torch.constant.int 32
%12523 = torch.aten.mul.int %3023, %int32_10150 : !torch.int, !torch.int -> !torch.int
%int2_10151 = torch.constant.int 2
%12524 = torch.aten.mul.int %12523, %int2_10151 : !torch.int, !torch.int -> !torch.int
%int16_10152 = torch.constant.int 16
%int1_10153 = torch.constant.int 1
%int128_10154 = torch.constant.int 128
%12525 = torch.prim.ListConstruct %12524, %int16_10152, %int1_10153, %int128_10154 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12526 = torch.aten.view %12522, %12525 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12526, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12527 = torch.prim.ListConstruct %12490 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10155 = torch.constant.bool false
%12528 = torch.aten.index_put %12526, %12527, %12506, %false_10155 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12528, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10156 = torch.constant.int 32
%int2_10157 = torch.constant.int 2
%int16_10158 = torch.constant.int 16
%int1_10159 = torch.constant.int 1
%int128_10160 = torch.constant.int 128
%12529 = torch.prim.ListConstruct %3023, %int32_10156, %int2_10157, %int16_10158, %int1_10159, %int128_10160 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12530 = torch.aten.view %12528, %12529 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12530, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10161 = torch.constant.int 131072
%12531 = torch.prim.ListConstruct %3023, %int131072_10161 : (!torch.int, !torch.int) -> !torch.list<int>
%12532 = torch.aten.view %12530, %12531 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12532, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10162 = torch.constant.int 32
%int2_10163 = torch.constant.int 2
%int16_10164 = torch.constant.int 16
%int1_10165 = torch.constant.int 1
%int128_10166 = torch.constant.int 128
%12533 = torch.prim.ListConstruct %3026, %int32_10162, %int2_10163, %int16_10164, %int1_10165, %int128_10166 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12534 = torch.aten.view %10683, %12533 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12534, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10167 = torch.constant.int 32
%12535 = torch.aten.mul.int %3026, %int32_10167 : !torch.int, !torch.int -> !torch.int
%int2_10168 = torch.constant.int 2
%12536 = torch.aten.mul.int %12535, %int2_10168 : !torch.int, !torch.int -> !torch.int
%int16_10169 = torch.constant.int 16
%int1_10170 = torch.constant.int 1
%int128_10171 = torch.constant.int 128
%12537 = torch.prim.ListConstruct %12536, %int16_10169, %int1_10170, %int128_10171 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12538 = torch.aten.view %12534, %12537 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12538, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12539 = torch.prim.ListConstruct %12492 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10172 = torch.constant.bool false
%12540 = torch.aten.index_put %12538, %12539, %12508, %false_10172 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12540, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10173 = torch.constant.int 32
%int2_10174 = torch.constant.int 2
%int16_10175 = torch.constant.int 16
%int1_10176 = torch.constant.int 1
%int128_10177 = torch.constant.int 128
%12541 = torch.prim.ListConstruct %3026, %int32_10173, %int2_10174, %int16_10175, %int1_10176, %int128_10177 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12542 = torch.aten.view %12540, %12541 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12542, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10178 = torch.constant.int 131072
%12543 = torch.prim.ListConstruct %3026, %int131072_10178 : (!torch.int, !torch.int) -> !torch.list<int>
%12544 = torch.aten.view %12542, %12543 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12544, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10179 = torch.constant.int 32
%int2_10180 = torch.constant.int 2
%int16_10181 = torch.constant.int 16
%int1_10182 = torch.constant.int 1
%int128_10183 = torch.constant.int 128
%12545 = torch.prim.ListConstruct %3029, %int32_10179, %int2_10180, %int16_10181, %int1_10182, %int128_10183 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12546 = torch.aten.view %10695, %12545 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12546, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10184 = torch.constant.int 32
%12547 = torch.aten.mul.int %3029, %int32_10184 : !torch.int, !torch.int -> !torch.int
%int2_10185 = torch.constant.int 2
%12548 = torch.aten.mul.int %12547, %int2_10185 : !torch.int, !torch.int -> !torch.int
%int16_10186 = torch.constant.int 16
%int1_10187 = torch.constant.int 1
%int128_10188 = torch.constant.int 128
%12549 = torch.prim.ListConstruct %12548, %int16_10186, %int1_10187, %int128_10188 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12550 = torch.aten.view %12546, %12549 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12550, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12551 = torch.prim.ListConstruct %12494 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10189 = torch.constant.bool false
%12552 = torch.aten.index_put %12550, %12551, %12510, %false_10189 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12552, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10190 = torch.constant.int 32
%int2_10191 = torch.constant.int 2
%int16_10192 = torch.constant.int 16
%int1_10193 = torch.constant.int 1
%int128_10194 = torch.constant.int 128
%12553 = torch.prim.ListConstruct %3029, %int32_10190, %int2_10191, %int16_10192, %int1_10193, %int128_10194 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12554 = torch.aten.view %12552, %12553 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12554, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10195 = torch.constant.int 131072
%12555 = torch.prim.ListConstruct %3029, %int131072_10195 : (!torch.int, !torch.int) -> !torch.list<int>
%12556 = torch.aten.view %12554, %12555 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12556, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10196 = torch.constant.int 32
%int2_10197 = torch.constant.int 2
%int16_10198 = torch.constant.int 16
%int1_10199 = torch.constant.int 1
%int128_10200 = torch.constant.int 128
%12557 = torch.prim.ListConstruct %3032, %int32_10196, %int2_10197, %int16_10198, %int1_10199, %int128_10200 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12558 = torch.aten.view %10707, %12557 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12558, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10201 = torch.constant.int 32
%12559 = torch.aten.mul.int %3032, %int32_10201 : !torch.int, !torch.int -> !torch.int
%int2_10202 = torch.constant.int 2
%12560 = torch.aten.mul.int %12559, %int2_10202 : !torch.int, !torch.int -> !torch.int
%int16_10203 = torch.constant.int 16
%int1_10204 = torch.constant.int 1
%int128_10205 = torch.constant.int 128
%12561 = torch.prim.ListConstruct %12560, %int16_10203, %int1_10204, %int128_10205 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12562 = torch.aten.view %12558, %12561 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12562, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12563 = torch.prim.ListConstruct %12496 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10206 = torch.constant.bool false
%12564 = torch.aten.index_put %12562, %12563, %12512, %false_10206 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12564, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10207 = torch.constant.int 32
%int2_10208 = torch.constant.int 2
%int16_10209 = torch.constant.int 16
%int1_10210 = torch.constant.int 1
%int128_10211 = torch.constant.int 128
%12565 = torch.prim.ListConstruct %3032, %int32_10207, %int2_10208, %int16_10209, %int1_10210, %int128_10211 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12566 = torch.aten.view %12564, %12565 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12566, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10212 = torch.constant.int 131072
%12567 = torch.prim.ListConstruct %3032, %int131072_10212 : (!torch.int, !torch.int) -> !torch.list<int>
%12568 = torch.aten.view %12566, %12567 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12568, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10213 = torch.constant.int 32
%int2_10214 = torch.constant.int 2
%int16_10215 = torch.constant.int 16
%int1_10216 = torch.constant.int 1
%int128_10217 = torch.constant.int 128
%12569 = torch.prim.ListConstruct %3035, %int32_10213, %int2_10214, %int16_10215, %int1_10216, %int128_10217 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12570 = torch.aten.view %10719, %12569 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12570, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10218 = torch.constant.int 32
%12571 = torch.aten.mul.int %3035, %int32_10218 : !torch.int, !torch.int -> !torch.int
%int2_10219 = torch.constant.int 2
%12572 = torch.aten.mul.int %12571, %int2_10219 : !torch.int, !torch.int -> !torch.int
%int16_10220 = torch.constant.int 16
%int1_10221 = torch.constant.int 1
%int128_10222 = torch.constant.int 128
%12573 = torch.prim.ListConstruct %12572, %int16_10220, %int1_10221, %int128_10222 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12574 = torch.aten.view %12570, %12573 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12574, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12575 = torch.prim.ListConstruct %12498 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10223 = torch.constant.bool false
%12576 = torch.aten.index_put %12574, %12575, %12514, %false_10223 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12576, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10224 = torch.constant.int 32
%int2_10225 = torch.constant.int 2
%int16_10226 = torch.constant.int 16
%int1_10227 = torch.constant.int 1
%int128_10228 = torch.constant.int 128
%12577 = torch.prim.ListConstruct %3035, %int32_10224, %int2_10225, %int16_10226, %int1_10227, %int128_10228 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12578 = torch.aten.view %12576, %12577 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12578, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10229 = torch.constant.int 131072
%12579 = torch.prim.ListConstruct %3035, %int131072_10229 : (!torch.int, !torch.int) -> !torch.list<int>
%12580 = torch.aten.view %12578, %12579 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12580, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10230 = torch.constant.int 32
%int2_10231 = torch.constant.int 2
%int16_10232 = torch.constant.int 16
%int1_10233 = torch.constant.int 1
%int128_10234 = torch.constant.int 128
%12581 = torch.prim.ListConstruct %3038, %int32_10230, %int2_10231, %int16_10232, %int1_10233, %int128_10234 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12582 = torch.aten.view %10731, %12581 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12582, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10235 = torch.constant.int 32
%12583 = torch.aten.mul.int %3038, %int32_10235 : !torch.int, !torch.int -> !torch.int
%int2_10236 = torch.constant.int 2
%12584 = torch.aten.mul.int %12583, %int2_10236 : !torch.int, !torch.int -> !torch.int
%int16_10237 = torch.constant.int 16
%int1_10238 = torch.constant.int 1
%int128_10239 = torch.constant.int 128
%12585 = torch.prim.ListConstruct %12584, %int16_10237, %int1_10238, %int128_10239 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12586 = torch.aten.view %12582, %12585 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12586, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12587 = torch.prim.ListConstruct %12500 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10240 = torch.constant.bool false
%12588 = torch.aten.index_put %12586, %12587, %12516, %false_10240 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12588, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10241 = torch.constant.int 32
%int2_10242 = torch.constant.int 2
%int16_10243 = torch.constant.int 16
%int1_10244 = torch.constant.int 1
%int128_10245 = torch.constant.int 128
%12589 = torch.prim.ListConstruct %3038, %int32_10241, %int2_10242, %int16_10243, %int1_10244, %int128_10245 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12590 = torch.aten.view %12588, %12589 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12590, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10246 = torch.constant.int 131072
%12591 = torch.prim.ListConstruct %3038, %int131072_10246 : (!torch.int, !torch.int) -> !torch.list<int>
%12592 = torch.aten.view %12590, %12591 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12592, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10247 = torch.constant.int 32
%int2_10248 = torch.constant.int 2
%int16_10249 = torch.constant.int 16
%int1_10250 = torch.constant.int 1
%int128_10251 = torch.constant.int 128
%12593 = torch.prim.ListConstruct %3041, %int32_10247, %int2_10248, %int16_10249, %int1_10250, %int128_10251 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12594 = torch.aten.view %10743, %12593 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12594, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10252 = torch.constant.int 32
%12595 = torch.aten.mul.int %3041, %int32_10252 : !torch.int, !torch.int -> !torch.int
%int2_10253 = torch.constant.int 2
%12596 = torch.aten.mul.int %12595, %int2_10253 : !torch.int, !torch.int -> !torch.int
%int16_10254 = torch.constant.int 16
%int1_10255 = torch.constant.int 1
%int128_10256 = torch.constant.int 128
%12597 = torch.prim.ListConstruct %12596, %int16_10254, %int1_10255, %int128_10256 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12598 = torch.aten.view %12594, %12597 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12598, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12599 = torch.prim.ListConstruct %12502 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10257 = torch.constant.bool false
%12600 = torch.aten.index_put %12598, %12599, %12518, %false_10257 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12600, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10258 = torch.constant.int 32
%int2_10259 = torch.constant.int 2
%int16_10260 = torch.constant.int 16
%int1_10261 = torch.constant.int 1
%int128_10262 = torch.constant.int 128
%12601 = torch.prim.ListConstruct %3041, %int32_10258, %int2_10259, %int16_10260, %int1_10261, %int128_10262 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12602 = torch.aten.view %12600, %12601 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12602, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10263 = torch.constant.int 131072
%12603 = torch.prim.ListConstruct %3041, %int131072_10263 : (!torch.int, !torch.int) -> !torch.list<int>
%12604 = torch.aten.view %12602, %12603 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12604, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_10264 = torch.constant.int 32
%int2_10265 = torch.constant.int 2
%int16_10266 = torch.constant.int 16
%int1_10267 = torch.constant.int 1
%int128_10268 = torch.constant.int 128
%12605 = torch.prim.ListConstruct %3044, %int32_10264, %int2_10265, %int16_10266, %int1_10267, %int128_10268 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12606 = torch.aten.view %10755, %12605 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12606, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_10269 = torch.constant.int 32
%12607 = torch.aten.mul.int %3044, %int32_10269 : !torch.int, !torch.int -> !torch.int
%int2_10270 = torch.constant.int 2
%12608 = torch.aten.mul.int %12607, %int2_10270 : !torch.int, !torch.int -> !torch.int
%int16_10271 = torch.constant.int 16
%int1_10272 = torch.constant.int 1
%int128_10273 = torch.constant.int 128
%12609 = torch.prim.ListConstruct %12608, %int16_10271, %int1_10272, %int128_10273 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12610 = torch.aten.view %12606, %12609 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12610, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%12611 = torch.prim.ListConstruct %12504 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_10274 = torch.constant.bool false
%12612 = torch.aten.index_put %12610, %12611, %12520, %false_10274 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %12612, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_10275 = torch.constant.int 32
%int2_10276 = torch.constant.int 2
%int16_10277 = torch.constant.int 16
%int1_10278 = torch.constant.int 1
%int128_10279 = torch.constant.int 128
%12613 = torch.prim.ListConstruct %3044, %int32_10275, %int2_10276, %int16_10277, %int1_10278, %int128_10279 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12614 = torch.aten.view %12612, %12613 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %12614, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_10280 = torch.constant.int 131072
%12615 = torch.prim.ListConstruct %3044, %int131072_10280 : (!torch.int, !torch.int) -> !torch.list<int>
%12616 = torch.aten.view %12614, %12615 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %12616, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_10281 = torch.constant.int -2
%12617 = torch.aten.unsqueeze %12231, %int-2_10281 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10282 = torch.constant.int -2
%12618 = torch.aten.unsqueeze %12246, %int-2_10282 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10283 = torch.constant.int -2
%12619 = torch.aten.unsqueeze %12261, %int-2_10283 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10284 = torch.constant.int -2
%12620 = torch.aten.unsqueeze %12276, %int-2_10284 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10285 = torch.constant.int -2
%12621 = torch.aten.unsqueeze %12291, %int-2_10285 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10286 = torch.constant.int -2
%12622 = torch.aten.unsqueeze %12306, %int-2_10286 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10287 = torch.constant.int -2
%12623 = torch.aten.unsqueeze %12321, %int-2_10287 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10288 = torch.constant.int -2
%12624 = torch.aten.unsqueeze %12336, %int-2_10288 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_10289 = torch.constant.int 4
%int1_10290 = torch.constant.int 1
%int4_10291 = torch.constant.int 4
%int128_10292 = torch.constant.int 128
%12625 = torch.prim.ListConstruct %int4_10289, %12217, %int1_10290, %int4_10291, %int128_10292 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10293 = torch.constant.bool false
%12626 = torch.aten.expand %12617, %12625, %false_10293 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10294 = torch.constant.int 4
%int1_10295 = torch.constant.int 1
%int4_10296 = torch.constant.int 4
%int128_10297 = torch.constant.int 128
%12627 = torch.prim.ListConstruct %int4_10294, %12217, %int1_10295, %int4_10296, %int128_10297 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10298 = torch.constant.bool false
%12628 = torch.aten.expand %12618, %12627, %false_10298 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10299 = torch.constant.int 4
%int1_10300 = torch.constant.int 1
%int4_10301 = torch.constant.int 4
%int128_10302 = torch.constant.int 128
%12629 = torch.prim.ListConstruct %int4_10299, %12217, %int1_10300, %int4_10301, %int128_10302 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10303 = torch.constant.bool false
%12630 = torch.aten.expand %12619, %12629, %false_10303 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10304 = torch.constant.int 4
%int1_10305 = torch.constant.int 1
%int4_10306 = torch.constant.int 4
%int128_10307 = torch.constant.int 128
%12631 = torch.prim.ListConstruct %int4_10304, %12217, %int1_10305, %int4_10306, %int128_10307 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10308 = torch.constant.bool false
%12632 = torch.aten.expand %12620, %12631, %false_10308 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10309 = torch.constant.int 4
%int1_10310 = torch.constant.int 1
%int4_10311 = torch.constant.int 4
%int128_10312 = torch.constant.int 128
%12633 = torch.prim.ListConstruct %int4_10309, %12217, %int1_10310, %int4_10311, %int128_10312 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10313 = torch.constant.bool false
%12634 = torch.aten.expand %12621, %12633, %false_10313 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10314 = torch.constant.int 4
%int1_10315 = torch.constant.int 1
%int4_10316 = torch.constant.int 4
%int128_10317 = torch.constant.int 128
%12635 = torch.prim.ListConstruct %int4_10314, %12217, %int1_10315, %int4_10316, %int128_10317 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10318 = torch.constant.bool false
%12636 = torch.aten.expand %12622, %12635, %false_10318 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10319 = torch.constant.int 4
%int1_10320 = torch.constant.int 1
%int4_10321 = torch.constant.int 4
%int128_10322 = torch.constant.int 128
%12637 = torch.prim.ListConstruct %int4_10319, %12217, %int1_10320, %int4_10321, %int128_10322 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10323 = torch.constant.bool false
%12638 = torch.aten.expand %12623, %12637, %false_10323 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10324 = torch.constant.int 4
%int1_10325 = torch.constant.int 1
%int4_10326 = torch.constant.int 4
%int128_10327 = torch.constant.int 128
%12639 = torch.prim.ListConstruct %int4_10324, %12217, %int1_10325, %int4_10326, %int128_10327 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10328 = torch.constant.bool false
%12640 = torch.aten.expand %12624, %12639, %false_10328 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10329 = torch.constant.int 4
%int4_10330 = torch.constant.int 4
%int128_10331 = torch.constant.int 128
%12641 = torch.prim.ListConstruct %int4_10329, %12217, %int4_10330, %int128_10331 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12642 = torch.aten.view %12626, %12641 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10332 = torch.constant.int 4
%int4_10333 = torch.constant.int 4
%int128_10334 = torch.constant.int 128
%12643 = torch.prim.ListConstruct %int4_10332, %12217, %int4_10333, %int128_10334 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12644 = torch.aten.view %12628, %12643 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10335 = torch.constant.int 4
%int4_10336 = torch.constant.int 4
%int128_10337 = torch.constant.int 128
%12645 = torch.prim.ListConstruct %int4_10335, %12217, %int4_10336, %int128_10337 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12646 = torch.aten.view %12630, %12645 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10338 = torch.constant.int 4
%int4_10339 = torch.constant.int 4
%int128_10340 = torch.constant.int 128
%12647 = torch.prim.ListConstruct %int4_10338, %12217, %int4_10339, %int128_10340 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12648 = torch.aten.view %12632, %12647 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10341 = torch.constant.int 4
%int4_10342 = torch.constant.int 4
%int128_10343 = torch.constant.int 128
%12649 = torch.prim.ListConstruct %int4_10341, %12217, %int4_10342, %int128_10343 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12650 = torch.aten.view %12634, %12649 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10344 = torch.constant.int 4
%int4_10345 = torch.constant.int 4
%int128_10346 = torch.constant.int 128
%12651 = torch.prim.ListConstruct %int4_10344, %12217, %int4_10345, %int128_10346 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12652 = torch.aten.view %12636, %12651 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10347 = torch.constant.int 4
%int4_10348 = torch.constant.int 4
%int128_10349 = torch.constant.int 128
%12653 = torch.prim.ListConstruct %int4_10347, %12217, %int4_10348, %int128_10349 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12654 = torch.aten.view %12638, %12653 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10350 = torch.constant.int 4
%int4_10351 = torch.constant.int 4
%int128_10352 = torch.constant.int 128
%12655 = torch.prim.ListConstruct %int4_10350, %12217, %int4_10351, %int128_10352 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12656 = torch.aten.view %12640, %12655 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_10353 = torch.constant.int -2
%12657 = torch.aten.unsqueeze %12006, %int-2_10353 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10354 = torch.constant.int -2
%12658 = torch.aten.unsqueeze %12008, %int-2_10354 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10355 = torch.constant.int -2
%12659 = torch.aten.unsqueeze %12010, %int-2_10355 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10356 = torch.constant.int -2
%12660 = torch.aten.unsqueeze %12012, %int-2_10356 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10357 = torch.constant.int -2
%12661 = torch.aten.unsqueeze %12014, %int-2_10357 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10358 = torch.constant.int -2
%12662 = torch.aten.unsqueeze %12016, %int-2_10358 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10359 = torch.constant.int -2
%12663 = torch.aten.unsqueeze %12018, %int-2_10359 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_10360 = torch.constant.int -2
%12664 = torch.aten.unsqueeze %12020, %int-2_10360 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %12664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_10361 = torch.constant.int 1
%12665 = torch.aten.size.int %11930, %int1_10361 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_10362 = torch.constant.int 4
%int1_10363 = torch.constant.int 1
%int4_10364 = torch.constant.int 4
%int128_10365 = torch.constant.int 128
%12666 = torch.prim.ListConstruct %int4_10362, %12665, %int1_10363, %int4_10364, %int128_10365 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10366 = torch.constant.bool false
%12667 = torch.aten.expand %12657, %12666, %false_10366 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10367 = torch.constant.int 4
%int1_10368 = torch.constant.int 1
%int4_10369 = torch.constant.int 4
%int128_10370 = torch.constant.int 128
%12668 = torch.prim.ListConstruct %int4_10367, %12665, %int1_10368, %int4_10369, %int128_10370 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10371 = torch.constant.bool false
%12669 = torch.aten.expand %12658, %12668, %false_10371 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10372 = torch.constant.int 4
%int1_10373 = torch.constant.int 1
%int4_10374 = torch.constant.int 4
%int128_10375 = torch.constant.int 128
%12670 = torch.prim.ListConstruct %int4_10372, %12665, %int1_10373, %int4_10374, %int128_10375 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10376 = torch.constant.bool false
%12671 = torch.aten.expand %12659, %12670, %false_10376 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10377 = torch.constant.int 4
%int1_10378 = torch.constant.int 1
%int4_10379 = torch.constant.int 4
%int128_10380 = torch.constant.int 128
%12672 = torch.prim.ListConstruct %int4_10377, %12665, %int1_10378, %int4_10379, %int128_10380 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10381 = torch.constant.bool false
%12673 = torch.aten.expand %12660, %12672, %false_10381 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10382 = torch.constant.int 4
%int1_10383 = torch.constant.int 1
%int4_10384 = torch.constant.int 4
%int128_10385 = torch.constant.int 128
%12674 = torch.prim.ListConstruct %int4_10382, %12665, %int1_10383, %int4_10384, %int128_10385 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10386 = torch.constant.bool false
%12675 = torch.aten.expand %12661, %12674, %false_10386 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10387 = torch.constant.int 4
%int1_10388 = torch.constant.int 1
%int4_10389 = torch.constant.int 4
%int128_10390 = torch.constant.int 128
%12676 = torch.prim.ListConstruct %int4_10387, %12665, %int1_10388, %int4_10389, %int128_10390 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10391 = torch.constant.bool false
%12677 = torch.aten.expand %12662, %12676, %false_10391 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10392 = torch.constant.int 4
%int1_10393 = torch.constant.int 1
%int4_10394 = torch.constant.int 4
%int128_10395 = torch.constant.int 128
%12678 = torch.prim.ListConstruct %int4_10392, %12665, %int1_10393, %int4_10394, %int128_10395 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10396 = torch.constant.bool false
%12679 = torch.aten.expand %12663, %12678, %false_10396 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10397 = torch.constant.int 4
%int1_10398 = torch.constant.int 1
%int4_10399 = torch.constant.int 4
%int128_10400 = torch.constant.int 128
%12680 = torch.prim.ListConstruct %int4_10397, %12665, %int1_10398, %int4_10399, %int128_10400 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_10401 = torch.constant.bool false
%12681 = torch.aten.expand %12664, %12680, %false_10401 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %12681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_10402 = torch.constant.int 4
%int4_10403 = torch.constant.int 4
%int128_10404 = torch.constant.int 128
%12682 = torch.prim.ListConstruct %int4_10402, %12665, %int4_10403, %int128_10404 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12683 = torch.aten.view %12667, %12682 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10405 = torch.constant.int 4
%int4_10406 = torch.constant.int 4
%int128_10407 = torch.constant.int 128
%12684 = torch.prim.ListConstruct %int4_10405, %12665, %int4_10406, %int128_10407 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12685 = torch.aten.view %12669, %12684 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10408 = torch.constant.int 4
%int4_10409 = torch.constant.int 4
%int128_10410 = torch.constant.int 128
%12686 = torch.prim.ListConstruct %int4_10408, %12665, %int4_10409, %int128_10410 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12687 = torch.aten.view %12671, %12686 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10411 = torch.constant.int 4
%int4_10412 = torch.constant.int 4
%int128_10413 = torch.constant.int 128
%12688 = torch.prim.ListConstruct %int4_10411, %12665, %int4_10412, %int128_10413 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12689 = torch.aten.view %12673, %12688 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10414 = torch.constant.int 4
%int4_10415 = torch.constant.int 4
%int128_10416 = torch.constant.int 128
%12690 = torch.prim.ListConstruct %int4_10414, %12665, %int4_10415, %int128_10416 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12691 = torch.aten.view %12675, %12690 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10417 = torch.constant.int 4
%int4_10418 = torch.constant.int 4
%int128_10419 = torch.constant.int 128
%12692 = torch.prim.ListConstruct %int4_10417, %12665, %int4_10418, %int128_10419 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12693 = torch.aten.view %12677, %12692 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10420 = torch.constant.int 4
%int4_10421 = torch.constant.int 4
%int128_10422 = torch.constant.int 128
%12694 = torch.prim.ListConstruct %int4_10420, %12665, %int4_10421, %int128_10422 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12695 = torch.aten.view %12679, %12694 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10423 = torch.constant.int 4
%int4_10424 = torch.constant.int 4
%int128_10425 = torch.constant.int 128
%12696 = torch.prim.ListConstruct %int4_10423, %12665, %int4_10424, %int128_10425 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12697 = torch.aten.view %12681, %12696 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10426 = torch.constant.int 1
%int2_10427 = torch.constant.int 2
%12698 = torch.aten.transpose.int %12073, %int1_10426, %int2_10427 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12698, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10428 = torch.constant.int 1
%int2_10429 = torch.constant.int 2
%12699 = torch.aten.transpose.int %12088, %int1_10428, %int2_10429 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12699, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10430 = torch.constant.int 1
%int2_10431 = torch.constant.int 2
%12700 = torch.aten.transpose.int %12103, %int1_10430, %int2_10431 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12700, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10432 = torch.constant.int 1
%int2_10433 = torch.constant.int 2
%12701 = torch.aten.transpose.int %12118, %int1_10432, %int2_10433 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12701, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10434 = torch.constant.int 1
%int2_10435 = torch.constant.int 2
%12702 = torch.aten.transpose.int %12133, %int1_10434, %int2_10435 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12702, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10436 = torch.constant.int 1
%int2_10437 = torch.constant.int 2
%12703 = torch.aten.transpose.int %12148, %int1_10436, %int2_10437 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12703, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10438 = torch.constant.int 1
%int2_10439 = torch.constant.int 2
%12704 = torch.aten.transpose.int %12163, %int1_10438, %int2_10439 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12704, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10440 = torch.constant.int 1
%int2_10441 = torch.constant.int 2
%12705 = torch.aten.transpose.int %12178, %int1_10440, %int2_10441 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12705, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10442 = torch.constant.int 1
%int2_10443 = torch.constant.int 2
%12706 = torch.aten.transpose.int %12642, %int1_10442, %int2_10443 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12706, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10444 = torch.constant.int 1
%int2_10445 = torch.constant.int 2
%12707 = torch.aten.transpose.int %12644, %int1_10444, %int2_10445 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12707, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10446 = torch.constant.int 1
%int2_10447 = torch.constant.int 2
%12708 = torch.aten.transpose.int %12646, %int1_10446, %int2_10447 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12708, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10448 = torch.constant.int 1
%int2_10449 = torch.constant.int 2
%12709 = torch.aten.transpose.int %12648, %int1_10448, %int2_10449 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12709, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10450 = torch.constant.int 1
%int2_10451 = torch.constant.int 2
%12710 = torch.aten.transpose.int %12650, %int1_10450, %int2_10451 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12710, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10452 = torch.constant.int 1
%int2_10453 = torch.constant.int 2
%12711 = torch.aten.transpose.int %12652, %int1_10452, %int2_10453 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12711, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10454 = torch.constant.int 1
%int2_10455 = torch.constant.int 2
%12712 = torch.aten.transpose.int %12654, %int1_10454, %int2_10455 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12712, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10456 = torch.constant.int 1
%int2_10457 = torch.constant.int 2
%12713 = torch.aten.transpose.int %12656, %int1_10456, %int2_10457 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12713, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10458 = torch.constant.int 1
%int2_10459 = torch.constant.int 2
%12714 = torch.aten.transpose.int %12683, %int1_10458, %int2_10459 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12714, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10460 = torch.constant.int 1
%int2_10461 = torch.constant.int 2
%12715 = torch.aten.transpose.int %12685, %int1_10460, %int2_10461 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12715, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10462 = torch.constant.int 1
%int2_10463 = torch.constant.int 2
%12716 = torch.aten.transpose.int %12687, %int1_10462, %int2_10463 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12716, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10464 = torch.constant.int 1
%int2_10465 = torch.constant.int 2
%12717 = torch.aten.transpose.int %12689, %int1_10464, %int2_10465 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12717, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10466 = torch.constant.int 1
%int2_10467 = torch.constant.int 2
%12718 = torch.aten.transpose.int %12691, %int1_10466, %int2_10467 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12718, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10468 = torch.constant.int 1
%int2_10469 = torch.constant.int 2
%12719 = torch.aten.transpose.int %12693, %int1_10468, %int2_10469 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12719, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10470 = torch.constant.int 1
%int2_10471 = torch.constant.int 2
%12720 = torch.aten.transpose.int %12695, %int1_10470, %int2_10471 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12720, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10472 = torch.constant.int 1
%int2_10473 = torch.constant.int 2
%12721 = torch.aten.transpose.int %12697, %int1_10472, %int2_10473 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %12721, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10474 = torch.constant.float 0.000000e+00
%true_10475 = torch.constant.bool true
%none_10476 = torch.constant.none
%none_10477 = torch.constant.none
%12722:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12698, %12706, %12714, %float0.000000e00_10474, %true_10475, %none_10476, %none_10477) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12722#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10478 = torch.constant.float 0.000000e+00
%true_10479 = torch.constant.bool true
%none_10480 = torch.constant.none
%none_10481 = torch.constant.none
%12723:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12699, %12707, %12715, %float0.000000e00_10478, %true_10479, %none_10480, %none_10481) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12723#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10482 = torch.constant.float 0.000000e+00
%true_10483 = torch.constant.bool true
%none_10484 = torch.constant.none
%none_10485 = torch.constant.none
%12724:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12700, %12708, %12716, %float0.000000e00_10482, %true_10483, %none_10484, %none_10485) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12724#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10486 = torch.constant.float 0.000000e+00
%true_10487 = torch.constant.bool true
%none_10488 = torch.constant.none
%none_10489 = torch.constant.none
%12725:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12701, %12709, %12717, %float0.000000e00_10486, %true_10487, %none_10488, %none_10489) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12725#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10490 = torch.constant.float 0.000000e+00
%true_10491 = torch.constant.bool true
%none_10492 = torch.constant.none
%none_10493 = torch.constant.none
%12726:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12702, %12710, %12718, %float0.000000e00_10490, %true_10491, %none_10492, %none_10493) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12726#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10494 = torch.constant.float 0.000000e+00
%true_10495 = torch.constant.bool true
%none_10496 = torch.constant.none
%none_10497 = torch.constant.none
%12727:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12703, %12711, %12719, %float0.000000e00_10494, %true_10495, %none_10496, %none_10497) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12727#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10498 = torch.constant.float 0.000000e+00
%true_10499 = torch.constant.bool true
%none_10500 = torch.constant.none
%none_10501 = torch.constant.none
%12728:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12704, %12712, %12720, %float0.000000e00_10498, %true_10499, %none_10500, %none_10501) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12728#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_10502 = torch.constant.float 0.000000e+00
%true_10503 = torch.constant.bool true
%none_10504 = torch.constant.none
%none_10505 = torch.constant.none
%12729:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%12705, %12713, %12721, %float0.000000e00_10502, %true_10503, %none_10504, %none_10505) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %12729#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_10506 = torch.constant.int 1
%int2_10507 = torch.constant.int 2
%12730 = torch.aten.transpose.int %12722#0, %int1_10506, %int2_10507 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10508 = torch.constant.int 1
%int2_10509 = torch.constant.int 2
%12731 = torch.aten.transpose.int %12723#0, %int1_10508, %int2_10509 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10510 = torch.constant.int 1
%int2_10511 = torch.constant.int 2
%12732 = torch.aten.transpose.int %12724#0, %int1_10510, %int2_10511 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10512 = torch.constant.int 1
%int2_10513 = torch.constant.int 2
%12733 = torch.aten.transpose.int %12725#0, %int1_10512, %int2_10513 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10514 = torch.constant.int 1
%int2_10515 = torch.constant.int 2
%12734 = torch.aten.transpose.int %12726#0, %int1_10514, %int2_10515 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10516 = torch.constant.int 1
%int2_10517 = torch.constant.int 2
%12735 = torch.aten.transpose.int %12727#0, %int1_10516, %int2_10517 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10518 = torch.constant.int 1
%int2_10519 = torch.constant.int 2
%12736 = torch.aten.transpose.int %12728#0, %int1_10518, %int2_10519 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_10520 = torch.constant.int 1
%int2_10521 = torch.constant.int 2
%12737 = torch.aten.transpose.int %12729#0, %int1_10520, %int2_10521 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %12737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_10522 = torch.constant.int 4
%int512_10523 = torch.constant.int 512
%12738 = torch.prim.ListConstruct %int4_10522, %12059, %int512_10523 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12739 = torch.aten.view %12730, %12738 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10524 = torch.constant.int 4
%int512_10525 = torch.constant.int 512
%12740 = torch.prim.ListConstruct %int4_10524, %12074, %int512_10525 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12741 = torch.aten.view %12731, %12740 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10526 = torch.constant.int 4
%int512_10527 = torch.constant.int 512
%12742 = torch.prim.ListConstruct %int4_10526, %12089, %int512_10527 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12743 = torch.aten.view %12732, %12742 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10528 = torch.constant.int 4
%int512_10529 = torch.constant.int 512
%12744 = torch.prim.ListConstruct %int4_10528, %12104, %int512_10529 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12745 = torch.aten.view %12733, %12744 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10530 = torch.constant.int 4
%int512_10531 = torch.constant.int 512
%12746 = torch.prim.ListConstruct %int4_10530, %12119, %int512_10531 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12747 = torch.aten.view %12734, %12746 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10532 = torch.constant.int 4
%int512_10533 = torch.constant.int 512
%12748 = torch.prim.ListConstruct %int4_10532, %12134, %int512_10533 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12749 = torch.aten.view %12735, %12748 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10534 = torch.constant.int 4
%int512_10535 = torch.constant.int 512
%12750 = torch.prim.ListConstruct %int4_10534, %12149, %int512_10535 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12751 = torch.aten.view %12736, %12750 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_10536 = torch.constant.int 4
%int512_10537 = torch.constant.int 512
%12752 = torch.prim.ListConstruct %int4_10536, %12164, %int512_10537 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12753 = torch.aten.view %12737, %12752 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %12753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_10538 = torch.constant.int 1
%int0_10539 = torch.constant.int 0
%12754 = torch.prim.ListConstruct %int1_10538, %int0_10539 : (!torch.int, !torch.int) -> !torch.list<int>
%12755 = torch.aten.permute %400, %12754 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10540 = torch.constant.int 1
%int0_10541 = torch.constant.int 0
%12756 = torch.prim.ListConstruct %int1_10540, %int0_10541 : (!torch.int, !torch.int) -> !torch.list<int>
%12757 = torch.aten.permute %401, %12756 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10542 = torch.constant.int 1
%int0_10543 = torch.constant.int 0
%12758 = torch.prim.ListConstruct %int1_10542, %int0_10543 : (!torch.int, !torch.int) -> !torch.list<int>
%12759 = torch.aten.permute %402, %12758 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10544 = torch.constant.int 1
%int0_10545 = torch.constant.int 0
%12760 = torch.prim.ListConstruct %int1_10544, %int0_10545 : (!torch.int, !torch.int) -> !torch.list<int>
%12761 = torch.aten.permute %403, %12760 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10546 = torch.constant.int 1
%int0_10547 = torch.constant.int 0
%12762 = torch.prim.ListConstruct %int1_10546, %int0_10547 : (!torch.int, !torch.int) -> !torch.list<int>
%12763 = torch.aten.permute %404, %12762 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10548 = torch.constant.int 1
%int0_10549 = torch.constant.int 0
%12764 = torch.prim.ListConstruct %int1_10548, %int0_10549 : (!torch.int, !torch.int) -> !torch.list<int>
%12765 = torch.aten.permute %405, %12764 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10550 = torch.constant.int 1
%int0_10551 = torch.constant.int 0
%12766 = torch.prim.ListConstruct %int1_10550, %int0_10551 : (!torch.int, !torch.int) -> !torch.list<int>
%12767 = torch.aten.permute %406, %12766 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_10552 = torch.constant.int 1
%int0_10553 = torch.constant.int 0
%12768 = torch.prim.ListConstruct %int1_10552, %int0_10553 : (!torch.int, !torch.int) -> !torch.list<int>
%12769 = torch.aten.permute %407, %12768 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_10554 = torch.constant.int 4
%12770 = torch.aten.mul.int %int4_10554, %12059 : !torch.int, !torch.int -> !torch.int
%int512_10555 = torch.constant.int 512
%12771 = torch.prim.ListConstruct %12770, %int512_10555 : (!torch.int, !torch.int) -> !torch.list<int>
%12772 = torch.aten.view %12739, %12771 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12772, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12773 = torch.aten.mm %12772, %12755 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12773, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10556 = torch.constant.int 4
%int4096_10557 = torch.constant.int 4096
%12774 = torch.prim.ListConstruct %int4_10556, %12059, %int4096_10557 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12775 = torch.aten.view %12773, %12774 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10558 = torch.constant.int 4
%12776 = torch.aten.mul.int %int4_10558, %12074 : !torch.int, !torch.int -> !torch.int
%int512_10559 = torch.constant.int 512
%12777 = torch.prim.ListConstruct %12776, %int512_10559 : (!torch.int, !torch.int) -> !torch.list<int>
%12778 = torch.aten.view %12741, %12777 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12778, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12779 = torch.aten.mm %12778, %12757 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12779, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10560 = torch.constant.int 4
%int4096_10561 = torch.constant.int 4096
%12780 = torch.prim.ListConstruct %int4_10560, %12074, %int4096_10561 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12781 = torch.aten.view %12779, %12780 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10562 = torch.constant.int 4
%12782 = torch.aten.mul.int %int4_10562, %12089 : !torch.int, !torch.int -> !torch.int
%int512_10563 = torch.constant.int 512
%12783 = torch.prim.ListConstruct %12782, %int512_10563 : (!torch.int, !torch.int) -> !torch.list<int>
%12784 = torch.aten.view %12743, %12783 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12784, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12785 = torch.aten.mm %12784, %12759 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12785, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10564 = torch.constant.int 4
%int4096_10565 = torch.constant.int 4096
%12786 = torch.prim.ListConstruct %int4_10564, %12089, %int4096_10565 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12787 = torch.aten.view %12785, %12786 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10566 = torch.constant.int 4
%12788 = torch.aten.mul.int %int4_10566, %12104 : !torch.int, !torch.int -> !torch.int
%int512_10567 = torch.constant.int 512
%12789 = torch.prim.ListConstruct %12788, %int512_10567 : (!torch.int, !torch.int) -> !torch.list<int>
%12790 = torch.aten.view %12745, %12789 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12790, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12791 = torch.aten.mm %12790, %12761 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12791, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10568 = torch.constant.int 4
%int4096_10569 = torch.constant.int 4096
%12792 = torch.prim.ListConstruct %int4_10568, %12104, %int4096_10569 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12793 = torch.aten.view %12791, %12792 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10570 = torch.constant.int 4
%12794 = torch.aten.mul.int %int4_10570, %12119 : !torch.int, !torch.int -> !torch.int
%int512_10571 = torch.constant.int 512
%12795 = torch.prim.ListConstruct %12794, %int512_10571 : (!torch.int, !torch.int) -> !torch.list<int>
%12796 = torch.aten.view %12747, %12795 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12796, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12797 = torch.aten.mm %12796, %12763 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12797, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10572 = torch.constant.int 4
%int4096_10573 = torch.constant.int 4096
%12798 = torch.prim.ListConstruct %int4_10572, %12119, %int4096_10573 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12799 = torch.aten.view %12797, %12798 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10574 = torch.constant.int 4
%12800 = torch.aten.mul.int %int4_10574, %12134 : !torch.int, !torch.int -> !torch.int
%int512_10575 = torch.constant.int 512
%12801 = torch.prim.ListConstruct %12800, %int512_10575 : (!torch.int, !torch.int) -> !torch.list<int>
%12802 = torch.aten.view %12749, %12801 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12802, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12803 = torch.aten.mm %12802, %12765 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12803, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10576 = torch.constant.int 4
%int4096_10577 = torch.constant.int 4096
%12804 = torch.prim.ListConstruct %int4_10576, %12134, %int4096_10577 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12805 = torch.aten.view %12803, %12804 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10578 = torch.constant.int 4
%12806 = torch.aten.mul.int %int4_10578, %12149 : !torch.int, !torch.int -> !torch.int
%int512_10579 = torch.constant.int 512
%12807 = torch.prim.ListConstruct %12806, %int512_10579 : (!torch.int, !torch.int) -> !torch.list<int>
%12808 = torch.aten.view %12751, %12807 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12808, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12809 = torch.aten.mm %12808, %12767 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12809, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10580 = torch.constant.int 4
%int4096_10581 = torch.constant.int 4096
%12810 = torch.prim.ListConstruct %int4_10580, %12149, %int4096_10581 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12811 = torch.aten.view %12809, %12810 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_10582 = torch.constant.int 4
%12812 = torch.aten.mul.int %int4_10582, %12164 : !torch.int, !torch.int -> !torch.int
%int512_10583 = torch.constant.int 512
%12813 = torch.prim.ListConstruct %12812, %int512_10583 : (!torch.int, !torch.int) -> !torch.list<int>
%12814 = torch.aten.view %12753, %12813 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %12814, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%12815 = torch.aten.mm %12814, %12769 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %12815, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10584 = torch.constant.int 4
%int4096_10585 = torch.constant.int 4096
%12816 = torch.prim.ListConstruct %int4_10584, %12164, %int4096_10585 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%12817 = torch.aten.view %12815, %12816 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12818 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10586 = arith.constant 1 : index
%dim_10587 = tensor.dim %12818, %c1_10586 : tensor<4x?x4096xf16>
%12819 = flow.tensor.transfer %12818 : tensor<4x?x4096xf16>{%dim_10587} to #hal.device.promise<@__device_0>
%12820 = torch_c.from_builtin_tensor %12819 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12821 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10588 = arith.constant 1 : index
%dim_10589 = tensor.dim %12821, %c1_10588 : tensor<4x?x4096xf16>
%12822 = flow.tensor.transfer %12821 : tensor<4x?x4096xf16>{%dim_10589} to #hal.device.promise<@__device_0>
%12823 = torch_c.from_builtin_tensor %12822 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12824 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10590 = arith.constant 1 : index
%dim_10591 = tensor.dim %12824, %c1_10590 : tensor<4x?x4096xf16>
%12825 = flow.tensor.transfer %12824 : tensor<4x?x4096xf16>{%dim_10591} to #hal.device.promise<@__device_0>
%12826 = torch_c.from_builtin_tensor %12825 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12827 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10592 = arith.constant 1 : index
%dim_10593 = tensor.dim %12827, %c1_10592 : tensor<4x?x4096xf16>
%12828 = flow.tensor.transfer %12827 : tensor<4x?x4096xf16>{%dim_10593} to #hal.device.promise<@__device_0>
%12829 = torch_c.from_builtin_tensor %12828 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12830 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10594 = arith.constant 1 : index
%dim_10595 = tensor.dim %12830, %c1_10594 : tensor<4x?x4096xf16>
%12831 = flow.tensor.transfer %12830 : tensor<4x?x4096xf16>{%dim_10595} to #hal.device.promise<@__device_0>
%12832 = torch_c.from_builtin_tensor %12831 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12833 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10596 = arith.constant 1 : index
%dim_10597 = tensor.dim %12833, %c1_10596 : tensor<4x?x4096xf16>
%12834 = flow.tensor.transfer %12833 : tensor<4x?x4096xf16>{%dim_10597} to #hal.device.promise<@__device_0>
%12835 = torch_c.from_builtin_tensor %12834 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12836 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10598 = arith.constant 1 : index
%dim_10599 = tensor.dim %12836, %c1_10598 : tensor<4x?x4096xf16>
%12837 = flow.tensor.transfer %12836 : tensor<4x?x4096xf16>{%dim_10599} to #hal.device.promise<@__device_0>
%12838 = torch_c.from_builtin_tensor %12837 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10600 = torch.constant.int 1
%12839 = torch.aten.add.Tensor %12775, %12820, %int1_10600 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10601 = torch.constant.int 1
%12840 = torch.aten.add.Tensor %12839, %12823, %int1_10601 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10602 = torch.constant.int 1
%12841 = torch.aten.add.Tensor %12840, %12826, %int1_10602 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10603 = torch.constant.int 1
%12842 = torch.aten.add.Tensor %12841, %12829, %int1_10603 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10604 = torch.constant.int 1
%12843 = torch.aten.add.Tensor %12842, %12832, %int1_10604 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10605 = torch.constant.int 1
%12844 = torch.aten.add.Tensor %12843, %12835, %int1_10605 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10606 = torch.constant.int 1
%12845 = torch.aten.add.Tensor %12844, %12838, %int1_10606 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12846 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10607 = arith.constant 1 : index
%dim_10608 = tensor.dim %12846, %c1_10607 : tensor<4x?x4096xf16>
%12847 = flow.tensor.transfer %12846 : tensor<4x?x4096xf16>{%dim_10608} to #hal.device.promise<@__device_1>
%12848 = torch_c.from_builtin_tensor %12847 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12849 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10609 = arith.constant 1 : index
%dim_10610 = tensor.dim %12849, %c1_10609 : tensor<4x?x4096xf16>
%12850 = flow.tensor.transfer %12849 : tensor<4x?x4096xf16>{%dim_10610} to #hal.device.promise<@__device_1>
%12851 = torch_c.from_builtin_tensor %12850 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12852 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10611 = arith.constant 1 : index
%dim_10612 = tensor.dim %12852, %c1_10611 : tensor<4x?x4096xf16>
%12853 = flow.tensor.transfer %12852 : tensor<4x?x4096xf16>{%dim_10612} to #hal.device.promise<@__device_1>
%12854 = torch_c.from_builtin_tensor %12853 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12855 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10613 = arith.constant 1 : index
%dim_10614 = tensor.dim %12855, %c1_10613 : tensor<4x?x4096xf16>
%12856 = flow.tensor.transfer %12855 : tensor<4x?x4096xf16>{%dim_10614} to #hal.device.promise<@__device_1>
%12857 = torch_c.from_builtin_tensor %12856 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12858 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10615 = arith.constant 1 : index
%dim_10616 = tensor.dim %12858, %c1_10615 : tensor<4x?x4096xf16>
%12859 = flow.tensor.transfer %12858 : tensor<4x?x4096xf16>{%dim_10616} to #hal.device.promise<@__device_1>
%12860 = torch_c.from_builtin_tensor %12859 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12861 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10617 = arith.constant 1 : index
%dim_10618 = tensor.dim %12861, %c1_10617 : tensor<4x?x4096xf16>
%12862 = flow.tensor.transfer %12861 : tensor<4x?x4096xf16>{%dim_10618} to #hal.device.promise<@__device_1>
%12863 = torch_c.from_builtin_tensor %12862 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12864 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10619 = arith.constant 1 : index
%dim_10620 = tensor.dim %12864, %c1_10619 : tensor<4x?x4096xf16>
%12865 = flow.tensor.transfer %12864 : tensor<4x?x4096xf16>{%dim_10620} to #hal.device.promise<@__device_1>
%12866 = torch_c.from_builtin_tensor %12865 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10621 = torch.constant.int 1
%12867 = torch.aten.add.Tensor %12848, %12781, %int1_10621 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10622 = torch.constant.int 1
%12868 = torch.aten.add.Tensor %12867, %12851, %int1_10622 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10623 = torch.constant.int 1
%12869 = torch.aten.add.Tensor %12868, %12854, %int1_10623 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10624 = torch.constant.int 1
%12870 = torch.aten.add.Tensor %12869, %12857, %int1_10624 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10625 = torch.constant.int 1
%12871 = torch.aten.add.Tensor %12870, %12860, %int1_10625 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10626 = torch.constant.int 1
%12872 = torch.aten.add.Tensor %12871, %12863, %int1_10626 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10627 = torch.constant.int 1
%12873 = torch.aten.add.Tensor %12872, %12866, %int1_10627 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12874 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10628 = arith.constant 1 : index
%dim_10629 = tensor.dim %12874, %c1_10628 : tensor<4x?x4096xf16>
%12875 = flow.tensor.transfer %12874 : tensor<4x?x4096xf16>{%dim_10629} to #hal.device.promise<@__device_2>
%12876 = torch_c.from_builtin_tensor %12875 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12877 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10630 = arith.constant 1 : index
%dim_10631 = tensor.dim %12877, %c1_10630 : tensor<4x?x4096xf16>
%12878 = flow.tensor.transfer %12877 : tensor<4x?x4096xf16>{%dim_10631} to #hal.device.promise<@__device_2>
%12879 = torch_c.from_builtin_tensor %12878 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12880 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10632 = arith.constant 1 : index
%dim_10633 = tensor.dim %12880, %c1_10632 : tensor<4x?x4096xf16>
%12881 = flow.tensor.transfer %12880 : tensor<4x?x4096xf16>{%dim_10633} to #hal.device.promise<@__device_2>
%12882 = torch_c.from_builtin_tensor %12881 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12883 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10634 = arith.constant 1 : index
%dim_10635 = tensor.dim %12883, %c1_10634 : tensor<4x?x4096xf16>
%12884 = flow.tensor.transfer %12883 : tensor<4x?x4096xf16>{%dim_10635} to #hal.device.promise<@__device_2>
%12885 = torch_c.from_builtin_tensor %12884 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12886 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10636 = arith.constant 1 : index
%dim_10637 = tensor.dim %12886, %c1_10636 : tensor<4x?x4096xf16>
%12887 = flow.tensor.transfer %12886 : tensor<4x?x4096xf16>{%dim_10637} to #hal.device.promise<@__device_2>
%12888 = torch_c.from_builtin_tensor %12887 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12889 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10638 = arith.constant 1 : index
%dim_10639 = tensor.dim %12889, %c1_10638 : tensor<4x?x4096xf16>
%12890 = flow.tensor.transfer %12889 : tensor<4x?x4096xf16>{%dim_10639} to #hal.device.promise<@__device_2>
%12891 = torch_c.from_builtin_tensor %12890 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12892 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10640 = arith.constant 1 : index
%dim_10641 = tensor.dim %12892, %c1_10640 : tensor<4x?x4096xf16>
%12893 = flow.tensor.transfer %12892 : tensor<4x?x4096xf16>{%dim_10641} to #hal.device.promise<@__device_2>
%12894 = torch_c.from_builtin_tensor %12893 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10642 = torch.constant.int 1
%12895 = torch.aten.add.Tensor %12876, %12879, %int1_10642 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10643 = torch.constant.int 1
%12896 = torch.aten.add.Tensor %12895, %12787, %int1_10643 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10644 = torch.constant.int 1
%12897 = torch.aten.add.Tensor %12896, %12882, %int1_10644 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10645 = torch.constant.int 1
%12898 = torch.aten.add.Tensor %12897, %12885, %int1_10645 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10646 = torch.constant.int 1
%12899 = torch.aten.add.Tensor %12898, %12888, %int1_10646 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10647 = torch.constant.int 1
%12900 = torch.aten.add.Tensor %12899, %12891, %int1_10647 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10648 = torch.constant.int 1
%12901 = torch.aten.add.Tensor %12900, %12894, %int1_10648 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12902 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10649 = arith.constant 1 : index
%dim_10650 = tensor.dim %12902, %c1_10649 : tensor<4x?x4096xf16>
%12903 = flow.tensor.transfer %12902 : tensor<4x?x4096xf16>{%dim_10650} to #hal.device.promise<@__device_3>
%12904 = torch_c.from_builtin_tensor %12903 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12905 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10651 = arith.constant 1 : index
%dim_10652 = tensor.dim %12905, %c1_10651 : tensor<4x?x4096xf16>
%12906 = flow.tensor.transfer %12905 : tensor<4x?x4096xf16>{%dim_10652} to #hal.device.promise<@__device_3>
%12907 = torch_c.from_builtin_tensor %12906 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12908 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10653 = arith.constant 1 : index
%dim_10654 = tensor.dim %12908, %c1_10653 : tensor<4x?x4096xf16>
%12909 = flow.tensor.transfer %12908 : tensor<4x?x4096xf16>{%dim_10654} to #hal.device.promise<@__device_3>
%12910 = torch_c.from_builtin_tensor %12909 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12911 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10655 = arith.constant 1 : index
%dim_10656 = tensor.dim %12911, %c1_10655 : tensor<4x?x4096xf16>
%12912 = flow.tensor.transfer %12911 : tensor<4x?x4096xf16>{%dim_10656} to #hal.device.promise<@__device_3>
%12913 = torch_c.from_builtin_tensor %12912 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12914 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10657 = arith.constant 1 : index
%dim_10658 = tensor.dim %12914, %c1_10657 : tensor<4x?x4096xf16>
%12915 = flow.tensor.transfer %12914 : tensor<4x?x4096xf16>{%dim_10658} to #hal.device.promise<@__device_3>
%12916 = torch_c.from_builtin_tensor %12915 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12917 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10659 = arith.constant 1 : index
%dim_10660 = tensor.dim %12917, %c1_10659 : tensor<4x?x4096xf16>
%12918 = flow.tensor.transfer %12917 : tensor<4x?x4096xf16>{%dim_10660} to #hal.device.promise<@__device_3>
%12919 = torch_c.from_builtin_tensor %12918 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12920 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10661 = arith.constant 1 : index
%dim_10662 = tensor.dim %12920, %c1_10661 : tensor<4x?x4096xf16>
%12921 = flow.tensor.transfer %12920 : tensor<4x?x4096xf16>{%dim_10662} to #hal.device.promise<@__device_3>
%12922 = torch_c.from_builtin_tensor %12921 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10663 = torch.constant.int 1
%12923 = torch.aten.add.Tensor %12904, %12907, %int1_10663 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10664 = torch.constant.int 1
%12924 = torch.aten.add.Tensor %12923, %12910, %int1_10664 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10665 = torch.constant.int 1
%12925 = torch.aten.add.Tensor %12924, %12793, %int1_10665 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10666 = torch.constant.int 1
%12926 = torch.aten.add.Tensor %12925, %12913, %int1_10666 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10667 = torch.constant.int 1
%12927 = torch.aten.add.Tensor %12926, %12916, %int1_10667 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10668 = torch.constant.int 1
%12928 = torch.aten.add.Tensor %12927, %12919, %int1_10668 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10669 = torch.constant.int 1
%12929 = torch.aten.add.Tensor %12928, %12922, %int1_10669 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12930 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10670 = arith.constant 1 : index
%dim_10671 = tensor.dim %12930, %c1_10670 : tensor<4x?x4096xf16>
%12931 = flow.tensor.transfer %12930 : tensor<4x?x4096xf16>{%dim_10671} to #hal.device.promise<@__device_4>
%12932 = torch_c.from_builtin_tensor %12931 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12933 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10672 = arith.constant 1 : index
%dim_10673 = tensor.dim %12933, %c1_10672 : tensor<4x?x4096xf16>
%12934 = flow.tensor.transfer %12933 : tensor<4x?x4096xf16>{%dim_10673} to #hal.device.promise<@__device_4>
%12935 = torch_c.from_builtin_tensor %12934 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12936 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10674 = arith.constant 1 : index
%dim_10675 = tensor.dim %12936, %c1_10674 : tensor<4x?x4096xf16>
%12937 = flow.tensor.transfer %12936 : tensor<4x?x4096xf16>{%dim_10675} to #hal.device.promise<@__device_4>
%12938 = torch_c.from_builtin_tensor %12937 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12939 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10676 = arith.constant 1 : index
%dim_10677 = tensor.dim %12939, %c1_10676 : tensor<4x?x4096xf16>
%12940 = flow.tensor.transfer %12939 : tensor<4x?x4096xf16>{%dim_10677} to #hal.device.promise<@__device_4>
%12941 = torch_c.from_builtin_tensor %12940 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12942 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10678 = arith.constant 1 : index
%dim_10679 = tensor.dim %12942, %c1_10678 : tensor<4x?x4096xf16>
%12943 = flow.tensor.transfer %12942 : tensor<4x?x4096xf16>{%dim_10679} to #hal.device.promise<@__device_4>
%12944 = torch_c.from_builtin_tensor %12943 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12945 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10680 = arith.constant 1 : index
%dim_10681 = tensor.dim %12945, %c1_10680 : tensor<4x?x4096xf16>
%12946 = flow.tensor.transfer %12945 : tensor<4x?x4096xf16>{%dim_10681} to #hal.device.promise<@__device_4>
%12947 = torch_c.from_builtin_tensor %12946 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12948 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10682 = arith.constant 1 : index
%dim_10683 = tensor.dim %12948, %c1_10682 : tensor<4x?x4096xf16>
%12949 = flow.tensor.transfer %12948 : tensor<4x?x4096xf16>{%dim_10683} to #hal.device.promise<@__device_4>
%12950 = torch_c.from_builtin_tensor %12949 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10684 = torch.constant.int 1
%12951 = torch.aten.add.Tensor %12932, %12935, %int1_10684 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10685 = torch.constant.int 1
%12952 = torch.aten.add.Tensor %12951, %12938, %int1_10685 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10686 = torch.constant.int 1
%12953 = torch.aten.add.Tensor %12952, %12941, %int1_10686 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10687 = torch.constant.int 1
%12954 = torch.aten.add.Tensor %12953, %12799, %int1_10687 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10688 = torch.constant.int 1
%12955 = torch.aten.add.Tensor %12954, %12944, %int1_10688 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10689 = torch.constant.int 1
%12956 = torch.aten.add.Tensor %12955, %12947, %int1_10689 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10690 = torch.constant.int 1
%12957 = torch.aten.add.Tensor %12956, %12950, %int1_10690 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12958 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10691 = arith.constant 1 : index
%dim_10692 = tensor.dim %12958, %c1_10691 : tensor<4x?x4096xf16>
%12959 = flow.tensor.transfer %12958 : tensor<4x?x4096xf16>{%dim_10692} to #hal.device.promise<@__device_5>
%12960 = torch_c.from_builtin_tensor %12959 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12961 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10693 = arith.constant 1 : index
%dim_10694 = tensor.dim %12961, %c1_10693 : tensor<4x?x4096xf16>
%12962 = flow.tensor.transfer %12961 : tensor<4x?x4096xf16>{%dim_10694} to #hal.device.promise<@__device_5>
%12963 = torch_c.from_builtin_tensor %12962 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12964 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10695 = arith.constant 1 : index
%dim_10696 = tensor.dim %12964, %c1_10695 : tensor<4x?x4096xf16>
%12965 = flow.tensor.transfer %12964 : tensor<4x?x4096xf16>{%dim_10696} to #hal.device.promise<@__device_5>
%12966 = torch_c.from_builtin_tensor %12965 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12967 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10697 = arith.constant 1 : index
%dim_10698 = tensor.dim %12967, %c1_10697 : tensor<4x?x4096xf16>
%12968 = flow.tensor.transfer %12967 : tensor<4x?x4096xf16>{%dim_10698} to #hal.device.promise<@__device_5>
%12969 = torch_c.from_builtin_tensor %12968 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12970 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10699 = arith.constant 1 : index
%dim_10700 = tensor.dim %12970, %c1_10699 : tensor<4x?x4096xf16>
%12971 = flow.tensor.transfer %12970 : tensor<4x?x4096xf16>{%dim_10700} to #hal.device.promise<@__device_5>
%12972 = torch_c.from_builtin_tensor %12971 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12973 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10701 = arith.constant 1 : index
%dim_10702 = tensor.dim %12973, %c1_10701 : tensor<4x?x4096xf16>
%12974 = flow.tensor.transfer %12973 : tensor<4x?x4096xf16>{%dim_10702} to #hal.device.promise<@__device_5>
%12975 = torch_c.from_builtin_tensor %12974 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12976 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10703 = arith.constant 1 : index
%dim_10704 = tensor.dim %12976, %c1_10703 : tensor<4x?x4096xf16>
%12977 = flow.tensor.transfer %12976 : tensor<4x?x4096xf16>{%dim_10704} to #hal.device.promise<@__device_5>
%12978 = torch_c.from_builtin_tensor %12977 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10705 = torch.constant.int 1
%12979 = torch.aten.add.Tensor %12960, %12963, %int1_10705 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10706 = torch.constant.int 1
%12980 = torch.aten.add.Tensor %12979, %12966, %int1_10706 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10707 = torch.constant.int 1
%12981 = torch.aten.add.Tensor %12980, %12969, %int1_10707 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10708 = torch.constant.int 1
%12982 = torch.aten.add.Tensor %12981, %12972, %int1_10708 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10709 = torch.constant.int 1
%12983 = torch.aten.add.Tensor %12982, %12805, %int1_10709 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10710 = torch.constant.int 1
%12984 = torch.aten.add.Tensor %12983, %12975, %int1_10710 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10711 = torch.constant.int 1
%12985 = torch.aten.add.Tensor %12984, %12978, %int1_10711 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12985, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12986 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10712 = arith.constant 1 : index
%dim_10713 = tensor.dim %12986, %c1_10712 : tensor<4x?x4096xf16>
%12987 = flow.tensor.transfer %12986 : tensor<4x?x4096xf16>{%dim_10713} to #hal.device.promise<@__device_6>
%12988 = torch_c.from_builtin_tensor %12987 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12989 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10714 = arith.constant 1 : index
%dim_10715 = tensor.dim %12989, %c1_10714 : tensor<4x?x4096xf16>
%12990 = flow.tensor.transfer %12989 : tensor<4x?x4096xf16>{%dim_10715} to #hal.device.promise<@__device_6>
%12991 = torch_c.from_builtin_tensor %12990 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12992 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10716 = arith.constant 1 : index
%dim_10717 = tensor.dim %12992, %c1_10716 : tensor<4x?x4096xf16>
%12993 = flow.tensor.transfer %12992 : tensor<4x?x4096xf16>{%dim_10717} to #hal.device.promise<@__device_6>
%12994 = torch_c.from_builtin_tensor %12993 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12995 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10718 = arith.constant 1 : index
%dim_10719 = tensor.dim %12995, %c1_10718 : tensor<4x?x4096xf16>
%12996 = flow.tensor.transfer %12995 : tensor<4x?x4096xf16>{%dim_10719} to #hal.device.promise<@__device_6>
%12997 = torch_c.from_builtin_tensor %12996 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %12997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%12998 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10720 = arith.constant 1 : index
%dim_10721 = tensor.dim %12998, %c1_10720 : tensor<4x?x4096xf16>
%12999 = flow.tensor.transfer %12998 : tensor<4x?x4096xf16>{%dim_10721} to #hal.device.promise<@__device_6>
%13000 = torch_c.from_builtin_tensor %12999 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13001 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10722 = arith.constant 1 : index
%dim_10723 = tensor.dim %13001, %c1_10722 : tensor<4x?x4096xf16>
%13002 = flow.tensor.transfer %13001 : tensor<4x?x4096xf16>{%dim_10723} to #hal.device.promise<@__device_6>
%13003 = torch_c.from_builtin_tensor %13002 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13004 = torch_c.to_builtin_tensor %12817 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10724 = arith.constant 1 : index
%dim_10725 = tensor.dim %13004, %c1_10724 : tensor<4x?x4096xf16>
%13005 = flow.tensor.transfer %13004 : tensor<4x?x4096xf16>{%dim_10725} to #hal.device.promise<@__device_6>
%13006 = torch_c.from_builtin_tensor %13005 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10726 = torch.constant.int 1
%13007 = torch.aten.add.Tensor %12988, %12991, %int1_10726 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10727 = torch.constant.int 1
%13008 = torch.aten.add.Tensor %13007, %12994, %int1_10727 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10728 = torch.constant.int 1
%13009 = torch.aten.add.Tensor %13008, %12997, %int1_10728 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10729 = torch.constant.int 1
%13010 = torch.aten.add.Tensor %13009, %13000, %int1_10729 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10730 = torch.constant.int 1
%13011 = torch.aten.add.Tensor %13010, %13003, %int1_10730 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10731 = torch.constant.int 1
%13012 = torch.aten.add.Tensor %13011, %12811, %int1_10731 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10732 = torch.constant.int 1
%13013 = torch.aten.add.Tensor %13012, %13006, %int1_10732 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13014 = torch_c.to_builtin_tensor %12775 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10733 = arith.constant 1 : index
%dim_10734 = tensor.dim %13014, %c1_10733 : tensor<4x?x4096xf16>
%13015 = flow.tensor.transfer %13014 : tensor<4x?x4096xf16>{%dim_10734} to #hal.device.promise<@__device_7>
%13016 = torch_c.from_builtin_tensor %13015 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13017 = torch_c.to_builtin_tensor %12781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10735 = arith.constant 1 : index
%dim_10736 = tensor.dim %13017, %c1_10735 : tensor<4x?x4096xf16>
%13018 = flow.tensor.transfer %13017 : tensor<4x?x4096xf16>{%dim_10736} to #hal.device.promise<@__device_7>
%13019 = torch_c.from_builtin_tensor %13018 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13020 = torch_c.to_builtin_tensor %12787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10737 = arith.constant 1 : index
%dim_10738 = tensor.dim %13020, %c1_10737 : tensor<4x?x4096xf16>
%13021 = flow.tensor.transfer %13020 : tensor<4x?x4096xf16>{%dim_10738} to #hal.device.promise<@__device_7>
%13022 = torch_c.from_builtin_tensor %13021 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13023 = torch_c.to_builtin_tensor %12793 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10739 = arith.constant 1 : index
%dim_10740 = tensor.dim %13023, %c1_10739 : tensor<4x?x4096xf16>
%13024 = flow.tensor.transfer %13023 : tensor<4x?x4096xf16>{%dim_10740} to #hal.device.promise<@__device_7>
%13025 = torch_c.from_builtin_tensor %13024 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13026 = torch_c.to_builtin_tensor %12799 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10741 = arith.constant 1 : index
%dim_10742 = tensor.dim %13026, %c1_10741 : tensor<4x?x4096xf16>
%13027 = flow.tensor.transfer %13026 : tensor<4x?x4096xf16>{%dim_10742} to #hal.device.promise<@__device_7>
%13028 = torch_c.from_builtin_tensor %13027 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13029 = torch_c.to_builtin_tensor %12805 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10743 = arith.constant 1 : index
%dim_10744 = tensor.dim %13029, %c1_10743 : tensor<4x?x4096xf16>
%13030 = flow.tensor.transfer %13029 : tensor<4x?x4096xf16>{%dim_10744} to #hal.device.promise<@__device_7>
%13031 = torch_c.from_builtin_tensor %13030 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13032 = torch_c.to_builtin_tensor %12811 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10745 = arith.constant 1 : index
%dim_10746 = tensor.dim %13032, %c1_10745 : tensor<4x?x4096xf16>
%13033 = flow.tensor.transfer %13032 : tensor<4x?x4096xf16>{%dim_10746} to #hal.device.promise<@__device_7>
%13034 = torch_c.from_builtin_tensor %13033 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10747 = torch.constant.int 1
%13035 = torch.aten.add.Tensor %13016, %13019, %int1_10747 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10748 = torch.constant.int 1
%13036 = torch.aten.add.Tensor %13035, %13022, %int1_10748 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10749 = torch.constant.int 1
%13037 = torch.aten.add.Tensor %13036, %13025, %int1_10749 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10750 = torch.constant.int 1
%13038 = torch.aten.add.Tensor %13037, %13028, %int1_10750 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10751 = torch.constant.int 1
%13039 = torch.aten.add.Tensor %13038, %13031, %int1_10751 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10752 = torch.constant.int 1
%13040 = torch.aten.add.Tensor %13039, %13034, %int1_10752 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10753 = torch.constant.int 1
%13041 = torch.aten.add.Tensor %13040, %12817, %int1_10753 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10754 = torch.constant.int 1
%13042 = torch.aten.add.Tensor %11701, %12845, %int1_10754 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10755 = torch.constant.int 1
%13043 = torch.aten.add.Tensor %11702, %12873, %int1_10755 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10756 = torch.constant.int 1
%13044 = torch.aten.add.Tensor %11703, %12901, %int1_10756 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10757 = torch.constant.int 1
%13045 = torch.aten.add.Tensor %11704, %12929, %int1_10757 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10758 = torch.constant.int 1
%13046 = torch.aten.add.Tensor %11705, %12957, %int1_10758 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10759 = torch.constant.int 1
%13047 = torch.aten.add.Tensor %11706, %12985, %int1_10759 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10760 = torch.constant.int 1
%13048 = torch.aten.add.Tensor %11707, %13013, %int1_10760 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10761 = torch.constant.int 1
%13049 = torch.aten.add.Tensor %11708, %13041, %int1_10761 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_10762 = torch.constant.int 6
%13050 = torch.prims.convert_element_type %13042, %int6_10762 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10763 = torch.constant.int 6
%13051 = torch.prims.convert_element_type %13043, %int6_10763 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10764 = torch.constant.int 6
%13052 = torch.prims.convert_element_type %13044, %int6_10764 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10765 = torch.constant.int 6
%13053 = torch.prims.convert_element_type %13045, %int6_10765 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10766 = torch.constant.int 6
%13054 = torch.prims.convert_element_type %13046, %int6_10766 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10767 = torch.constant.int 6
%13055 = torch.prims.convert_element_type %13047, %int6_10767 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10768 = torch.constant.int 6
%13056 = torch.prims.convert_element_type %13048, %int6_10768 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_10769 = torch.constant.int 6
%13057 = torch.prims.convert_element_type %13049, %int6_10769 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10770 = torch.constant.int 2
%13058 = torch.aten.pow.Tensor_Scalar %13050, %int2_10770 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10771 = torch.constant.int 2
%13059 = torch.aten.pow.Tensor_Scalar %13051, %int2_10771 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10772 = torch.constant.int 2
%13060 = torch.aten.pow.Tensor_Scalar %13052, %int2_10772 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10773 = torch.constant.int 2
%13061 = torch.aten.pow.Tensor_Scalar %13053, %int2_10773 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10774 = torch.constant.int 2
%13062 = torch.aten.pow.Tensor_Scalar %13054, %int2_10774 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10775 = torch.constant.int 2
%13063 = torch.aten.pow.Tensor_Scalar %13055, %int2_10775 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10776 = torch.constant.int 2
%13064 = torch.aten.pow.Tensor_Scalar %13056, %int2_10776 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_10777 = torch.constant.int 2
%13065 = torch.aten.pow.Tensor_Scalar %13057, %int2_10777 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_10778 = torch.constant.int -1
%13066 = torch.prim.ListConstruct %int-1_10778 : (!torch.int) -> !torch.list<int>
%true_10779 = torch.constant.bool true
%none_10780 = torch.constant.none
%13067 = torch.aten.mean.dim %13058, %13066, %true_10779, %none_10780 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10781 = torch.constant.int -1
%13068 = torch.prim.ListConstruct %int-1_10781 : (!torch.int) -> !torch.list<int>
%true_10782 = torch.constant.bool true
%none_10783 = torch.constant.none
%13069 = torch.aten.mean.dim %13059, %13068, %true_10782, %none_10783 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10784 = torch.constant.int -1
%13070 = torch.prim.ListConstruct %int-1_10784 : (!torch.int) -> !torch.list<int>
%true_10785 = torch.constant.bool true
%none_10786 = torch.constant.none
%13071 = torch.aten.mean.dim %13060, %13070, %true_10785, %none_10786 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10787 = torch.constant.int -1
%13072 = torch.prim.ListConstruct %int-1_10787 : (!torch.int) -> !torch.list<int>
%true_10788 = torch.constant.bool true
%none_10789 = torch.constant.none
%13073 = torch.aten.mean.dim %13061, %13072, %true_10788, %none_10789 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10790 = torch.constant.int -1
%13074 = torch.prim.ListConstruct %int-1_10790 : (!torch.int) -> !torch.list<int>
%true_10791 = torch.constant.bool true
%none_10792 = torch.constant.none
%13075 = torch.aten.mean.dim %13062, %13074, %true_10791, %none_10792 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10793 = torch.constant.int -1
%13076 = torch.prim.ListConstruct %int-1_10793 : (!torch.int) -> !torch.list<int>
%true_10794 = torch.constant.bool true
%none_10795 = torch.constant.none
%13077 = torch.aten.mean.dim %13063, %13076, %true_10794, %none_10795 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10796 = torch.constant.int -1
%13078 = torch.prim.ListConstruct %int-1_10796 : (!torch.int) -> !torch.list<int>
%true_10797 = torch.constant.bool true
%none_10798 = torch.constant.none
%13079 = torch.aten.mean.dim %13064, %13078, %true_10797, %none_10798 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_10799 = torch.constant.int -1
%13080 = torch.prim.ListConstruct %int-1_10799 : (!torch.int) -> !torch.list<int>
%true_10800 = torch.constant.bool true
%none_10801 = torch.constant.none
%13081 = torch.aten.mean.dim %13065, %13080, %true_10800, %none_10801 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10802 = torch.constant.float 9.9999997473787516E-6
%int1_10803 = torch.constant.int 1
%13082 = torch.aten.add.Scalar %13067, %float9.999990e-06_10802, %int1_10803 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10804 = torch.constant.float 9.9999997473787516E-6
%int1_10805 = torch.constant.int 1
%13083 = torch.aten.add.Scalar %13069, %float9.999990e-06_10804, %int1_10805 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10806 = torch.constant.float 9.9999997473787516E-6
%int1_10807 = torch.constant.int 1
%13084 = torch.aten.add.Scalar %13071, %float9.999990e-06_10806, %int1_10807 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10808 = torch.constant.float 9.9999997473787516E-6
%int1_10809 = torch.constant.int 1
%13085 = torch.aten.add.Scalar %13073, %float9.999990e-06_10808, %int1_10809 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10810 = torch.constant.float 9.9999997473787516E-6
%int1_10811 = torch.constant.int 1
%13086 = torch.aten.add.Scalar %13075, %float9.999990e-06_10810, %int1_10811 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10812 = torch.constant.float 9.9999997473787516E-6
%int1_10813 = torch.constant.int 1
%13087 = torch.aten.add.Scalar %13077, %float9.999990e-06_10812, %int1_10813 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10814 = torch.constant.float 9.9999997473787516E-6
%int1_10815 = torch.constant.int 1
%13088 = torch.aten.add.Scalar %13079, %float9.999990e-06_10814, %int1_10815 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_10816 = torch.constant.float 9.9999997473787516E-6
%int1_10817 = torch.constant.int 1
%13089 = torch.aten.add.Scalar %13081, %float9.999990e-06_10816, %int1_10817 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13090 = torch.aten.rsqrt %13082 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13091 = torch.aten.rsqrt %13083 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13092 = torch.aten.rsqrt %13084 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13093 = torch.aten.rsqrt %13085 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13094 = torch.aten.rsqrt %13086 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13095 = torch.aten.rsqrt %13087 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13096 = torch.aten.rsqrt %13088 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13097 = torch.aten.rsqrt %13089 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13098 = torch.aten.mul.Tensor %13050, %13090 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13099 = torch.aten.mul.Tensor %13051, %13091 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13100 = torch.aten.mul.Tensor %13052, %13092 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13101 = torch.aten.mul.Tensor %13053, %13093 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13102 = torch.aten.mul.Tensor %13054, %13094 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13103 = torch.aten.mul.Tensor %13055, %13095 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13104 = torch.aten.mul.Tensor %13056, %13096 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13105 = torch.aten.mul.Tensor %13057, %13097 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13106 = torch.aten.mul.Tensor %408, %13098 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13107 = torch.aten.mul.Tensor %409, %13099 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13108 = torch.aten.mul.Tensor %410, %13100 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13109 = torch.aten.mul.Tensor %411, %13101 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13110 = torch.aten.mul.Tensor %412, %13102 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13111 = torch.aten.mul.Tensor %413, %13103 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13112 = torch.aten.mul.Tensor %414, %13104 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13113 = torch.aten.mul.Tensor %415, %13105 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_10818 = torch.constant.int 5
%13114 = torch.prims.convert_element_type %13106, %int5_10818 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10819 = torch.constant.int 5
%13115 = torch.prims.convert_element_type %13107, %int5_10819 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10820 = torch.constant.int 5
%13116 = torch.prims.convert_element_type %13108, %int5_10820 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10821 = torch.constant.int 5
%13117 = torch.prims.convert_element_type %13109, %int5_10821 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10822 = torch.constant.int 5
%13118 = torch.prims.convert_element_type %13110, %int5_10822 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10823 = torch.constant.int 5
%13119 = torch.prims.convert_element_type %13111, %int5_10823 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10824 = torch.constant.int 5
%13120 = torch.prims.convert_element_type %13112, %int5_10824 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_10825 = torch.constant.int 5
%13121 = torch.prims.convert_element_type %13113, %int5_10825 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10826 = torch.constant.int 1
%int0_10827 = torch.constant.int 0
%13122 = torch.prim.ListConstruct %int1_10826, %int0_10827 : (!torch.int, !torch.int) -> !torch.list<int>
%13123 = torch.aten.permute %416, %13122 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10828 = torch.constant.int 1
%int0_10829 = torch.constant.int 0
%13124 = torch.prim.ListConstruct %int1_10828, %int0_10829 : (!torch.int, !torch.int) -> !torch.list<int>
%13125 = torch.aten.permute %417, %13124 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10830 = torch.constant.int 1
%int0_10831 = torch.constant.int 0
%13126 = torch.prim.ListConstruct %int1_10830, %int0_10831 : (!torch.int, !torch.int) -> !torch.list<int>
%13127 = torch.aten.permute %418, %13126 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10832 = torch.constant.int 1
%int0_10833 = torch.constant.int 0
%13128 = torch.prim.ListConstruct %int1_10832, %int0_10833 : (!torch.int, !torch.int) -> !torch.list<int>
%13129 = torch.aten.permute %419, %13128 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10834 = torch.constant.int 1
%int0_10835 = torch.constant.int 0
%13130 = torch.prim.ListConstruct %int1_10834, %int0_10835 : (!torch.int, !torch.int) -> !torch.list<int>
%13131 = torch.aten.permute %420, %13130 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10836 = torch.constant.int 1
%int0_10837 = torch.constant.int 0
%13132 = torch.prim.ListConstruct %int1_10836, %int0_10837 : (!torch.int, !torch.int) -> !torch.list<int>
%13133 = torch.aten.permute %421, %13132 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10838 = torch.constant.int 1
%int0_10839 = torch.constant.int 0
%13134 = torch.prim.ListConstruct %int1_10838, %int0_10839 : (!torch.int, !torch.int) -> !torch.list<int>
%13135 = torch.aten.permute %422, %13134 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10840 = torch.constant.int 1
%int0_10841 = torch.constant.int 0
%13136 = torch.prim.ListConstruct %int1_10840, %int0_10841 : (!torch.int, !torch.int) -> !torch.list<int>
%13137 = torch.aten.permute %423, %13136 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_10842 = torch.constant.int 4
%13138 = torch.aten.mul.int %int4_10842, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10843 = torch.constant.int 4096
%13139 = torch.prim.ListConstruct %13138, %int4096_10843 : (!torch.int, !torch.int) -> !torch.list<int>
%13140 = torch.aten.view %13114, %13139 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13140, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13141 = torch.aten.mm %13140, %13123 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13141, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10844 = torch.constant.int 4
%int1792_10845 = torch.constant.int 1792
%13142 = torch.prim.ListConstruct %int4_10844, %2482, %int1792_10845 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13143 = torch.aten.view %13141, %13142 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10846 = torch.constant.int 4
%13144 = torch.aten.mul.int %int4_10846, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10847 = torch.constant.int 4096
%13145 = torch.prim.ListConstruct %13144, %int4096_10847 : (!torch.int, !torch.int) -> !torch.list<int>
%13146 = torch.aten.view %13115, %13145 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13146, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13147 = torch.aten.mm %13146, %13125 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13147, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10848 = torch.constant.int 4
%int1792_10849 = torch.constant.int 1792
%13148 = torch.prim.ListConstruct %int4_10848, %2482, %int1792_10849 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13149 = torch.aten.view %13147, %13148 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10850 = torch.constant.int 4
%13150 = torch.aten.mul.int %int4_10850, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10851 = torch.constant.int 4096
%13151 = torch.prim.ListConstruct %13150, %int4096_10851 : (!torch.int, !torch.int) -> !torch.list<int>
%13152 = torch.aten.view %13116, %13151 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13152, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13153 = torch.aten.mm %13152, %13127 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13153, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10852 = torch.constant.int 4
%int1792_10853 = torch.constant.int 1792
%13154 = torch.prim.ListConstruct %int4_10852, %2482, %int1792_10853 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13155 = torch.aten.view %13153, %13154 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10854 = torch.constant.int 4
%13156 = torch.aten.mul.int %int4_10854, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10855 = torch.constant.int 4096
%13157 = torch.prim.ListConstruct %13156, %int4096_10855 : (!torch.int, !torch.int) -> !torch.list<int>
%13158 = torch.aten.view %13117, %13157 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13158, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13159 = torch.aten.mm %13158, %13129 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13159, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10856 = torch.constant.int 4
%int1792_10857 = torch.constant.int 1792
%13160 = torch.prim.ListConstruct %int4_10856, %2482, %int1792_10857 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13161 = torch.aten.view %13159, %13160 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10858 = torch.constant.int 4
%13162 = torch.aten.mul.int %int4_10858, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10859 = torch.constant.int 4096
%13163 = torch.prim.ListConstruct %13162, %int4096_10859 : (!torch.int, !torch.int) -> !torch.list<int>
%13164 = torch.aten.view %13118, %13163 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13164, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13165 = torch.aten.mm %13164, %13131 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13165, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10860 = torch.constant.int 4
%int1792_10861 = torch.constant.int 1792
%13166 = torch.prim.ListConstruct %int4_10860, %2482, %int1792_10861 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13167 = torch.aten.view %13165, %13166 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10862 = torch.constant.int 4
%13168 = torch.aten.mul.int %int4_10862, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10863 = torch.constant.int 4096
%13169 = torch.prim.ListConstruct %13168, %int4096_10863 : (!torch.int, !torch.int) -> !torch.list<int>
%13170 = torch.aten.view %13119, %13169 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13170, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13171 = torch.aten.mm %13170, %13133 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13171, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10864 = torch.constant.int 4
%int1792_10865 = torch.constant.int 1792
%13172 = torch.prim.ListConstruct %int4_10864, %2482, %int1792_10865 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13173 = torch.aten.view %13171, %13172 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10866 = torch.constant.int 4
%13174 = torch.aten.mul.int %int4_10866, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10867 = torch.constant.int 4096
%13175 = torch.prim.ListConstruct %13174, %int4096_10867 : (!torch.int, !torch.int) -> !torch.list<int>
%13176 = torch.aten.view %13120, %13175 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13176, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13177 = torch.aten.mm %13176, %13135 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13177, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10868 = torch.constant.int 4
%int1792_10869 = torch.constant.int 1792
%13178 = torch.prim.ListConstruct %int4_10868, %2482, %int1792_10869 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13179 = torch.aten.view %13177, %13178 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10870 = torch.constant.int 4
%13180 = torch.aten.mul.int %int4_10870, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10871 = torch.constant.int 4096
%13181 = torch.prim.ListConstruct %13180, %int4096_10871 : (!torch.int, !torch.int) -> !torch.list<int>
%13182 = torch.aten.view %13121, %13181 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13182, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13183 = torch.aten.mm %13182, %13137 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13183, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10872 = torch.constant.int 4
%int1792_10873 = torch.constant.int 1792
%13184 = torch.prim.ListConstruct %int4_10872, %2482, %int1792_10873 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13185 = torch.aten.view %13183, %13184 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13186 = torch.aten.silu %13143 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13187 = torch.aten.silu %13149 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13188 = torch.aten.silu %13155 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13189 = torch.aten.silu %13161 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13190 = torch.aten.silu %13167 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13191 = torch.aten.silu %13173 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13192 = torch.aten.silu %13179 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13193 = torch.aten.silu %13185 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_10874 = torch.constant.int 1
%int0_10875 = torch.constant.int 0
%13194 = torch.prim.ListConstruct %int1_10874, %int0_10875 : (!torch.int, !torch.int) -> !torch.list<int>
%13195 = torch.aten.permute %424, %13194 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10876 = torch.constant.int 1
%int0_10877 = torch.constant.int 0
%13196 = torch.prim.ListConstruct %int1_10876, %int0_10877 : (!torch.int, !torch.int) -> !torch.list<int>
%13197 = torch.aten.permute %425, %13196 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10878 = torch.constant.int 1
%int0_10879 = torch.constant.int 0
%13198 = torch.prim.ListConstruct %int1_10878, %int0_10879 : (!torch.int, !torch.int) -> !torch.list<int>
%13199 = torch.aten.permute %426, %13198 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10880 = torch.constant.int 1
%int0_10881 = torch.constant.int 0
%13200 = torch.prim.ListConstruct %int1_10880, %int0_10881 : (!torch.int, !torch.int) -> !torch.list<int>
%13201 = torch.aten.permute %427, %13200 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10882 = torch.constant.int 1
%int0_10883 = torch.constant.int 0
%13202 = torch.prim.ListConstruct %int1_10882, %int0_10883 : (!torch.int, !torch.int) -> !torch.list<int>
%13203 = torch.aten.permute %428, %13202 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10884 = torch.constant.int 1
%int0_10885 = torch.constant.int 0
%13204 = torch.prim.ListConstruct %int1_10884, %int0_10885 : (!torch.int, !torch.int) -> !torch.list<int>
%13205 = torch.aten.permute %429, %13204 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10886 = torch.constant.int 1
%int0_10887 = torch.constant.int 0
%13206 = torch.prim.ListConstruct %int1_10886, %int0_10887 : (!torch.int, !torch.int) -> !torch.list<int>
%13207 = torch.aten.permute %430, %13206 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_10888 = torch.constant.int 1
%int0_10889 = torch.constant.int 0
%13208 = torch.prim.ListConstruct %int1_10888, %int0_10889 : (!torch.int, !torch.int) -> !torch.list<int>
%13209 = torch.aten.permute %431, %13208 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_10890 = torch.constant.int 4
%13210 = torch.aten.mul.int %int4_10890, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10891 = torch.constant.int 4096
%13211 = torch.prim.ListConstruct %13210, %int4096_10891 : (!torch.int, !torch.int) -> !torch.list<int>
%13212 = torch.aten.view %13114, %13211 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13212, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13213 = torch.aten.mm %13212, %13195 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13213, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10892 = torch.constant.int 4
%int1792_10893 = torch.constant.int 1792
%13214 = torch.prim.ListConstruct %int4_10892, %2482, %int1792_10893 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13215 = torch.aten.view %13213, %13214 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10894 = torch.constant.int 4
%13216 = torch.aten.mul.int %int4_10894, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10895 = torch.constant.int 4096
%13217 = torch.prim.ListConstruct %13216, %int4096_10895 : (!torch.int, !torch.int) -> !torch.list<int>
%13218 = torch.aten.view %13115, %13217 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13218, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13219 = torch.aten.mm %13218, %13197 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13219, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10896 = torch.constant.int 4
%int1792_10897 = torch.constant.int 1792
%13220 = torch.prim.ListConstruct %int4_10896, %2482, %int1792_10897 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13221 = torch.aten.view %13219, %13220 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10898 = torch.constant.int 4
%13222 = torch.aten.mul.int %int4_10898, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10899 = torch.constant.int 4096
%13223 = torch.prim.ListConstruct %13222, %int4096_10899 : (!torch.int, !torch.int) -> !torch.list<int>
%13224 = torch.aten.view %13116, %13223 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13224, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13225 = torch.aten.mm %13224, %13199 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13225, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10900 = torch.constant.int 4
%int1792_10901 = torch.constant.int 1792
%13226 = torch.prim.ListConstruct %int4_10900, %2482, %int1792_10901 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13227 = torch.aten.view %13225, %13226 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10902 = torch.constant.int 4
%13228 = torch.aten.mul.int %int4_10902, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10903 = torch.constant.int 4096
%13229 = torch.prim.ListConstruct %13228, %int4096_10903 : (!torch.int, !torch.int) -> !torch.list<int>
%13230 = torch.aten.view %13117, %13229 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13230, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13231 = torch.aten.mm %13230, %13201 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13231, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10904 = torch.constant.int 4
%int1792_10905 = torch.constant.int 1792
%13232 = torch.prim.ListConstruct %int4_10904, %2482, %int1792_10905 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13233 = torch.aten.view %13231, %13232 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10906 = torch.constant.int 4
%13234 = torch.aten.mul.int %int4_10906, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10907 = torch.constant.int 4096
%13235 = torch.prim.ListConstruct %13234, %int4096_10907 : (!torch.int, !torch.int) -> !torch.list<int>
%13236 = torch.aten.view %13118, %13235 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13236, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13237 = torch.aten.mm %13236, %13203 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13237, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10908 = torch.constant.int 4
%int1792_10909 = torch.constant.int 1792
%13238 = torch.prim.ListConstruct %int4_10908, %2482, %int1792_10909 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13239 = torch.aten.view %13237, %13238 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10910 = torch.constant.int 4
%13240 = torch.aten.mul.int %int4_10910, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10911 = torch.constant.int 4096
%13241 = torch.prim.ListConstruct %13240, %int4096_10911 : (!torch.int, !torch.int) -> !torch.list<int>
%13242 = torch.aten.view %13119, %13241 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13242, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13243 = torch.aten.mm %13242, %13205 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13243, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10912 = torch.constant.int 4
%int1792_10913 = torch.constant.int 1792
%13244 = torch.prim.ListConstruct %int4_10912, %2482, %int1792_10913 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13245 = torch.aten.view %13243, %13244 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10914 = torch.constant.int 4
%13246 = torch.aten.mul.int %int4_10914, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10915 = torch.constant.int 4096
%13247 = torch.prim.ListConstruct %13246, %int4096_10915 : (!torch.int, !torch.int) -> !torch.list<int>
%13248 = torch.aten.view %13120, %13247 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13248, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13249 = torch.aten.mm %13248, %13207 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13249, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10916 = torch.constant.int 4
%int1792_10917 = torch.constant.int 1792
%13250 = torch.prim.ListConstruct %int4_10916, %2482, %int1792_10917 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13251 = torch.aten.view %13249, %13250 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_10918 = torch.constant.int 4
%13252 = torch.aten.mul.int %int4_10918, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_10919 = torch.constant.int 4096
%13253 = torch.prim.ListConstruct %13252, %int4096_10919 : (!torch.int, !torch.int) -> !torch.list<int>
%13254 = torch.aten.view %13121, %13253 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13254, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13255 = torch.aten.mm %13254, %13209 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13255, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_10920 = torch.constant.int 4
%int1792_10921 = torch.constant.int 1792
%13256 = torch.prim.ListConstruct %int4_10920, %2482, %int1792_10921 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13257 = torch.aten.view %13255, %13256 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13258 = torch.aten.mul.Tensor %13186, %13215 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13259 = torch.aten.mul.Tensor %13187, %13221 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13260 = torch.aten.mul.Tensor %13188, %13227 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13261 = torch.aten.mul.Tensor %13189, %13233 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13262 = torch.aten.mul.Tensor %13190, %13239 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13263 = torch.aten.mul.Tensor %13191, %13245 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13263, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13264 = torch.aten.mul.Tensor %13192, %13251 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%13265 = torch.aten.mul.Tensor %13193, %13257 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %13265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_10922 = torch.constant.int 1
%int0_10923 = torch.constant.int 0
%13266 = torch.prim.ListConstruct %int1_10922, %int0_10923 : (!torch.int, !torch.int) -> !torch.list<int>
%13267 = torch.aten.permute %432, %13266 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10924 = torch.constant.int 1
%int0_10925 = torch.constant.int 0
%13268 = torch.prim.ListConstruct %int1_10924, %int0_10925 : (!torch.int, !torch.int) -> !torch.list<int>
%13269 = torch.aten.permute %433, %13268 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10926 = torch.constant.int 1
%int0_10927 = torch.constant.int 0
%13270 = torch.prim.ListConstruct %int1_10926, %int0_10927 : (!torch.int, !torch.int) -> !torch.list<int>
%13271 = torch.aten.permute %434, %13270 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10928 = torch.constant.int 1
%int0_10929 = torch.constant.int 0
%13272 = torch.prim.ListConstruct %int1_10928, %int0_10929 : (!torch.int, !torch.int) -> !torch.list<int>
%13273 = torch.aten.permute %435, %13272 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10930 = torch.constant.int 1
%int0_10931 = torch.constant.int 0
%13274 = torch.prim.ListConstruct %int1_10930, %int0_10931 : (!torch.int, !torch.int) -> !torch.list<int>
%13275 = torch.aten.permute %436, %13274 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10932 = torch.constant.int 1
%int0_10933 = torch.constant.int 0
%13276 = torch.prim.ListConstruct %int1_10932, %int0_10933 : (!torch.int, !torch.int) -> !torch.list<int>
%13277 = torch.aten.permute %437, %13276 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10934 = torch.constant.int 1
%int0_10935 = torch.constant.int 0
%13278 = torch.prim.ListConstruct %int1_10934, %int0_10935 : (!torch.int, !torch.int) -> !torch.list<int>
%13279 = torch.aten.permute %438, %13278 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10936 = torch.constant.int 1
%int0_10937 = torch.constant.int 0
%13280 = torch.prim.ListConstruct %int1_10936, %int0_10937 : (!torch.int, !torch.int) -> !torch.list<int>
%13281 = torch.aten.permute %439, %13280 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_10938 = torch.constant.int 1
%13282 = torch.aten.size.int %13143, %int1_10938 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10939 = torch.constant.int 4
%13283 = torch.aten.mul.int %int4_10939, %13282 : !torch.int, !torch.int -> !torch.int
%int1792_10940 = torch.constant.int 1792
%13284 = torch.prim.ListConstruct %13283, %int1792_10940 : (!torch.int, !torch.int) -> !torch.list<int>
%13285 = torch.aten.view %13258, %13284 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13285, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13286 = torch.aten.mm %13285, %13267 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13286, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10941 = torch.constant.int 4
%int4096_10942 = torch.constant.int 4096
%13287 = torch.prim.ListConstruct %int4_10941, %13282, %int4096_10942 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13288 = torch.aten.view %13286, %13287 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10943 = torch.constant.int 1
%13289 = torch.aten.size.int %13149, %int1_10943 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10944 = torch.constant.int 4
%13290 = torch.aten.mul.int %int4_10944, %13289 : !torch.int, !torch.int -> !torch.int
%int1792_10945 = torch.constant.int 1792
%13291 = torch.prim.ListConstruct %13290, %int1792_10945 : (!torch.int, !torch.int) -> !torch.list<int>
%13292 = torch.aten.view %13259, %13291 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13292, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13293 = torch.aten.mm %13292, %13269 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13293, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10946 = torch.constant.int 4
%int4096_10947 = torch.constant.int 4096
%13294 = torch.prim.ListConstruct %int4_10946, %13289, %int4096_10947 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13295 = torch.aten.view %13293, %13294 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10948 = torch.constant.int 1
%13296 = torch.aten.size.int %13155, %int1_10948 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10949 = torch.constant.int 4
%13297 = torch.aten.mul.int %int4_10949, %13296 : !torch.int, !torch.int -> !torch.int
%int1792_10950 = torch.constant.int 1792
%13298 = torch.prim.ListConstruct %13297, %int1792_10950 : (!torch.int, !torch.int) -> !torch.list<int>
%13299 = torch.aten.view %13260, %13298 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13299, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13300 = torch.aten.mm %13299, %13271 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13300, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10951 = torch.constant.int 4
%int4096_10952 = torch.constant.int 4096
%13301 = torch.prim.ListConstruct %int4_10951, %13296, %int4096_10952 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13302 = torch.aten.view %13300, %13301 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10953 = torch.constant.int 1
%13303 = torch.aten.size.int %13161, %int1_10953 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10954 = torch.constant.int 4
%13304 = torch.aten.mul.int %int4_10954, %13303 : !torch.int, !torch.int -> !torch.int
%int1792_10955 = torch.constant.int 1792
%13305 = torch.prim.ListConstruct %13304, %int1792_10955 : (!torch.int, !torch.int) -> !torch.list<int>
%13306 = torch.aten.view %13261, %13305 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13306, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13307 = torch.aten.mm %13306, %13273 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13307, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10956 = torch.constant.int 4
%int4096_10957 = torch.constant.int 4096
%13308 = torch.prim.ListConstruct %int4_10956, %13303, %int4096_10957 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13309 = torch.aten.view %13307, %13308 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10958 = torch.constant.int 1
%13310 = torch.aten.size.int %13167, %int1_10958 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10959 = torch.constant.int 4
%13311 = torch.aten.mul.int %int4_10959, %13310 : !torch.int, !torch.int -> !torch.int
%int1792_10960 = torch.constant.int 1792
%13312 = torch.prim.ListConstruct %13311, %int1792_10960 : (!torch.int, !torch.int) -> !torch.list<int>
%13313 = torch.aten.view %13262, %13312 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13313, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13314 = torch.aten.mm %13313, %13275 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13314, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10961 = torch.constant.int 4
%int4096_10962 = torch.constant.int 4096
%13315 = torch.prim.ListConstruct %int4_10961, %13310, %int4096_10962 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13316 = torch.aten.view %13314, %13315 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10963 = torch.constant.int 1
%13317 = torch.aten.size.int %13173, %int1_10963 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10964 = torch.constant.int 4
%13318 = torch.aten.mul.int %int4_10964, %13317 : !torch.int, !torch.int -> !torch.int
%int1792_10965 = torch.constant.int 1792
%13319 = torch.prim.ListConstruct %13318, %int1792_10965 : (!torch.int, !torch.int) -> !torch.list<int>
%13320 = torch.aten.view %13263, %13319 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13320, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13321 = torch.aten.mm %13320, %13277 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13321, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10966 = torch.constant.int 4
%int4096_10967 = torch.constant.int 4096
%13322 = torch.prim.ListConstruct %int4_10966, %13317, %int4096_10967 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13323 = torch.aten.view %13321, %13322 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10968 = torch.constant.int 1
%13324 = torch.aten.size.int %13179, %int1_10968 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10969 = torch.constant.int 4
%13325 = torch.aten.mul.int %int4_10969, %13324 : !torch.int, !torch.int -> !torch.int
%int1792_10970 = torch.constant.int 1792
%13326 = torch.prim.ListConstruct %13325, %int1792_10970 : (!torch.int, !torch.int) -> !torch.list<int>
%13327 = torch.aten.view %13264, %13326 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13327, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13328 = torch.aten.mm %13327, %13279 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13328, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10971 = torch.constant.int 4
%int4096_10972 = torch.constant.int 4096
%13329 = torch.prim.ListConstruct %int4_10971, %13324, %int4096_10972 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13330 = torch.aten.view %13328, %13329 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10973 = torch.constant.int 1
%13331 = torch.aten.size.int %13185, %int1_10973 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_10974 = torch.constant.int 4
%13332 = torch.aten.mul.int %int4_10974, %13331 : !torch.int, !torch.int -> !torch.int
%int1792_10975 = torch.constant.int 1792
%13333 = torch.prim.ListConstruct %13332, %int1792_10975 : (!torch.int, !torch.int) -> !torch.list<int>
%13334 = torch.aten.view %13265, %13333 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %13334, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%13335 = torch.aten.mm %13334, %13281 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13335, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_10976 = torch.constant.int 4
%int4096_10977 = torch.constant.int 4096
%13336 = torch.prim.ListConstruct %int4_10976, %13331, %int4096_10977 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13337 = torch.aten.view %13335, %13336 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13338 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10978 = arith.constant 1 : index
%dim_10979 = tensor.dim %13338, %c1_10978 : tensor<4x?x4096xf16>
%13339 = flow.tensor.transfer %13338 : tensor<4x?x4096xf16>{%dim_10979} to #hal.device.promise<@__device_0>
%13340 = torch_c.from_builtin_tensor %13339 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13341 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10980 = arith.constant 1 : index
%dim_10981 = tensor.dim %13341, %c1_10980 : tensor<4x?x4096xf16>
%13342 = flow.tensor.transfer %13341 : tensor<4x?x4096xf16>{%dim_10981} to #hal.device.promise<@__device_0>
%13343 = torch_c.from_builtin_tensor %13342 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13344 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10982 = arith.constant 1 : index
%dim_10983 = tensor.dim %13344, %c1_10982 : tensor<4x?x4096xf16>
%13345 = flow.tensor.transfer %13344 : tensor<4x?x4096xf16>{%dim_10983} to #hal.device.promise<@__device_0>
%13346 = torch_c.from_builtin_tensor %13345 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13347 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10984 = arith.constant 1 : index
%dim_10985 = tensor.dim %13347, %c1_10984 : tensor<4x?x4096xf16>
%13348 = flow.tensor.transfer %13347 : tensor<4x?x4096xf16>{%dim_10985} to #hal.device.promise<@__device_0>
%13349 = torch_c.from_builtin_tensor %13348 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13350 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10986 = arith.constant 1 : index
%dim_10987 = tensor.dim %13350, %c1_10986 : tensor<4x?x4096xf16>
%13351 = flow.tensor.transfer %13350 : tensor<4x?x4096xf16>{%dim_10987} to #hal.device.promise<@__device_0>
%13352 = torch_c.from_builtin_tensor %13351 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13353 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10988 = arith.constant 1 : index
%dim_10989 = tensor.dim %13353, %c1_10988 : tensor<4x?x4096xf16>
%13354 = flow.tensor.transfer %13353 : tensor<4x?x4096xf16>{%dim_10989} to #hal.device.promise<@__device_0>
%13355 = torch_c.from_builtin_tensor %13354 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13356 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10990 = arith.constant 1 : index
%dim_10991 = tensor.dim %13356, %c1_10990 : tensor<4x?x4096xf16>
%13357 = flow.tensor.transfer %13356 : tensor<4x?x4096xf16>{%dim_10991} to #hal.device.promise<@__device_0>
%13358 = torch_c.from_builtin_tensor %13357 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10992 = torch.constant.int 1
%13359 = torch.aten.add.Tensor %13288, %13340, %int1_10992 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10993 = torch.constant.int 1
%13360 = torch.aten.add.Tensor %13359, %13343, %int1_10993 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10994 = torch.constant.int 1
%13361 = torch.aten.add.Tensor %13360, %13346, %int1_10994 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10995 = torch.constant.int 1
%13362 = torch.aten.add.Tensor %13361, %13349, %int1_10995 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10996 = torch.constant.int 1
%13363 = torch.aten.add.Tensor %13362, %13352, %int1_10996 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10997 = torch.constant.int 1
%13364 = torch.aten.add.Tensor %13363, %13355, %int1_10997 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_10998 = torch.constant.int 1
%13365 = torch.aten.add.Tensor %13364, %13358, %int1_10998 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13366 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_10999 = arith.constant 1 : index
%dim_11000 = tensor.dim %13366, %c1_10999 : tensor<4x?x4096xf16>
%13367 = flow.tensor.transfer %13366 : tensor<4x?x4096xf16>{%dim_11000} to #hal.device.promise<@__device_1>
%13368 = torch_c.from_builtin_tensor %13367 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13369 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11001 = arith.constant 1 : index
%dim_11002 = tensor.dim %13369, %c1_11001 : tensor<4x?x4096xf16>
%13370 = flow.tensor.transfer %13369 : tensor<4x?x4096xf16>{%dim_11002} to #hal.device.promise<@__device_1>
%13371 = torch_c.from_builtin_tensor %13370 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13372 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11003 = arith.constant 1 : index
%dim_11004 = tensor.dim %13372, %c1_11003 : tensor<4x?x4096xf16>
%13373 = flow.tensor.transfer %13372 : tensor<4x?x4096xf16>{%dim_11004} to #hal.device.promise<@__device_1>
%13374 = torch_c.from_builtin_tensor %13373 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13375 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11005 = arith.constant 1 : index
%dim_11006 = tensor.dim %13375, %c1_11005 : tensor<4x?x4096xf16>
%13376 = flow.tensor.transfer %13375 : tensor<4x?x4096xf16>{%dim_11006} to #hal.device.promise<@__device_1>
%13377 = torch_c.from_builtin_tensor %13376 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13378 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11007 = arith.constant 1 : index
%dim_11008 = tensor.dim %13378, %c1_11007 : tensor<4x?x4096xf16>
%13379 = flow.tensor.transfer %13378 : tensor<4x?x4096xf16>{%dim_11008} to #hal.device.promise<@__device_1>
%13380 = torch_c.from_builtin_tensor %13379 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13381 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11009 = arith.constant 1 : index
%dim_11010 = tensor.dim %13381, %c1_11009 : tensor<4x?x4096xf16>
%13382 = flow.tensor.transfer %13381 : tensor<4x?x4096xf16>{%dim_11010} to #hal.device.promise<@__device_1>
%13383 = torch_c.from_builtin_tensor %13382 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13384 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11011 = arith.constant 1 : index
%dim_11012 = tensor.dim %13384, %c1_11011 : tensor<4x?x4096xf16>
%13385 = flow.tensor.transfer %13384 : tensor<4x?x4096xf16>{%dim_11012} to #hal.device.promise<@__device_1>
%13386 = torch_c.from_builtin_tensor %13385 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11013 = torch.constant.int 1
%13387 = torch.aten.add.Tensor %13368, %13295, %int1_11013 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11014 = torch.constant.int 1
%13388 = torch.aten.add.Tensor %13387, %13371, %int1_11014 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11015 = torch.constant.int 1
%13389 = torch.aten.add.Tensor %13388, %13374, %int1_11015 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11016 = torch.constant.int 1
%13390 = torch.aten.add.Tensor %13389, %13377, %int1_11016 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11017 = torch.constant.int 1
%13391 = torch.aten.add.Tensor %13390, %13380, %int1_11017 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11018 = torch.constant.int 1
%13392 = torch.aten.add.Tensor %13391, %13383, %int1_11018 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11019 = torch.constant.int 1
%13393 = torch.aten.add.Tensor %13392, %13386, %int1_11019 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13394 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11020 = arith.constant 1 : index
%dim_11021 = tensor.dim %13394, %c1_11020 : tensor<4x?x4096xf16>
%13395 = flow.tensor.transfer %13394 : tensor<4x?x4096xf16>{%dim_11021} to #hal.device.promise<@__device_2>
%13396 = torch_c.from_builtin_tensor %13395 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13397 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11022 = arith.constant 1 : index
%dim_11023 = tensor.dim %13397, %c1_11022 : tensor<4x?x4096xf16>
%13398 = flow.tensor.transfer %13397 : tensor<4x?x4096xf16>{%dim_11023} to #hal.device.promise<@__device_2>
%13399 = torch_c.from_builtin_tensor %13398 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13400 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11024 = arith.constant 1 : index
%dim_11025 = tensor.dim %13400, %c1_11024 : tensor<4x?x4096xf16>
%13401 = flow.tensor.transfer %13400 : tensor<4x?x4096xf16>{%dim_11025} to #hal.device.promise<@__device_2>
%13402 = torch_c.from_builtin_tensor %13401 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13402, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13403 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11026 = arith.constant 1 : index
%dim_11027 = tensor.dim %13403, %c1_11026 : tensor<4x?x4096xf16>
%13404 = flow.tensor.transfer %13403 : tensor<4x?x4096xf16>{%dim_11027} to #hal.device.promise<@__device_2>
%13405 = torch_c.from_builtin_tensor %13404 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13406 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11028 = arith.constant 1 : index
%dim_11029 = tensor.dim %13406, %c1_11028 : tensor<4x?x4096xf16>
%13407 = flow.tensor.transfer %13406 : tensor<4x?x4096xf16>{%dim_11029} to #hal.device.promise<@__device_2>
%13408 = torch_c.from_builtin_tensor %13407 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13409 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11030 = arith.constant 1 : index
%dim_11031 = tensor.dim %13409, %c1_11030 : tensor<4x?x4096xf16>
%13410 = flow.tensor.transfer %13409 : tensor<4x?x4096xf16>{%dim_11031} to #hal.device.promise<@__device_2>
%13411 = torch_c.from_builtin_tensor %13410 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13412 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11032 = arith.constant 1 : index
%dim_11033 = tensor.dim %13412, %c1_11032 : tensor<4x?x4096xf16>
%13413 = flow.tensor.transfer %13412 : tensor<4x?x4096xf16>{%dim_11033} to #hal.device.promise<@__device_2>
%13414 = torch_c.from_builtin_tensor %13413 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11034 = torch.constant.int 1
%13415 = torch.aten.add.Tensor %13396, %13399, %int1_11034 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11035 = torch.constant.int 1
%13416 = torch.aten.add.Tensor %13415, %13302, %int1_11035 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11036 = torch.constant.int 1
%13417 = torch.aten.add.Tensor %13416, %13402, %int1_11036 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11037 = torch.constant.int 1
%13418 = torch.aten.add.Tensor %13417, %13405, %int1_11037 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11038 = torch.constant.int 1
%13419 = torch.aten.add.Tensor %13418, %13408, %int1_11038 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11039 = torch.constant.int 1
%13420 = torch.aten.add.Tensor %13419, %13411, %int1_11039 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11040 = torch.constant.int 1
%13421 = torch.aten.add.Tensor %13420, %13414, %int1_11040 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13422 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11041 = arith.constant 1 : index
%dim_11042 = tensor.dim %13422, %c1_11041 : tensor<4x?x4096xf16>
%13423 = flow.tensor.transfer %13422 : tensor<4x?x4096xf16>{%dim_11042} to #hal.device.promise<@__device_3>
%13424 = torch_c.from_builtin_tensor %13423 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13425 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11043 = arith.constant 1 : index
%dim_11044 = tensor.dim %13425, %c1_11043 : tensor<4x?x4096xf16>
%13426 = flow.tensor.transfer %13425 : tensor<4x?x4096xf16>{%dim_11044} to #hal.device.promise<@__device_3>
%13427 = torch_c.from_builtin_tensor %13426 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13428 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11045 = arith.constant 1 : index
%dim_11046 = tensor.dim %13428, %c1_11045 : tensor<4x?x4096xf16>
%13429 = flow.tensor.transfer %13428 : tensor<4x?x4096xf16>{%dim_11046} to #hal.device.promise<@__device_3>
%13430 = torch_c.from_builtin_tensor %13429 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13431 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11047 = arith.constant 1 : index
%dim_11048 = tensor.dim %13431, %c1_11047 : tensor<4x?x4096xf16>
%13432 = flow.tensor.transfer %13431 : tensor<4x?x4096xf16>{%dim_11048} to #hal.device.promise<@__device_3>
%13433 = torch_c.from_builtin_tensor %13432 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13434 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11049 = arith.constant 1 : index
%dim_11050 = tensor.dim %13434, %c1_11049 : tensor<4x?x4096xf16>
%13435 = flow.tensor.transfer %13434 : tensor<4x?x4096xf16>{%dim_11050} to #hal.device.promise<@__device_3>
%13436 = torch_c.from_builtin_tensor %13435 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13437 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11051 = arith.constant 1 : index
%dim_11052 = tensor.dim %13437, %c1_11051 : tensor<4x?x4096xf16>
%13438 = flow.tensor.transfer %13437 : tensor<4x?x4096xf16>{%dim_11052} to #hal.device.promise<@__device_3>
%13439 = torch_c.from_builtin_tensor %13438 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13440 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11053 = arith.constant 1 : index
%dim_11054 = tensor.dim %13440, %c1_11053 : tensor<4x?x4096xf16>
%13441 = flow.tensor.transfer %13440 : tensor<4x?x4096xf16>{%dim_11054} to #hal.device.promise<@__device_3>
%13442 = torch_c.from_builtin_tensor %13441 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11055 = torch.constant.int 1
%13443 = torch.aten.add.Tensor %13424, %13427, %int1_11055 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11056 = torch.constant.int 1
%13444 = torch.aten.add.Tensor %13443, %13430, %int1_11056 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11057 = torch.constant.int 1
%13445 = torch.aten.add.Tensor %13444, %13309, %int1_11057 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11058 = torch.constant.int 1
%13446 = torch.aten.add.Tensor %13445, %13433, %int1_11058 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11059 = torch.constant.int 1
%13447 = torch.aten.add.Tensor %13446, %13436, %int1_11059 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11060 = torch.constant.int 1
%13448 = torch.aten.add.Tensor %13447, %13439, %int1_11060 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11061 = torch.constant.int 1
%13449 = torch.aten.add.Tensor %13448, %13442, %int1_11061 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13450 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11062 = arith.constant 1 : index
%dim_11063 = tensor.dim %13450, %c1_11062 : tensor<4x?x4096xf16>
%13451 = flow.tensor.transfer %13450 : tensor<4x?x4096xf16>{%dim_11063} to #hal.device.promise<@__device_4>
%13452 = torch_c.from_builtin_tensor %13451 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13453 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11064 = arith.constant 1 : index
%dim_11065 = tensor.dim %13453, %c1_11064 : tensor<4x?x4096xf16>
%13454 = flow.tensor.transfer %13453 : tensor<4x?x4096xf16>{%dim_11065} to #hal.device.promise<@__device_4>
%13455 = torch_c.from_builtin_tensor %13454 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13456 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11066 = arith.constant 1 : index
%dim_11067 = tensor.dim %13456, %c1_11066 : tensor<4x?x4096xf16>
%13457 = flow.tensor.transfer %13456 : tensor<4x?x4096xf16>{%dim_11067} to #hal.device.promise<@__device_4>
%13458 = torch_c.from_builtin_tensor %13457 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13459 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11068 = arith.constant 1 : index
%dim_11069 = tensor.dim %13459, %c1_11068 : tensor<4x?x4096xf16>
%13460 = flow.tensor.transfer %13459 : tensor<4x?x4096xf16>{%dim_11069} to #hal.device.promise<@__device_4>
%13461 = torch_c.from_builtin_tensor %13460 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13462 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11070 = arith.constant 1 : index
%dim_11071 = tensor.dim %13462, %c1_11070 : tensor<4x?x4096xf16>
%13463 = flow.tensor.transfer %13462 : tensor<4x?x4096xf16>{%dim_11071} to #hal.device.promise<@__device_4>
%13464 = torch_c.from_builtin_tensor %13463 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13465 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11072 = arith.constant 1 : index
%dim_11073 = tensor.dim %13465, %c1_11072 : tensor<4x?x4096xf16>
%13466 = flow.tensor.transfer %13465 : tensor<4x?x4096xf16>{%dim_11073} to #hal.device.promise<@__device_4>
%13467 = torch_c.from_builtin_tensor %13466 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13468 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11074 = arith.constant 1 : index
%dim_11075 = tensor.dim %13468, %c1_11074 : tensor<4x?x4096xf16>
%13469 = flow.tensor.transfer %13468 : tensor<4x?x4096xf16>{%dim_11075} to #hal.device.promise<@__device_4>
%13470 = torch_c.from_builtin_tensor %13469 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11076 = torch.constant.int 1
%13471 = torch.aten.add.Tensor %13452, %13455, %int1_11076 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11077 = torch.constant.int 1
%13472 = torch.aten.add.Tensor %13471, %13458, %int1_11077 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11078 = torch.constant.int 1
%13473 = torch.aten.add.Tensor %13472, %13461, %int1_11078 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11079 = torch.constant.int 1
%13474 = torch.aten.add.Tensor %13473, %13316, %int1_11079 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11080 = torch.constant.int 1
%13475 = torch.aten.add.Tensor %13474, %13464, %int1_11080 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11081 = torch.constant.int 1
%13476 = torch.aten.add.Tensor %13475, %13467, %int1_11081 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11082 = torch.constant.int 1
%13477 = torch.aten.add.Tensor %13476, %13470, %int1_11082 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13478 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11083 = arith.constant 1 : index
%dim_11084 = tensor.dim %13478, %c1_11083 : tensor<4x?x4096xf16>
%13479 = flow.tensor.transfer %13478 : tensor<4x?x4096xf16>{%dim_11084} to #hal.device.promise<@__device_5>
%13480 = torch_c.from_builtin_tensor %13479 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13481 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11085 = arith.constant 1 : index
%dim_11086 = tensor.dim %13481, %c1_11085 : tensor<4x?x4096xf16>
%13482 = flow.tensor.transfer %13481 : tensor<4x?x4096xf16>{%dim_11086} to #hal.device.promise<@__device_5>
%13483 = torch_c.from_builtin_tensor %13482 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13484 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11087 = arith.constant 1 : index
%dim_11088 = tensor.dim %13484, %c1_11087 : tensor<4x?x4096xf16>
%13485 = flow.tensor.transfer %13484 : tensor<4x?x4096xf16>{%dim_11088} to #hal.device.promise<@__device_5>
%13486 = torch_c.from_builtin_tensor %13485 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13487 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11089 = arith.constant 1 : index
%dim_11090 = tensor.dim %13487, %c1_11089 : tensor<4x?x4096xf16>
%13488 = flow.tensor.transfer %13487 : tensor<4x?x4096xf16>{%dim_11090} to #hal.device.promise<@__device_5>
%13489 = torch_c.from_builtin_tensor %13488 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13490 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11091 = arith.constant 1 : index
%dim_11092 = tensor.dim %13490, %c1_11091 : tensor<4x?x4096xf16>
%13491 = flow.tensor.transfer %13490 : tensor<4x?x4096xf16>{%dim_11092} to #hal.device.promise<@__device_5>
%13492 = torch_c.from_builtin_tensor %13491 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13493 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11093 = arith.constant 1 : index
%dim_11094 = tensor.dim %13493, %c1_11093 : tensor<4x?x4096xf16>
%13494 = flow.tensor.transfer %13493 : tensor<4x?x4096xf16>{%dim_11094} to #hal.device.promise<@__device_5>
%13495 = torch_c.from_builtin_tensor %13494 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13496 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11095 = arith.constant 1 : index
%dim_11096 = tensor.dim %13496, %c1_11095 : tensor<4x?x4096xf16>
%13497 = flow.tensor.transfer %13496 : tensor<4x?x4096xf16>{%dim_11096} to #hal.device.promise<@__device_5>
%13498 = torch_c.from_builtin_tensor %13497 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11097 = torch.constant.int 1
%13499 = torch.aten.add.Tensor %13480, %13483, %int1_11097 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11098 = torch.constant.int 1
%13500 = torch.aten.add.Tensor %13499, %13486, %int1_11098 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11099 = torch.constant.int 1
%13501 = torch.aten.add.Tensor %13500, %13489, %int1_11099 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11100 = torch.constant.int 1
%13502 = torch.aten.add.Tensor %13501, %13492, %int1_11100 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11101 = torch.constant.int 1
%13503 = torch.aten.add.Tensor %13502, %13323, %int1_11101 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11102 = torch.constant.int 1
%13504 = torch.aten.add.Tensor %13503, %13495, %int1_11102 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11103 = torch.constant.int 1
%13505 = torch.aten.add.Tensor %13504, %13498, %int1_11103 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13506 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11104 = arith.constant 1 : index
%dim_11105 = tensor.dim %13506, %c1_11104 : tensor<4x?x4096xf16>
%13507 = flow.tensor.transfer %13506 : tensor<4x?x4096xf16>{%dim_11105} to #hal.device.promise<@__device_6>
%13508 = torch_c.from_builtin_tensor %13507 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13509 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11106 = arith.constant 1 : index
%dim_11107 = tensor.dim %13509, %c1_11106 : tensor<4x?x4096xf16>
%13510 = flow.tensor.transfer %13509 : tensor<4x?x4096xf16>{%dim_11107} to #hal.device.promise<@__device_6>
%13511 = torch_c.from_builtin_tensor %13510 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13512 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11108 = arith.constant 1 : index
%dim_11109 = tensor.dim %13512, %c1_11108 : tensor<4x?x4096xf16>
%13513 = flow.tensor.transfer %13512 : tensor<4x?x4096xf16>{%dim_11109} to #hal.device.promise<@__device_6>
%13514 = torch_c.from_builtin_tensor %13513 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13515 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11110 = arith.constant 1 : index
%dim_11111 = tensor.dim %13515, %c1_11110 : tensor<4x?x4096xf16>
%13516 = flow.tensor.transfer %13515 : tensor<4x?x4096xf16>{%dim_11111} to #hal.device.promise<@__device_6>
%13517 = torch_c.from_builtin_tensor %13516 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13518 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11112 = arith.constant 1 : index
%dim_11113 = tensor.dim %13518, %c1_11112 : tensor<4x?x4096xf16>
%13519 = flow.tensor.transfer %13518 : tensor<4x?x4096xf16>{%dim_11113} to #hal.device.promise<@__device_6>
%13520 = torch_c.from_builtin_tensor %13519 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13521 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11114 = arith.constant 1 : index
%dim_11115 = tensor.dim %13521, %c1_11114 : tensor<4x?x4096xf16>
%13522 = flow.tensor.transfer %13521 : tensor<4x?x4096xf16>{%dim_11115} to #hal.device.promise<@__device_6>
%13523 = torch_c.from_builtin_tensor %13522 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13524 = torch_c.to_builtin_tensor %13337 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11116 = arith.constant 1 : index
%dim_11117 = tensor.dim %13524, %c1_11116 : tensor<4x?x4096xf16>
%13525 = flow.tensor.transfer %13524 : tensor<4x?x4096xf16>{%dim_11117} to #hal.device.promise<@__device_6>
%13526 = torch_c.from_builtin_tensor %13525 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11118 = torch.constant.int 1
%13527 = torch.aten.add.Tensor %13508, %13511, %int1_11118 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11119 = torch.constant.int 1
%13528 = torch.aten.add.Tensor %13527, %13514, %int1_11119 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11120 = torch.constant.int 1
%13529 = torch.aten.add.Tensor %13528, %13517, %int1_11120 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11121 = torch.constant.int 1
%13530 = torch.aten.add.Tensor %13529, %13520, %int1_11121 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11122 = torch.constant.int 1
%13531 = torch.aten.add.Tensor %13530, %13523, %int1_11122 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11123 = torch.constant.int 1
%13532 = torch.aten.add.Tensor %13531, %13330, %int1_11123 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11124 = torch.constant.int 1
%13533 = torch.aten.add.Tensor %13532, %13526, %int1_11124 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13534 = torch_c.to_builtin_tensor %13288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11125 = arith.constant 1 : index
%dim_11126 = tensor.dim %13534, %c1_11125 : tensor<4x?x4096xf16>
%13535 = flow.tensor.transfer %13534 : tensor<4x?x4096xf16>{%dim_11126} to #hal.device.promise<@__device_7>
%13536 = torch_c.from_builtin_tensor %13535 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13537 = torch_c.to_builtin_tensor %13295 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11127 = arith.constant 1 : index
%dim_11128 = tensor.dim %13537, %c1_11127 : tensor<4x?x4096xf16>
%13538 = flow.tensor.transfer %13537 : tensor<4x?x4096xf16>{%dim_11128} to #hal.device.promise<@__device_7>
%13539 = torch_c.from_builtin_tensor %13538 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13540 = torch_c.to_builtin_tensor %13302 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11129 = arith.constant 1 : index
%dim_11130 = tensor.dim %13540, %c1_11129 : tensor<4x?x4096xf16>
%13541 = flow.tensor.transfer %13540 : tensor<4x?x4096xf16>{%dim_11130} to #hal.device.promise<@__device_7>
%13542 = torch_c.from_builtin_tensor %13541 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13543 = torch_c.to_builtin_tensor %13309 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11131 = arith.constant 1 : index
%dim_11132 = tensor.dim %13543, %c1_11131 : tensor<4x?x4096xf16>
%13544 = flow.tensor.transfer %13543 : tensor<4x?x4096xf16>{%dim_11132} to #hal.device.promise<@__device_7>
%13545 = torch_c.from_builtin_tensor %13544 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13546 = torch_c.to_builtin_tensor %13316 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11133 = arith.constant 1 : index
%dim_11134 = tensor.dim %13546, %c1_11133 : tensor<4x?x4096xf16>
%13547 = flow.tensor.transfer %13546 : tensor<4x?x4096xf16>{%dim_11134} to #hal.device.promise<@__device_7>
%13548 = torch_c.from_builtin_tensor %13547 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13549 = torch_c.to_builtin_tensor %13323 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11135 = arith.constant 1 : index
%dim_11136 = tensor.dim %13549, %c1_11135 : tensor<4x?x4096xf16>
%13550 = flow.tensor.transfer %13549 : tensor<4x?x4096xf16>{%dim_11136} to #hal.device.promise<@__device_7>
%13551 = torch_c.from_builtin_tensor %13550 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%13552 = torch_c.to_builtin_tensor %13330 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_11137 = arith.constant 1 : index
%dim_11138 = tensor.dim %13552, %c1_11137 : tensor<4x?x4096xf16>
%13553 = flow.tensor.transfer %13552 : tensor<4x?x4096xf16>{%dim_11138} to #hal.device.promise<@__device_7>
%13554 = torch_c.from_builtin_tensor %13553 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11139 = torch.constant.int 1
%13555 = torch.aten.add.Tensor %13536, %13539, %int1_11139 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11140 = torch.constant.int 1
%13556 = torch.aten.add.Tensor %13555, %13542, %int1_11140 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11141 = torch.constant.int 1
%13557 = torch.aten.add.Tensor %13556, %13545, %int1_11141 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11142 = torch.constant.int 1
%13558 = torch.aten.add.Tensor %13557, %13548, %int1_11142 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11143 = torch.constant.int 1
%13559 = torch.aten.add.Tensor %13558, %13551, %int1_11143 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11144 = torch.constant.int 1
%13560 = torch.aten.add.Tensor %13559, %13554, %int1_11144 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11145 = torch.constant.int 1
%13561 = torch.aten.add.Tensor %13560, %13337, %int1_11145 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11146 = torch.constant.int 1
%13562 = torch.aten.add.Tensor %13042, %13365, %int1_11146 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11147 = torch.constant.int 1
%13563 = torch.aten.add.Tensor %13043, %13393, %int1_11147 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11148 = torch.constant.int 1
%13564 = torch.aten.add.Tensor %13044, %13421, %int1_11148 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11149 = torch.constant.int 1
%13565 = torch.aten.add.Tensor %13045, %13449, %int1_11149 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11150 = torch.constant.int 1
%13566 = torch.aten.add.Tensor %13046, %13477, %int1_11150 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11151 = torch.constant.int 1
%13567 = torch.aten.add.Tensor %13047, %13505, %int1_11151 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11152 = torch.constant.int 1
%13568 = torch.aten.add.Tensor %13048, %13533, %int1_11152 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11153 = torch.constant.int 1
%13569 = torch.aten.add.Tensor %13049, %13561, %int1_11153 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_11154 = torch.constant.int 6
%13570 = torch.prims.convert_element_type %13562, %int6_11154 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11155 = torch.constant.int 6
%13571 = torch.prims.convert_element_type %13563, %int6_11155 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11156 = torch.constant.int 6
%13572 = torch.prims.convert_element_type %13564, %int6_11156 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11157 = torch.constant.int 6
%13573 = torch.prims.convert_element_type %13565, %int6_11157 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11158 = torch.constant.int 6
%13574 = torch.prims.convert_element_type %13566, %int6_11158 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11159 = torch.constant.int 6
%13575 = torch.prims.convert_element_type %13567, %int6_11159 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11160 = torch.constant.int 6
%13576 = torch.prims.convert_element_type %13568, %int6_11160 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_11161 = torch.constant.int 6
%13577 = torch.prims.convert_element_type %13569, %int6_11161 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11162 = torch.constant.int 2
%13578 = torch.aten.pow.Tensor_Scalar %13570, %int2_11162 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11163 = torch.constant.int 2
%13579 = torch.aten.pow.Tensor_Scalar %13571, %int2_11163 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11164 = torch.constant.int 2
%13580 = torch.aten.pow.Tensor_Scalar %13572, %int2_11164 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11165 = torch.constant.int 2
%13581 = torch.aten.pow.Tensor_Scalar %13573, %int2_11165 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11166 = torch.constant.int 2
%13582 = torch.aten.pow.Tensor_Scalar %13574, %int2_11166 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11167 = torch.constant.int 2
%13583 = torch.aten.pow.Tensor_Scalar %13575, %int2_11167 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11168 = torch.constant.int 2
%13584 = torch.aten.pow.Tensor_Scalar %13576, %int2_11168 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_11169 = torch.constant.int 2
%13585 = torch.aten.pow.Tensor_Scalar %13577, %int2_11169 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_11170 = torch.constant.int -1
%13586 = torch.prim.ListConstruct %int-1_11170 : (!torch.int) -> !torch.list<int>
%true_11171 = torch.constant.bool true
%none_11172 = torch.constant.none
%13587 = torch.aten.mean.dim %13578, %13586, %true_11171, %none_11172 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11173 = torch.constant.int -1
%13588 = torch.prim.ListConstruct %int-1_11173 : (!torch.int) -> !torch.list<int>
%true_11174 = torch.constant.bool true
%none_11175 = torch.constant.none
%13589 = torch.aten.mean.dim %13579, %13588, %true_11174, %none_11175 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11176 = torch.constant.int -1
%13590 = torch.prim.ListConstruct %int-1_11176 : (!torch.int) -> !torch.list<int>
%true_11177 = torch.constant.bool true
%none_11178 = torch.constant.none
%13591 = torch.aten.mean.dim %13580, %13590, %true_11177, %none_11178 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11179 = torch.constant.int -1
%13592 = torch.prim.ListConstruct %int-1_11179 : (!torch.int) -> !torch.list<int>
%true_11180 = torch.constant.bool true
%none_11181 = torch.constant.none
%13593 = torch.aten.mean.dim %13581, %13592, %true_11180, %none_11181 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11182 = torch.constant.int -1
%13594 = torch.prim.ListConstruct %int-1_11182 : (!torch.int) -> !torch.list<int>
%true_11183 = torch.constant.bool true
%none_11184 = torch.constant.none
%13595 = torch.aten.mean.dim %13582, %13594, %true_11183, %none_11184 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11185 = torch.constant.int -1
%13596 = torch.prim.ListConstruct %int-1_11185 : (!torch.int) -> !torch.list<int>
%true_11186 = torch.constant.bool true
%none_11187 = torch.constant.none
%13597 = torch.aten.mean.dim %13583, %13596, %true_11186, %none_11187 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11188 = torch.constant.int -1
%13598 = torch.prim.ListConstruct %int-1_11188 : (!torch.int) -> !torch.list<int>
%true_11189 = torch.constant.bool true
%none_11190 = torch.constant.none
%13599 = torch.aten.mean.dim %13584, %13598, %true_11189, %none_11190 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_11191 = torch.constant.int -1
%13600 = torch.prim.ListConstruct %int-1_11191 : (!torch.int) -> !torch.list<int>
%true_11192 = torch.constant.bool true
%none_11193 = torch.constant.none
%13601 = torch.aten.mean.dim %13585, %13600, %true_11192, %none_11193 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11194 = torch.constant.float 9.9999997473787516E-6
%int1_11195 = torch.constant.int 1
%13602 = torch.aten.add.Scalar %13587, %float9.999990e-06_11194, %int1_11195 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11196 = torch.constant.float 9.9999997473787516E-6
%int1_11197 = torch.constant.int 1
%13603 = torch.aten.add.Scalar %13589, %float9.999990e-06_11196, %int1_11197 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11198 = torch.constant.float 9.9999997473787516E-6
%int1_11199 = torch.constant.int 1
%13604 = torch.aten.add.Scalar %13591, %float9.999990e-06_11198, %int1_11199 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11200 = torch.constant.float 9.9999997473787516E-6
%int1_11201 = torch.constant.int 1
%13605 = torch.aten.add.Scalar %13593, %float9.999990e-06_11200, %int1_11201 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11202 = torch.constant.float 9.9999997473787516E-6
%int1_11203 = torch.constant.int 1
%13606 = torch.aten.add.Scalar %13595, %float9.999990e-06_11202, %int1_11203 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11204 = torch.constant.float 9.9999997473787516E-6
%int1_11205 = torch.constant.int 1
%13607 = torch.aten.add.Scalar %13597, %float9.999990e-06_11204, %int1_11205 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11206 = torch.constant.float 9.9999997473787516E-6
%int1_11207 = torch.constant.int 1
%13608 = torch.aten.add.Scalar %13599, %float9.999990e-06_11206, %int1_11207 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_11208 = torch.constant.float 9.9999997473787516E-6
%int1_11209 = torch.constant.int 1
%13609 = torch.aten.add.Scalar %13601, %float9.999990e-06_11208, %int1_11209 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13610 = torch.aten.rsqrt %13602 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13611 = torch.aten.rsqrt %13603 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13612 = torch.aten.rsqrt %13604 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13613 = torch.aten.rsqrt %13605 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13614 = torch.aten.rsqrt %13606 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13615 = torch.aten.rsqrt %13607 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13616 = torch.aten.rsqrt %13608 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13617 = torch.aten.rsqrt %13609 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %13617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%13618 = torch.aten.mul.Tensor %13570, %13610 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13619 = torch.aten.mul.Tensor %13571, %13611 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13620 = torch.aten.mul.Tensor %13572, %13612 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13621 = torch.aten.mul.Tensor %13573, %13613 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13622 = torch.aten.mul.Tensor %13574, %13614 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13623 = torch.aten.mul.Tensor %13575, %13615 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13624 = torch.aten.mul.Tensor %13576, %13616 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13625 = torch.aten.mul.Tensor %13577, %13617 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13626 = torch.aten.mul.Tensor %440, %13618 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13627 = torch.aten.mul.Tensor %441, %13619 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13628 = torch.aten.mul.Tensor %442, %13620 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13629 = torch.aten.mul.Tensor %443, %13621 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13630 = torch.aten.mul.Tensor %444, %13622 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13631 = torch.aten.mul.Tensor %445, %13623 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13632 = torch.aten.mul.Tensor %446, %13624 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%13633 = torch.aten.mul.Tensor %447, %13625 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %13633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_11210 = torch.constant.int 5
%13634 = torch.prims.convert_element_type %13626, %int5_11210 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11211 = torch.constant.int 5
%13635 = torch.prims.convert_element_type %13627, %int5_11211 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11212 = torch.constant.int 5
%13636 = torch.prims.convert_element_type %13628, %int5_11212 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11213 = torch.constant.int 5
%13637 = torch.prims.convert_element_type %13629, %int5_11213 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11214 = torch.constant.int 5
%13638 = torch.prims.convert_element_type %13630, %int5_11214 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11215 = torch.constant.int 5
%13639 = torch.prims.convert_element_type %13631, %int5_11215 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11216 = torch.constant.int 5
%13640 = torch.prims.convert_element_type %13632, %int5_11216 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_11217 = torch.constant.int 5
%13641 = torch.prims.convert_element_type %13633, %int5_11217 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %13641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_11218 = torch.constant.int 1
%int0_11219 = torch.constant.int 0
%13642 = torch.prim.ListConstruct %int1_11218, %int0_11219 : (!torch.int, !torch.int) -> !torch.list<int>
%13643 = torch.aten.permute %448, %13642 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11220 = torch.constant.int 1
%int0_11221 = torch.constant.int 0
%13644 = torch.prim.ListConstruct %int1_11220, %int0_11221 : (!torch.int, !torch.int) -> !torch.list<int>
%13645 = torch.aten.permute %449, %13644 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11222 = torch.constant.int 1
%int0_11223 = torch.constant.int 0
%13646 = torch.prim.ListConstruct %int1_11222, %int0_11223 : (!torch.int, !torch.int) -> !torch.list<int>
%13647 = torch.aten.permute %450, %13646 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11224 = torch.constant.int 1
%int0_11225 = torch.constant.int 0
%13648 = torch.prim.ListConstruct %int1_11224, %int0_11225 : (!torch.int, !torch.int) -> !torch.list<int>
%13649 = torch.aten.permute %451, %13648 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11226 = torch.constant.int 1
%int0_11227 = torch.constant.int 0
%13650 = torch.prim.ListConstruct %int1_11226, %int0_11227 : (!torch.int, !torch.int) -> !torch.list<int>
%13651 = torch.aten.permute %452, %13650 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11228 = torch.constant.int 1
%int0_11229 = torch.constant.int 0
%13652 = torch.prim.ListConstruct %int1_11228, %int0_11229 : (!torch.int, !torch.int) -> !torch.list<int>
%13653 = torch.aten.permute %453, %13652 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11230 = torch.constant.int 1
%int0_11231 = torch.constant.int 0
%13654 = torch.prim.ListConstruct %int1_11230, %int0_11231 : (!torch.int, !torch.int) -> !torch.list<int>
%13655 = torch.aten.permute %454, %13654 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_11232 = torch.constant.int 1
%int0_11233 = torch.constant.int 0
%13656 = torch.prim.ListConstruct %int1_11232, %int0_11233 : (!torch.int, !torch.int) -> !torch.list<int>
%13657 = torch.aten.permute %455, %13656 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_11234 = torch.constant.int 4
%13658 = torch.aten.mul.int %int4_11234, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11235 = torch.constant.int 4096
%13659 = torch.prim.ListConstruct %13658, %int4096_11235 : (!torch.int, !torch.int) -> !torch.list<int>
%13660 = torch.aten.view %13634, %13659 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13660, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13661 = torch.aten.mm %13660, %13643 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13661, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11236 = torch.constant.int 4
%int512_11237 = torch.constant.int 512
%13662 = torch.prim.ListConstruct %int4_11236, %2482, %int512_11237 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13663 = torch.aten.view %13661, %13662 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11238 = torch.constant.int 4
%13664 = torch.aten.mul.int %int4_11238, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11239 = torch.constant.int 4096
%13665 = torch.prim.ListConstruct %13664, %int4096_11239 : (!torch.int, !torch.int) -> !torch.list<int>
%13666 = torch.aten.view %13635, %13665 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13666, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13667 = torch.aten.mm %13666, %13645 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13667, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11240 = torch.constant.int 4
%int512_11241 = torch.constant.int 512
%13668 = torch.prim.ListConstruct %int4_11240, %2482, %int512_11241 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13669 = torch.aten.view %13667, %13668 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11242 = torch.constant.int 4
%13670 = torch.aten.mul.int %int4_11242, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11243 = torch.constant.int 4096
%13671 = torch.prim.ListConstruct %13670, %int4096_11243 : (!torch.int, !torch.int) -> !torch.list<int>
%13672 = torch.aten.view %13636, %13671 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13672, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13673 = torch.aten.mm %13672, %13647 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13673, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11244 = torch.constant.int 4
%int512_11245 = torch.constant.int 512
%13674 = torch.prim.ListConstruct %int4_11244, %2482, %int512_11245 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13675 = torch.aten.view %13673, %13674 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11246 = torch.constant.int 4
%13676 = torch.aten.mul.int %int4_11246, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11247 = torch.constant.int 4096
%13677 = torch.prim.ListConstruct %13676, %int4096_11247 : (!torch.int, !torch.int) -> !torch.list<int>
%13678 = torch.aten.view %13637, %13677 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13678, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13679 = torch.aten.mm %13678, %13649 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13679, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11248 = torch.constant.int 4
%int512_11249 = torch.constant.int 512
%13680 = torch.prim.ListConstruct %int4_11248, %2482, %int512_11249 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13681 = torch.aten.view %13679, %13680 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11250 = torch.constant.int 4
%13682 = torch.aten.mul.int %int4_11250, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11251 = torch.constant.int 4096
%13683 = torch.prim.ListConstruct %13682, %int4096_11251 : (!torch.int, !torch.int) -> !torch.list<int>
%13684 = torch.aten.view %13638, %13683 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13684, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13685 = torch.aten.mm %13684, %13651 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13685, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11252 = torch.constant.int 4
%int512_11253 = torch.constant.int 512
%13686 = torch.prim.ListConstruct %int4_11252, %2482, %int512_11253 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13687 = torch.aten.view %13685, %13686 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11254 = torch.constant.int 4
%13688 = torch.aten.mul.int %int4_11254, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11255 = torch.constant.int 4096
%13689 = torch.prim.ListConstruct %13688, %int4096_11255 : (!torch.int, !torch.int) -> !torch.list<int>
%13690 = torch.aten.view %13639, %13689 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13690, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13691 = torch.aten.mm %13690, %13653 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13691, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11256 = torch.constant.int 4
%int512_11257 = torch.constant.int 512
%13692 = torch.prim.ListConstruct %int4_11256, %2482, %int512_11257 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13693 = torch.aten.view %13691, %13692 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11258 = torch.constant.int 4
%13694 = torch.aten.mul.int %int4_11258, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11259 = torch.constant.int 4096
%13695 = torch.prim.ListConstruct %13694, %int4096_11259 : (!torch.int, !torch.int) -> !torch.list<int>
%13696 = torch.aten.view %13640, %13695 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13696, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13697 = torch.aten.mm %13696, %13655 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13697, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11260 = torch.constant.int 4
%int512_11261 = torch.constant.int 512
%13698 = torch.prim.ListConstruct %int4_11260, %2482, %int512_11261 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13699 = torch.aten.view %13697, %13698 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_11262 = torch.constant.int 4
%13700 = torch.aten.mul.int %int4_11262, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11263 = torch.constant.int 4096
%13701 = torch.prim.ListConstruct %13700, %int4096_11263 : (!torch.int, !torch.int) -> !torch.list<int>
%13702 = torch.aten.view %13641, %13701 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13702, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13703 = torch.aten.mm %13702, %13657 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %13703, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_11264 = torch.constant.int 4
%int512_11265 = torch.constant.int 512
%13704 = torch.prim.ListConstruct %int4_11264, %2482, %int512_11265 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13705 = torch.aten.view %13703, %13704 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %13705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_11266 = torch.constant.int 1
%int0_11267 = torch.constant.int 0
%13706 = torch.prim.ListConstruct %int1_11266, %int0_11267 : (!torch.int, !torch.int) -> !torch.list<int>
%13707 = torch.aten.permute %456, %13706 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11268 = torch.constant.int 1
%int0_11269 = torch.constant.int 0
%13708 = torch.prim.ListConstruct %int1_11268, %int0_11269 : (!torch.int, !torch.int) -> !torch.list<int>
%13709 = torch.aten.permute %457, %13708 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11270 = torch.constant.int 1
%int0_11271 = torch.constant.int 0
%13710 = torch.prim.ListConstruct %int1_11270, %int0_11271 : (!torch.int, !torch.int) -> !torch.list<int>
%13711 = torch.aten.permute %458, %13710 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11272 = torch.constant.int 1
%int0_11273 = torch.constant.int 0
%13712 = torch.prim.ListConstruct %int1_11272, %int0_11273 : (!torch.int, !torch.int) -> !torch.list<int>
%13713 = torch.aten.permute %459, %13712 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11274 = torch.constant.int 1
%int0_11275 = torch.constant.int 0
%13714 = torch.prim.ListConstruct %int1_11274, %int0_11275 : (!torch.int, !torch.int) -> !torch.list<int>
%13715 = torch.aten.permute %460, %13714 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11276 = torch.constant.int 1
%int0_11277 = torch.constant.int 0
%13716 = torch.prim.ListConstruct %int1_11276, %int0_11277 : (!torch.int, !torch.int) -> !torch.list<int>
%13717 = torch.aten.permute %461, %13716 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11278 = torch.constant.int 1
%int0_11279 = torch.constant.int 0
%13718 = torch.prim.ListConstruct %int1_11278, %int0_11279 : (!torch.int, !torch.int) -> !torch.list<int>
%13719 = torch.aten.permute %462, %13718 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11280 = torch.constant.int 1
%int0_11281 = torch.constant.int 0
%13720 = torch.prim.ListConstruct %int1_11280, %int0_11281 : (!torch.int, !torch.int) -> !torch.list<int>
%13721 = torch.aten.permute %463, %13720 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_11282 = torch.constant.int 4
%13722 = torch.aten.mul.int %int4_11282, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11283 = torch.constant.int 4096
%13723 = torch.prim.ListConstruct %13722, %int4096_11283 : (!torch.int, !torch.int) -> !torch.list<int>
%13724 = torch.aten.view %13634, %13723 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13724, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13725 = torch.aten.mm %13724, %13707 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13725, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11284 = torch.constant.int 4
%int128_11285 = torch.constant.int 128
%13726 = torch.prim.ListConstruct %int4_11284, %2482, %int128_11285 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13727 = torch.aten.view %13725, %13726 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11286 = torch.constant.int 4
%13728 = torch.aten.mul.int %int4_11286, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11287 = torch.constant.int 4096
%13729 = torch.prim.ListConstruct %13728, %int4096_11287 : (!torch.int, !torch.int) -> !torch.list<int>
%13730 = torch.aten.view %13635, %13729 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13730, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13731 = torch.aten.mm %13730, %13709 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13731, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11288 = torch.constant.int 4
%int128_11289 = torch.constant.int 128
%13732 = torch.prim.ListConstruct %int4_11288, %2482, %int128_11289 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13733 = torch.aten.view %13731, %13732 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11290 = torch.constant.int 4
%13734 = torch.aten.mul.int %int4_11290, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11291 = torch.constant.int 4096
%13735 = torch.prim.ListConstruct %13734, %int4096_11291 : (!torch.int, !torch.int) -> !torch.list<int>
%13736 = torch.aten.view %13636, %13735 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13736, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13737 = torch.aten.mm %13736, %13711 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13737, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11292 = torch.constant.int 4
%int128_11293 = torch.constant.int 128
%13738 = torch.prim.ListConstruct %int4_11292, %2482, %int128_11293 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13739 = torch.aten.view %13737, %13738 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11294 = torch.constant.int 4
%13740 = torch.aten.mul.int %int4_11294, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11295 = torch.constant.int 4096
%13741 = torch.prim.ListConstruct %13740, %int4096_11295 : (!torch.int, !torch.int) -> !torch.list<int>
%13742 = torch.aten.view %13637, %13741 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13742, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13743 = torch.aten.mm %13742, %13713 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13743, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11296 = torch.constant.int 4
%int128_11297 = torch.constant.int 128
%13744 = torch.prim.ListConstruct %int4_11296, %2482, %int128_11297 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13745 = torch.aten.view %13743, %13744 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11298 = torch.constant.int 4
%13746 = torch.aten.mul.int %int4_11298, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11299 = torch.constant.int 4096
%13747 = torch.prim.ListConstruct %13746, %int4096_11299 : (!torch.int, !torch.int) -> !torch.list<int>
%13748 = torch.aten.view %13638, %13747 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13748, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13749 = torch.aten.mm %13748, %13715 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13749, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11300 = torch.constant.int 4
%int128_11301 = torch.constant.int 128
%13750 = torch.prim.ListConstruct %int4_11300, %2482, %int128_11301 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13751 = torch.aten.view %13749, %13750 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11302 = torch.constant.int 4
%13752 = torch.aten.mul.int %int4_11302, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11303 = torch.constant.int 4096
%13753 = torch.prim.ListConstruct %13752, %int4096_11303 : (!torch.int, !torch.int) -> !torch.list<int>
%13754 = torch.aten.view %13639, %13753 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13754, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13755 = torch.aten.mm %13754, %13717 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13755, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11304 = torch.constant.int 4
%int128_11305 = torch.constant.int 128
%13756 = torch.prim.ListConstruct %int4_11304, %2482, %int128_11305 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13757 = torch.aten.view %13755, %13756 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11306 = torch.constant.int 4
%13758 = torch.aten.mul.int %int4_11306, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11307 = torch.constant.int 4096
%13759 = torch.prim.ListConstruct %13758, %int4096_11307 : (!torch.int, !torch.int) -> !torch.list<int>
%13760 = torch.aten.view %13640, %13759 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13760, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13761 = torch.aten.mm %13760, %13719 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13761, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11308 = torch.constant.int 4
%int128_11309 = torch.constant.int 128
%13762 = torch.prim.ListConstruct %int4_11308, %2482, %int128_11309 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13763 = torch.aten.view %13761, %13762 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11310 = torch.constant.int 4
%13764 = torch.aten.mul.int %int4_11310, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11311 = torch.constant.int 4096
%13765 = torch.prim.ListConstruct %13764, %int4096_11311 : (!torch.int, !torch.int) -> !torch.list<int>
%13766 = torch.aten.view %13641, %13765 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13766, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13767 = torch.aten.mm %13766, %13721 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13767, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11312 = torch.constant.int 4
%int128_11313 = torch.constant.int 128
%13768 = torch.prim.ListConstruct %int4_11312, %2482, %int128_11313 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13769 = torch.aten.view %13767, %13768 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_11314 = torch.constant.int 1
%int0_11315 = torch.constant.int 0
%13770 = torch.prim.ListConstruct %int1_11314, %int0_11315 : (!torch.int, !torch.int) -> !torch.list<int>
%13771 = torch.aten.permute %464, %13770 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11316 = torch.constant.int 1
%int0_11317 = torch.constant.int 0
%13772 = torch.prim.ListConstruct %int1_11316, %int0_11317 : (!torch.int, !torch.int) -> !torch.list<int>
%13773 = torch.aten.permute %465, %13772 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11318 = torch.constant.int 1
%int0_11319 = torch.constant.int 0
%13774 = torch.prim.ListConstruct %int1_11318, %int0_11319 : (!torch.int, !torch.int) -> !torch.list<int>
%13775 = torch.aten.permute %466, %13774 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11320 = torch.constant.int 1
%int0_11321 = torch.constant.int 0
%13776 = torch.prim.ListConstruct %int1_11320, %int0_11321 : (!torch.int, !torch.int) -> !torch.list<int>
%13777 = torch.aten.permute %467, %13776 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11322 = torch.constant.int 1
%int0_11323 = torch.constant.int 0
%13778 = torch.prim.ListConstruct %int1_11322, %int0_11323 : (!torch.int, !torch.int) -> !torch.list<int>
%13779 = torch.aten.permute %468, %13778 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11324 = torch.constant.int 1
%int0_11325 = torch.constant.int 0
%13780 = torch.prim.ListConstruct %int1_11324, %int0_11325 : (!torch.int, !torch.int) -> !torch.list<int>
%13781 = torch.aten.permute %469, %13780 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11326 = torch.constant.int 1
%int0_11327 = torch.constant.int 0
%13782 = torch.prim.ListConstruct %int1_11326, %int0_11327 : (!torch.int, !torch.int) -> !torch.list<int>
%13783 = torch.aten.permute %470, %13782 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_11328 = torch.constant.int 1
%int0_11329 = torch.constant.int 0
%13784 = torch.prim.ListConstruct %int1_11328, %int0_11329 : (!torch.int, !torch.int) -> !torch.list<int>
%13785 = torch.aten.permute %471, %13784 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_11330 = torch.constant.int 4
%13786 = torch.aten.mul.int %int4_11330, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11331 = torch.constant.int 4096
%13787 = torch.prim.ListConstruct %13786, %int4096_11331 : (!torch.int, !torch.int) -> !torch.list<int>
%13788 = torch.aten.view %13634, %13787 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13788, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13789 = torch.aten.mm %13788, %13771 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13789, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11332 = torch.constant.int 4
%int128_11333 = torch.constant.int 128
%13790 = torch.prim.ListConstruct %int4_11332, %2482, %int128_11333 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13791 = torch.aten.view %13789, %13790 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11334 = torch.constant.int 4
%13792 = torch.aten.mul.int %int4_11334, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11335 = torch.constant.int 4096
%13793 = torch.prim.ListConstruct %13792, %int4096_11335 : (!torch.int, !torch.int) -> !torch.list<int>
%13794 = torch.aten.view %13635, %13793 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13794, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13795 = torch.aten.mm %13794, %13773 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13795, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11336 = torch.constant.int 4
%int128_11337 = torch.constant.int 128
%13796 = torch.prim.ListConstruct %int4_11336, %2482, %int128_11337 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13797 = torch.aten.view %13795, %13796 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11338 = torch.constant.int 4
%13798 = torch.aten.mul.int %int4_11338, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11339 = torch.constant.int 4096
%13799 = torch.prim.ListConstruct %13798, %int4096_11339 : (!torch.int, !torch.int) -> !torch.list<int>
%13800 = torch.aten.view %13636, %13799 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13800, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13801 = torch.aten.mm %13800, %13775 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13801, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11340 = torch.constant.int 4
%int128_11341 = torch.constant.int 128
%13802 = torch.prim.ListConstruct %int4_11340, %2482, %int128_11341 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13803 = torch.aten.view %13801, %13802 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11342 = torch.constant.int 4
%13804 = torch.aten.mul.int %int4_11342, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11343 = torch.constant.int 4096
%13805 = torch.prim.ListConstruct %13804, %int4096_11343 : (!torch.int, !torch.int) -> !torch.list<int>
%13806 = torch.aten.view %13637, %13805 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13806, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13807 = torch.aten.mm %13806, %13777 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13807, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11344 = torch.constant.int 4
%int128_11345 = torch.constant.int 128
%13808 = torch.prim.ListConstruct %int4_11344, %2482, %int128_11345 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13809 = torch.aten.view %13807, %13808 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11346 = torch.constant.int 4
%13810 = torch.aten.mul.int %int4_11346, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11347 = torch.constant.int 4096
%13811 = torch.prim.ListConstruct %13810, %int4096_11347 : (!torch.int, !torch.int) -> !torch.list<int>
%13812 = torch.aten.view %13638, %13811 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13812, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13813 = torch.aten.mm %13812, %13779 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13813, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11348 = torch.constant.int 4
%int128_11349 = torch.constant.int 128
%13814 = torch.prim.ListConstruct %int4_11348, %2482, %int128_11349 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13815 = torch.aten.view %13813, %13814 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11350 = torch.constant.int 4
%13816 = torch.aten.mul.int %int4_11350, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11351 = torch.constant.int 4096
%13817 = torch.prim.ListConstruct %13816, %int4096_11351 : (!torch.int, !torch.int) -> !torch.list<int>
%13818 = torch.aten.view %13639, %13817 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13818, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13819 = torch.aten.mm %13818, %13781 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13819, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11352 = torch.constant.int 4
%int128_11353 = torch.constant.int 128
%13820 = torch.prim.ListConstruct %int4_11352, %2482, %int128_11353 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13821 = torch.aten.view %13819, %13820 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11354 = torch.constant.int 4
%13822 = torch.aten.mul.int %int4_11354, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11355 = torch.constant.int 4096
%13823 = torch.prim.ListConstruct %13822, %int4096_11355 : (!torch.int, !torch.int) -> !torch.list<int>
%13824 = torch.aten.view %13640, %13823 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13824, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13825 = torch.aten.mm %13824, %13783 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13825, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11356 = torch.constant.int 4
%int128_11357 = torch.constant.int 128
%13826 = torch.prim.ListConstruct %int4_11356, %2482, %int128_11357 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13827 = torch.aten.view %13825, %13826 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11358 = torch.constant.int 4
%13828 = torch.aten.mul.int %int4_11358, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_11359 = torch.constant.int 4096
%13829 = torch.prim.ListConstruct %13828, %int4096_11359 : (!torch.int, !torch.int) -> !torch.list<int>
%13830 = torch.aten.view %13641, %13829 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %13830, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%13831 = torch.aten.mm %13830, %13785 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %13831, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_11360 = torch.constant.int 4
%int128_11361 = torch.constant.int 128
%13832 = torch.prim.ListConstruct %int4_11360, %2482, %int128_11361 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13833 = torch.aten.view %13831, %13832 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %13833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_11362 = torch.constant.int 4
%int4_11363 = torch.constant.int 4
%int128_11364 = torch.constant.int 128
%13834 = torch.prim.ListConstruct %int4_11362, %2482, %int4_11363, %int128_11364 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13835 = torch.aten.view %13663, %13834 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11365 = torch.constant.int 4
%int4_11366 = torch.constant.int 4
%int128_11367 = torch.constant.int 128
%13836 = torch.prim.ListConstruct %int4_11365, %2482, %int4_11366, %int128_11367 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13837 = torch.aten.view %13669, %13836 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11368 = torch.constant.int 4
%int4_11369 = torch.constant.int 4
%int128_11370 = torch.constant.int 128
%13838 = torch.prim.ListConstruct %int4_11368, %2482, %int4_11369, %int128_11370 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13839 = torch.aten.view %13675, %13838 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11371 = torch.constant.int 4
%int4_11372 = torch.constant.int 4
%int128_11373 = torch.constant.int 128
%13840 = torch.prim.ListConstruct %int4_11371, %2482, %int4_11372, %int128_11373 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13841 = torch.aten.view %13681, %13840 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11374 = torch.constant.int 4
%int4_11375 = torch.constant.int 4
%int128_11376 = torch.constant.int 128
%13842 = torch.prim.ListConstruct %int4_11374, %2482, %int4_11375, %int128_11376 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13843 = torch.aten.view %13687, %13842 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11377 = torch.constant.int 4
%int4_11378 = torch.constant.int 4
%int128_11379 = torch.constant.int 128
%13844 = torch.prim.ListConstruct %int4_11377, %2482, %int4_11378, %int128_11379 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13845 = torch.aten.view %13693, %13844 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11380 = torch.constant.int 4
%int4_11381 = torch.constant.int 4
%int128_11382 = torch.constant.int 128
%13846 = torch.prim.ListConstruct %int4_11380, %2482, %int4_11381, %int128_11382 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13847 = torch.aten.view %13699, %13846 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11383 = torch.constant.int 4
%int4_11384 = torch.constant.int 4
%int128_11385 = torch.constant.int 128
%13848 = torch.prim.ListConstruct %int4_11383, %2482, %int4_11384, %int128_11385 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13849 = torch.aten.view %13705, %13848 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_11386 = torch.constant.int 4
%int1_11387 = torch.constant.int 1
%int128_11388 = torch.constant.int 128
%13850 = torch.prim.ListConstruct %int4_11386, %2482, %int1_11387, %int128_11388 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13851 = torch.aten.view %13727, %13850 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11389 = torch.constant.int 4
%int1_11390 = torch.constant.int 1
%int128_11391 = torch.constant.int 128
%13852 = torch.prim.ListConstruct %int4_11389, %2482, %int1_11390, %int128_11391 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13853 = torch.aten.view %13733, %13852 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11392 = torch.constant.int 4
%int1_11393 = torch.constant.int 1
%int128_11394 = torch.constant.int 128
%13854 = torch.prim.ListConstruct %int4_11392, %2482, %int1_11393, %int128_11394 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13855 = torch.aten.view %13739, %13854 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11395 = torch.constant.int 4
%int1_11396 = torch.constant.int 1
%int128_11397 = torch.constant.int 128
%13856 = torch.prim.ListConstruct %int4_11395, %2482, %int1_11396, %int128_11397 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13857 = torch.aten.view %13745, %13856 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11398 = torch.constant.int 4
%int1_11399 = torch.constant.int 1
%int128_11400 = torch.constant.int 128
%13858 = torch.prim.ListConstruct %int4_11398, %2482, %int1_11399, %int128_11400 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13859 = torch.aten.view %13751, %13858 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11401 = torch.constant.int 4
%int1_11402 = torch.constant.int 1
%int128_11403 = torch.constant.int 128
%13860 = torch.prim.ListConstruct %int4_11401, %2482, %int1_11402, %int128_11403 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13861 = torch.aten.view %13757, %13860 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11404 = torch.constant.int 4
%int1_11405 = torch.constant.int 1
%int128_11406 = torch.constant.int 128
%13862 = torch.prim.ListConstruct %int4_11404, %2482, %int1_11405, %int128_11406 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13863 = torch.aten.view %13763, %13862 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11407 = torch.constant.int 4
%int1_11408 = torch.constant.int 1
%int128_11409 = torch.constant.int 128
%13864 = torch.prim.ListConstruct %int4_11407, %2482, %int1_11408, %int128_11409 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13865 = torch.aten.view %13769, %13864 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11410 = torch.constant.int 4
%int1_11411 = torch.constant.int 1
%int128_11412 = torch.constant.int 128
%13866 = torch.prim.ListConstruct %int4_11410, %2482, %int1_11411, %int128_11412 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13867 = torch.aten.view %13791, %13866 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11413 = torch.constant.int 4
%int1_11414 = torch.constant.int 1
%int128_11415 = torch.constant.int 128
%13868 = torch.prim.ListConstruct %int4_11413, %2482, %int1_11414, %int128_11415 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13869 = torch.aten.view %13797, %13868 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11416 = torch.constant.int 4
%int1_11417 = torch.constant.int 1
%int128_11418 = torch.constant.int 128
%13870 = torch.prim.ListConstruct %int4_11416, %2482, %int1_11417, %int128_11418 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13871 = torch.aten.view %13803, %13870 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11419 = torch.constant.int 4
%int1_11420 = torch.constant.int 1
%int128_11421 = torch.constant.int 128
%13872 = torch.prim.ListConstruct %int4_11419, %2482, %int1_11420, %int128_11421 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13873 = torch.aten.view %13809, %13872 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11422 = torch.constant.int 4
%int1_11423 = torch.constant.int 1
%int128_11424 = torch.constant.int 128
%13874 = torch.prim.ListConstruct %int4_11422, %2482, %int1_11423, %int128_11424 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13875 = torch.aten.view %13815, %13874 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11425 = torch.constant.int 4
%int1_11426 = torch.constant.int 1
%int128_11427 = torch.constant.int 128
%13876 = torch.prim.ListConstruct %int4_11425, %2482, %int1_11426, %int128_11427 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13877 = torch.aten.view %13821, %13876 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11428 = torch.constant.int 4
%int1_11429 = torch.constant.int 1
%int128_11430 = torch.constant.int 128
%13878 = torch.prim.ListConstruct %int4_11428, %2482, %int1_11429, %int128_11430 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13879 = torch.aten.view %13827, %13878 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_11431 = torch.constant.int 4
%int1_11432 = torch.constant.int 1
%int128_11433 = torch.constant.int 128
%13880 = torch.prim.ListConstruct %int4_11431, %2482, %int1_11432, %int128_11433 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%13881 = torch.aten.view %13833, %13880 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %13881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_11434 = torch.constant.int 131072
%none_11435 = torch.constant.none
%none_11436 = torch.constant.none
%cpu_11437 = torch.constant.device "cpu"
%false_11438 = torch.constant.bool false
%13882 = torch.aten.arange %int131072_11434, %none_11435, %none_11436, %cpu_11437, %false_11438 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_11439 = torch.constant.int 0
%int128_11440 = torch.constant.int 128
%int2_11441 = torch.constant.int 2
%none_11442 = torch.constant.none
%none_11443 = torch.constant.none
%cpu_11444 = torch.constant.device "cpu"
%false_11445 = torch.constant.bool false
%13883 = torch.aten.arange.start_step %int0_11439, %int128_11440, %int2_11441, %none_11442, %none_11443, %cpu_11444, %false_11445 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_11446 = torch.constant.int 0
%int0_11447 = torch.constant.int 0
%int64_11448 = torch.constant.int 64
%int1_11449 = torch.constant.int 1
%13884 = torch.aten.slice.Tensor %13883, %int0_11446, %int0_11447, %int64_11448, %int1_11449 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_11450 = torch.constant.int 6
%13885 = torch.prims.convert_element_type %13884, %int6_11450 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_11451 = torch.constant.int 128
%13886 = torch.aten.div.Scalar %13885, %int128_11451 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_11452 = torch.constant.float 5.000000e+05
%13887 = torch.aten.pow.Scalar %float5.000000e05_11452, %13886 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%13888 = torch.aten.reciprocal %13887 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_11453 = torch.constant.float 1.000000e+00
%13889 = torch.aten.mul.Scalar %13888, %float1.000000e00_11453 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_11454 = torch.constant.int 131072
%int1_11455 = torch.constant.int 1
%13890 = torch.prim.ListConstruct %int131072_11454, %int1_11455 : (!torch.int, !torch.int) -> !torch.list<int>
%13891 = torch.aten.view %13882, %13890 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%13892 = torch.aten.mul.Tensor %13891, %13889 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%13893 = torch.aten.cos %13892 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%13894 = torch.aten.sin %13892 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%13895 = torch.aten.complex %13893, %13894 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%13896 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13897 = flow.tensor.transfer %13896 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%13898 = torch_c.from_builtin_tensor %13897 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13899 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13900 = flow.tensor.transfer %13899 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%13901 = torch_c.from_builtin_tensor %13900 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13902 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13903 = flow.tensor.transfer %13902 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%13904 = torch_c.from_builtin_tensor %13903 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13905 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13906 = flow.tensor.transfer %13905 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%13907 = torch_c.from_builtin_tensor %13906 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13908 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13909 = flow.tensor.transfer %13908 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%13910 = torch_c.from_builtin_tensor %13909 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13911 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13912 = flow.tensor.transfer %13911 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%13913 = torch_c.from_builtin_tensor %13912 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13914 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13915 = flow.tensor.transfer %13914 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%13916 = torch_c.from_builtin_tensor %13915 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%13917 = torch_c.to_builtin_tensor %13895 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%13918 = flow.tensor.transfer %13917 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%13919 = torch_c.from_builtin_tensor %13918 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_11456 = torch.constant.int 1
%13920 = torch.aten.size.int %13663, %int1_11456 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11457 = torch.constant.int 0
%13921 = torch.aten.add.int %int0_11457, %13920 : !torch.int, !torch.int -> !torch.int
%int0_11458 = torch.constant.int 0
%int0_11459 = torch.constant.int 0
%int1_11460 = torch.constant.int 1
%13922 = torch.aten.slice.Tensor %13898, %int0_11458, %int0_11459, %13921, %int1_11460 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13922, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11461 = torch.constant.int 1
%int0_11462 = torch.constant.int 0
%int9223372036854775807_11463 = torch.constant.int 9223372036854775807
%int1_11464 = torch.constant.int 1
%13923 = torch.aten.slice.Tensor %13922, %int1_11461, %int0_11462, %int9223372036854775807_11463, %int1_11464 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13923, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11465 = torch.constant.int 0
%13924 = torch.aten.unsqueeze %13923, %int0_11465 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %13924, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11466 = torch.constant.int 2
%13925 = torch.aten.unsqueeze %13924, %int2_11466 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13925, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11467 = torch.constant.int 3
%int0_11468 = torch.constant.int 0
%int9223372036854775807_11469 = torch.constant.int 9223372036854775807
%int1_11470 = torch.constant.int 1
%13926 = torch.aten.slice.Tensor %13925, %int3_11467, %int0_11468, %int9223372036854775807_11469, %int1_11470 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13926, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%13927 = torch_c.to_builtin_tensor %13835 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11471 = arith.constant 1 : index
%dim_11472 = tensor.dim %13927, %c1_11471 : tensor<4x?x4x128xf16>
%13928 = flow.tensor.bitcast %13927 : tensor<4x?x4x128xf16>{%dim_11472} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11472}
%13929 = torch_c.from_builtin_tensor %13928 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %13929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%13930 = torch.aten.mul.Tensor %13929, %13926 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %13930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%13931 = torch_c.to_builtin_tensor %13930 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11473 = arith.constant 1 : index
%dim_11474 = tensor.dim %13931, %c1_11473 : tensor<4x?x4x64xcomplex<f32>>
%13932 = flow.tensor.bitcast %13931 : tensor<4x?x4x64xcomplex<f32>>{%dim_11474} -> tensor<4x?x4x128xf32>{%dim_11474}
%13933 = torch_c.from_builtin_tensor %13932 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %13933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11475 = torch.constant.int 5
%13934 = torch.prims.convert_element_type %13933, %int5_11475 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11476 = torch.constant.int 1
%13935 = torch.aten.size.int %13669, %int1_11476 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11477 = torch.constant.int 0
%13936 = torch.aten.add.int %int0_11477, %13935 : !torch.int, !torch.int -> !torch.int
%int0_11478 = torch.constant.int 0
%int0_11479 = torch.constant.int 0
%int1_11480 = torch.constant.int 1
%13937 = torch.aten.slice.Tensor %13901, %int0_11478, %int0_11479, %13936, %int1_11480 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13937, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11481 = torch.constant.int 1
%int0_11482 = torch.constant.int 0
%int9223372036854775807_11483 = torch.constant.int 9223372036854775807
%int1_11484 = torch.constant.int 1
%13938 = torch.aten.slice.Tensor %13937, %int1_11481, %int0_11482, %int9223372036854775807_11483, %int1_11484 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13938, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11485 = torch.constant.int 0
%13939 = torch.aten.unsqueeze %13938, %int0_11485 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %13939, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11486 = torch.constant.int 2
%13940 = torch.aten.unsqueeze %13939, %int2_11486 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13940, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11487 = torch.constant.int 3
%int0_11488 = torch.constant.int 0
%int9223372036854775807_11489 = torch.constant.int 9223372036854775807
%int1_11490 = torch.constant.int 1
%13941 = torch.aten.slice.Tensor %13940, %int3_11487, %int0_11488, %int9223372036854775807_11489, %int1_11490 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13941, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%13942 = torch_c.to_builtin_tensor %13837 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11491 = arith.constant 1 : index
%dim_11492 = tensor.dim %13942, %c1_11491 : tensor<4x?x4x128xf16>
%13943 = flow.tensor.bitcast %13942 : tensor<4x?x4x128xf16>{%dim_11492} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11492}
%13944 = torch_c.from_builtin_tensor %13943 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %13944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%13945 = torch.aten.mul.Tensor %13944, %13941 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %13945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%13946 = torch_c.to_builtin_tensor %13945 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11493 = arith.constant 1 : index
%dim_11494 = tensor.dim %13946, %c1_11493 : tensor<4x?x4x64xcomplex<f32>>
%13947 = flow.tensor.bitcast %13946 : tensor<4x?x4x64xcomplex<f32>>{%dim_11494} -> tensor<4x?x4x128xf32>{%dim_11494}
%13948 = torch_c.from_builtin_tensor %13947 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %13948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11495 = torch.constant.int 5
%13949 = torch.prims.convert_element_type %13948, %int5_11495 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11496 = torch.constant.int 1
%13950 = torch.aten.size.int %13675, %int1_11496 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11497 = torch.constant.int 0
%13951 = torch.aten.add.int %int0_11497, %13950 : !torch.int, !torch.int -> !torch.int
%int0_11498 = torch.constant.int 0
%int0_11499 = torch.constant.int 0
%int1_11500 = torch.constant.int 1
%13952 = torch.aten.slice.Tensor %13904, %int0_11498, %int0_11499, %13951, %int1_11500 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13952, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11501 = torch.constant.int 1
%int0_11502 = torch.constant.int 0
%int9223372036854775807_11503 = torch.constant.int 9223372036854775807
%int1_11504 = torch.constant.int 1
%13953 = torch.aten.slice.Tensor %13952, %int1_11501, %int0_11502, %int9223372036854775807_11503, %int1_11504 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13953, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11505 = torch.constant.int 0
%13954 = torch.aten.unsqueeze %13953, %int0_11505 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %13954, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11506 = torch.constant.int 2
%13955 = torch.aten.unsqueeze %13954, %int2_11506 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13955, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11507 = torch.constant.int 3
%int0_11508 = torch.constant.int 0
%int9223372036854775807_11509 = torch.constant.int 9223372036854775807
%int1_11510 = torch.constant.int 1
%13956 = torch.aten.slice.Tensor %13955, %int3_11507, %int0_11508, %int9223372036854775807_11509, %int1_11510 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13956, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%13957 = torch_c.to_builtin_tensor %13839 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11511 = arith.constant 1 : index
%dim_11512 = tensor.dim %13957, %c1_11511 : tensor<4x?x4x128xf16>
%13958 = flow.tensor.bitcast %13957 : tensor<4x?x4x128xf16>{%dim_11512} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11512}
%13959 = torch_c.from_builtin_tensor %13958 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %13959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%13960 = torch.aten.mul.Tensor %13959, %13956 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %13960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%13961 = torch_c.to_builtin_tensor %13960 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11513 = arith.constant 1 : index
%dim_11514 = tensor.dim %13961, %c1_11513 : tensor<4x?x4x64xcomplex<f32>>
%13962 = flow.tensor.bitcast %13961 : tensor<4x?x4x64xcomplex<f32>>{%dim_11514} -> tensor<4x?x4x128xf32>{%dim_11514}
%13963 = torch_c.from_builtin_tensor %13962 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %13963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11515 = torch.constant.int 5
%13964 = torch.prims.convert_element_type %13963, %int5_11515 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11516 = torch.constant.int 1
%13965 = torch.aten.size.int %13681, %int1_11516 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11517 = torch.constant.int 0
%13966 = torch.aten.add.int %int0_11517, %13965 : !torch.int, !torch.int -> !torch.int
%int0_11518 = torch.constant.int 0
%int0_11519 = torch.constant.int 0
%int1_11520 = torch.constant.int 1
%13967 = torch.aten.slice.Tensor %13907, %int0_11518, %int0_11519, %13966, %int1_11520 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13967, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11521 = torch.constant.int 1
%int0_11522 = torch.constant.int 0
%int9223372036854775807_11523 = torch.constant.int 9223372036854775807
%int1_11524 = torch.constant.int 1
%13968 = torch.aten.slice.Tensor %13967, %int1_11521, %int0_11522, %int9223372036854775807_11523, %int1_11524 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13968, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11525 = torch.constant.int 0
%13969 = torch.aten.unsqueeze %13968, %int0_11525 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %13969, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11526 = torch.constant.int 2
%13970 = torch.aten.unsqueeze %13969, %int2_11526 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13970, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11527 = torch.constant.int 3
%int0_11528 = torch.constant.int 0
%int9223372036854775807_11529 = torch.constant.int 9223372036854775807
%int1_11530 = torch.constant.int 1
%13971 = torch.aten.slice.Tensor %13970, %int3_11527, %int0_11528, %int9223372036854775807_11529, %int1_11530 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13971, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%13972 = torch_c.to_builtin_tensor %13841 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11531 = arith.constant 1 : index
%dim_11532 = tensor.dim %13972, %c1_11531 : tensor<4x?x4x128xf16>
%13973 = flow.tensor.bitcast %13972 : tensor<4x?x4x128xf16>{%dim_11532} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11532}
%13974 = torch_c.from_builtin_tensor %13973 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %13974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%13975 = torch.aten.mul.Tensor %13974, %13971 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %13975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%13976 = torch_c.to_builtin_tensor %13975 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11533 = arith.constant 1 : index
%dim_11534 = tensor.dim %13976, %c1_11533 : tensor<4x?x4x64xcomplex<f32>>
%13977 = flow.tensor.bitcast %13976 : tensor<4x?x4x64xcomplex<f32>>{%dim_11534} -> tensor<4x?x4x128xf32>{%dim_11534}
%13978 = torch_c.from_builtin_tensor %13977 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %13978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11535 = torch.constant.int 5
%13979 = torch.prims.convert_element_type %13978, %int5_11535 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11536 = torch.constant.int 1
%13980 = torch.aten.size.int %13687, %int1_11536 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11537 = torch.constant.int 0
%13981 = torch.aten.add.int %int0_11537, %13980 : !torch.int, !torch.int -> !torch.int
%int0_11538 = torch.constant.int 0
%int0_11539 = torch.constant.int 0
%int1_11540 = torch.constant.int 1
%13982 = torch.aten.slice.Tensor %13910, %int0_11538, %int0_11539, %13981, %int1_11540 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13982, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11541 = torch.constant.int 1
%int0_11542 = torch.constant.int 0
%int9223372036854775807_11543 = torch.constant.int 9223372036854775807
%int1_11544 = torch.constant.int 1
%13983 = torch.aten.slice.Tensor %13982, %int1_11541, %int0_11542, %int9223372036854775807_11543, %int1_11544 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13983, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11545 = torch.constant.int 0
%13984 = torch.aten.unsqueeze %13983, %int0_11545 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %13984, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11546 = torch.constant.int 2
%13985 = torch.aten.unsqueeze %13984, %int2_11546 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13985, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11547 = torch.constant.int 3
%int0_11548 = torch.constant.int 0
%int9223372036854775807_11549 = torch.constant.int 9223372036854775807
%int1_11550 = torch.constant.int 1
%13986 = torch.aten.slice.Tensor %13985, %int3_11547, %int0_11548, %int9223372036854775807_11549, %int1_11550 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %13986, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%13987 = torch_c.to_builtin_tensor %13843 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11551 = arith.constant 1 : index
%dim_11552 = tensor.dim %13987, %c1_11551 : tensor<4x?x4x128xf16>
%13988 = flow.tensor.bitcast %13987 : tensor<4x?x4x128xf16>{%dim_11552} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11552}
%13989 = torch_c.from_builtin_tensor %13988 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %13989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%13990 = torch.aten.mul.Tensor %13989, %13986 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %13990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%13991 = torch_c.to_builtin_tensor %13990 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11553 = arith.constant 1 : index
%dim_11554 = tensor.dim %13991, %c1_11553 : tensor<4x?x4x64xcomplex<f32>>
%13992 = flow.tensor.bitcast %13991 : tensor<4x?x4x64xcomplex<f32>>{%dim_11554} -> tensor<4x?x4x128xf32>{%dim_11554}
%13993 = torch_c.from_builtin_tensor %13992 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %13993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11555 = torch.constant.int 5
%13994 = torch.prims.convert_element_type %13993, %int5_11555 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %13994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11556 = torch.constant.int 1
%13995 = torch.aten.size.int %13693, %int1_11556 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11557 = torch.constant.int 0
%13996 = torch.aten.add.int %int0_11557, %13995 : !torch.int, !torch.int -> !torch.int
%int0_11558 = torch.constant.int 0
%int0_11559 = torch.constant.int 0
%int1_11560 = torch.constant.int 1
%13997 = torch.aten.slice.Tensor %13913, %int0_11558, %int0_11559, %13996, %int1_11560 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13997, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11561 = torch.constant.int 1
%int0_11562 = torch.constant.int 0
%int9223372036854775807_11563 = torch.constant.int 9223372036854775807
%int1_11564 = torch.constant.int 1
%13998 = torch.aten.slice.Tensor %13997, %int1_11561, %int0_11562, %int9223372036854775807_11563, %int1_11564 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %13998, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11565 = torch.constant.int 0
%13999 = torch.aten.unsqueeze %13998, %int0_11565 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %13999, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11566 = torch.constant.int 2
%14000 = torch.aten.unsqueeze %13999, %int2_11566 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14000, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11567 = torch.constant.int 3
%int0_11568 = torch.constant.int 0
%int9223372036854775807_11569 = torch.constant.int 9223372036854775807
%int1_11570 = torch.constant.int 1
%14001 = torch.aten.slice.Tensor %14000, %int3_11567, %int0_11568, %int9223372036854775807_11569, %int1_11570 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14001, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14002 = torch_c.to_builtin_tensor %13845 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11571 = arith.constant 1 : index
%dim_11572 = tensor.dim %14002, %c1_11571 : tensor<4x?x4x128xf16>
%14003 = flow.tensor.bitcast %14002 : tensor<4x?x4x128xf16>{%dim_11572} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11572}
%14004 = torch_c.from_builtin_tensor %14003 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %14004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%14005 = torch.aten.mul.Tensor %14004, %14001 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %14005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%14006 = torch_c.to_builtin_tensor %14005 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11573 = arith.constant 1 : index
%dim_11574 = tensor.dim %14006, %c1_11573 : tensor<4x?x4x64xcomplex<f32>>
%14007 = flow.tensor.bitcast %14006 : tensor<4x?x4x64xcomplex<f32>>{%dim_11574} -> tensor<4x?x4x128xf32>{%dim_11574}
%14008 = torch_c.from_builtin_tensor %14007 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %14008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11575 = torch.constant.int 5
%14009 = torch.prims.convert_element_type %14008, %int5_11575 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11576 = torch.constant.int 1
%14010 = torch.aten.size.int %13699, %int1_11576 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11577 = torch.constant.int 0
%14011 = torch.aten.add.int %int0_11577, %14010 : !torch.int, !torch.int -> !torch.int
%int0_11578 = torch.constant.int 0
%int0_11579 = torch.constant.int 0
%int1_11580 = torch.constant.int 1
%14012 = torch.aten.slice.Tensor %13916, %int0_11578, %int0_11579, %14011, %int1_11580 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14012, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11581 = torch.constant.int 1
%int0_11582 = torch.constant.int 0
%int9223372036854775807_11583 = torch.constant.int 9223372036854775807
%int1_11584 = torch.constant.int 1
%14013 = torch.aten.slice.Tensor %14012, %int1_11581, %int0_11582, %int9223372036854775807_11583, %int1_11584 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14013, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11585 = torch.constant.int 0
%14014 = torch.aten.unsqueeze %14013, %int0_11585 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14014, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11586 = torch.constant.int 2
%14015 = torch.aten.unsqueeze %14014, %int2_11586 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14015, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11587 = torch.constant.int 3
%int0_11588 = torch.constant.int 0
%int9223372036854775807_11589 = torch.constant.int 9223372036854775807
%int1_11590 = torch.constant.int 1
%14016 = torch.aten.slice.Tensor %14015, %int3_11587, %int0_11588, %int9223372036854775807_11589, %int1_11590 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14016, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14017 = torch_c.to_builtin_tensor %13847 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11591 = arith.constant 1 : index
%dim_11592 = tensor.dim %14017, %c1_11591 : tensor<4x?x4x128xf16>
%14018 = flow.tensor.bitcast %14017 : tensor<4x?x4x128xf16>{%dim_11592} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11592}
%14019 = torch_c.from_builtin_tensor %14018 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %14019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%14020 = torch.aten.mul.Tensor %14019, %14016 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %14020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%14021 = torch_c.to_builtin_tensor %14020 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11593 = arith.constant 1 : index
%dim_11594 = tensor.dim %14021, %c1_11593 : tensor<4x?x4x64xcomplex<f32>>
%14022 = flow.tensor.bitcast %14021 : tensor<4x?x4x64xcomplex<f32>>{%dim_11594} -> tensor<4x?x4x128xf32>{%dim_11594}
%14023 = torch_c.from_builtin_tensor %14022 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %14023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11595 = torch.constant.int 5
%14024 = torch.prims.convert_element_type %14023, %int5_11595 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_11596 = torch.constant.int 1
%14025 = torch.aten.size.int %13705, %int1_11596 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_11597 = torch.constant.int 0
%14026 = torch.aten.add.int %int0_11597, %14025 : !torch.int, !torch.int -> !torch.int
%int0_11598 = torch.constant.int 0
%int0_11599 = torch.constant.int 0
%int1_11600 = torch.constant.int 1
%14027 = torch.aten.slice.Tensor %13919, %int0_11598, %int0_11599, %14026, %int1_11600 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14027, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11601 = torch.constant.int 1
%int0_11602 = torch.constant.int 0
%int9223372036854775807_11603 = torch.constant.int 9223372036854775807
%int1_11604 = torch.constant.int 1
%14028 = torch.aten.slice.Tensor %14027, %int1_11601, %int0_11602, %int9223372036854775807_11603, %int1_11604 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14028, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11605 = torch.constant.int 0
%14029 = torch.aten.unsqueeze %14028, %int0_11605 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14029, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11606 = torch.constant.int 2
%14030 = torch.aten.unsqueeze %14029, %int2_11606 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14030, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11607 = torch.constant.int 3
%int0_11608 = torch.constant.int 0
%int9223372036854775807_11609 = torch.constant.int 9223372036854775807
%int1_11610 = torch.constant.int 1
%14031 = torch.aten.slice.Tensor %14030, %int3_11607, %int0_11608, %int9223372036854775807_11609, %int1_11610 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14031, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14032 = torch_c.to_builtin_tensor %13849 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_11611 = arith.constant 1 : index
%dim_11612 = tensor.dim %14032, %c1_11611 : tensor<4x?x4x128xf16>
%14033 = flow.tensor.bitcast %14032 : tensor<4x?x4x128xf16>{%dim_11612} -> tensor<4x?x4x64xcomplex<f16>>{%dim_11612}
%14034 = torch_c.from_builtin_tensor %14033 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %14034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%14035 = torch.aten.mul.Tensor %14034, %14031 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %14035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%14036 = torch_c.to_builtin_tensor %14035 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_11613 = arith.constant 1 : index
%dim_11614 = tensor.dim %14036, %c1_11613 : tensor<4x?x4x64xcomplex<f32>>
%14037 = flow.tensor.bitcast %14036 : tensor<4x?x4x64xcomplex<f32>>{%dim_11614} -> tensor<4x?x4x128xf32>{%dim_11614}
%14038 = torch_c.from_builtin_tensor %14037 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %14038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_11615 = torch.constant.int 5
%14039 = torch.prims.convert_element_type %14038, %int5_11615 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_11616 = torch.constant.int 131072
%none_11617 = torch.constant.none
%none_11618 = torch.constant.none
%cpu_11619 = torch.constant.device "cpu"
%false_11620 = torch.constant.bool false
%14040 = torch.aten.arange %int131072_11616, %none_11617, %none_11618, %cpu_11619, %false_11620 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_11621 = torch.constant.int 0
%int128_11622 = torch.constant.int 128
%int2_11623 = torch.constant.int 2
%none_11624 = torch.constant.none
%none_11625 = torch.constant.none
%cpu_11626 = torch.constant.device "cpu"
%false_11627 = torch.constant.bool false
%14041 = torch.aten.arange.start_step %int0_11621, %int128_11622, %int2_11623, %none_11624, %none_11625, %cpu_11626, %false_11627 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_11628 = torch.constant.int 0
%int0_11629 = torch.constant.int 0
%int64_11630 = torch.constant.int 64
%int1_11631 = torch.constant.int 1
%14042 = torch.aten.slice.Tensor %14041, %int0_11628, %int0_11629, %int64_11630, %int1_11631 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_11632 = torch.constant.int 6
%14043 = torch.prims.convert_element_type %14042, %int6_11632 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_11633 = torch.constant.int 128
%14044 = torch.aten.div.Scalar %14043, %int128_11633 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_11634 = torch.constant.float 5.000000e+05
%14045 = torch.aten.pow.Scalar %float5.000000e05_11634, %14044 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%14046 = torch.aten.reciprocal %14045 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_11635 = torch.constant.float 1.000000e+00
%14047 = torch.aten.mul.Scalar %14046, %float1.000000e00_11635 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_11636 = torch.constant.int 131072
%int1_11637 = torch.constant.int 1
%14048 = torch.prim.ListConstruct %int131072_11636, %int1_11637 : (!torch.int, !torch.int) -> !torch.list<int>
%14049 = torch.aten.view %14040, %14048 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%14050 = torch.aten.mul.Tensor %14049, %14047 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%14051 = torch.aten.cos %14050 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%14052 = torch.aten.sin %14050 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%14053 = torch.aten.complex %14051, %14052 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%14054 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14055 = flow.tensor.transfer %14054 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%14056 = torch_c.from_builtin_tensor %14055 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14057 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14058 = flow.tensor.transfer %14057 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%14059 = torch_c.from_builtin_tensor %14058 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14060 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14061 = flow.tensor.transfer %14060 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%14062 = torch_c.from_builtin_tensor %14061 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14063 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14064 = flow.tensor.transfer %14063 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%14065 = torch_c.from_builtin_tensor %14064 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14066 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14067 = flow.tensor.transfer %14066 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%14068 = torch_c.from_builtin_tensor %14067 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14069 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14070 = flow.tensor.transfer %14069 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%14071 = torch_c.from_builtin_tensor %14070 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14072 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14073 = flow.tensor.transfer %14072 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%14074 = torch_c.from_builtin_tensor %14073 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%14075 = torch_c.to_builtin_tensor %14053 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%14076 = flow.tensor.transfer %14075 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%14077 = torch_c.from_builtin_tensor %14076 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_11638 = torch.constant.int 1
%14078 = torch.aten.size.int %13727, %int1_11638 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11639 = torch.constant.int 0
%14079 = torch.aten.add.int %int0_11639, %14078 : !torch.int, !torch.int -> !torch.int
%int0_11640 = torch.constant.int 0
%int0_11641 = torch.constant.int 0
%int1_11642 = torch.constant.int 1
%14080 = torch.aten.slice.Tensor %14056, %int0_11640, %int0_11641, %14079, %int1_11642 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14080, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11643 = torch.constant.int 1
%int0_11644 = torch.constant.int 0
%int9223372036854775807_11645 = torch.constant.int 9223372036854775807
%int1_11646 = torch.constant.int 1
%14081 = torch.aten.slice.Tensor %14080, %int1_11643, %int0_11644, %int9223372036854775807_11645, %int1_11646 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14081, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11647 = torch.constant.int 0
%14082 = torch.aten.unsqueeze %14081, %int0_11647 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14082, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11648 = torch.constant.int 2
%14083 = torch.aten.unsqueeze %14082, %int2_11648 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14083, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11649 = torch.constant.int 3
%int0_11650 = torch.constant.int 0
%int9223372036854775807_11651 = torch.constant.int 9223372036854775807
%int1_11652 = torch.constant.int 1
%14084 = torch.aten.slice.Tensor %14083, %int3_11649, %int0_11650, %int9223372036854775807_11651, %int1_11652 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14084, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14085 = torch_c.to_builtin_tensor %13851 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11653 = arith.constant 1 : index
%dim_11654 = tensor.dim %14085, %c1_11653 : tensor<4x?x1x128xf16>
%14086 = flow.tensor.bitcast %14085 : tensor<4x?x1x128xf16>{%dim_11654} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11654}
%14087 = torch_c.from_builtin_tensor %14086 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14088 = torch.aten.mul.Tensor %14087, %14084 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14089 = torch_c.to_builtin_tensor %14088 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11655 = arith.constant 1 : index
%dim_11656 = tensor.dim %14089, %c1_11655 : tensor<4x?x1x64xcomplex<f32>>
%14090 = flow.tensor.bitcast %14089 : tensor<4x?x1x64xcomplex<f32>>{%dim_11656} -> tensor<4x?x1x128xf32>{%dim_11656}
%14091 = torch_c.from_builtin_tensor %14090 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11657 = torch.constant.int 5
%14092 = torch.prims.convert_element_type %14091, %int5_11657 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11658 = torch.constant.int 1
%14093 = torch.aten.size.int %13733, %int1_11658 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11659 = torch.constant.int 0
%14094 = torch.aten.add.int %int0_11659, %14093 : !torch.int, !torch.int -> !torch.int
%int0_11660 = torch.constant.int 0
%int0_11661 = torch.constant.int 0
%int1_11662 = torch.constant.int 1
%14095 = torch.aten.slice.Tensor %14059, %int0_11660, %int0_11661, %14094, %int1_11662 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14095, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11663 = torch.constant.int 1
%int0_11664 = torch.constant.int 0
%int9223372036854775807_11665 = torch.constant.int 9223372036854775807
%int1_11666 = torch.constant.int 1
%14096 = torch.aten.slice.Tensor %14095, %int1_11663, %int0_11664, %int9223372036854775807_11665, %int1_11666 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14096, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11667 = torch.constant.int 0
%14097 = torch.aten.unsqueeze %14096, %int0_11667 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14097, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11668 = torch.constant.int 2
%14098 = torch.aten.unsqueeze %14097, %int2_11668 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14098, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11669 = torch.constant.int 3
%int0_11670 = torch.constant.int 0
%int9223372036854775807_11671 = torch.constant.int 9223372036854775807
%int1_11672 = torch.constant.int 1
%14099 = torch.aten.slice.Tensor %14098, %int3_11669, %int0_11670, %int9223372036854775807_11671, %int1_11672 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14099, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14100 = torch_c.to_builtin_tensor %13853 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11673 = arith.constant 1 : index
%dim_11674 = tensor.dim %14100, %c1_11673 : tensor<4x?x1x128xf16>
%14101 = flow.tensor.bitcast %14100 : tensor<4x?x1x128xf16>{%dim_11674} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11674}
%14102 = torch_c.from_builtin_tensor %14101 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14103 = torch.aten.mul.Tensor %14102, %14099 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14104 = torch_c.to_builtin_tensor %14103 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11675 = arith.constant 1 : index
%dim_11676 = tensor.dim %14104, %c1_11675 : tensor<4x?x1x64xcomplex<f32>>
%14105 = flow.tensor.bitcast %14104 : tensor<4x?x1x64xcomplex<f32>>{%dim_11676} -> tensor<4x?x1x128xf32>{%dim_11676}
%14106 = torch_c.from_builtin_tensor %14105 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11677 = torch.constant.int 5
%14107 = torch.prims.convert_element_type %14106, %int5_11677 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11678 = torch.constant.int 1
%14108 = torch.aten.size.int %13739, %int1_11678 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11679 = torch.constant.int 0
%14109 = torch.aten.add.int %int0_11679, %14108 : !torch.int, !torch.int -> !torch.int
%int0_11680 = torch.constant.int 0
%int0_11681 = torch.constant.int 0
%int1_11682 = torch.constant.int 1
%14110 = torch.aten.slice.Tensor %14062, %int0_11680, %int0_11681, %14109, %int1_11682 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14110, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11683 = torch.constant.int 1
%int0_11684 = torch.constant.int 0
%int9223372036854775807_11685 = torch.constant.int 9223372036854775807
%int1_11686 = torch.constant.int 1
%14111 = torch.aten.slice.Tensor %14110, %int1_11683, %int0_11684, %int9223372036854775807_11685, %int1_11686 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14111, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11687 = torch.constant.int 0
%14112 = torch.aten.unsqueeze %14111, %int0_11687 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14112, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11688 = torch.constant.int 2
%14113 = torch.aten.unsqueeze %14112, %int2_11688 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14113, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11689 = torch.constant.int 3
%int0_11690 = torch.constant.int 0
%int9223372036854775807_11691 = torch.constant.int 9223372036854775807
%int1_11692 = torch.constant.int 1
%14114 = torch.aten.slice.Tensor %14113, %int3_11689, %int0_11690, %int9223372036854775807_11691, %int1_11692 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14114, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14115 = torch_c.to_builtin_tensor %13855 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11693 = arith.constant 1 : index
%dim_11694 = tensor.dim %14115, %c1_11693 : tensor<4x?x1x128xf16>
%14116 = flow.tensor.bitcast %14115 : tensor<4x?x1x128xf16>{%dim_11694} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11694}
%14117 = torch_c.from_builtin_tensor %14116 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14118 = torch.aten.mul.Tensor %14117, %14114 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14119 = torch_c.to_builtin_tensor %14118 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11695 = arith.constant 1 : index
%dim_11696 = tensor.dim %14119, %c1_11695 : tensor<4x?x1x64xcomplex<f32>>
%14120 = flow.tensor.bitcast %14119 : tensor<4x?x1x64xcomplex<f32>>{%dim_11696} -> tensor<4x?x1x128xf32>{%dim_11696}
%14121 = torch_c.from_builtin_tensor %14120 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11697 = torch.constant.int 5
%14122 = torch.prims.convert_element_type %14121, %int5_11697 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11698 = torch.constant.int 1
%14123 = torch.aten.size.int %13745, %int1_11698 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11699 = torch.constant.int 0
%14124 = torch.aten.add.int %int0_11699, %14123 : !torch.int, !torch.int -> !torch.int
%int0_11700 = torch.constant.int 0
%int0_11701 = torch.constant.int 0
%int1_11702 = torch.constant.int 1
%14125 = torch.aten.slice.Tensor %14065, %int0_11700, %int0_11701, %14124, %int1_11702 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14125, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11703 = torch.constant.int 1
%int0_11704 = torch.constant.int 0
%int9223372036854775807_11705 = torch.constant.int 9223372036854775807
%int1_11706 = torch.constant.int 1
%14126 = torch.aten.slice.Tensor %14125, %int1_11703, %int0_11704, %int9223372036854775807_11705, %int1_11706 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14126, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11707 = torch.constant.int 0
%14127 = torch.aten.unsqueeze %14126, %int0_11707 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14127, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11708 = torch.constant.int 2
%14128 = torch.aten.unsqueeze %14127, %int2_11708 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14128, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11709 = torch.constant.int 3
%int0_11710 = torch.constant.int 0
%int9223372036854775807_11711 = torch.constant.int 9223372036854775807
%int1_11712 = torch.constant.int 1
%14129 = torch.aten.slice.Tensor %14128, %int3_11709, %int0_11710, %int9223372036854775807_11711, %int1_11712 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14129, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14130 = torch_c.to_builtin_tensor %13857 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11713 = arith.constant 1 : index
%dim_11714 = tensor.dim %14130, %c1_11713 : tensor<4x?x1x128xf16>
%14131 = flow.tensor.bitcast %14130 : tensor<4x?x1x128xf16>{%dim_11714} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11714}
%14132 = torch_c.from_builtin_tensor %14131 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14133 = torch.aten.mul.Tensor %14132, %14129 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14134 = torch_c.to_builtin_tensor %14133 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11715 = arith.constant 1 : index
%dim_11716 = tensor.dim %14134, %c1_11715 : tensor<4x?x1x64xcomplex<f32>>
%14135 = flow.tensor.bitcast %14134 : tensor<4x?x1x64xcomplex<f32>>{%dim_11716} -> tensor<4x?x1x128xf32>{%dim_11716}
%14136 = torch_c.from_builtin_tensor %14135 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11717 = torch.constant.int 5
%14137 = torch.prims.convert_element_type %14136, %int5_11717 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11718 = torch.constant.int 1
%14138 = torch.aten.size.int %13751, %int1_11718 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11719 = torch.constant.int 0
%14139 = torch.aten.add.int %int0_11719, %14138 : !torch.int, !torch.int -> !torch.int
%int0_11720 = torch.constant.int 0
%int0_11721 = torch.constant.int 0
%int1_11722 = torch.constant.int 1
%14140 = torch.aten.slice.Tensor %14068, %int0_11720, %int0_11721, %14139, %int1_11722 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14140, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11723 = torch.constant.int 1
%int0_11724 = torch.constant.int 0
%int9223372036854775807_11725 = torch.constant.int 9223372036854775807
%int1_11726 = torch.constant.int 1
%14141 = torch.aten.slice.Tensor %14140, %int1_11723, %int0_11724, %int9223372036854775807_11725, %int1_11726 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14141, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11727 = torch.constant.int 0
%14142 = torch.aten.unsqueeze %14141, %int0_11727 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14142, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11728 = torch.constant.int 2
%14143 = torch.aten.unsqueeze %14142, %int2_11728 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14143, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11729 = torch.constant.int 3
%int0_11730 = torch.constant.int 0
%int9223372036854775807_11731 = torch.constant.int 9223372036854775807
%int1_11732 = torch.constant.int 1
%14144 = torch.aten.slice.Tensor %14143, %int3_11729, %int0_11730, %int9223372036854775807_11731, %int1_11732 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14144, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14145 = torch_c.to_builtin_tensor %13859 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11733 = arith.constant 1 : index
%dim_11734 = tensor.dim %14145, %c1_11733 : tensor<4x?x1x128xf16>
%14146 = flow.tensor.bitcast %14145 : tensor<4x?x1x128xf16>{%dim_11734} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11734}
%14147 = torch_c.from_builtin_tensor %14146 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14148 = torch.aten.mul.Tensor %14147, %14144 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14149 = torch_c.to_builtin_tensor %14148 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11735 = arith.constant 1 : index
%dim_11736 = tensor.dim %14149, %c1_11735 : tensor<4x?x1x64xcomplex<f32>>
%14150 = flow.tensor.bitcast %14149 : tensor<4x?x1x64xcomplex<f32>>{%dim_11736} -> tensor<4x?x1x128xf32>{%dim_11736}
%14151 = torch_c.from_builtin_tensor %14150 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11737 = torch.constant.int 5
%14152 = torch.prims.convert_element_type %14151, %int5_11737 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11738 = torch.constant.int 1
%14153 = torch.aten.size.int %13757, %int1_11738 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11739 = torch.constant.int 0
%14154 = torch.aten.add.int %int0_11739, %14153 : !torch.int, !torch.int -> !torch.int
%int0_11740 = torch.constant.int 0
%int0_11741 = torch.constant.int 0
%int1_11742 = torch.constant.int 1
%14155 = torch.aten.slice.Tensor %14071, %int0_11740, %int0_11741, %14154, %int1_11742 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14155, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11743 = torch.constant.int 1
%int0_11744 = torch.constant.int 0
%int9223372036854775807_11745 = torch.constant.int 9223372036854775807
%int1_11746 = torch.constant.int 1
%14156 = torch.aten.slice.Tensor %14155, %int1_11743, %int0_11744, %int9223372036854775807_11745, %int1_11746 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14156, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11747 = torch.constant.int 0
%14157 = torch.aten.unsqueeze %14156, %int0_11747 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14157, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11748 = torch.constant.int 2
%14158 = torch.aten.unsqueeze %14157, %int2_11748 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14158, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11749 = torch.constant.int 3
%int0_11750 = torch.constant.int 0
%int9223372036854775807_11751 = torch.constant.int 9223372036854775807
%int1_11752 = torch.constant.int 1
%14159 = torch.aten.slice.Tensor %14158, %int3_11749, %int0_11750, %int9223372036854775807_11751, %int1_11752 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14159, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14160 = torch_c.to_builtin_tensor %13861 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11753 = arith.constant 1 : index
%dim_11754 = tensor.dim %14160, %c1_11753 : tensor<4x?x1x128xf16>
%14161 = flow.tensor.bitcast %14160 : tensor<4x?x1x128xf16>{%dim_11754} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11754}
%14162 = torch_c.from_builtin_tensor %14161 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14163 = torch.aten.mul.Tensor %14162, %14159 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14164 = torch_c.to_builtin_tensor %14163 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11755 = arith.constant 1 : index
%dim_11756 = tensor.dim %14164, %c1_11755 : tensor<4x?x1x64xcomplex<f32>>
%14165 = flow.tensor.bitcast %14164 : tensor<4x?x1x64xcomplex<f32>>{%dim_11756} -> tensor<4x?x1x128xf32>{%dim_11756}
%14166 = torch_c.from_builtin_tensor %14165 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11757 = torch.constant.int 5
%14167 = torch.prims.convert_element_type %14166, %int5_11757 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11758 = torch.constant.int 1
%14168 = torch.aten.size.int %13763, %int1_11758 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11759 = torch.constant.int 0
%14169 = torch.aten.add.int %int0_11759, %14168 : !torch.int, !torch.int -> !torch.int
%int0_11760 = torch.constant.int 0
%int0_11761 = torch.constant.int 0
%int1_11762 = torch.constant.int 1
%14170 = torch.aten.slice.Tensor %14074, %int0_11760, %int0_11761, %14169, %int1_11762 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14170, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11763 = torch.constant.int 1
%int0_11764 = torch.constant.int 0
%int9223372036854775807_11765 = torch.constant.int 9223372036854775807
%int1_11766 = torch.constant.int 1
%14171 = torch.aten.slice.Tensor %14170, %int1_11763, %int0_11764, %int9223372036854775807_11765, %int1_11766 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14171, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11767 = torch.constant.int 0
%14172 = torch.aten.unsqueeze %14171, %int0_11767 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14172, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11768 = torch.constant.int 2
%14173 = torch.aten.unsqueeze %14172, %int2_11768 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14173, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11769 = torch.constant.int 3
%int0_11770 = torch.constant.int 0
%int9223372036854775807_11771 = torch.constant.int 9223372036854775807
%int1_11772 = torch.constant.int 1
%14174 = torch.aten.slice.Tensor %14173, %int3_11769, %int0_11770, %int9223372036854775807_11771, %int1_11772 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14174, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14175 = torch_c.to_builtin_tensor %13863 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11773 = arith.constant 1 : index
%dim_11774 = tensor.dim %14175, %c1_11773 : tensor<4x?x1x128xf16>
%14176 = flow.tensor.bitcast %14175 : tensor<4x?x1x128xf16>{%dim_11774} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11774}
%14177 = torch_c.from_builtin_tensor %14176 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14178 = torch.aten.mul.Tensor %14177, %14174 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14179 = torch_c.to_builtin_tensor %14178 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11775 = arith.constant 1 : index
%dim_11776 = tensor.dim %14179, %c1_11775 : tensor<4x?x1x64xcomplex<f32>>
%14180 = flow.tensor.bitcast %14179 : tensor<4x?x1x64xcomplex<f32>>{%dim_11776} -> tensor<4x?x1x128xf32>{%dim_11776}
%14181 = torch_c.from_builtin_tensor %14180 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11777 = torch.constant.int 5
%14182 = torch.prims.convert_element_type %14181, %int5_11777 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_11778 = torch.constant.int 1
%14183 = torch.aten.size.int %13769, %int1_11778 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_11779 = torch.constant.int 0
%14184 = torch.aten.add.int %int0_11779, %14183 : !torch.int, !torch.int -> !torch.int
%int0_11780 = torch.constant.int 0
%int0_11781 = torch.constant.int 0
%int1_11782 = torch.constant.int 1
%14185 = torch.aten.slice.Tensor %14077, %int0_11780, %int0_11781, %14184, %int1_11782 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14185, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_11783 = torch.constant.int 1
%int0_11784 = torch.constant.int 0
%int9223372036854775807_11785 = torch.constant.int 9223372036854775807
%int1_11786 = torch.constant.int 1
%14186 = torch.aten.slice.Tensor %14185, %int1_11783, %int0_11784, %int9223372036854775807_11785, %int1_11786 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %14186, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_11787 = torch.constant.int 0
%14187 = torch.aten.unsqueeze %14186, %int0_11787 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %14187, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_11788 = torch.constant.int 2
%14188 = torch.aten.unsqueeze %14187, %int2_11788 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14188, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_11789 = torch.constant.int 3
%int0_11790 = torch.constant.int 0
%int9223372036854775807_11791 = torch.constant.int 9223372036854775807
%int1_11792 = torch.constant.int 1
%14189 = torch.aten.slice.Tensor %14188, %int3_11789, %int0_11790, %int9223372036854775807_11791, %int1_11792 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14189, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%14190 = torch_c.to_builtin_tensor %13865 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_11793 = arith.constant 1 : index
%dim_11794 = tensor.dim %14190, %c1_11793 : tensor<4x?x1x128xf16>
%14191 = flow.tensor.bitcast %14190 : tensor<4x?x1x128xf16>{%dim_11794} -> tensor<4x?x1x64xcomplex<f16>>{%dim_11794}
%14192 = torch_c.from_builtin_tensor %14191 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %14192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%14193 = torch.aten.mul.Tensor %14192, %14189 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %14193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%14194 = torch_c.to_builtin_tensor %14193 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_11795 = arith.constant 1 : index
%dim_11796 = tensor.dim %14194, %c1_11795 : tensor<4x?x1x64xcomplex<f32>>
%14195 = flow.tensor.bitcast %14194 : tensor<4x?x1x64xcomplex<f32>>{%dim_11796} -> tensor<4x?x1x128xf32>{%dim_11796}
%14196 = torch_c.from_builtin_tensor %14195 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %14196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_11797 = torch.constant.int 5
%14197 = torch.prims.convert_element_type %14196, %int5_11797 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %14197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_11798 = torch.constant.int 64
%14198 = torch.aten.mul.Scalar %2364, %int64_11798 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14198, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11799 = torch.constant.int 64
%14199 = torch.aten.mul.Scalar %2367, %int64_11799 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14199, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11800 = torch.constant.int 64
%14200 = torch.aten.mul.Scalar %2370, %int64_11800 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14200, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11801 = torch.constant.int 64
%14201 = torch.aten.mul.Scalar %2373, %int64_11801 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14201, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11802 = torch.constant.int 64
%14202 = torch.aten.mul.Scalar %2376, %int64_11802 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14202, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11803 = torch.constant.int 64
%14203 = torch.aten.mul.Scalar %2379, %int64_11803 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14203, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11804 = torch.constant.int 64
%14204 = torch.aten.mul.Scalar %2382, %int64_11804 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14204, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_11805 = torch.constant.int 64
%14205 = torch.aten.mul.Scalar %2385, %int64_11805 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14205, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12 = torch.constant.int 12
%int1_11806 = torch.constant.int 1
%14206 = torch.aten.add.Scalar %14198, %int12, %int1_11806 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14206, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11807 = torch.constant.int 12
%int1_11808 = torch.constant.int 1
%14207 = torch.aten.add.Scalar %14199, %int12_11807, %int1_11808 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14207, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11809 = torch.constant.int 12
%int1_11810 = torch.constant.int 1
%14208 = torch.aten.add.Scalar %14200, %int12_11809, %int1_11810 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14208, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11811 = torch.constant.int 12
%int1_11812 = torch.constant.int 1
%14209 = torch.aten.add.Scalar %14201, %int12_11811, %int1_11812 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14209, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11813 = torch.constant.int 12
%int1_11814 = torch.constant.int 1
%14210 = torch.aten.add.Scalar %14202, %int12_11813, %int1_11814 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14210, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11815 = torch.constant.int 12
%int1_11816 = torch.constant.int 1
%14211 = torch.aten.add.Scalar %14203, %int12_11815, %int1_11816 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14211, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11817 = torch.constant.int 12
%int1_11818 = torch.constant.int 1
%14212 = torch.aten.add.Scalar %14204, %int12_11817, %int1_11818 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14212, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int12_11819 = torch.constant.int 12
%int1_11820 = torch.constant.int 1
%14213 = torch.aten.add.Scalar %14205, %int12_11819, %int1_11820 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14213, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_11821 = torch.constant.int 4
%int16_11822 = torch.constant.int 16
%int1_11823 = torch.constant.int 1
%int128_11824 = torch.constant.int 128
%14214 = torch.prim.ListConstruct %int4_11821, %3095, %int16_11822, %int1_11823, %int128_11824 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14215 = torch.aten.view %14092, %14214 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14215, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11825 = torch.constant.int 4
%int16_11826 = torch.constant.int 16
%int1_11827 = torch.constant.int 1
%int128_11828 = torch.constant.int 128
%14216 = torch.prim.ListConstruct %int4_11825, %3095, %int16_11826, %int1_11827, %int128_11828 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14217 = torch.aten.view %14107, %14216 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14217, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11829 = torch.constant.int 4
%int16_11830 = torch.constant.int 16
%int1_11831 = torch.constant.int 1
%int128_11832 = torch.constant.int 128
%14218 = torch.prim.ListConstruct %int4_11829, %3095, %int16_11830, %int1_11831, %int128_11832 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14219 = torch.aten.view %14122, %14218 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14219, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11833 = torch.constant.int 4
%int16_11834 = torch.constant.int 16
%int1_11835 = torch.constant.int 1
%int128_11836 = torch.constant.int 128
%14220 = torch.prim.ListConstruct %int4_11833, %3095, %int16_11834, %int1_11835, %int128_11836 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14221 = torch.aten.view %14137, %14220 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14221, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11837 = torch.constant.int 4
%int16_11838 = torch.constant.int 16
%int1_11839 = torch.constant.int 1
%int128_11840 = torch.constant.int 128
%14222 = torch.prim.ListConstruct %int4_11837, %3095, %int16_11838, %int1_11839, %int128_11840 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14223 = torch.aten.view %14152, %14222 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14223, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11841 = torch.constant.int 4
%int16_11842 = torch.constant.int 16
%int1_11843 = torch.constant.int 1
%int128_11844 = torch.constant.int 128
%14224 = torch.prim.ListConstruct %int4_11841, %3095, %int16_11842, %int1_11843, %int128_11844 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14225 = torch.aten.view %14167, %14224 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14225, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11845 = torch.constant.int 4
%int16_11846 = torch.constant.int 16
%int1_11847 = torch.constant.int 1
%int128_11848 = torch.constant.int 128
%14226 = torch.prim.ListConstruct %int4_11845, %3095, %int16_11846, %int1_11847, %int128_11848 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14227 = torch.aten.view %14182, %14226 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14227, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11849 = torch.constant.int 4
%int16_11850 = torch.constant.int 16
%int1_11851 = torch.constant.int 1
%int128_11852 = torch.constant.int 128
%14228 = torch.prim.ListConstruct %int4_11849, %3095, %int16_11850, %int1_11851, %int128_11852 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14229 = torch.aten.view %14197, %14228 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14229, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11853 = torch.constant.int 4
%14230 = torch.aten.mul.int %int4_11853, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11854 = torch.constant.int 16
%int1_11855 = torch.constant.int 1
%int128_11856 = torch.constant.int 128
%14231 = torch.prim.ListConstruct %14230, %int16_11854, %int1_11855, %int128_11856 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14232 = torch.aten.view %14215, %14231 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14232, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11857 = torch.constant.int 4
%14233 = torch.aten.mul.int %int4_11857, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11858 = torch.constant.int 16
%int1_11859 = torch.constant.int 1
%int128_11860 = torch.constant.int 128
%14234 = torch.prim.ListConstruct %14233, %int16_11858, %int1_11859, %int128_11860 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14235 = torch.aten.view %14217, %14234 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14235, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11861 = torch.constant.int 4
%14236 = torch.aten.mul.int %int4_11861, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11862 = torch.constant.int 16
%int1_11863 = torch.constant.int 1
%int128_11864 = torch.constant.int 128
%14237 = torch.prim.ListConstruct %14236, %int16_11862, %int1_11863, %int128_11864 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14238 = torch.aten.view %14219, %14237 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14238, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11865 = torch.constant.int 4
%14239 = torch.aten.mul.int %int4_11865, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11866 = torch.constant.int 16
%int1_11867 = torch.constant.int 1
%int128_11868 = torch.constant.int 128
%14240 = torch.prim.ListConstruct %14239, %int16_11866, %int1_11867, %int128_11868 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14241 = torch.aten.view %14221, %14240 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14241, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11869 = torch.constant.int 4
%14242 = torch.aten.mul.int %int4_11869, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11870 = torch.constant.int 16
%int1_11871 = torch.constant.int 1
%int128_11872 = torch.constant.int 128
%14243 = torch.prim.ListConstruct %14242, %int16_11870, %int1_11871, %int128_11872 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14244 = torch.aten.view %14223, %14243 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14244, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11873 = torch.constant.int 4
%14245 = torch.aten.mul.int %int4_11873, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11874 = torch.constant.int 16
%int1_11875 = torch.constant.int 1
%int128_11876 = torch.constant.int 128
%14246 = torch.prim.ListConstruct %14245, %int16_11874, %int1_11875, %int128_11876 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14247 = torch.aten.view %14225, %14246 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14247, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11877 = torch.constant.int 4
%14248 = torch.aten.mul.int %int4_11877, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11878 = torch.constant.int 16
%int1_11879 = torch.constant.int 1
%int128_11880 = torch.constant.int 128
%14249 = torch.prim.ListConstruct %14248, %int16_11878, %int1_11879, %int128_11880 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14250 = torch.aten.view %14227, %14249 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14250, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11881 = torch.constant.int 4
%14251 = torch.aten.mul.int %int4_11881, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11882 = torch.constant.int 16
%int1_11883 = torch.constant.int 1
%int128_11884 = torch.constant.int 128
%14252 = torch.prim.ListConstruct %14251, %int16_11882, %int1_11883, %int128_11884 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14253 = torch.aten.view %14229, %14252 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14253, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11885 = torch.constant.int 4
%14254 = torch.aten.mul.int %int4_11885, %3095 : !torch.int, !torch.int -> !torch.int
%14255 = torch.prim.ListConstruct %14254 : (!torch.int) -> !torch.list<int>
%14256 = torch.aten.view %14206, %14255 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14256, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11886 = torch.constant.int 4
%14257 = torch.aten.mul.int %int4_11886, %3095 : !torch.int, !torch.int -> !torch.int
%14258 = torch.prim.ListConstruct %14257 : (!torch.int) -> !torch.list<int>
%14259 = torch.aten.view %14207, %14258 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14259, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11887 = torch.constant.int 4
%14260 = torch.aten.mul.int %int4_11887, %3095 : !torch.int, !torch.int -> !torch.int
%14261 = torch.prim.ListConstruct %14260 : (!torch.int) -> !torch.list<int>
%14262 = torch.aten.view %14208, %14261 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14262, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11888 = torch.constant.int 4
%14263 = torch.aten.mul.int %int4_11888, %3095 : !torch.int, !torch.int -> !torch.int
%14264 = torch.prim.ListConstruct %14263 : (!torch.int) -> !torch.list<int>
%14265 = torch.aten.view %14209, %14264 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14265, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11889 = torch.constant.int 4
%14266 = torch.aten.mul.int %int4_11889, %3095 : !torch.int, !torch.int -> !torch.int
%14267 = torch.prim.ListConstruct %14266 : (!torch.int) -> !torch.list<int>
%14268 = torch.aten.view %14210, %14267 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14268, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11890 = torch.constant.int 4
%14269 = torch.aten.mul.int %int4_11890, %3095 : !torch.int, !torch.int -> !torch.int
%14270 = torch.prim.ListConstruct %14269 : (!torch.int) -> !torch.list<int>
%14271 = torch.aten.view %14211, %14270 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14271, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11891 = torch.constant.int 4
%14272 = torch.aten.mul.int %int4_11891, %3095 : !torch.int, !torch.int -> !torch.int
%14273 = torch.prim.ListConstruct %14272 : (!torch.int) -> !torch.list<int>
%14274 = torch.aten.view %14212, %14273 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14274, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11892 = torch.constant.int 4
%14275 = torch.aten.mul.int %int4_11892, %3095 : !torch.int, !torch.int -> !torch.int
%14276 = torch.prim.ListConstruct %14275 : (!torch.int) -> !torch.list<int>
%14277 = torch.aten.view %14213, %14276 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14277, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11893 = torch.constant.int 4
%int16_11894 = torch.constant.int 16
%int1_11895 = torch.constant.int 1
%int128_11896 = torch.constant.int 128
%14278 = torch.prim.ListConstruct %int4_11893, %3095, %int16_11894, %int1_11895, %int128_11896 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14279 = torch.aten.view %13867, %14278 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14279, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11897 = torch.constant.int 4
%int16_11898 = torch.constant.int 16
%int1_11899 = torch.constant.int 1
%int128_11900 = torch.constant.int 128
%14280 = torch.prim.ListConstruct %int4_11897, %3095, %int16_11898, %int1_11899, %int128_11900 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14281 = torch.aten.view %13869, %14280 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14281, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11901 = torch.constant.int 4
%int16_11902 = torch.constant.int 16
%int1_11903 = torch.constant.int 1
%int128_11904 = torch.constant.int 128
%14282 = torch.prim.ListConstruct %int4_11901, %3095, %int16_11902, %int1_11903, %int128_11904 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14283 = torch.aten.view %13871, %14282 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14283, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11905 = torch.constant.int 4
%int16_11906 = torch.constant.int 16
%int1_11907 = torch.constant.int 1
%int128_11908 = torch.constant.int 128
%14284 = torch.prim.ListConstruct %int4_11905, %3095, %int16_11906, %int1_11907, %int128_11908 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14285 = torch.aten.view %13873, %14284 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14285, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11909 = torch.constant.int 4
%int16_11910 = torch.constant.int 16
%int1_11911 = torch.constant.int 1
%int128_11912 = torch.constant.int 128
%14286 = torch.prim.ListConstruct %int4_11909, %3095, %int16_11910, %int1_11911, %int128_11912 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14287 = torch.aten.view %13875, %14286 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14287, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11913 = torch.constant.int 4
%int16_11914 = torch.constant.int 16
%int1_11915 = torch.constant.int 1
%int128_11916 = torch.constant.int 128
%14288 = torch.prim.ListConstruct %int4_11913, %3095, %int16_11914, %int1_11915, %int128_11916 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14289 = torch.aten.view %13877, %14288 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14289, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11917 = torch.constant.int 4
%int16_11918 = torch.constant.int 16
%int1_11919 = torch.constant.int 1
%int128_11920 = torch.constant.int 128
%14290 = torch.prim.ListConstruct %int4_11917, %3095, %int16_11918, %int1_11919, %int128_11920 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14291 = torch.aten.view %13879, %14290 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14291, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11921 = torch.constant.int 4
%int16_11922 = torch.constant.int 16
%int1_11923 = torch.constant.int 1
%int128_11924 = torch.constant.int 128
%14292 = torch.prim.ListConstruct %int4_11921, %3095, %int16_11922, %int1_11923, %int128_11924 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14293 = torch.aten.view %13881, %14292 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %14293, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_11925 = torch.constant.int 4
%14294 = torch.aten.mul.int %int4_11925, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11926 = torch.constant.int 16
%int1_11927 = torch.constant.int 1
%int128_11928 = torch.constant.int 128
%14295 = torch.prim.ListConstruct %14294, %int16_11926, %int1_11927, %int128_11928 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14296 = torch.aten.view %14279, %14295 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14296, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11929 = torch.constant.int 4
%14297 = torch.aten.mul.int %int4_11929, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11930 = torch.constant.int 16
%int1_11931 = torch.constant.int 1
%int128_11932 = torch.constant.int 128
%14298 = torch.prim.ListConstruct %14297, %int16_11930, %int1_11931, %int128_11932 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14299 = torch.aten.view %14281, %14298 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14299, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11933 = torch.constant.int 4
%14300 = torch.aten.mul.int %int4_11933, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11934 = torch.constant.int 16
%int1_11935 = torch.constant.int 1
%int128_11936 = torch.constant.int 128
%14301 = torch.prim.ListConstruct %14300, %int16_11934, %int1_11935, %int128_11936 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14302 = torch.aten.view %14283, %14301 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14302, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11937 = torch.constant.int 4
%14303 = torch.aten.mul.int %int4_11937, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11938 = torch.constant.int 16
%int1_11939 = torch.constant.int 1
%int128_11940 = torch.constant.int 128
%14304 = torch.prim.ListConstruct %14303, %int16_11938, %int1_11939, %int128_11940 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14305 = torch.aten.view %14285, %14304 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14305, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11941 = torch.constant.int 4
%14306 = torch.aten.mul.int %int4_11941, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11942 = torch.constant.int 16
%int1_11943 = torch.constant.int 1
%int128_11944 = torch.constant.int 128
%14307 = torch.prim.ListConstruct %14306, %int16_11942, %int1_11943, %int128_11944 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14308 = torch.aten.view %14287, %14307 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14308, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11945 = torch.constant.int 4
%14309 = torch.aten.mul.int %int4_11945, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11946 = torch.constant.int 16
%int1_11947 = torch.constant.int 1
%int128_11948 = torch.constant.int 128
%14310 = torch.prim.ListConstruct %14309, %int16_11946, %int1_11947, %int128_11948 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14311 = torch.aten.view %14289, %14310 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14311, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11949 = torch.constant.int 4
%14312 = torch.aten.mul.int %int4_11949, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11950 = torch.constant.int 16
%int1_11951 = torch.constant.int 1
%int128_11952 = torch.constant.int 128
%14313 = torch.prim.ListConstruct %14312, %int16_11950, %int1_11951, %int128_11952 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14314 = torch.aten.view %14291, %14313 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14314, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_11953 = torch.constant.int 4
%14315 = torch.aten.mul.int %int4_11953, %3095 : !torch.int, !torch.int -> !torch.int
%int16_11954 = torch.constant.int 16
%int1_11955 = torch.constant.int 1
%int128_11956 = torch.constant.int 128
%14316 = torch.prim.ListConstruct %14315, %int16_11954, %int1_11955, %int128_11956 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14317 = torch.aten.view %14293, %14316 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14317, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_11957 = torch.constant.int 1
%int1_11958 = torch.constant.int 1
%14318 = torch.aten.add.Scalar %14206, %int1_11957, %int1_11958 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14318, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11959 = torch.constant.int 1
%int1_11960 = torch.constant.int 1
%14319 = torch.aten.add.Scalar %14207, %int1_11959, %int1_11960 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14319, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11961 = torch.constant.int 1
%int1_11962 = torch.constant.int 1
%14320 = torch.aten.add.Scalar %14208, %int1_11961, %int1_11962 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14320, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11963 = torch.constant.int 1
%int1_11964 = torch.constant.int 1
%14321 = torch.aten.add.Scalar %14209, %int1_11963, %int1_11964 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14321, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11965 = torch.constant.int 1
%int1_11966 = torch.constant.int 1
%14322 = torch.aten.add.Scalar %14210, %int1_11965, %int1_11966 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14322, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11967 = torch.constant.int 1
%int1_11968 = torch.constant.int 1
%14323 = torch.aten.add.Scalar %14211, %int1_11967, %int1_11968 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14323, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11969 = torch.constant.int 1
%int1_11970 = torch.constant.int 1
%14324 = torch.aten.add.Scalar %14212, %int1_11969, %int1_11970 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14324, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_11971 = torch.constant.int 1
%int1_11972 = torch.constant.int 1
%14325 = torch.aten.add.Scalar %14213, %int1_11971, %int1_11972 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %14325, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_11973 = torch.constant.int 4
%14326 = torch.aten.mul.int %int4_11973, %3095 : !torch.int, !torch.int -> !torch.int
%14327 = torch.prim.ListConstruct %14326 : (!torch.int) -> !torch.list<int>
%14328 = torch.aten.view %14318, %14327 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14328, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11974 = torch.constant.int 4
%14329 = torch.aten.mul.int %int4_11974, %3095 : !torch.int, !torch.int -> !torch.int
%14330 = torch.prim.ListConstruct %14329 : (!torch.int) -> !torch.list<int>
%14331 = torch.aten.view %14319, %14330 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14331, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11975 = torch.constant.int 4
%14332 = torch.aten.mul.int %int4_11975, %3095 : !torch.int, !torch.int -> !torch.int
%14333 = torch.prim.ListConstruct %14332 : (!torch.int) -> !torch.list<int>
%14334 = torch.aten.view %14320, %14333 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14334, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11976 = torch.constant.int 4
%14335 = torch.aten.mul.int %int4_11976, %3095 : !torch.int, !torch.int -> !torch.int
%14336 = torch.prim.ListConstruct %14335 : (!torch.int) -> !torch.list<int>
%14337 = torch.aten.view %14321, %14336 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14337, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11977 = torch.constant.int 4
%14338 = torch.aten.mul.int %int4_11977, %3095 : !torch.int, !torch.int -> !torch.int
%14339 = torch.prim.ListConstruct %14338 : (!torch.int) -> !torch.list<int>
%14340 = torch.aten.view %14322, %14339 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14340, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11978 = torch.constant.int 4
%14341 = torch.aten.mul.int %int4_11978, %3095 : !torch.int, !torch.int -> !torch.int
%14342 = torch.prim.ListConstruct %14341 : (!torch.int) -> !torch.list<int>
%14343 = torch.aten.view %14323, %14342 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14343, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11979 = torch.constant.int 4
%14344 = torch.aten.mul.int %int4_11979, %3095 : !torch.int, !torch.int -> !torch.int
%14345 = torch.prim.ListConstruct %14344 : (!torch.int) -> !torch.list<int>
%14346 = torch.aten.view %14324, %14345 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14346, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_11980 = torch.constant.int 4
%14347 = torch.aten.mul.int %int4_11980, %3095 : !torch.int, !torch.int -> !torch.int
%14348 = torch.prim.ListConstruct %14347 : (!torch.int) -> !torch.list<int>
%14349 = torch.aten.view %14325, %14348 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14349, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%14350 = torch.prim.ListConstruct %14256, %14328 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11981 = torch.constant.int 0
%14351 = torch.aten.cat %14350, %int0_11981 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14351, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14352 = torch.prim.ListConstruct %14259, %14331 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11982 = torch.constant.int 0
%14353 = torch.aten.cat %14352, %int0_11982 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14353, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14354 = torch.prim.ListConstruct %14262, %14334 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11983 = torch.constant.int 0
%14355 = torch.aten.cat %14354, %int0_11983 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14355, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14356 = torch.prim.ListConstruct %14265, %14337 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11984 = torch.constant.int 0
%14357 = torch.aten.cat %14356, %int0_11984 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14357, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14358 = torch.prim.ListConstruct %14268, %14340 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11985 = torch.constant.int 0
%14359 = torch.aten.cat %14358, %int0_11985 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14359, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14360 = torch.prim.ListConstruct %14271, %14343 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11986 = torch.constant.int 0
%14361 = torch.aten.cat %14360, %int0_11986 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14361, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14362 = torch.prim.ListConstruct %14274, %14346 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11987 = torch.constant.int 0
%14363 = torch.aten.cat %14362, %int0_11987 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14363, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14364 = torch.prim.ListConstruct %14277, %14349 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_11988 = torch.constant.int 0
%14365 = torch.aten.cat %14364, %int0_11988 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %14365, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%14366 = torch.prim.ListConstruct %14232, %14296 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11989 = torch.constant.int 0
%14367 = torch.aten.cat %14366, %int0_11989 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14367, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14368 = torch.prim.ListConstruct %14235, %14299 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11990 = torch.constant.int 0
%14369 = torch.aten.cat %14368, %int0_11990 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14369, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14370 = torch.prim.ListConstruct %14238, %14302 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11991 = torch.constant.int 0
%14371 = torch.aten.cat %14370, %int0_11991 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14371, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14372 = torch.prim.ListConstruct %14241, %14305 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11992 = torch.constant.int 0
%14373 = torch.aten.cat %14372, %int0_11992 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14373, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14374 = torch.prim.ListConstruct %14244, %14308 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11993 = torch.constant.int 0
%14375 = torch.aten.cat %14374, %int0_11993 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14375, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14376 = torch.prim.ListConstruct %14247, %14311 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11994 = torch.constant.int 0
%14377 = torch.aten.cat %14376, %int0_11994 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14377, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14378 = torch.prim.ListConstruct %14250, %14314 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11995 = torch.constant.int 0
%14379 = torch.aten.cat %14378, %int0_11995 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14379, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14380 = torch.prim.ListConstruct %14253, %14317 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_11996 = torch.constant.int 0
%14381 = torch.aten.cat %14380, %int0_11996 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14381, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_11997 = torch.constant.int 32
%int2_11998 = torch.constant.int 2
%int16_11999 = torch.constant.int 16
%int1_12000 = torch.constant.int 1
%int128_12001 = torch.constant.int 128
%14382 = torch.prim.ListConstruct %3023, %int32_11997, %int2_11998, %int16_11999, %int1_12000, %int128_12001 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14383 = torch.aten.view %12532, %14382 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14383, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12002 = torch.constant.int 32
%14384 = torch.aten.mul.int %3023, %int32_12002 : !torch.int, !torch.int -> !torch.int
%int2_12003 = torch.constant.int 2
%14385 = torch.aten.mul.int %14384, %int2_12003 : !torch.int, !torch.int -> !torch.int
%int16_12004 = torch.constant.int 16
%int1_12005 = torch.constant.int 1
%int128_12006 = torch.constant.int 128
%14386 = torch.prim.ListConstruct %14385, %int16_12004, %int1_12005, %int128_12006 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14387 = torch.aten.view %14383, %14386 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14387, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14388 = torch.prim.ListConstruct %14351 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12007 = torch.constant.bool false
%14389 = torch.aten.index_put %14387, %14388, %14367, %false_12007 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14389, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12008 = torch.constant.int 32
%int2_12009 = torch.constant.int 2
%int16_12010 = torch.constant.int 16
%int1_12011 = torch.constant.int 1
%int128_12012 = torch.constant.int 128
%14390 = torch.prim.ListConstruct %3023, %int32_12008, %int2_12009, %int16_12010, %int1_12011, %int128_12012 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14391 = torch.aten.view %14389, %14390 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14391, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12013 = torch.constant.int 131072
%14392 = torch.prim.ListConstruct %3023, %int131072_12013 : (!torch.int, !torch.int) -> !torch.list<int>
%14393 = torch.aten.view %14391, %14392 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14393, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12014 = torch.constant.int 32
%int2_12015 = torch.constant.int 2
%int16_12016 = torch.constant.int 16
%int1_12017 = torch.constant.int 1
%int128_12018 = torch.constant.int 128
%14394 = torch.prim.ListConstruct %3026, %int32_12014, %int2_12015, %int16_12016, %int1_12017, %int128_12018 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14395 = torch.aten.view %12544, %14394 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14395, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12019 = torch.constant.int 32
%14396 = torch.aten.mul.int %3026, %int32_12019 : !torch.int, !torch.int -> !torch.int
%int2_12020 = torch.constant.int 2
%14397 = torch.aten.mul.int %14396, %int2_12020 : !torch.int, !torch.int -> !torch.int
%int16_12021 = torch.constant.int 16
%int1_12022 = torch.constant.int 1
%int128_12023 = torch.constant.int 128
%14398 = torch.prim.ListConstruct %14397, %int16_12021, %int1_12022, %int128_12023 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14399 = torch.aten.view %14395, %14398 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14399, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14400 = torch.prim.ListConstruct %14353 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12024 = torch.constant.bool false
%14401 = torch.aten.index_put %14399, %14400, %14369, %false_12024 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14401, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12025 = torch.constant.int 32
%int2_12026 = torch.constant.int 2
%int16_12027 = torch.constant.int 16
%int1_12028 = torch.constant.int 1
%int128_12029 = torch.constant.int 128
%14402 = torch.prim.ListConstruct %3026, %int32_12025, %int2_12026, %int16_12027, %int1_12028, %int128_12029 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14403 = torch.aten.view %14401, %14402 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14403, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12030 = torch.constant.int 131072
%14404 = torch.prim.ListConstruct %3026, %int131072_12030 : (!torch.int, !torch.int) -> !torch.list<int>
%14405 = torch.aten.view %14403, %14404 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14405, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12031 = torch.constant.int 32
%int2_12032 = torch.constant.int 2
%int16_12033 = torch.constant.int 16
%int1_12034 = torch.constant.int 1
%int128_12035 = torch.constant.int 128
%14406 = torch.prim.ListConstruct %3029, %int32_12031, %int2_12032, %int16_12033, %int1_12034, %int128_12035 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14407 = torch.aten.view %12556, %14406 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14407, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12036 = torch.constant.int 32
%14408 = torch.aten.mul.int %3029, %int32_12036 : !torch.int, !torch.int -> !torch.int
%int2_12037 = torch.constant.int 2
%14409 = torch.aten.mul.int %14408, %int2_12037 : !torch.int, !torch.int -> !torch.int
%int16_12038 = torch.constant.int 16
%int1_12039 = torch.constant.int 1
%int128_12040 = torch.constant.int 128
%14410 = torch.prim.ListConstruct %14409, %int16_12038, %int1_12039, %int128_12040 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14411 = torch.aten.view %14407, %14410 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14411, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14412 = torch.prim.ListConstruct %14355 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12041 = torch.constant.bool false
%14413 = torch.aten.index_put %14411, %14412, %14371, %false_12041 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14413, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12042 = torch.constant.int 32
%int2_12043 = torch.constant.int 2
%int16_12044 = torch.constant.int 16
%int1_12045 = torch.constant.int 1
%int128_12046 = torch.constant.int 128
%14414 = torch.prim.ListConstruct %3029, %int32_12042, %int2_12043, %int16_12044, %int1_12045, %int128_12046 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14415 = torch.aten.view %14413, %14414 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14415, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12047 = torch.constant.int 131072
%14416 = torch.prim.ListConstruct %3029, %int131072_12047 : (!torch.int, !torch.int) -> !torch.list<int>
%14417 = torch.aten.view %14415, %14416 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14417, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12048 = torch.constant.int 32
%int2_12049 = torch.constant.int 2
%int16_12050 = torch.constant.int 16
%int1_12051 = torch.constant.int 1
%int128_12052 = torch.constant.int 128
%14418 = torch.prim.ListConstruct %3032, %int32_12048, %int2_12049, %int16_12050, %int1_12051, %int128_12052 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14419 = torch.aten.view %12568, %14418 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14419, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12053 = torch.constant.int 32
%14420 = torch.aten.mul.int %3032, %int32_12053 : !torch.int, !torch.int -> !torch.int
%int2_12054 = torch.constant.int 2
%14421 = torch.aten.mul.int %14420, %int2_12054 : !torch.int, !torch.int -> !torch.int
%int16_12055 = torch.constant.int 16
%int1_12056 = torch.constant.int 1
%int128_12057 = torch.constant.int 128
%14422 = torch.prim.ListConstruct %14421, %int16_12055, %int1_12056, %int128_12057 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14423 = torch.aten.view %14419, %14422 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14423, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14424 = torch.prim.ListConstruct %14357 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12058 = torch.constant.bool false
%14425 = torch.aten.index_put %14423, %14424, %14373, %false_12058 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14425, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12059 = torch.constant.int 32
%int2_12060 = torch.constant.int 2
%int16_12061 = torch.constant.int 16
%int1_12062 = torch.constant.int 1
%int128_12063 = torch.constant.int 128
%14426 = torch.prim.ListConstruct %3032, %int32_12059, %int2_12060, %int16_12061, %int1_12062, %int128_12063 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14427 = torch.aten.view %14425, %14426 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14427, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12064 = torch.constant.int 131072
%14428 = torch.prim.ListConstruct %3032, %int131072_12064 : (!torch.int, !torch.int) -> !torch.list<int>
%14429 = torch.aten.view %14427, %14428 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14429, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12065 = torch.constant.int 32
%int2_12066 = torch.constant.int 2
%int16_12067 = torch.constant.int 16
%int1_12068 = torch.constant.int 1
%int128_12069 = torch.constant.int 128
%14430 = torch.prim.ListConstruct %3035, %int32_12065, %int2_12066, %int16_12067, %int1_12068, %int128_12069 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14431 = torch.aten.view %12580, %14430 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14431, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12070 = torch.constant.int 32
%14432 = torch.aten.mul.int %3035, %int32_12070 : !torch.int, !torch.int -> !torch.int
%int2_12071 = torch.constant.int 2
%14433 = torch.aten.mul.int %14432, %int2_12071 : !torch.int, !torch.int -> !torch.int
%int16_12072 = torch.constant.int 16
%int1_12073 = torch.constant.int 1
%int128_12074 = torch.constant.int 128
%14434 = torch.prim.ListConstruct %14433, %int16_12072, %int1_12073, %int128_12074 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14435 = torch.aten.view %14431, %14434 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14435, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14436 = torch.prim.ListConstruct %14359 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12075 = torch.constant.bool false
%14437 = torch.aten.index_put %14435, %14436, %14375, %false_12075 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14437, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12076 = torch.constant.int 32
%int2_12077 = torch.constant.int 2
%int16_12078 = torch.constant.int 16
%int1_12079 = torch.constant.int 1
%int128_12080 = torch.constant.int 128
%14438 = torch.prim.ListConstruct %3035, %int32_12076, %int2_12077, %int16_12078, %int1_12079, %int128_12080 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14439 = torch.aten.view %14437, %14438 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14439, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12081 = torch.constant.int 131072
%14440 = torch.prim.ListConstruct %3035, %int131072_12081 : (!torch.int, !torch.int) -> !torch.list<int>
%14441 = torch.aten.view %14439, %14440 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14441, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12082 = torch.constant.int 32
%int2_12083 = torch.constant.int 2
%int16_12084 = torch.constant.int 16
%int1_12085 = torch.constant.int 1
%int128_12086 = torch.constant.int 128
%14442 = torch.prim.ListConstruct %3038, %int32_12082, %int2_12083, %int16_12084, %int1_12085, %int128_12086 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14443 = torch.aten.view %12592, %14442 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14443, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12087 = torch.constant.int 32
%14444 = torch.aten.mul.int %3038, %int32_12087 : !torch.int, !torch.int -> !torch.int
%int2_12088 = torch.constant.int 2
%14445 = torch.aten.mul.int %14444, %int2_12088 : !torch.int, !torch.int -> !torch.int
%int16_12089 = torch.constant.int 16
%int1_12090 = torch.constant.int 1
%int128_12091 = torch.constant.int 128
%14446 = torch.prim.ListConstruct %14445, %int16_12089, %int1_12090, %int128_12091 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14447 = torch.aten.view %14443, %14446 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14447, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14448 = torch.prim.ListConstruct %14361 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12092 = torch.constant.bool false
%14449 = torch.aten.index_put %14447, %14448, %14377, %false_12092 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14449, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12093 = torch.constant.int 32
%int2_12094 = torch.constant.int 2
%int16_12095 = torch.constant.int 16
%int1_12096 = torch.constant.int 1
%int128_12097 = torch.constant.int 128
%14450 = torch.prim.ListConstruct %3038, %int32_12093, %int2_12094, %int16_12095, %int1_12096, %int128_12097 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14451 = torch.aten.view %14449, %14450 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14451, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12098 = torch.constant.int 131072
%14452 = torch.prim.ListConstruct %3038, %int131072_12098 : (!torch.int, !torch.int) -> !torch.list<int>
%14453 = torch.aten.view %14451, %14452 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14453, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12099 = torch.constant.int 32
%int2_12100 = torch.constant.int 2
%int16_12101 = torch.constant.int 16
%int1_12102 = torch.constant.int 1
%int128_12103 = torch.constant.int 128
%14454 = torch.prim.ListConstruct %3041, %int32_12099, %int2_12100, %int16_12101, %int1_12102, %int128_12103 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14455 = torch.aten.view %12604, %14454 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14455, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12104 = torch.constant.int 32
%14456 = torch.aten.mul.int %3041, %int32_12104 : !torch.int, !torch.int -> !torch.int
%int2_12105 = torch.constant.int 2
%14457 = torch.aten.mul.int %14456, %int2_12105 : !torch.int, !torch.int -> !torch.int
%int16_12106 = torch.constant.int 16
%int1_12107 = torch.constant.int 1
%int128_12108 = torch.constant.int 128
%14458 = torch.prim.ListConstruct %14457, %int16_12106, %int1_12107, %int128_12108 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14459 = torch.aten.view %14455, %14458 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14459, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14460 = torch.prim.ListConstruct %14363 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12109 = torch.constant.bool false
%14461 = torch.aten.index_put %14459, %14460, %14379, %false_12109 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14461, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12110 = torch.constant.int 32
%int2_12111 = torch.constant.int 2
%int16_12112 = torch.constant.int 16
%int1_12113 = torch.constant.int 1
%int128_12114 = torch.constant.int 128
%14462 = torch.prim.ListConstruct %3041, %int32_12110, %int2_12111, %int16_12112, %int1_12113, %int128_12114 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14463 = torch.aten.view %14461, %14462 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14463, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12115 = torch.constant.int 131072
%14464 = torch.prim.ListConstruct %3041, %int131072_12115 : (!torch.int, !torch.int) -> !torch.list<int>
%14465 = torch.aten.view %14463, %14464 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14465, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_12116 = torch.constant.int 32
%int2_12117 = torch.constant.int 2
%int16_12118 = torch.constant.int 16
%int1_12119 = torch.constant.int 1
%int128_12120 = torch.constant.int 128
%14466 = torch.prim.ListConstruct %3044, %int32_12116, %int2_12117, %int16_12118, %int1_12119, %int128_12120 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14467 = torch.aten.view %12616, %14466 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14467, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_12121 = torch.constant.int 32
%14468 = torch.aten.mul.int %3044, %int32_12121 : !torch.int, !torch.int -> !torch.int
%int2_12122 = torch.constant.int 2
%14469 = torch.aten.mul.int %14468, %int2_12122 : !torch.int, !torch.int -> !torch.int
%int16_12123 = torch.constant.int 16
%int1_12124 = torch.constant.int 1
%int128_12125 = torch.constant.int 128
%14470 = torch.prim.ListConstruct %14469, %int16_12123, %int1_12124, %int128_12125 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14471 = torch.aten.view %14467, %14470 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14471, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%14472 = torch.prim.ListConstruct %14365 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_12126 = torch.constant.bool false
%14473 = torch.aten.index_put %14471, %14472, %14381, %false_12126 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %14473, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_12127 = torch.constant.int 32
%int2_12128 = torch.constant.int 2
%int16_12129 = torch.constant.int 16
%int1_12130 = torch.constant.int 1
%int128_12131 = torch.constant.int 128
%14474 = torch.prim.ListConstruct %3044, %int32_12127, %int2_12128, %int16_12129, %int1_12130, %int128_12131 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14475 = torch.aten.view %14473, %14474 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %14475, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_12132 = torch.constant.int 131072
%14476 = torch.prim.ListConstruct %3044, %int131072_12132 : (!torch.int, !torch.int) -> !torch.list<int>
%14477 = torch.aten.view %14475, %14476 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %14477, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_12133 = torch.constant.int -2
%14478 = torch.aten.unsqueeze %14092, %int-2_12133 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12134 = torch.constant.int -2
%14479 = torch.aten.unsqueeze %14107, %int-2_12134 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12135 = torch.constant.int -2
%14480 = torch.aten.unsqueeze %14122, %int-2_12135 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12136 = torch.constant.int -2
%14481 = torch.aten.unsqueeze %14137, %int-2_12136 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12137 = torch.constant.int -2
%14482 = torch.aten.unsqueeze %14152, %int-2_12137 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12138 = torch.constant.int -2
%14483 = torch.aten.unsqueeze %14167, %int-2_12138 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12139 = torch.constant.int -2
%14484 = torch.aten.unsqueeze %14182, %int-2_12139 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12140 = torch.constant.int -2
%14485 = torch.aten.unsqueeze %14197, %int-2_12140 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_12141 = torch.constant.int 4
%int1_12142 = torch.constant.int 1
%int4_12143 = torch.constant.int 4
%int128_12144 = torch.constant.int 128
%14486 = torch.prim.ListConstruct %int4_12141, %14078, %int1_12142, %int4_12143, %int128_12144 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12145 = torch.constant.bool false
%14487 = torch.aten.expand %14478, %14486, %false_12145 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12146 = torch.constant.int 4
%int1_12147 = torch.constant.int 1
%int4_12148 = torch.constant.int 4
%int128_12149 = torch.constant.int 128
%14488 = torch.prim.ListConstruct %int4_12146, %14078, %int1_12147, %int4_12148, %int128_12149 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12150 = torch.constant.bool false
%14489 = torch.aten.expand %14479, %14488, %false_12150 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12151 = torch.constant.int 4
%int1_12152 = torch.constant.int 1
%int4_12153 = torch.constant.int 4
%int128_12154 = torch.constant.int 128
%14490 = torch.prim.ListConstruct %int4_12151, %14078, %int1_12152, %int4_12153, %int128_12154 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12155 = torch.constant.bool false
%14491 = torch.aten.expand %14480, %14490, %false_12155 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12156 = torch.constant.int 4
%int1_12157 = torch.constant.int 1
%int4_12158 = torch.constant.int 4
%int128_12159 = torch.constant.int 128
%14492 = torch.prim.ListConstruct %int4_12156, %14078, %int1_12157, %int4_12158, %int128_12159 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12160 = torch.constant.bool false
%14493 = torch.aten.expand %14481, %14492, %false_12160 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12161 = torch.constant.int 4
%int1_12162 = torch.constant.int 1
%int4_12163 = torch.constant.int 4
%int128_12164 = torch.constant.int 128
%14494 = torch.prim.ListConstruct %int4_12161, %14078, %int1_12162, %int4_12163, %int128_12164 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12165 = torch.constant.bool false
%14495 = torch.aten.expand %14482, %14494, %false_12165 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12166 = torch.constant.int 4
%int1_12167 = torch.constant.int 1
%int4_12168 = torch.constant.int 4
%int128_12169 = torch.constant.int 128
%14496 = torch.prim.ListConstruct %int4_12166, %14078, %int1_12167, %int4_12168, %int128_12169 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12170 = torch.constant.bool false
%14497 = torch.aten.expand %14483, %14496, %false_12170 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12171 = torch.constant.int 4
%int1_12172 = torch.constant.int 1
%int4_12173 = torch.constant.int 4
%int128_12174 = torch.constant.int 128
%14498 = torch.prim.ListConstruct %int4_12171, %14078, %int1_12172, %int4_12173, %int128_12174 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12175 = torch.constant.bool false
%14499 = torch.aten.expand %14484, %14498, %false_12175 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12176 = torch.constant.int 4
%int1_12177 = torch.constant.int 1
%int4_12178 = torch.constant.int 4
%int128_12179 = torch.constant.int 128
%14500 = torch.prim.ListConstruct %int4_12176, %14078, %int1_12177, %int4_12178, %int128_12179 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12180 = torch.constant.bool false
%14501 = torch.aten.expand %14485, %14500, %false_12180 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12181 = torch.constant.int 4
%int4_12182 = torch.constant.int 4
%int128_12183 = torch.constant.int 128
%14502 = torch.prim.ListConstruct %int4_12181, %14078, %int4_12182, %int128_12183 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14503 = torch.aten.view %14487, %14502 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12184 = torch.constant.int 4
%int4_12185 = torch.constant.int 4
%int128_12186 = torch.constant.int 128
%14504 = torch.prim.ListConstruct %int4_12184, %14078, %int4_12185, %int128_12186 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14505 = torch.aten.view %14489, %14504 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12187 = torch.constant.int 4
%int4_12188 = torch.constant.int 4
%int128_12189 = torch.constant.int 128
%14506 = torch.prim.ListConstruct %int4_12187, %14078, %int4_12188, %int128_12189 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14507 = torch.aten.view %14491, %14506 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12190 = torch.constant.int 4
%int4_12191 = torch.constant.int 4
%int128_12192 = torch.constant.int 128
%14508 = torch.prim.ListConstruct %int4_12190, %14078, %int4_12191, %int128_12192 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14509 = torch.aten.view %14493, %14508 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12193 = torch.constant.int 4
%int4_12194 = torch.constant.int 4
%int128_12195 = torch.constant.int 128
%14510 = torch.prim.ListConstruct %int4_12193, %14078, %int4_12194, %int128_12195 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14511 = torch.aten.view %14495, %14510 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12196 = torch.constant.int 4
%int4_12197 = torch.constant.int 4
%int128_12198 = torch.constant.int 128
%14512 = torch.prim.ListConstruct %int4_12196, %14078, %int4_12197, %int128_12198 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14513 = torch.aten.view %14497, %14512 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12199 = torch.constant.int 4
%int4_12200 = torch.constant.int 4
%int128_12201 = torch.constant.int 128
%14514 = torch.prim.ListConstruct %int4_12199, %14078, %int4_12200, %int128_12201 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14515 = torch.aten.view %14499, %14514 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12202 = torch.constant.int 4
%int4_12203 = torch.constant.int 4
%int128_12204 = torch.constant.int 128
%14516 = torch.prim.ListConstruct %int4_12202, %14078, %int4_12203, %int128_12204 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14517 = torch.aten.view %14501, %14516 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_12205 = torch.constant.int -2
%14518 = torch.aten.unsqueeze %13867, %int-2_12205 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12206 = torch.constant.int -2
%14519 = torch.aten.unsqueeze %13869, %int-2_12206 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12207 = torch.constant.int -2
%14520 = torch.aten.unsqueeze %13871, %int-2_12207 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12208 = torch.constant.int -2
%14521 = torch.aten.unsqueeze %13873, %int-2_12208 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12209 = torch.constant.int -2
%14522 = torch.aten.unsqueeze %13875, %int-2_12209 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12210 = torch.constant.int -2
%14523 = torch.aten.unsqueeze %13877, %int-2_12210 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12211 = torch.constant.int -2
%14524 = torch.aten.unsqueeze %13879, %int-2_12211 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_12212 = torch.constant.int -2
%14525 = torch.aten.unsqueeze %13881, %int-2_12212 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %14525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_12213 = torch.constant.int 1
%14526 = torch.aten.size.int %13791, %int1_12213 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_12214 = torch.constant.int 4
%int1_12215 = torch.constant.int 1
%int4_12216 = torch.constant.int 4
%int128_12217 = torch.constant.int 128
%14527 = torch.prim.ListConstruct %int4_12214, %14526, %int1_12215, %int4_12216, %int128_12217 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12218 = torch.constant.bool false
%14528 = torch.aten.expand %14518, %14527, %false_12218 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12219 = torch.constant.int 4
%int1_12220 = torch.constant.int 1
%int4_12221 = torch.constant.int 4
%int128_12222 = torch.constant.int 128
%14529 = torch.prim.ListConstruct %int4_12219, %14526, %int1_12220, %int4_12221, %int128_12222 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12223 = torch.constant.bool false
%14530 = torch.aten.expand %14519, %14529, %false_12223 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12224 = torch.constant.int 4
%int1_12225 = torch.constant.int 1
%int4_12226 = torch.constant.int 4
%int128_12227 = torch.constant.int 128
%14531 = torch.prim.ListConstruct %int4_12224, %14526, %int1_12225, %int4_12226, %int128_12227 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12228 = torch.constant.bool false
%14532 = torch.aten.expand %14520, %14531, %false_12228 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12229 = torch.constant.int 4
%int1_12230 = torch.constant.int 1
%int4_12231 = torch.constant.int 4
%int128_12232 = torch.constant.int 128
%14533 = torch.prim.ListConstruct %int4_12229, %14526, %int1_12230, %int4_12231, %int128_12232 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12233 = torch.constant.bool false
%14534 = torch.aten.expand %14521, %14533, %false_12233 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12234 = torch.constant.int 4
%int1_12235 = torch.constant.int 1
%int4_12236 = torch.constant.int 4
%int128_12237 = torch.constant.int 128
%14535 = torch.prim.ListConstruct %int4_12234, %14526, %int1_12235, %int4_12236, %int128_12237 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12238 = torch.constant.bool false
%14536 = torch.aten.expand %14522, %14535, %false_12238 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12239 = torch.constant.int 4
%int1_12240 = torch.constant.int 1
%int4_12241 = torch.constant.int 4
%int128_12242 = torch.constant.int 128
%14537 = torch.prim.ListConstruct %int4_12239, %14526, %int1_12240, %int4_12241, %int128_12242 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12243 = torch.constant.bool false
%14538 = torch.aten.expand %14523, %14537, %false_12243 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12244 = torch.constant.int 4
%int1_12245 = torch.constant.int 1
%int4_12246 = torch.constant.int 4
%int128_12247 = torch.constant.int 128
%14539 = torch.prim.ListConstruct %int4_12244, %14526, %int1_12245, %int4_12246, %int128_12247 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12248 = torch.constant.bool false
%14540 = torch.aten.expand %14524, %14539, %false_12248 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12249 = torch.constant.int 4
%int1_12250 = torch.constant.int 1
%int4_12251 = torch.constant.int 4
%int128_12252 = torch.constant.int 128
%14541 = torch.prim.ListConstruct %int4_12249, %14526, %int1_12250, %int4_12251, %int128_12252 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_12253 = torch.constant.bool false
%14542 = torch.aten.expand %14525, %14541, %false_12253 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %14542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_12254 = torch.constant.int 4
%int4_12255 = torch.constant.int 4
%int128_12256 = torch.constant.int 128
%14543 = torch.prim.ListConstruct %int4_12254, %14526, %int4_12255, %int128_12256 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14544 = torch.aten.view %14528, %14543 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12257 = torch.constant.int 4
%int4_12258 = torch.constant.int 4
%int128_12259 = torch.constant.int 128
%14545 = torch.prim.ListConstruct %int4_12257, %14526, %int4_12258, %int128_12259 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14546 = torch.aten.view %14530, %14545 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12260 = torch.constant.int 4
%int4_12261 = torch.constant.int 4
%int128_12262 = torch.constant.int 128
%14547 = torch.prim.ListConstruct %int4_12260, %14526, %int4_12261, %int128_12262 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14548 = torch.aten.view %14532, %14547 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12263 = torch.constant.int 4
%int4_12264 = torch.constant.int 4
%int128_12265 = torch.constant.int 128
%14549 = torch.prim.ListConstruct %int4_12263, %14526, %int4_12264, %int128_12265 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14550 = torch.aten.view %14534, %14549 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12266 = torch.constant.int 4
%int4_12267 = torch.constant.int 4
%int128_12268 = torch.constant.int 128
%14551 = torch.prim.ListConstruct %int4_12266, %14526, %int4_12267, %int128_12268 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14552 = torch.aten.view %14536, %14551 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12269 = torch.constant.int 4
%int4_12270 = torch.constant.int 4
%int128_12271 = torch.constant.int 128
%14553 = torch.prim.ListConstruct %int4_12269, %14526, %int4_12270, %int128_12271 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14554 = torch.aten.view %14538, %14553 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12272 = torch.constant.int 4
%int4_12273 = torch.constant.int 4
%int128_12274 = torch.constant.int 128
%14555 = torch.prim.ListConstruct %int4_12272, %14526, %int4_12273, %int128_12274 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14556 = torch.aten.view %14540, %14555 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12275 = torch.constant.int 4
%int4_12276 = torch.constant.int 4
%int128_12277 = torch.constant.int 128
%14557 = torch.prim.ListConstruct %int4_12275, %14526, %int4_12276, %int128_12277 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14558 = torch.aten.view %14542, %14557 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12278 = torch.constant.int 1
%int2_12279 = torch.constant.int 2
%14559 = torch.aten.transpose.int %13934, %int1_12278, %int2_12279 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14559, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12280 = torch.constant.int 1
%int2_12281 = torch.constant.int 2
%14560 = torch.aten.transpose.int %13949, %int1_12280, %int2_12281 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14560, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12282 = torch.constant.int 1
%int2_12283 = torch.constant.int 2
%14561 = torch.aten.transpose.int %13964, %int1_12282, %int2_12283 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14561, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12284 = torch.constant.int 1
%int2_12285 = torch.constant.int 2
%14562 = torch.aten.transpose.int %13979, %int1_12284, %int2_12285 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14562, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12286 = torch.constant.int 1
%int2_12287 = torch.constant.int 2
%14563 = torch.aten.transpose.int %13994, %int1_12286, %int2_12287 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14563, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12288 = torch.constant.int 1
%int2_12289 = torch.constant.int 2
%14564 = torch.aten.transpose.int %14009, %int1_12288, %int2_12289 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14564, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12290 = torch.constant.int 1
%int2_12291 = torch.constant.int 2
%14565 = torch.aten.transpose.int %14024, %int1_12290, %int2_12291 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14565, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12292 = torch.constant.int 1
%int2_12293 = torch.constant.int 2
%14566 = torch.aten.transpose.int %14039, %int1_12292, %int2_12293 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14566, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12294 = torch.constant.int 1
%int2_12295 = torch.constant.int 2
%14567 = torch.aten.transpose.int %14503, %int1_12294, %int2_12295 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14567, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12296 = torch.constant.int 1
%int2_12297 = torch.constant.int 2
%14568 = torch.aten.transpose.int %14505, %int1_12296, %int2_12297 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14568, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12298 = torch.constant.int 1
%int2_12299 = torch.constant.int 2
%14569 = torch.aten.transpose.int %14507, %int1_12298, %int2_12299 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14569, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12300 = torch.constant.int 1
%int2_12301 = torch.constant.int 2
%14570 = torch.aten.transpose.int %14509, %int1_12300, %int2_12301 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14570, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12302 = torch.constant.int 1
%int2_12303 = torch.constant.int 2
%14571 = torch.aten.transpose.int %14511, %int1_12302, %int2_12303 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14571, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12304 = torch.constant.int 1
%int2_12305 = torch.constant.int 2
%14572 = torch.aten.transpose.int %14513, %int1_12304, %int2_12305 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14572, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12306 = torch.constant.int 1
%int2_12307 = torch.constant.int 2
%14573 = torch.aten.transpose.int %14515, %int1_12306, %int2_12307 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14573, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12308 = torch.constant.int 1
%int2_12309 = torch.constant.int 2
%14574 = torch.aten.transpose.int %14517, %int1_12308, %int2_12309 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14574, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12310 = torch.constant.int 1
%int2_12311 = torch.constant.int 2
%14575 = torch.aten.transpose.int %14544, %int1_12310, %int2_12311 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14575, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12312 = torch.constant.int 1
%int2_12313 = torch.constant.int 2
%14576 = torch.aten.transpose.int %14546, %int1_12312, %int2_12313 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14576, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12314 = torch.constant.int 1
%int2_12315 = torch.constant.int 2
%14577 = torch.aten.transpose.int %14548, %int1_12314, %int2_12315 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14577, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12316 = torch.constant.int 1
%int2_12317 = torch.constant.int 2
%14578 = torch.aten.transpose.int %14550, %int1_12316, %int2_12317 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14578, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12318 = torch.constant.int 1
%int2_12319 = torch.constant.int 2
%14579 = torch.aten.transpose.int %14552, %int1_12318, %int2_12319 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14579, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12320 = torch.constant.int 1
%int2_12321 = torch.constant.int 2
%14580 = torch.aten.transpose.int %14554, %int1_12320, %int2_12321 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14580, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12322 = torch.constant.int 1
%int2_12323 = torch.constant.int 2
%14581 = torch.aten.transpose.int %14556, %int1_12322, %int2_12323 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14581, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12324 = torch.constant.int 1
%int2_12325 = torch.constant.int 2
%14582 = torch.aten.transpose.int %14558, %int1_12324, %int2_12325 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %14582, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12326 = torch.constant.float 0.000000e+00
%true_12327 = torch.constant.bool true
%none_12328 = torch.constant.none
%none_12329 = torch.constant.none
%14583:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14559, %14567, %14575, %float0.000000e00_12326, %true_12327, %none_12328, %none_12329) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14583#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12330 = torch.constant.float 0.000000e+00
%true_12331 = torch.constant.bool true
%none_12332 = torch.constant.none
%none_12333 = torch.constant.none
%14584:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14560, %14568, %14576, %float0.000000e00_12330, %true_12331, %none_12332, %none_12333) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14584#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12334 = torch.constant.float 0.000000e+00
%true_12335 = torch.constant.bool true
%none_12336 = torch.constant.none
%none_12337 = torch.constant.none
%14585:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14561, %14569, %14577, %float0.000000e00_12334, %true_12335, %none_12336, %none_12337) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14585#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12338 = torch.constant.float 0.000000e+00
%true_12339 = torch.constant.bool true
%none_12340 = torch.constant.none
%none_12341 = torch.constant.none
%14586:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14562, %14570, %14578, %float0.000000e00_12338, %true_12339, %none_12340, %none_12341) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14586#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12342 = torch.constant.float 0.000000e+00
%true_12343 = torch.constant.bool true
%none_12344 = torch.constant.none
%none_12345 = torch.constant.none
%14587:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14563, %14571, %14579, %float0.000000e00_12342, %true_12343, %none_12344, %none_12345) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14587#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12346 = torch.constant.float 0.000000e+00
%true_12347 = torch.constant.bool true
%none_12348 = torch.constant.none
%none_12349 = torch.constant.none
%14588:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14564, %14572, %14580, %float0.000000e00_12346, %true_12347, %none_12348, %none_12349) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14588#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12350 = torch.constant.float 0.000000e+00
%true_12351 = torch.constant.bool true
%none_12352 = torch.constant.none
%none_12353 = torch.constant.none
%14589:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14565, %14573, %14581, %float0.000000e00_12350, %true_12351, %none_12352, %none_12353) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14589#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_12354 = torch.constant.float 0.000000e+00
%true_12355 = torch.constant.bool true
%none_12356 = torch.constant.none
%none_12357 = torch.constant.none
%14590:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%14566, %14574, %14582, %float0.000000e00_12354, %true_12355, %none_12356, %none_12357) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %14590#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_12358 = torch.constant.int 1
%int2_12359 = torch.constant.int 2
%14591 = torch.aten.transpose.int %14583#0, %int1_12358, %int2_12359 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12360 = torch.constant.int 1
%int2_12361 = torch.constant.int 2
%14592 = torch.aten.transpose.int %14584#0, %int1_12360, %int2_12361 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12362 = torch.constant.int 1
%int2_12363 = torch.constant.int 2
%14593 = torch.aten.transpose.int %14585#0, %int1_12362, %int2_12363 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12364 = torch.constant.int 1
%int2_12365 = torch.constant.int 2
%14594 = torch.aten.transpose.int %14586#0, %int1_12364, %int2_12365 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12366 = torch.constant.int 1
%int2_12367 = torch.constant.int 2
%14595 = torch.aten.transpose.int %14587#0, %int1_12366, %int2_12367 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12368 = torch.constant.int 1
%int2_12369 = torch.constant.int 2
%14596 = torch.aten.transpose.int %14588#0, %int1_12368, %int2_12369 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12370 = torch.constant.int 1
%int2_12371 = torch.constant.int 2
%14597 = torch.aten.transpose.int %14589#0, %int1_12370, %int2_12371 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_12372 = torch.constant.int 1
%int2_12373 = torch.constant.int 2
%14598 = torch.aten.transpose.int %14590#0, %int1_12372, %int2_12373 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %14598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_12374 = torch.constant.int 4
%int512_12375 = torch.constant.int 512
%14599 = torch.prim.ListConstruct %int4_12374, %13920, %int512_12375 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14600 = torch.aten.view %14591, %14599 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12376 = torch.constant.int 4
%int512_12377 = torch.constant.int 512
%14601 = torch.prim.ListConstruct %int4_12376, %13935, %int512_12377 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14602 = torch.aten.view %14592, %14601 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12378 = torch.constant.int 4
%int512_12379 = torch.constant.int 512
%14603 = torch.prim.ListConstruct %int4_12378, %13950, %int512_12379 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14604 = torch.aten.view %14593, %14603 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12380 = torch.constant.int 4
%int512_12381 = torch.constant.int 512
%14605 = torch.prim.ListConstruct %int4_12380, %13965, %int512_12381 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14606 = torch.aten.view %14594, %14605 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12382 = torch.constant.int 4
%int512_12383 = torch.constant.int 512
%14607 = torch.prim.ListConstruct %int4_12382, %13980, %int512_12383 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14608 = torch.aten.view %14595, %14607 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12384 = torch.constant.int 4
%int512_12385 = torch.constant.int 512
%14609 = torch.prim.ListConstruct %int4_12384, %13995, %int512_12385 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14610 = torch.aten.view %14596, %14609 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12386 = torch.constant.int 4
%int512_12387 = torch.constant.int 512
%14611 = torch.prim.ListConstruct %int4_12386, %14010, %int512_12387 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14612 = torch.aten.view %14597, %14611 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_12388 = torch.constant.int 4
%int512_12389 = torch.constant.int 512
%14613 = torch.prim.ListConstruct %int4_12388, %14025, %int512_12389 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14614 = torch.aten.view %14598, %14613 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %14614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_12390 = torch.constant.int 1
%int0_12391 = torch.constant.int 0
%14615 = torch.prim.ListConstruct %int1_12390, %int0_12391 : (!torch.int, !torch.int) -> !torch.list<int>
%14616 = torch.aten.permute %472, %14615 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12392 = torch.constant.int 1
%int0_12393 = torch.constant.int 0
%14617 = torch.prim.ListConstruct %int1_12392, %int0_12393 : (!torch.int, !torch.int) -> !torch.list<int>
%14618 = torch.aten.permute %473, %14617 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12394 = torch.constant.int 1
%int0_12395 = torch.constant.int 0
%14619 = torch.prim.ListConstruct %int1_12394, %int0_12395 : (!torch.int, !torch.int) -> !torch.list<int>
%14620 = torch.aten.permute %474, %14619 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12396 = torch.constant.int 1
%int0_12397 = torch.constant.int 0
%14621 = torch.prim.ListConstruct %int1_12396, %int0_12397 : (!torch.int, !torch.int) -> !torch.list<int>
%14622 = torch.aten.permute %475, %14621 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12398 = torch.constant.int 1
%int0_12399 = torch.constant.int 0
%14623 = torch.prim.ListConstruct %int1_12398, %int0_12399 : (!torch.int, !torch.int) -> !torch.list<int>
%14624 = torch.aten.permute %476, %14623 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12400 = torch.constant.int 1
%int0_12401 = torch.constant.int 0
%14625 = torch.prim.ListConstruct %int1_12400, %int0_12401 : (!torch.int, !torch.int) -> !torch.list<int>
%14626 = torch.aten.permute %477, %14625 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12402 = torch.constant.int 1
%int0_12403 = torch.constant.int 0
%14627 = torch.prim.ListConstruct %int1_12402, %int0_12403 : (!torch.int, !torch.int) -> !torch.list<int>
%14628 = torch.aten.permute %478, %14627 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_12404 = torch.constant.int 1
%int0_12405 = torch.constant.int 0
%14629 = torch.prim.ListConstruct %int1_12404, %int0_12405 : (!torch.int, !torch.int) -> !torch.list<int>
%14630 = torch.aten.permute %479, %14629 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_12406 = torch.constant.int 4
%14631 = torch.aten.mul.int %int4_12406, %13920 : !torch.int, !torch.int -> !torch.int
%int512_12407 = torch.constant.int 512
%14632 = torch.prim.ListConstruct %14631, %int512_12407 : (!torch.int, !torch.int) -> !torch.list<int>
%14633 = torch.aten.view %14600, %14632 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14633, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14634 = torch.aten.mm %14633, %14616 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14634, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12408 = torch.constant.int 4
%int4096_12409 = torch.constant.int 4096
%14635 = torch.prim.ListConstruct %int4_12408, %13920, %int4096_12409 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14636 = torch.aten.view %14634, %14635 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12410 = torch.constant.int 4
%14637 = torch.aten.mul.int %int4_12410, %13935 : !torch.int, !torch.int -> !torch.int
%int512_12411 = torch.constant.int 512
%14638 = torch.prim.ListConstruct %14637, %int512_12411 : (!torch.int, !torch.int) -> !torch.list<int>
%14639 = torch.aten.view %14602, %14638 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14639, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14640 = torch.aten.mm %14639, %14618 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14640, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12412 = torch.constant.int 4
%int4096_12413 = torch.constant.int 4096
%14641 = torch.prim.ListConstruct %int4_12412, %13935, %int4096_12413 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14642 = torch.aten.view %14640, %14641 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12414 = torch.constant.int 4
%14643 = torch.aten.mul.int %int4_12414, %13950 : !torch.int, !torch.int -> !torch.int
%int512_12415 = torch.constant.int 512
%14644 = torch.prim.ListConstruct %14643, %int512_12415 : (!torch.int, !torch.int) -> !torch.list<int>
%14645 = torch.aten.view %14604, %14644 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14645, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14646 = torch.aten.mm %14645, %14620 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14646, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12416 = torch.constant.int 4
%int4096_12417 = torch.constant.int 4096
%14647 = torch.prim.ListConstruct %int4_12416, %13950, %int4096_12417 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14648 = torch.aten.view %14646, %14647 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12418 = torch.constant.int 4
%14649 = torch.aten.mul.int %int4_12418, %13965 : !torch.int, !torch.int -> !torch.int
%int512_12419 = torch.constant.int 512
%14650 = torch.prim.ListConstruct %14649, %int512_12419 : (!torch.int, !torch.int) -> !torch.list<int>
%14651 = torch.aten.view %14606, %14650 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14651, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14652 = torch.aten.mm %14651, %14622 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14652, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12420 = torch.constant.int 4
%int4096_12421 = torch.constant.int 4096
%14653 = torch.prim.ListConstruct %int4_12420, %13965, %int4096_12421 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14654 = torch.aten.view %14652, %14653 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12422 = torch.constant.int 4
%14655 = torch.aten.mul.int %int4_12422, %13980 : !torch.int, !torch.int -> !torch.int
%int512_12423 = torch.constant.int 512
%14656 = torch.prim.ListConstruct %14655, %int512_12423 : (!torch.int, !torch.int) -> !torch.list<int>
%14657 = torch.aten.view %14608, %14656 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14657, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14658 = torch.aten.mm %14657, %14624 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14658, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12424 = torch.constant.int 4
%int4096_12425 = torch.constant.int 4096
%14659 = torch.prim.ListConstruct %int4_12424, %13980, %int4096_12425 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14660 = torch.aten.view %14658, %14659 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12426 = torch.constant.int 4
%14661 = torch.aten.mul.int %int4_12426, %13995 : !torch.int, !torch.int -> !torch.int
%int512_12427 = torch.constant.int 512
%14662 = torch.prim.ListConstruct %14661, %int512_12427 : (!torch.int, !torch.int) -> !torch.list<int>
%14663 = torch.aten.view %14610, %14662 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14663, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14664 = torch.aten.mm %14663, %14626 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14664, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12428 = torch.constant.int 4
%int4096_12429 = torch.constant.int 4096
%14665 = torch.prim.ListConstruct %int4_12428, %13995, %int4096_12429 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14666 = torch.aten.view %14664, %14665 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12430 = torch.constant.int 4
%14667 = torch.aten.mul.int %int4_12430, %14010 : !torch.int, !torch.int -> !torch.int
%int512_12431 = torch.constant.int 512
%14668 = torch.prim.ListConstruct %14667, %int512_12431 : (!torch.int, !torch.int) -> !torch.list<int>
%14669 = torch.aten.view %14612, %14668 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14669, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14670 = torch.aten.mm %14669, %14628 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14670, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12432 = torch.constant.int 4
%int4096_12433 = torch.constant.int 4096
%14671 = torch.prim.ListConstruct %int4_12432, %14010, %int4096_12433 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14672 = torch.aten.view %14670, %14671 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_12434 = torch.constant.int 4
%14673 = torch.aten.mul.int %int4_12434, %14025 : !torch.int, !torch.int -> !torch.int
%int512_12435 = torch.constant.int 512
%14674 = torch.prim.ListConstruct %14673, %int512_12435 : (!torch.int, !torch.int) -> !torch.list<int>
%14675 = torch.aten.view %14614, %14674 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %14675, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%14676 = torch.aten.mm %14675, %14630 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %14676, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12436 = torch.constant.int 4
%int4096_12437 = torch.constant.int 4096
%14677 = torch.prim.ListConstruct %int4_12436, %14025, %int4096_12437 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%14678 = torch.aten.view %14676, %14677 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14679 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12438 = arith.constant 1 : index
%dim_12439 = tensor.dim %14679, %c1_12438 : tensor<4x?x4096xf16>
%14680 = flow.tensor.transfer %14679 : tensor<4x?x4096xf16>{%dim_12439} to #hal.device.promise<@__device_0>
%14681 = torch_c.from_builtin_tensor %14680 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14682 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12440 = arith.constant 1 : index
%dim_12441 = tensor.dim %14682, %c1_12440 : tensor<4x?x4096xf16>
%14683 = flow.tensor.transfer %14682 : tensor<4x?x4096xf16>{%dim_12441} to #hal.device.promise<@__device_0>
%14684 = torch_c.from_builtin_tensor %14683 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14685 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12442 = arith.constant 1 : index
%dim_12443 = tensor.dim %14685, %c1_12442 : tensor<4x?x4096xf16>
%14686 = flow.tensor.transfer %14685 : tensor<4x?x4096xf16>{%dim_12443} to #hal.device.promise<@__device_0>
%14687 = torch_c.from_builtin_tensor %14686 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14688 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12444 = arith.constant 1 : index
%dim_12445 = tensor.dim %14688, %c1_12444 : tensor<4x?x4096xf16>
%14689 = flow.tensor.transfer %14688 : tensor<4x?x4096xf16>{%dim_12445} to #hal.device.promise<@__device_0>
%14690 = torch_c.from_builtin_tensor %14689 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14691 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12446 = arith.constant 1 : index
%dim_12447 = tensor.dim %14691, %c1_12446 : tensor<4x?x4096xf16>
%14692 = flow.tensor.transfer %14691 : tensor<4x?x4096xf16>{%dim_12447} to #hal.device.promise<@__device_0>
%14693 = torch_c.from_builtin_tensor %14692 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14694 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12448 = arith.constant 1 : index
%dim_12449 = tensor.dim %14694, %c1_12448 : tensor<4x?x4096xf16>
%14695 = flow.tensor.transfer %14694 : tensor<4x?x4096xf16>{%dim_12449} to #hal.device.promise<@__device_0>
%14696 = torch_c.from_builtin_tensor %14695 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14697 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12450 = arith.constant 1 : index
%dim_12451 = tensor.dim %14697, %c1_12450 : tensor<4x?x4096xf16>
%14698 = flow.tensor.transfer %14697 : tensor<4x?x4096xf16>{%dim_12451} to #hal.device.promise<@__device_0>
%14699 = torch_c.from_builtin_tensor %14698 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12452 = torch.constant.int 1
%14700 = torch.aten.add.Tensor %14636, %14681, %int1_12452 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12453 = torch.constant.int 1
%14701 = torch.aten.add.Tensor %14700, %14684, %int1_12453 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12454 = torch.constant.int 1
%14702 = torch.aten.add.Tensor %14701, %14687, %int1_12454 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12455 = torch.constant.int 1
%14703 = torch.aten.add.Tensor %14702, %14690, %int1_12455 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12456 = torch.constant.int 1
%14704 = torch.aten.add.Tensor %14703, %14693, %int1_12456 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12457 = torch.constant.int 1
%14705 = torch.aten.add.Tensor %14704, %14696, %int1_12457 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12458 = torch.constant.int 1
%14706 = torch.aten.add.Tensor %14705, %14699, %int1_12458 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14707 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12459 = arith.constant 1 : index
%dim_12460 = tensor.dim %14707, %c1_12459 : tensor<4x?x4096xf16>
%14708 = flow.tensor.transfer %14707 : tensor<4x?x4096xf16>{%dim_12460} to #hal.device.promise<@__device_1>
%14709 = torch_c.from_builtin_tensor %14708 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14710 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12461 = arith.constant 1 : index
%dim_12462 = tensor.dim %14710, %c1_12461 : tensor<4x?x4096xf16>
%14711 = flow.tensor.transfer %14710 : tensor<4x?x4096xf16>{%dim_12462} to #hal.device.promise<@__device_1>
%14712 = torch_c.from_builtin_tensor %14711 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14713 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12463 = arith.constant 1 : index
%dim_12464 = tensor.dim %14713, %c1_12463 : tensor<4x?x4096xf16>
%14714 = flow.tensor.transfer %14713 : tensor<4x?x4096xf16>{%dim_12464} to #hal.device.promise<@__device_1>
%14715 = torch_c.from_builtin_tensor %14714 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14716 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12465 = arith.constant 1 : index
%dim_12466 = tensor.dim %14716, %c1_12465 : tensor<4x?x4096xf16>
%14717 = flow.tensor.transfer %14716 : tensor<4x?x4096xf16>{%dim_12466} to #hal.device.promise<@__device_1>
%14718 = torch_c.from_builtin_tensor %14717 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14719 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12467 = arith.constant 1 : index
%dim_12468 = tensor.dim %14719, %c1_12467 : tensor<4x?x4096xf16>
%14720 = flow.tensor.transfer %14719 : tensor<4x?x4096xf16>{%dim_12468} to #hal.device.promise<@__device_1>
%14721 = torch_c.from_builtin_tensor %14720 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14722 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12469 = arith.constant 1 : index
%dim_12470 = tensor.dim %14722, %c1_12469 : tensor<4x?x4096xf16>
%14723 = flow.tensor.transfer %14722 : tensor<4x?x4096xf16>{%dim_12470} to #hal.device.promise<@__device_1>
%14724 = torch_c.from_builtin_tensor %14723 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14725 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12471 = arith.constant 1 : index
%dim_12472 = tensor.dim %14725, %c1_12471 : tensor<4x?x4096xf16>
%14726 = flow.tensor.transfer %14725 : tensor<4x?x4096xf16>{%dim_12472} to #hal.device.promise<@__device_1>
%14727 = torch_c.from_builtin_tensor %14726 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12473 = torch.constant.int 1
%14728 = torch.aten.add.Tensor %14709, %14642, %int1_12473 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12474 = torch.constant.int 1
%14729 = torch.aten.add.Tensor %14728, %14712, %int1_12474 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12475 = torch.constant.int 1
%14730 = torch.aten.add.Tensor %14729, %14715, %int1_12475 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12476 = torch.constant.int 1
%14731 = torch.aten.add.Tensor %14730, %14718, %int1_12476 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12477 = torch.constant.int 1
%14732 = torch.aten.add.Tensor %14731, %14721, %int1_12477 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12478 = torch.constant.int 1
%14733 = torch.aten.add.Tensor %14732, %14724, %int1_12478 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12479 = torch.constant.int 1
%14734 = torch.aten.add.Tensor %14733, %14727, %int1_12479 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14735 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12480 = arith.constant 1 : index
%dim_12481 = tensor.dim %14735, %c1_12480 : tensor<4x?x4096xf16>
%14736 = flow.tensor.transfer %14735 : tensor<4x?x4096xf16>{%dim_12481} to #hal.device.promise<@__device_2>
%14737 = torch_c.from_builtin_tensor %14736 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14738 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12482 = arith.constant 1 : index
%dim_12483 = tensor.dim %14738, %c1_12482 : tensor<4x?x4096xf16>
%14739 = flow.tensor.transfer %14738 : tensor<4x?x4096xf16>{%dim_12483} to #hal.device.promise<@__device_2>
%14740 = torch_c.from_builtin_tensor %14739 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14741 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12484 = arith.constant 1 : index
%dim_12485 = tensor.dim %14741, %c1_12484 : tensor<4x?x4096xf16>
%14742 = flow.tensor.transfer %14741 : tensor<4x?x4096xf16>{%dim_12485} to #hal.device.promise<@__device_2>
%14743 = torch_c.from_builtin_tensor %14742 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14744 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12486 = arith.constant 1 : index
%dim_12487 = tensor.dim %14744, %c1_12486 : tensor<4x?x4096xf16>
%14745 = flow.tensor.transfer %14744 : tensor<4x?x4096xf16>{%dim_12487} to #hal.device.promise<@__device_2>
%14746 = torch_c.from_builtin_tensor %14745 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14747 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12488 = arith.constant 1 : index
%dim_12489 = tensor.dim %14747, %c1_12488 : tensor<4x?x4096xf16>
%14748 = flow.tensor.transfer %14747 : tensor<4x?x4096xf16>{%dim_12489} to #hal.device.promise<@__device_2>
%14749 = torch_c.from_builtin_tensor %14748 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14750 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12490 = arith.constant 1 : index
%dim_12491 = tensor.dim %14750, %c1_12490 : tensor<4x?x4096xf16>
%14751 = flow.tensor.transfer %14750 : tensor<4x?x4096xf16>{%dim_12491} to #hal.device.promise<@__device_2>
%14752 = torch_c.from_builtin_tensor %14751 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14753 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12492 = arith.constant 1 : index
%dim_12493 = tensor.dim %14753, %c1_12492 : tensor<4x?x4096xf16>
%14754 = flow.tensor.transfer %14753 : tensor<4x?x4096xf16>{%dim_12493} to #hal.device.promise<@__device_2>
%14755 = torch_c.from_builtin_tensor %14754 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12494 = torch.constant.int 1
%14756 = torch.aten.add.Tensor %14737, %14740, %int1_12494 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12495 = torch.constant.int 1
%14757 = torch.aten.add.Tensor %14756, %14648, %int1_12495 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12496 = torch.constant.int 1
%14758 = torch.aten.add.Tensor %14757, %14743, %int1_12496 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12497 = torch.constant.int 1
%14759 = torch.aten.add.Tensor %14758, %14746, %int1_12497 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12498 = torch.constant.int 1
%14760 = torch.aten.add.Tensor %14759, %14749, %int1_12498 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12499 = torch.constant.int 1
%14761 = torch.aten.add.Tensor %14760, %14752, %int1_12499 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12500 = torch.constant.int 1
%14762 = torch.aten.add.Tensor %14761, %14755, %int1_12500 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14763 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12501 = arith.constant 1 : index
%dim_12502 = tensor.dim %14763, %c1_12501 : tensor<4x?x4096xf16>
%14764 = flow.tensor.transfer %14763 : tensor<4x?x4096xf16>{%dim_12502} to #hal.device.promise<@__device_3>
%14765 = torch_c.from_builtin_tensor %14764 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14766 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12503 = arith.constant 1 : index
%dim_12504 = tensor.dim %14766, %c1_12503 : tensor<4x?x4096xf16>
%14767 = flow.tensor.transfer %14766 : tensor<4x?x4096xf16>{%dim_12504} to #hal.device.promise<@__device_3>
%14768 = torch_c.from_builtin_tensor %14767 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14769 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12505 = arith.constant 1 : index
%dim_12506 = tensor.dim %14769, %c1_12505 : tensor<4x?x4096xf16>
%14770 = flow.tensor.transfer %14769 : tensor<4x?x4096xf16>{%dim_12506} to #hal.device.promise<@__device_3>
%14771 = torch_c.from_builtin_tensor %14770 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14772 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12507 = arith.constant 1 : index
%dim_12508 = tensor.dim %14772, %c1_12507 : tensor<4x?x4096xf16>
%14773 = flow.tensor.transfer %14772 : tensor<4x?x4096xf16>{%dim_12508} to #hal.device.promise<@__device_3>
%14774 = torch_c.from_builtin_tensor %14773 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14775 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12509 = arith.constant 1 : index
%dim_12510 = tensor.dim %14775, %c1_12509 : tensor<4x?x4096xf16>
%14776 = flow.tensor.transfer %14775 : tensor<4x?x4096xf16>{%dim_12510} to #hal.device.promise<@__device_3>
%14777 = torch_c.from_builtin_tensor %14776 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14778 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12511 = arith.constant 1 : index
%dim_12512 = tensor.dim %14778, %c1_12511 : tensor<4x?x4096xf16>
%14779 = flow.tensor.transfer %14778 : tensor<4x?x4096xf16>{%dim_12512} to #hal.device.promise<@__device_3>
%14780 = torch_c.from_builtin_tensor %14779 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14781 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12513 = arith.constant 1 : index
%dim_12514 = tensor.dim %14781, %c1_12513 : tensor<4x?x4096xf16>
%14782 = flow.tensor.transfer %14781 : tensor<4x?x4096xf16>{%dim_12514} to #hal.device.promise<@__device_3>
%14783 = torch_c.from_builtin_tensor %14782 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12515 = torch.constant.int 1
%14784 = torch.aten.add.Tensor %14765, %14768, %int1_12515 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12516 = torch.constant.int 1
%14785 = torch.aten.add.Tensor %14784, %14771, %int1_12516 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12517 = torch.constant.int 1
%14786 = torch.aten.add.Tensor %14785, %14654, %int1_12517 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12518 = torch.constant.int 1
%14787 = torch.aten.add.Tensor %14786, %14774, %int1_12518 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12519 = torch.constant.int 1
%14788 = torch.aten.add.Tensor %14787, %14777, %int1_12519 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12520 = torch.constant.int 1
%14789 = torch.aten.add.Tensor %14788, %14780, %int1_12520 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12521 = torch.constant.int 1
%14790 = torch.aten.add.Tensor %14789, %14783, %int1_12521 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14791 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12522 = arith.constant 1 : index
%dim_12523 = tensor.dim %14791, %c1_12522 : tensor<4x?x4096xf16>
%14792 = flow.tensor.transfer %14791 : tensor<4x?x4096xf16>{%dim_12523} to #hal.device.promise<@__device_4>
%14793 = torch_c.from_builtin_tensor %14792 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14794 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12524 = arith.constant 1 : index
%dim_12525 = tensor.dim %14794, %c1_12524 : tensor<4x?x4096xf16>
%14795 = flow.tensor.transfer %14794 : tensor<4x?x4096xf16>{%dim_12525} to #hal.device.promise<@__device_4>
%14796 = torch_c.from_builtin_tensor %14795 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14797 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12526 = arith.constant 1 : index
%dim_12527 = tensor.dim %14797, %c1_12526 : tensor<4x?x4096xf16>
%14798 = flow.tensor.transfer %14797 : tensor<4x?x4096xf16>{%dim_12527} to #hal.device.promise<@__device_4>
%14799 = torch_c.from_builtin_tensor %14798 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14800 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12528 = arith.constant 1 : index
%dim_12529 = tensor.dim %14800, %c1_12528 : tensor<4x?x4096xf16>
%14801 = flow.tensor.transfer %14800 : tensor<4x?x4096xf16>{%dim_12529} to #hal.device.promise<@__device_4>
%14802 = torch_c.from_builtin_tensor %14801 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14803 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12530 = arith.constant 1 : index
%dim_12531 = tensor.dim %14803, %c1_12530 : tensor<4x?x4096xf16>
%14804 = flow.tensor.transfer %14803 : tensor<4x?x4096xf16>{%dim_12531} to #hal.device.promise<@__device_4>
%14805 = torch_c.from_builtin_tensor %14804 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14806 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12532 = arith.constant 1 : index
%dim_12533 = tensor.dim %14806, %c1_12532 : tensor<4x?x4096xf16>
%14807 = flow.tensor.transfer %14806 : tensor<4x?x4096xf16>{%dim_12533} to #hal.device.promise<@__device_4>
%14808 = torch_c.from_builtin_tensor %14807 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14809 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12534 = arith.constant 1 : index
%dim_12535 = tensor.dim %14809, %c1_12534 : tensor<4x?x4096xf16>
%14810 = flow.tensor.transfer %14809 : tensor<4x?x4096xf16>{%dim_12535} to #hal.device.promise<@__device_4>
%14811 = torch_c.from_builtin_tensor %14810 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12536 = torch.constant.int 1
%14812 = torch.aten.add.Tensor %14793, %14796, %int1_12536 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12537 = torch.constant.int 1
%14813 = torch.aten.add.Tensor %14812, %14799, %int1_12537 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12538 = torch.constant.int 1
%14814 = torch.aten.add.Tensor %14813, %14802, %int1_12538 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12539 = torch.constant.int 1
%14815 = torch.aten.add.Tensor %14814, %14660, %int1_12539 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12540 = torch.constant.int 1
%14816 = torch.aten.add.Tensor %14815, %14805, %int1_12540 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12541 = torch.constant.int 1
%14817 = torch.aten.add.Tensor %14816, %14808, %int1_12541 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12542 = torch.constant.int 1
%14818 = torch.aten.add.Tensor %14817, %14811, %int1_12542 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14819 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12543 = arith.constant 1 : index
%dim_12544 = tensor.dim %14819, %c1_12543 : tensor<4x?x4096xf16>
%14820 = flow.tensor.transfer %14819 : tensor<4x?x4096xf16>{%dim_12544} to #hal.device.promise<@__device_5>
%14821 = torch_c.from_builtin_tensor %14820 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14822 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12545 = arith.constant 1 : index
%dim_12546 = tensor.dim %14822, %c1_12545 : tensor<4x?x4096xf16>
%14823 = flow.tensor.transfer %14822 : tensor<4x?x4096xf16>{%dim_12546} to #hal.device.promise<@__device_5>
%14824 = torch_c.from_builtin_tensor %14823 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14825 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12547 = arith.constant 1 : index
%dim_12548 = tensor.dim %14825, %c1_12547 : tensor<4x?x4096xf16>
%14826 = flow.tensor.transfer %14825 : tensor<4x?x4096xf16>{%dim_12548} to #hal.device.promise<@__device_5>
%14827 = torch_c.from_builtin_tensor %14826 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14828 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12549 = arith.constant 1 : index
%dim_12550 = tensor.dim %14828, %c1_12549 : tensor<4x?x4096xf16>
%14829 = flow.tensor.transfer %14828 : tensor<4x?x4096xf16>{%dim_12550} to #hal.device.promise<@__device_5>
%14830 = torch_c.from_builtin_tensor %14829 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14831 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12551 = arith.constant 1 : index
%dim_12552 = tensor.dim %14831, %c1_12551 : tensor<4x?x4096xf16>
%14832 = flow.tensor.transfer %14831 : tensor<4x?x4096xf16>{%dim_12552} to #hal.device.promise<@__device_5>
%14833 = torch_c.from_builtin_tensor %14832 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14834 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12553 = arith.constant 1 : index
%dim_12554 = tensor.dim %14834, %c1_12553 : tensor<4x?x4096xf16>
%14835 = flow.tensor.transfer %14834 : tensor<4x?x4096xf16>{%dim_12554} to #hal.device.promise<@__device_5>
%14836 = torch_c.from_builtin_tensor %14835 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14837 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12555 = arith.constant 1 : index
%dim_12556 = tensor.dim %14837, %c1_12555 : tensor<4x?x4096xf16>
%14838 = flow.tensor.transfer %14837 : tensor<4x?x4096xf16>{%dim_12556} to #hal.device.promise<@__device_5>
%14839 = torch_c.from_builtin_tensor %14838 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12557 = torch.constant.int 1
%14840 = torch.aten.add.Tensor %14821, %14824, %int1_12557 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12558 = torch.constant.int 1
%14841 = torch.aten.add.Tensor %14840, %14827, %int1_12558 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12559 = torch.constant.int 1
%14842 = torch.aten.add.Tensor %14841, %14830, %int1_12559 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12560 = torch.constant.int 1
%14843 = torch.aten.add.Tensor %14842, %14833, %int1_12560 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12561 = torch.constant.int 1
%14844 = torch.aten.add.Tensor %14843, %14666, %int1_12561 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12562 = torch.constant.int 1
%14845 = torch.aten.add.Tensor %14844, %14836, %int1_12562 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12563 = torch.constant.int 1
%14846 = torch.aten.add.Tensor %14845, %14839, %int1_12563 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14847 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12564 = arith.constant 1 : index
%dim_12565 = tensor.dim %14847, %c1_12564 : tensor<4x?x4096xf16>
%14848 = flow.tensor.transfer %14847 : tensor<4x?x4096xf16>{%dim_12565} to #hal.device.promise<@__device_6>
%14849 = torch_c.from_builtin_tensor %14848 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14850 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12566 = arith.constant 1 : index
%dim_12567 = tensor.dim %14850, %c1_12566 : tensor<4x?x4096xf16>
%14851 = flow.tensor.transfer %14850 : tensor<4x?x4096xf16>{%dim_12567} to #hal.device.promise<@__device_6>
%14852 = torch_c.from_builtin_tensor %14851 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14853 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12568 = arith.constant 1 : index
%dim_12569 = tensor.dim %14853, %c1_12568 : tensor<4x?x4096xf16>
%14854 = flow.tensor.transfer %14853 : tensor<4x?x4096xf16>{%dim_12569} to #hal.device.promise<@__device_6>
%14855 = torch_c.from_builtin_tensor %14854 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14856 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12570 = arith.constant 1 : index
%dim_12571 = tensor.dim %14856, %c1_12570 : tensor<4x?x4096xf16>
%14857 = flow.tensor.transfer %14856 : tensor<4x?x4096xf16>{%dim_12571} to #hal.device.promise<@__device_6>
%14858 = torch_c.from_builtin_tensor %14857 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14859 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12572 = arith.constant 1 : index
%dim_12573 = tensor.dim %14859, %c1_12572 : tensor<4x?x4096xf16>
%14860 = flow.tensor.transfer %14859 : tensor<4x?x4096xf16>{%dim_12573} to #hal.device.promise<@__device_6>
%14861 = torch_c.from_builtin_tensor %14860 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14862 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12574 = arith.constant 1 : index
%dim_12575 = tensor.dim %14862, %c1_12574 : tensor<4x?x4096xf16>
%14863 = flow.tensor.transfer %14862 : tensor<4x?x4096xf16>{%dim_12575} to #hal.device.promise<@__device_6>
%14864 = torch_c.from_builtin_tensor %14863 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14865 = torch_c.to_builtin_tensor %14678 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12576 = arith.constant 1 : index
%dim_12577 = tensor.dim %14865, %c1_12576 : tensor<4x?x4096xf16>
%14866 = flow.tensor.transfer %14865 : tensor<4x?x4096xf16>{%dim_12577} to #hal.device.promise<@__device_6>
%14867 = torch_c.from_builtin_tensor %14866 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12578 = torch.constant.int 1
%14868 = torch.aten.add.Tensor %14849, %14852, %int1_12578 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12579 = torch.constant.int 1
%14869 = torch.aten.add.Tensor %14868, %14855, %int1_12579 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12580 = torch.constant.int 1
%14870 = torch.aten.add.Tensor %14869, %14858, %int1_12580 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12581 = torch.constant.int 1
%14871 = torch.aten.add.Tensor %14870, %14861, %int1_12581 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12582 = torch.constant.int 1
%14872 = torch.aten.add.Tensor %14871, %14864, %int1_12582 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12583 = torch.constant.int 1
%14873 = torch.aten.add.Tensor %14872, %14672, %int1_12583 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12584 = torch.constant.int 1
%14874 = torch.aten.add.Tensor %14873, %14867, %int1_12584 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14875 = torch_c.to_builtin_tensor %14636 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12585 = arith.constant 1 : index
%dim_12586 = tensor.dim %14875, %c1_12585 : tensor<4x?x4096xf16>
%14876 = flow.tensor.transfer %14875 : tensor<4x?x4096xf16>{%dim_12586} to #hal.device.promise<@__device_7>
%14877 = torch_c.from_builtin_tensor %14876 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14878 = torch_c.to_builtin_tensor %14642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12587 = arith.constant 1 : index
%dim_12588 = tensor.dim %14878, %c1_12587 : tensor<4x?x4096xf16>
%14879 = flow.tensor.transfer %14878 : tensor<4x?x4096xf16>{%dim_12588} to #hal.device.promise<@__device_7>
%14880 = torch_c.from_builtin_tensor %14879 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14881 = torch_c.to_builtin_tensor %14648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12589 = arith.constant 1 : index
%dim_12590 = tensor.dim %14881, %c1_12589 : tensor<4x?x4096xf16>
%14882 = flow.tensor.transfer %14881 : tensor<4x?x4096xf16>{%dim_12590} to #hal.device.promise<@__device_7>
%14883 = torch_c.from_builtin_tensor %14882 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14884 = torch_c.to_builtin_tensor %14654 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12591 = arith.constant 1 : index
%dim_12592 = tensor.dim %14884, %c1_12591 : tensor<4x?x4096xf16>
%14885 = flow.tensor.transfer %14884 : tensor<4x?x4096xf16>{%dim_12592} to #hal.device.promise<@__device_7>
%14886 = torch_c.from_builtin_tensor %14885 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14887 = torch_c.to_builtin_tensor %14660 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12593 = arith.constant 1 : index
%dim_12594 = tensor.dim %14887, %c1_12593 : tensor<4x?x4096xf16>
%14888 = flow.tensor.transfer %14887 : tensor<4x?x4096xf16>{%dim_12594} to #hal.device.promise<@__device_7>
%14889 = torch_c.from_builtin_tensor %14888 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14890 = torch_c.to_builtin_tensor %14666 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12595 = arith.constant 1 : index
%dim_12596 = tensor.dim %14890, %c1_12595 : tensor<4x?x4096xf16>
%14891 = flow.tensor.transfer %14890 : tensor<4x?x4096xf16>{%dim_12596} to #hal.device.promise<@__device_7>
%14892 = torch_c.from_builtin_tensor %14891 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%14893 = torch_c.to_builtin_tensor %14672 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12597 = arith.constant 1 : index
%dim_12598 = tensor.dim %14893, %c1_12597 : tensor<4x?x4096xf16>
%14894 = flow.tensor.transfer %14893 : tensor<4x?x4096xf16>{%dim_12598} to #hal.device.promise<@__device_7>
%14895 = torch_c.from_builtin_tensor %14894 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12599 = torch.constant.int 1
%14896 = torch.aten.add.Tensor %14877, %14880, %int1_12599 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12600 = torch.constant.int 1
%14897 = torch.aten.add.Tensor %14896, %14883, %int1_12600 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12601 = torch.constant.int 1
%14898 = torch.aten.add.Tensor %14897, %14886, %int1_12601 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12602 = torch.constant.int 1
%14899 = torch.aten.add.Tensor %14898, %14889, %int1_12602 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12603 = torch.constant.int 1
%14900 = torch.aten.add.Tensor %14899, %14892, %int1_12603 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12604 = torch.constant.int 1
%14901 = torch.aten.add.Tensor %14900, %14895, %int1_12604 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12605 = torch.constant.int 1
%14902 = torch.aten.add.Tensor %14901, %14678, %int1_12605 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12606 = torch.constant.int 1
%14903 = torch.aten.add.Tensor %13562, %14706, %int1_12606 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12607 = torch.constant.int 1
%14904 = torch.aten.add.Tensor %13563, %14734, %int1_12607 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12608 = torch.constant.int 1
%14905 = torch.aten.add.Tensor %13564, %14762, %int1_12608 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12609 = torch.constant.int 1
%14906 = torch.aten.add.Tensor %13565, %14790, %int1_12609 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12610 = torch.constant.int 1
%14907 = torch.aten.add.Tensor %13566, %14818, %int1_12610 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12611 = torch.constant.int 1
%14908 = torch.aten.add.Tensor %13567, %14846, %int1_12611 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12612 = torch.constant.int 1
%14909 = torch.aten.add.Tensor %13568, %14874, %int1_12612 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12613 = torch.constant.int 1
%14910 = torch.aten.add.Tensor %13569, %14902, %int1_12613 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_12614 = torch.constant.int 6
%14911 = torch.prims.convert_element_type %14903, %int6_12614 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12615 = torch.constant.int 6
%14912 = torch.prims.convert_element_type %14904, %int6_12615 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12616 = torch.constant.int 6
%14913 = torch.prims.convert_element_type %14905, %int6_12616 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12617 = torch.constant.int 6
%14914 = torch.prims.convert_element_type %14906, %int6_12617 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12618 = torch.constant.int 6
%14915 = torch.prims.convert_element_type %14907, %int6_12618 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12619 = torch.constant.int 6
%14916 = torch.prims.convert_element_type %14908, %int6_12619 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12620 = torch.constant.int 6
%14917 = torch.prims.convert_element_type %14909, %int6_12620 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_12621 = torch.constant.int 6
%14918 = torch.prims.convert_element_type %14910, %int6_12621 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12622 = torch.constant.int 2
%14919 = torch.aten.pow.Tensor_Scalar %14911, %int2_12622 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12623 = torch.constant.int 2
%14920 = torch.aten.pow.Tensor_Scalar %14912, %int2_12623 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12624 = torch.constant.int 2
%14921 = torch.aten.pow.Tensor_Scalar %14913, %int2_12624 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12625 = torch.constant.int 2
%14922 = torch.aten.pow.Tensor_Scalar %14914, %int2_12625 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12626 = torch.constant.int 2
%14923 = torch.aten.pow.Tensor_Scalar %14915, %int2_12626 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12627 = torch.constant.int 2
%14924 = torch.aten.pow.Tensor_Scalar %14916, %int2_12627 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12628 = torch.constant.int 2
%14925 = torch.aten.pow.Tensor_Scalar %14917, %int2_12628 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_12629 = torch.constant.int 2
%14926 = torch.aten.pow.Tensor_Scalar %14918, %int2_12629 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_12630 = torch.constant.int -1
%14927 = torch.prim.ListConstruct %int-1_12630 : (!torch.int) -> !torch.list<int>
%true_12631 = torch.constant.bool true
%none_12632 = torch.constant.none
%14928 = torch.aten.mean.dim %14919, %14927, %true_12631, %none_12632 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12633 = torch.constant.int -1
%14929 = torch.prim.ListConstruct %int-1_12633 : (!torch.int) -> !torch.list<int>
%true_12634 = torch.constant.bool true
%none_12635 = torch.constant.none
%14930 = torch.aten.mean.dim %14920, %14929, %true_12634, %none_12635 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12636 = torch.constant.int -1
%14931 = torch.prim.ListConstruct %int-1_12636 : (!torch.int) -> !torch.list<int>
%true_12637 = torch.constant.bool true
%none_12638 = torch.constant.none
%14932 = torch.aten.mean.dim %14921, %14931, %true_12637, %none_12638 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12639 = torch.constant.int -1
%14933 = torch.prim.ListConstruct %int-1_12639 : (!torch.int) -> !torch.list<int>
%true_12640 = torch.constant.bool true
%none_12641 = torch.constant.none
%14934 = torch.aten.mean.dim %14922, %14933, %true_12640, %none_12641 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12642 = torch.constant.int -1
%14935 = torch.prim.ListConstruct %int-1_12642 : (!torch.int) -> !torch.list<int>
%true_12643 = torch.constant.bool true
%none_12644 = torch.constant.none
%14936 = torch.aten.mean.dim %14923, %14935, %true_12643, %none_12644 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12645 = torch.constant.int -1
%14937 = torch.prim.ListConstruct %int-1_12645 : (!torch.int) -> !torch.list<int>
%true_12646 = torch.constant.bool true
%none_12647 = torch.constant.none
%14938 = torch.aten.mean.dim %14924, %14937, %true_12646, %none_12647 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12648 = torch.constant.int -1
%14939 = torch.prim.ListConstruct %int-1_12648 : (!torch.int) -> !torch.list<int>
%true_12649 = torch.constant.bool true
%none_12650 = torch.constant.none
%14940 = torch.aten.mean.dim %14925, %14939, %true_12649, %none_12650 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_12651 = torch.constant.int -1
%14941 = torch.prim.ListConstruct %int-1_12651 : (!torch.int) -> !torch.list<int>
%true_12652 = torch.constant.bool true
%none_12653 = torch.constant.none
%14942 = torch.aten.mean.dim %14926, %14941, %true_12652, %none_12653 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12654 = torch.constant.float 9.9999997473787516E-6
%int1_12655 = torch.constant.int 1
%14943 = torch.aten.add.Scalar %14928, %float9.999990e-06_12654, %int1_12655 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12656 = torch.constant.float 9.9999997473787516E-6
%int1_12657 = torch.constant.int 1
%14944 = torch.aten.add.Scalar %14930, %float9.999990e-06_12656, %int1_12657 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12658 = torch.constant.float 9.9999997473787516E-6
%int1_12659 = torch.constant.int 1
%14945 = torch.aten.add.Scalar %14932, %float9.999990e-06_12658, %int1_12659 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12660 = torch.constant.float 9.9999997473787516E-6
%int1_12661 = torch.constant.int 1
%14946 = torch.aten.add.Scalar %14934, %float9.999990e-06_12660, %int1_12661 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12662 = torch.constant.float 9.9999997473787516E-6
%int1_12663 = torch.constant.int 1
%14947 = torch.aten.add.Scalar %14936, %float9.999990e-06_12662, %int1_12663 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12664 = torch.constant.float 9.9999997473787516E-6
%int1_12665 = torch.constant.int 1
%14948 = torch.aten.add.Scalar %14938, %float9.999990e-06_12664, %int1_12665 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12666 = torch.constant.float 9.9999997473787516E-6
%int1_12667 = torch.constant.int 1
%14949 = torch.aten.add.Scalar %14940, %float9.999990e-06_12666, %int1_12667 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_12668 = torch.constant.float 9.9999997473787516E-6
%int1_12669 = torch.constant.int 1
%14950 = torch.aten.add.Scalar %14942, %float9.999990e-06_12668, %int1_12669 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14951 = torch.aten.rsqrt %14943 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14952 = torch.aten.rsqrt %14944 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14953 = torch.aten.rsqrt %14945 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14954 = torch.aten.rsqrt %14946 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14955 = torch.aten.rsqrt %14947 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14956 = torch.aten.rsqrt %14948 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14957 = torch.aten.rsqrt %14949 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14958 = torch.aten.rsqrt %14950 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %14958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%14959 = torch.aten.mul.Tensor %14911, %14951 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14960 = torch.aten.mul.Tensor %14912, %14952 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14961 = torch.aten.mul.Tensor %14913, %14953 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14962 = torch.aten.mul.Tensor %14914, %14954 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14963 = torch.aten.mul.Tensor %14915, %14955 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14964 = torch.aten.mul.Tensor %14916, %14956 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14965 = torch.aten.mul.Tensor %14917, %14957 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14966 = torch.aten.mul.Tensor %14918, %14958 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14967 = torch.aten.mul.Tensor %480, %14959 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14968 = torch.aten.mul.Tensor %481, %14960 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14969 = torch.aten.mul.Tensor %482, %14961 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14970 = torch.aten.mul.Tensor %483, %14962 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14971 = torch.aten.mul.Tensor %484, %14963 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14972 = torch.aten.mul.Tensor %485, %14964 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14973 = torch.aten.mul.Tensor %486, %14965 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%14974 = torch.aten.mul.Tensor %487, %14966 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %14974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_12670 = torch.constant.int 5
%14975 = torch.prims.convert_element_type %14967, %int5_12670 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12671 = torch.constant.int 5
%14976 = torch.prims.convert_element_type %14968, %int5_12671 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12672 = torch.constant.int 5
%14977 = torch.prims.convert_element_type %14969, %int5_12672 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12673 = torch.constant.int 5
%14978 = torch.prims.convert_element_type %14970, %int5_12673 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12674 = torch.constant.int 5
%14979 = torch.prims.convert_element_type %14971, %int5_12674 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12675 = torch.constant.int 5
%14980 = torch.prims.convert_element_type %14972, %int5_12675 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12676 = torch.constant.int 5
%14981 = torch.prims.convert_element_type %14973, %int5_12676 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_12677 = torch.constant.int 5
%14982 = torch.prims.convert_element_type %14974, %int5_12677 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %14982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12678 = torch.constant.int 1
%int0_12679 = torch.constant.int 0
%14983 = torch.prim.ListConstruct %int1_12678, %int0_12679 : (!torch.int, !torch.int) -> !torch.list<int>
%14984 = torch.aten.permute %488, %14983 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12680 = torch.constant.int 1
%int0_12681 = torch.constant.int 0
%14985 = torch.prim.ListConstruct %int1_12680, %int0_12681 : (!torch.int, !torch.int) -> !torch.list<int>
%14986 = torch.aten.permute %489, %14985 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12682 = torch.constant.int 1
%int0_12683 = torch.constant.int 0
%14987 = torch.prim.ListConstruct %int1_12682, %int0_12683 : (!torch.int, !torch.int) -> !torch.list<int>
%14988 = torch.aten.permute %490, %14987 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12684 = torch.constant.int 1
%int0_12685 = torch.constant.int 0
%14989 = torch.prim.ListConstruct %int1_12684, %int0_12685 : (!torch.int, !torch.int) -> !torch.list<int>
%14990 = torch.aten.permute %491, %14989 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12686 = torch.constant.int 1
%int0_12687 = torch.constant.int 0
%14991 = torch.prim.ListConstruct %int1_12686, %int0_12687 : (!torch.int, !torch.int) -> !torch.list<int>
%14992 = torch.aten.permute %492, %14991 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12688 = torch.constant.int 1
%int0_12689 = torch.constant.int 0
%14993 = torch.prim.ListConstruct %int1_12688, %int0_12689 : (!torch.int, !torch.int) -> !torch.list<int>
%14994 = torch.aten.permute %493, %14993 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12690 = torch.constant.int 1
%int0_12691 = torch.constant.int 0
%14995 = torch.prim.ListConstruct %int1_12690, %int0_12691 : (!torch.int, !torch.int) -> !torch.list<int>
%14996 = torch.aten.permute %494, %14995 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12692 = torch.constant.int 1
%int0_12693 = torch.constant.int 0
%14997 = torch.prim.ListConstruct %int1_12692, %int0_12693 : (!torch.int, !torch.int) -> !torch.list<int>
%14998 = torch.aten.permute %495, %14997 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_12694 = torch.constant.int 4
%14999 = torch.aten.mul.int %int4_12694, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12695 = torch.constant.int 4096
%15000 = torch.prim.ListConstruct %14999, %int4096_12695 : (!torch.int, !torch.int) -> !torch.list<int>
%15001 = torch.aten.view %14975, %15000 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15001, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15002 = torch.aten.mm %15001, %14984 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15002, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12696 = torch.constant.int 4
%int1792_12697 = torch.constant.int 1792
%15003 = torch.prim.ListConstruct %int4_12696, %2482, %int1792_12697 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15004 = torch.aten.view %15002, %15003 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12698 = torch.constant.int 4
%15005 = torch.aten.mul.int %int4_12698, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12699 = torch.constant.int 4096
%15006 = torch.prim.ListConstruct %15005, %int4096_12699 : (!torch.int, !torch.int) -> !torch.list<int>
%15007 = torch.aten.view %14976, %15006 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15007, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15008 = torch.aten.mm %15007, %14986 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15008, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12700 = torch.constant.int 4
%int1792_12701 = torch.constant.int 1792
%15009 = torch.prim.ListConstruct %int4_12700, %2482, %int1792_12701 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15010 = torch.aten.view %15008, %15009 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12702 = torch.constant.int 4
%15011 = torch.aten.mul.int %int4_12702, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12703 = torch.constant.int 4096
%15012 = torch.prim.ListConstruct %15011, %int4096_12703 : (!torch.int, !torch.int) -> !torch.list<int>
%15013 = torch.aten.view %14977, %15012 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15013, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15014 = torch.aten.mm %15013, %14988 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15014, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12704 = torch.constant.int 4
%int1792_12705 = torch.constant.int 1792
%15015 = torch.prim.ListConstruct %int4_12704, %2482, %int1792_12705 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15016 = torch.aten.view %15014, %15015 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12706 = torch.constant.int 4
%15017 = torch.aten.mul.int %int4_12706, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12707 = torch.constant.int 4096
%15018 = torch.prim.ListConstruct %15017, %int4096_12707 : (!torch.int, !torch.int) -> !torch.list<int>
%15019 = torch.aten.view %14978, %15018 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15019, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15020 = torch.aten.mm %15019, %14990 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15020, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12708 = torch.constant.int 4
%int1792_12709 = torch.constant.int 1792
%15021 = torch.prim.ListConstruct %int4_12708, %2482, %int1792_12709 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15022 = torch.aten.view %15020, %15021 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12710 = torch.constant.int 4
%15023 = torch.aten.mul.int %int4_12710, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12711 = torch.constant.int 4096
%15024 = torch.prim.ListConstruct %15023, %int4096_12711 : (!torch.int, !torch.int) -> !torch.list<int>
%15025 = torch.aten.view %14979, %15024 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15025, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15026 = torch.aten.mm %15025, %14992 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15026, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12712 = torch.constant.int 4
%int1792_12713 = torch.constant.int 1792
%15027 = torch.prim.ListConstruct %int4_12712, %2482, %int1792_12713 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15028 = torch.aten.view %15026, %15027 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12714 = torch.constant.int 4
%15029 = torch.aten.mul.int %int4_12714, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12715 = torch.constant.int 4096
%15030 = torch.prim.ListConstruct %15029, %int4096_12715 : (!torch.int, !torch.int) -> !torch.list<int>
%15031 = torch.aten.view %14980, %15030 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15031, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15032 = torch.aten.mm %15031, %14994 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15032, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12716 = torch.constant.int 4
%int1792_12717 = torch.constant.int 1792
%15033 = torch.prim.ListConstruct %int4_12716, %2482, %int1792_12717 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15034 = torch.aten.view %15032, %15033 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12718 = torch.constant.int 4
%15035 = torch.aten.mul.int %int4_12718, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12719 = torch.constant.int 4096
%15036 = torch.prim.ListConstruct %15035, %int4096_12719 : (!torch.int, !torch.int) -> !torch.list<int>
%15037 = torch.aten.view %14981, %15036 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15037, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15038 = torch.aten.mm %15037, %14996 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15038, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12720 = torch.constant.int 4
%int1792_12721 = torch.constant.int 1792
%15039 = torch.prim.ListConstruct %int4_12720, %2482, %int1792_12721 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15040 = torch.aten.view %15038, %15039 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12722 = torch.constant.int 4
%15041 = torch.aten.mul.int %int4_12722, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12723 = torch.constant.int 4096
%15042 = torch.prim.ListConstruct %15041, %int4096_12723 : (!torch.int, !torch.int) -> !torch.list<int>
%15043 = torch.aten.view %14982, %15042 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15043, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15044 = torch.aten.mm %15043, %14998 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15044, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12724 = torch.constant.int 4
%int1792_12725 = torch.constant.int 1792
%15045 = torch.prim.ListConstruct %int4_12724, %2482, %int1792_12725 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15046 = torch.aten.view %15044, %15045 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15047 = torch.aten.silu %15004 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15048 = torch.aten.silu %15010 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15049 = torch.aten.silu %15016 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15050 = torch.aten.silu %15022 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15051 = torch.aten.silu %15028 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15052 = torch.aten.silu %15034 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15053 = torch.aten.silu %15040 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15054 = torch.aten.silu %15046 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_12726 = torch.constant.int 1
%int0_12727 = torch.constant.int 0
%15055 = torch.prim.ListConstruct %int1_12726, %int0_12727 : (!torch.int, !torch.int) -> !torch.list<int>
%15056 = torch.aten.permute %496, %15055 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12728 = torch.constant.int 1
%int0_12729 = torch.constant.int 0
%15057 = torch.prim.ListConstruct %int1_12728, %int0_12729 : (!torch.int, !torch.int) -> !torch.list<int>
%15058 = torch.aten.permute %497, %15057 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12730 = torch.constant.int 1
%int0_12731 = torch.constant.int 0
%15059 = torch.prim.ListConstruct %int1_12730, %int0_12731 : (!torch.int, !torch.int) -> !torch.list<int>
%15060 = torch.aten.permute %498, %15059 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12732 = torch.constant.int 1
%int0_12733 = torch.constant.int 0
%15061 = torch.prim.ListConstruct %int1_12732, %int0_12733 : (!torch.int, !torch.int) -> !torch.list<int>
%15062 = torch.aten.permute %499, %15061 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12734 = torch.constant.int 1
%int0_12735 = torch.constant.int 0
%15063 = torch.prim.ListConstruct %int1_12734, %int0_12735 : (!torch.int, !torch.int) -> !torch.list<int>
%15064 = torch.aten.permute %500, %15063 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12736 = torch.constant.int 1
%int0_12737 = torch.constant.int 0
%15065 = torch.prim.ListConstruct %int1_12736, %int0_12737 : (!torch.int, !torch.int) -> !torch.list<int>
%15066 = torch.aten.permute %501, %15065 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12738 = torch.constant.int 1
%int0_12739 = torch.constant.int 0
%15067 = torch.prim.ListConstruct %int1_12738, %int0_12739 : (!torch.int, !torch.int) -> !torch.list<int>
%15068 = torch.aten.permute %502, %15067 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_12740 = torch.constant.int 1
%int0_12741 = torch.constant.int 0
%15069 = torch.prim.ListConstruct %int1_12740, %int0_12741 : (!torch.int, !torch.int) -> !torch.list<int>
%15070 = torch.aten.permute %503, %15069 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_12742 = torch.constant.int 4
%15071 = torch.aten.mul.int %int4_12742, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12743 = torch.constant.int 4096
%15072 = torch.prim.ListConstruct %15071, %int4096_12743 : (!torch.int, !torch.int) -> !torch.list<int>
%15073 = torch.aten.view %14975, %15072 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15073, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15074 = torch.aten.mm %15073, %15056 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15074, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12744 = torch.constant.int 4
%int1792_12745 = torch.constant.int 1792
%15075 = torch.prim.ListConstruct %int4_12744, %2482, %int1792_12745 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15076 = torch.aten.view %15074, %15075 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12746 = torch.constant.int 4
%15077 = torch.aten.mul.int %int4_12746, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12747 = torch.constant.int 4096
%15078 = torch.prim.ListConstruct %15077, %int4096_12747 : (!torch.int, !torch.int) -> !torch.list<int>
%15079 = torch.aten.view %14976, %15078 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15079, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15080 = torch.aten.mm %15079, %15058 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15080, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12748 = torch.constant.int 4
%int1792_12749 = torch.constant.int 1792
%15081 = torch.prim.ListConstruct %int4_12748, %2482, %int1792_12749 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15082 = torch.aten.view %15080, %15081 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12750 = torch.constant.int 4
%15083 = torch.aten.mul.int %int4_12750, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12751 = torch.constant.int 4096
%15084 = torch.prim.ListConstruct %15083, %int4096_12751 : (!torch.int, !torch.int) -> !torch.list<int>
%15085 = torch.aten.view %14977, %15084 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15085, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15086 = torch.aten.mm %15085, %15060 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15086, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12752 = torch.constant.int 4
%int1792_12753 = torch.constant.int 1792
%15087 = torch.prim.ListConstruct %int4_12752, %2482, %int1792_12753 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15088 = torch.aten.view %15086, %15087 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12754 = torch.constant.int 4
%15089 = torch.aten.mul.int %int4_12754, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12755 = torch.constant.int 4096
%15090 = torch.prim.ListConstruct %15089, %int4096_12755 : (!torch.int, !torch.int) -> !torch.list<int>
%15091 = torch.aten.view %14978, %15090 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15091, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15092 = torch.aten.mm %15091, %15062 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15092, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12756 = torch.constant.int 4
%int1792_12757 = torch.constant.int 1792
%15093 = torch.prim.ListConstruct %int4_12756, %2482, %int1792_12757 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15094 = torch.aten.view %15092, %15093 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12758 = torch.constant.int 4
%15095 = torch.aten.mul.int %int4_12758, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12759 = torch.constant.int 4096
%15096 = torch.prim.ListConstruct %15095, %int4096_12759 : (!torch.int, !torch.int) -> !torch.list<int>
%15097 = torch.aten.view %14979, %15096 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15097, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15098 = torch.aten.mm %15097, %15064 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15098, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12760 = torch.constant.int 4
%int1792_12761 = torch.constant.int 1792
%15099 = torch.prim.ListConstruct %int4_12760, %2482, %int1792_12761 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15100 = torch.aten.view %15098, %15099 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12762 = torch.constant.int 4
%15101 = torch.aten.mul.int %int4_12762, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12763 = torch.constant.int 4096
%15102 = torch.prim.ListConstruct %15101, %int4096_12763 : (!torch.int, !torch.int) -> !torch.list<int>
%15103 = torch.aten.view %14980, %15102 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15103, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15104 = torch.aten.mm %15103, %15066 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15104, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12764 = torch.constant.int 4
%int1792_12765 = torch.constant.int 1792
%15105 = torch.prim.ListConstruct %int4_12764, %2482, %int1792_12765 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15106 = torch.aten.view %15104, %15105 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12766 = torch.constant.int 4
%15107 = torch.aten.mul.int %int4_12766, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12767 = torch.constant.int 4096
%15108 = torch.prim.ListConstruct %15107, %int4096_12767 : (!torch.int, !torch.int) -> !torch.list<int>
%15109 = torch.aten.view %14981, %15108 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15109, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15110 = torch.aten.mm %15109, %15068 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15110, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12768 = torch.constant.int 4
%int1792_12769 = torch.constant.int 1792
%15111 = torch.prim.ListConstruct %int4_12768, %2482, %int1792_12769 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15112 = torch.aten.view %15110, %15111 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_12770 = torch.constant.int 4
%15113 = torch.aten.mul.int %int4_12770, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_12771 = torch.constant.int 4096
%15114 = torch.prim.ListConstruct %15113, %int4096_12771 : (!torch.int, !torch.int) -> !torch.list<int>
%15115 = torch.aten.view %14982, %15114 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15115, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15116 = torch.aten.mm %15115, %15070 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15116, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_12772 = torch.constant.int 4
%int1792_12773 = torch.constant.int 1792
%15117 = torch.prim.ListConstruct %int4_12772, %2482, %int1792_12773 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15118 = torch.aten.view %15116, %15117 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15119 = torch.aten.mul.Tensor %15047, %15076 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15120 = torch.aten.mul.Tensor %15048, %15082 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15121 = torch.aten.mul.Tensor %15049, %15088 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15122 = torch.aten.mul.Tensor %15050, %15094 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15123 = torch.aten.mul.Tensor %15051, %15100 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15124 = torch.aten.mul.Tensor %15052, %15106 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15124, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15125 = torch.aten.mul.Tensor %15053, %15112 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%15126 = torch.aten.mul.Tensor %15054, %15118 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %15126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_12774 = torch.constant.int 1
%int0_12775 = torch.constant.int 0
%15127 = torch.prim.ListConstruct %int1_12774, %int0_12775 : (!torch.int, !torch.int) -> !torch.list<int>
%15128 = torch.aten.permute %504, %15127 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12776 = torch.constant.int 1
%int0_12777 = torch.constant.int 0
%15129 = torch.prim.ListConstruct %int1_12776, %int0_12777 : (!torch.int, !torch.int) -> !torch.list<int>
%15130 = torch.aten.permute %505, %15129 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12778 = torch.constant.int 1
%int0_12779 = torch.constant.int 0
%15131 = torch.prim.ListConstruct %int1_12778, %int0_12779 : (!torch.int, !torch.int) -> !torch.list<int>
%15132 = torch.aten.permute %506, %15131 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12780 = torch.constant.int 1
%int0_12781 = torch.constant.int 0
%15133 = torch.prim.ListConstruct %int1_12780, %int0_12781 : (!torch.int, !torch.int) -> !torch.list<int>
%15134 = torch.aten.permute %507, %15133 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12782 = torch.constant.int 1
%int0_12783 = torch.constant.int 0
%15135 = torch.prim.ListConstruct %int1_12782, %int0_12783 : (!torch.int, !torch.int) -> !torch.list<int>
%15136 = torch.aten.permute %508, %15135 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12784 = torch.constant.int 1
%int0_12785 = torch.constant.int 0
%15137 = torch.prim.ListConstruct %int1_12784, %int0_12785 : (!torch.int, !torch.int) -> !torch.list<int>
%15138 = torch.aten.permute %509, %15137 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12786 = torch.constant.int 1
%int0_12787 = torch.constant.int 0
%15139 = torch.prim.ListConstruct %int1_12786, %int0_12787 : (!torch.int, !torch.int) -> !torch.list<int>
%15140 = torch.aten.permute %510, %15139 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12788 = torch.constant.int 1
%int0_12789 = torch.constant.int 0
%15141 = torch.prim.ListConstruct %int1_12788, %int0_12789 : (!torch.int, !torch.int) -> !torch.list<int>
%15142 = torch.aten.permute %511, %15141 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_12790 = torch.constant.int 1
%15143 = torch.aten.size.int %15004, %int1_12790 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12791 = torch.constant.int 4
%15144 = torch.aten.mul.int %int4_12791, %15143 : !torch.int, !torch.int -> !torch.int
%int1792_12792 = torch.constant.int 1792
%15145 = torch.prim.ListConstruct %15144, %int1792_12792 : (!torch.int, !torch.int) -> !torch.list<int>
%15146 = torch.aten.view %15119, %15145 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15146, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15147 = torch.aten.mm %15146, %15128 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15147, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12793 = torch.constant.int 4
%int4096_12794 = torch.constant.int 4096
%15148 = torch.prim.ListConstruct %int4_12793, %15143, %int4096_12794 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15149 = torch.aten.view %15147, %15148 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12795 = torch.constant.int 1
%15150 = torch.aten.size.int %15010, %int1_12795 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12796 = torch.constant.int 4
%15151 = torch.aten.mul.int %int4_12796, %15150 : !torch.int, !torch.int -> !torch.int
%int1792_12797 = torch.constant.int 1792
%15152 = torch.prim.ListConstruct %15151, %int1792_12797 : (!torch.int, !torch.int) -> !torch.list<int>
%15153 = torch.aten.view %15120, %15152 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15153, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15154 = torch.aten.mm %15153, %15130 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15154, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12798 = torch.constant.int 4
%int4096_12799 = torch.constant.int 4096
%15155 = torch.prim.ListConstruct %int4_12798, %15150, %int4096_12799 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15156 = torch.aten.view %15154, %15155 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12800 = torch.constant.int 1
%15157 = torch.aten.size.int %15016, %int1_12800 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12801 = torch.constant.int 4
%15158 = torch.aten.mul.int %int4_12801, %15157 : !torch.int, !torch.int -> !torch.int
%int1792_12802 = torch.constant.int 1792
%15159 = torch.prim.ListConstruct %15158, %int1792_12802 : (!torch.int, !torch.int) -> !torch.list<int>
%15160 = torch.aten.view %15121, %15159 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15160, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15161 = torch.aten.mm %15160, %15132 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15161, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12803 = torch.constant.int 4
%int4096_12804 = torch.constant.int 4096
%15162 = torch.prim.ListConstruct %int4_12803, %15157, %int4096_12804 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15163 = torch.aten.view %15161, %15162 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12805 = torch.constant.int 1
%15164 = torch.aten.size.int %15022, %int1_12805 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12806 = torch.constant.int 4
%15165 = torch.aten.mul.int %int4_12806, %15164 : !torch.int, !torch.int -> !torch.int
%int1792_12807 = torch.constant.int 1792
%15166 = torch.prim.ListConstruct %15165, %int1792_12807 : (!torch.int, !torch.int) -> !torch.list<int>
%15167 = torch.aten.view %15122, %15166 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15167, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15168 = torch.aten.mm %15167, %15134 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15168, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12808 = torch.constant.int 4
%int4096_12809 = torch.constant.int 4096
%15169 = torch.prim.ListConstruct %int4_12808, %15164, %int4096_12809 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15170 = torch.aten.view %15168, %15169 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12810 = torch.constant.int 1
%15171 = torch.aten.size.int %15028, %int1_12810 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12811 = torch.constant.int 4
%15172 = torch.aten.mul.int %int4_12811, %15171 : !torch.int, !torch.int -> !torch.int
%int1792_12812 = torch.constant.int 1792
%15173 = torch.prim.ListConstruct %15172, %int1792_12812 : (!torch.int, !torch.int) -> !torch.list<int>
%15174 = torch.aten.view %15123, %15173 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15174, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15175 = torch.aten.mm %15174, %15136 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15175, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12813 = torch.constant.int 4
%int4096_12814 = torch.constant.int 4096
%15176 = torch.prim.ListConstruct %int4_12813, %15171, %int4096_12814 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15177 = torch.aten.view %15175, %15176 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12815 = torch.constant.int 1
%15178 = torch.aten.size.int %15034, %int1_12815 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12816 = torch.constant.int 4
%15179 = torch.aten.mul.int %int4_12816, %15178 : !torch.int, !torch.int -> !torch.int
%int1792_12817 = torch.constant.int 1792
%15180 = torch.prim.ListConstruct %15179, %int1792_12817 : (!torch.int, !torch.int) -> !torch.list<int>
%15181 = torch.aten.view %15124, %15180 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15181, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15182 = torch.aten.mm %15181, %15138 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15182, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12818 = torch.constant.int 4
%int4096_12819 = torch.constant.int 4096
%15183 = torch.prim.ListConstruct %int4_12818, %15178, %int4096_12819 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15184 = torch.aten.view %15182, %15183 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12820 = torch.constant.int 1
%15185 = torch.aten.size.int %15040, %int1_12820 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12821 = torch.constant.int 4
%15186 = torch.aten.mul.int %int4_12821, %15185 : !torch.int, !torch.int -> !torch.int
%int1792_12822 = torch.constant.int 1792
%15187 = torch.prim.ListConstruct %15186, %int1792_12822 : (!torch.int, !torch.int) -> !torch.list<int>
%15188 = torch.aten.view %15125, %15187 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15188, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15189 = torch.aten.mm %15188, %15140 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15189, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12823 = torch.constant.int 4
%int4096_12824 = torch.constant.int 4096
%15190 = torch.prim.ListConstruct %int4_12823, %15185, %int4096_12824 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15191 = torch.aten.view %15189, %15190 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12825 = torch.constant.int 1
%15192 = torch.aten.size.int %15046, %int1_12825 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_12826 = torch.constant.int 4
%15193 = torch.aten.mul.int %int4_12826, %15192 : !torch.int, !torch.int -> !torch.int
%int1792_12827 = torch.constant.int 1792
%15194 = torch.prim.ListConstruct %15193, %int1792_12827 : (!torch.int, !torch.int) -> !torch.list<int>
%15195 = torch.aten.view %15126, %15194 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %15195, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%15196 = torch.aten.mm %15195, %15142 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15196, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_12828 = torch.constant.int 4
%int4096_12829 = torch.constant.int 4096
%15197 = torch.prim.ListConstruct %int4_12828, %15192, %int4096_12829 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15198 = torch.aten.view %15196, %15197 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15199 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12830 = arith.constant 1 : index
%dim_12831 = tensor.dim %15199, %c1_12830 : tensor<4x?x4096xf16>
%15200 = flow.tensor.transfer %15199 : tensor<4x?x4096xf16>{%dim_12831} to #hal.device.promise<@__device_0>
%15201 = torch_c.from_builtin_tensor %15200 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15202 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12832 = arith.constant 1 : index
%dim_12833 = tensor.dim %15202, %c1_12832 : tensor<4x?x4096xf16>
%15203 = flow.tensor.transfer %15202 : tensor<4x?x4096xf16>{%dim_12833} to #hal.device.promise<@__device_0>
%15204 = torch_c.from_builtin_tensor %15203 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15205 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12834 = arith.constant 1 : index
%dim_12835 = tensor.dim %15205, %c1_12834 : tensor<4x?x4096xf16>
%15206 = flow.tensor.transfer %15205 : tensor<4x?x4096xf16>{%dim_12835} to #hal.device.promise<@__device_0>
%15207 = torch_c.from_builtin_tensor %15206 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15208 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12836 = arith.constant 1 : index
%dim_12837 = tensor.dim %15208, %c1_12836 : tensor<4x?x4096xf16>
%15209 = flow.tensor.transfer %15208 : tensor<4x?x4096xf16>{%dim_12837} to #hal.device.promise<@__device_0>
%15210 = torch_c.from_builtin_tensor %15209 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15211 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12838 = arith.constant 1 : index
%dim_12839 = tensor.dim %15211, %c1_12838 : tensor<4x?x4096xf16>
%15212 = flow.tensor.transfer %15211 : tensor<4x?x4096xf16>{%dim_12839} to #hal.device.promise<@__device_0>
%15213 = torch_c.from_builtin_tensor %15212 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15214 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12840 = arith.constant 1 : index
%dim_12841 = tensor.dim %15214, %c1_12840 : tensor<4x?x4096xf16>
%15215 = flow.tensor.transfer %15214 : tensor<4x?x4096xf16>{%dim_12841} to #hal.device.promise<@__device_0>
%15216 = torch_c.from_builtin_tensor %15215 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15217 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12842 = arith.constant 1 : index
%dim_12843 = tensor.dim %15217, %c1_12842 : tensor<4x?x4096xf16>
%15218 = flow.tensor.transfer %15217 : tensor<4x?x4096xf16>{%dim_12843} to #hal.device.promise<@__device_0>
%15219 = torch_c.from_builtin_tensor %15218 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12844 = torch.constant.int 1
%15220 = torch.aten.add.Tensor %15149, %15201, %int1_12844 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12845 = torch.constant.int 1
%15221 = torch.aten.add.Tensor %15220, %15204, %int1_12845 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12846 = torch.constant.int 1
%15222 = torch.aten.add.Tensor %15221, %15207, %int1_12846 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12847 = torch.constant.int 1
%15223 = torch.aten.add.Tensor %15222, %15210, %int1_12847 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12848 = torch.constant.int 1
%15224 = torch.aten.add.Tensor %15223, %15213, %int1_12848 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12849 = torch.constant.int 1
%15225 = torch.aten.add.Tensor %15224, %15216, %int1_12849 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12850 = torch.constant.int 1
%15226 = torch.aten.add.Tensor %15225, %15219, %int1_12850 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15227 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12851 = arith.constant 1 : index
%dim_12852 = tensor.dim %15227, %c1_12851 : tensor<4x?x4096xf16>
%15228 = flow.tensor.transfer %15227 : tensor<4x?x4096xf16>{%dim_12852} to #hal.device.promise<@__device_1>
%15229 = torch_c.from_builtin_tensor %15228 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15230 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12853 = arith.constant 1 : index
%dim_12854 = tensor.dim %15230, %c1_12853 : tensor<4x?x4096xf16>
%15231 = flow.tensor.transfer %15230 : tensor<4x?x4096xf16>{%dim_12854} to #hal.device.promise<@__device_1>
%15232 = torch_c.from_builtin_tensor %15231 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15233 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12855 = arith.constant 1 : index
%dim_12856 = tensor.dim %15233, %c1_12855 : tensor<4x?x4096xf16>
%15234 = flow.tensor.transfer %15233 : tensor<4x?x4096xf16>{%dim_12856} to #hal.device.promise<@__device_1>
%15235 = torch_c.from_builtin_tensor %15234 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15236 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12857 = arith.constant 1 : index
%dim_12858 = tensor.dim %15236, %c1_12857 : tensor<4x?x4096xf16>
%15237 = flow.tensor.transfer %15236 : tensor<4x?x4096xf16>{%dim_12858} to #hal.device.promise<@__device_1>
%15238 = torch_c.from_builtin_tensor %15237 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15239 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12859 = arith.constant 1 : index
%dim_12860 = tensor.dim %15239, %c1_12859 : tensor<4x?x4096xf16>
%15240 = flow.tensor.transfer %15239 : tensor<4x?x4096xf16>{%dim_12860} to #hal.device.promise<@__device_1>
%15241 = torch_c.from_builtin_tensor %15240 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15242 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12861 = arith.constant 1 : index
%dim_12862 = tensor.dim %15242, %c1_12861 : tensor<4x?x4096xf16>
%15243 = flow.tensor.transfer %15242 : tensor<4x?x4096xf16>{%dim_12862} to #hal.device.promise<@__device_1>
%15244 = torch_c.from_builtin_tensor %15243 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15245 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12863 = arith.constant 1 : index
%dim_12864 = tensor.dim %15245, %c1_12863 : tensor<4x?x4096xf16>
%15246 = flow.tensor.transfer %15245 : tensor<4x?x4096xf16>{%dim_12864} to #hal.device.promise<@__device_1>
%15247 = torch_c.from_builtin_tensor %15246 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12865 = torch.constant.int 1
%15248 = torch.aten.add.Tensor %15229, %15156, %int1_12865 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12866 = torch.constant.int 1
%15249 = torch.aten.add.Tensor %15248, %15232, %int1_12866 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12867 = torch.constant.int 1
%15250 = torch.aten.add.Tensor %15249, %15235, %int1_12867 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12868 = torch.constant.int 1
%15251 = torch.aten.add.Tensor %15250, %15238, %int1_12868 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12869 = torch.constant.int 1
%15252 = torch.aten.add.Tensor %15251, %15241, %int1_12869 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12870 = torch.constant.int 1
%15253 = torch.aten.add.Tensor %15252, %15244, %int1_12870 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12871 = torch.constant.int 1
%15254 = torch.aten.add.Tensor %15253, %15247, %int1_12871 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15255 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12872 = arith.constant 1 : index
%dim_12873 = tensor.dim %15255, %c1_12872 : tensor<4x?x4096xf16>
%15256 = flow.tensor.transfer %15255 : tensor<4x?x4096xf16>{%dim_12873} to #hal.device.promise<@__device_2>
%15257 = torch_c.from_builtin_tensor %15256 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15258 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12874 = arith.constant 1 : index
%dim_12875 = tensor.dim %15258, %c1_12874 : tensor<4x?x4096xf16>
%15259 = flow.tensor.transfer %15258 : tensor<4x?x4096xf16>{%dim_12875} to #hal.device.promise<@__device_2>
%15260 = torch_c.from_builtin_tensor %15259 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15261 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12876 = arith.constant 1 : index
%dim_12877 = tensor.dim %15261, %c1_12876 : tensor<4x?x4096xf16>
%15262 = flow.tensor.transfer %15261 : tensor<4x?x4096xf16>{%dim_12877} to #hal.device.promise<@__device_2>
%15263 = torch_c.from_builtin_tensor %15262 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15263, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15264 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12878 = arith.constant 1 : index
%dim_12879 = tensor.dim %15264, %c1_12878 : tensor<4x?x4096xf16>
%15265 = flow.tensor.transfer %15264 : tensor<4x?x4096xf16>{%dim_12879} to #hal.device.promise<@__device_2>
%15266 = torch_c.from_builtin_tensor %15265 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15267 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12880 = arith.constant 1 : index
%dim_12881 = tensor.dim %15267, %c1_12880 : tensor<4x?x4096xf16>
%15268 = flow.tensor.transfer %15267 : tensor<4x?x4096xf16>{%dim_12881} to #hal.device.promise<@__device_2>
%15269 = torch_c.from_builtin_tensor %15268 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15270 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12882 = arith.constant 1 : index
%dim_12883 = tensor.dim %15270, %c1_12882 : tensor<4x?x4096xf16>
%15271 = flow.tensor.transfer %15270 : tensor<4x?x4096xf16>{%dim_12883} to #hal.device.promise<@__device_2>
%15272 = torch_c.from_builtin_tensor %15271 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15273 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12884 = arith.constant 1 : index
%dim_12885 = tensor.dim %15273, %c1_12884 : tensor<4x?x4096xf16>
%15274 = flow.tensor.transfer %15273 : tensor<4x?x4096xf16>{%dim_12885} to #hal.device.promise<@__device_2>
%15275 = torch_c.from_builtin_tensor %15274 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12886 = torch.constant.int 1
%15276 = torch.aten.add.Tensor %15257, %15260, %int1_12886 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12887 = torch.constant.int 1
%15277 = torch.aten.add.Tensor %15276, %15163, %int1_12887 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12888 = torch.constant.int 1
%15278 = torch.aten.add.Tensor %15277, %15263, %int1_12888 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12889 = torch.constant.int 1
%15279 = torch.aten.add.Tensor %15278, %15266, %int1_12889 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12890 = torch.constant.int 1
%15280 = torch.aten.add.Tensor %15279, %15269, %int1_12890 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12891 = torch.constant.int 1
%15281 = torch.aten.add.Tensor %15280, %15272, %int1_12891 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12892 = torch.constant.int 1
%15282 = torch.aten.add.Tensor %15281, %15275, %int1_12892 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15283 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12893 = arith.constant 1 : index
%dim_12894 = tensor.dim %15283, %c1_12893 : tensor<4x?x4096xf16>
%15284 = flow.tensor.transfer %15283 : tensor<4x?x4096xf16>{%dim_12894} to #hal.device.promise<@__device_3>
%15285 = torch_c.from_builtin_tensor %15284 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15286 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12895 = arith.constant 1 : index
%dim_12896 = tensor.dim %15286, %c1_12895 : tensor<4x?x4096xf16>
%15287 = flow.tensor.transfer %15286 : tensor<4x?x4096xf16>{%dim_12896} to #hal.device.promise<@__device_3>
%15288 = torch_c.from_builtin_tensor %15287 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15289 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12897 = arith.constant 1 : index
%dim_12898 = tensor.dim %15289, %c1_12897 : tensor<4x?x4096xf16>
%15290 = flow.tensor.transfer %15289 : tensor<4x?x4096xf16>{%dim_12898} to #hal.device.promise<@__device_3>
%15291 = torch_c.from_builtin_tensor %15290 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15292 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12899 = arith.constant 1 : index
%dim_12900 = tensor.dim %15292, %c1_12899 : tensor<4x?x4096xf16>
%15293 = flow.tensor.transfer %15292 : tensor<4x?x4096xf16>{%dim_12900} to #hal.device.promise<@__device_3>
%15294 = torch_c.from_builtin_tensor %15293 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15295 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12901 = arith.constant 1 : index
%dim_12902 = tensor.dim %15295, %c1_12901 : tensor<4x?x4096xf16>
%15296 = flow.tensor.transfer %15295 : tensor<4x?x4096xf16>{%dim_12902} to #hal.device.promise<@__device_3>
%15297 = torch_c.from_builtin_tensor %15296 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15298 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12903 = arith.constant 1 : index
%dim_12904 = tensor.dim %15298, %c1_12903 : tensor<4x?x4096xf16>
%15299 = flow.tensor.transfer %15298 : tensor<4x?x4096xf16>{%dim_12904} to #hal.device.promise<@__device_3>
%15300 = torch_c.from_builtin_tensor %15299 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15301 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12905 = arith.constant 1 : index
%dim_12906 = tensor.dim %15301, %c1_12905 : tensor<4x?x4096xf16>
%15302 = flow.tensor.transfer %15301 : tensor<4x?x4096xf16>{%dim_12906} to #hal.device.promise<@__device_3>
%15303 = torch_c.from_builtin_tensor %15302 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12907 = torch.constant.int 1
%15304 = torch.aten.add.Tensor %15285, %15288, %int1_12907 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12908 = torch.constant.int 1
%15305 = torch.aten.add.Tensor %15304, %15291, %int1_12908 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12909 = torch.constant.int 1
%15306 = torch.aten.add.Tensor %15305, %15170, %int1_12909 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12910 = torch.constant.int 1
%15307 = torch.aten.add.Tensor %15306, %15294, %int1_12910 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12911 = torch.constant.int 1
%15308 = torch.aten.add.Tensor %15307, %15297, %int1_12911 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12912 = torch.constant.int 1
%15309 = torch.aten.add.Tensor %15308, %15300, %int1_12912 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12913 = torch.constant.int 1
%15310 = torch.aten.add.Tensor %15309, %15303, %int1_12913 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15311 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12914 = arith.constant 1 : index
%dim_12915 = tensor.dim %15311, %c1_12914 : tensor<4x?x4096xf16>
%15312 = flow.tensor.transfer %15311 : tensor<4x?x4096xf16>{%dim_12915} to #hal.device.promise<@__device_4>
%15313 = torch_c.from_builtin_tensor %15312 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15314 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12916 = arith.constant 1 : index
%dim_12917 = tensor.dim %15314, %c1_12916 : tensor<4x?x4096xf16>
%15315 = flow.tensor.transfer %15314 : tensor<4x?x4096xf16>{%dim_12917} to #hal.device.promise<@__device_4>
%15316 = torch_c.from_builtin_tensor %15315 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15317 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12918 = arith.constant 1 : index
%dim_12919 = tensor.dim %15317, %c1_12918 : tensor<4x?x4096xf16>
%15318 = flow.tensor.transfer %15317 : tensor<4x?x4096xf16>{%dim_12919} to #hal.device.promise<@__device_4>
%15319 = torch_c.from_builtin_tensor %15318 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15320 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12920 = arith.constant 1 : index
%dim_12921 = tensor.dim %15320, %c1_12920 : tensor<4x?x4096xf16>
%15321 = flow.tensor.transfer %15320 : tensor<4x?x4096xf16>{%dim_12921} to #hal.device.promise<@__device_4>
%15322 = torch_c.from_builtin_tensor %15321 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15323 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12922 = arith.constant 1 : index
%dim_12923 = tensor.dim %15323, %c1_12922 : tensor<4x?x4096xf16>
%15324 = flow.tensor.transfer %15323 : tensor<4x?x4096xf16>{%dim_12923} to #hal.device.promise<@__device_4>
%15325 = torch_c.from_builtin_tensor %15324 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15326 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12924 = arith.constant 1 : index
%dim_12925 = tensor.dim %15326, %c1_12924 : tensor<4x?x4096xf16>
%15327 = flow.tensor.transfer %15326 : tensor<4x?x4096xf16>{%dim_12925} to #hal.device.promise<@__device_4>
%15328 = torch_c.from_builtin_tensor %15327 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15329 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12926 = arith.constant 1 : index
%dim_12927 = tensor.dim %15329, %c1_12926 : tensor<4x?x4096xf16>
%15330 = flow.tensor.transfer %15329 : tensor<4x?x4096xf16>{%dim_12927} to #hal.device.promise<@__device_4>
%15331 = torch_c.from_builtin_tensor %15330 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12928 = torch.constant.int 1
%15332 = torch.aten.add.Tensor %15313, %15316, %int1_12928 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12929 = torch.constant.int 1
%15333 = torch.aten.add.Tensor %15332, %15319, %int1_12929 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12930 = torch.constant.int 1
%15334 = torch.aten.add.Tensor %15333, %15322, %int1_12930 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12931 = torch.constant.int 1
%15335 = torch.aten.add.Tensor %15334, %15177, %int1_12931 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12932 = torch.constant.int 1
%15336 = torch.aten.add.Tensor %15335, %15325, %int1_12932 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12933 = torch.constant.int 1
%15337 = torch.aten.add.Tensor %15336, %15328, %int1_12933 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12934 = torch.constant.int 1
%15338 = torch.aten.add.Tensor %15337, %15331, %int1_12934 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15339 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12935 = arith.constant 1 : index
%dim_12936 = tensor.dim %15339, %c1_12935 : tensor<4x?x4096xf16>
%15340 = flow.tensor.transfer %15339 : tensor<4x?x4096xf16>{%dim_12936} to #hal.device.promise<@__device_5>
%15341 = torch_c.from_builtin_tensor %15340 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15342 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12937 = arith.constant 1 : index
%dim_12938 = tensor.dim %15342, %c1_12937 : tensor<4x?x4096xf16>
%15343 = flow.tensor.transfer %15342 : tensor<4x?x4096xf16>{%dim_12938} to #hal.device.promise<@__device_5>
%15344 = torch_c.from_builtin_tensor %15343 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15345 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12939 = arith.constant 1 : index
%dim_12940 = tensor.dim %15345, %c1_12939 : tensor<4x?x4096xf16>
%15346 = flow.tensor.transfer %15345 : tensor<4x?x4096xf16>{%dim_12940} to #hal.device.promise<@__device_5>
%15347 = torch_c.from_builtin_tensor %15346 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15348 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12941 = arith.constant 1 : index
%dim_12942 = tensor.dim %15348, %c1_12941 : tensor<4x?x4096xf16>
%15349 = flow.tensor.transfer %15348 : tensor<4x?x4096xf16>{%dim_12942} to #hal.device.promise<@__device_5>
%15350 = torch_c.from_builtin_tensor %15349 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15351 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12943 = arith.constant 1 : index
%dim_12944 = tensor.dim %15351, %c1_12943 : tensor<4x?x4096xf16>
%15352 = flow.tensor.transfer %15351 : tensor<4x?x4096xf16>{%dim_12944} to #hal.device.promise<@__device_5>
%15353 = torch_c.from_builtin_tensor %15352 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15354 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12945 = arith.constant 1 : index
%dim_12946 = tensor.dim %15354, %c1_12945 : tensor<4x?x4096xf16>
%15355 = flow.tensor.transfer %15354 : tensor<4x?x4096xf16>{%dim_12946} to #hal.device.promise<@__device_5>
%15356 = torch_c.from_builtin_tensor %15355 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15357 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12947 = arith.constant 1 : index
%dim_12948 = tensor.dim %15357, %c1_12947 : tensor<4x?x4096xf16>
%15358 = flow.tensor.transfer %15357 : tensor<4x?x4096xf16>{%dim_12948} to #hal.device.promise<@__device_5>
%15359 = torch_c.from_builtin_tensor %15358 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12949 = torch.constant.int 1
%15360 = torch.aten.add.Tensor %15341, %15344, %int1_12949 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12950 = torch.constant.int 1
%15361 = torch.aten.add.Tensor %15360, %15347, %int1_12950 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12951 = torch.constant.int 1
%15362 = torch.aten.add.Tensor %15361, %15350, %int1_12951 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12952 = torch.constant.int 1
%15363 = torch.aten.add.Tensor %15362, %15353, %int1_12952 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12953 = torch.constant.int 1
%15364 = torch.aten.add.Tensor %15363, %15184, %int1_12953 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12954 = torch.constant.int 1
%15365 = torch.aten.add.Tensor %15364, %15356, %int1_12954 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12955 = torch.constant.int 1
%15366 = torch.aten.add.Tensor %15365, %15359, %int1_12955 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15367 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12956 = arith.constant 1 : index
%dim_12957 = tensor.dim %15367, %c1_12956 : tensor<4x?x4096xf16>
%15368 = flow.tensor.transfer %15367 : tensor<4x?x4096xf16>{%dim_12957} to #hal.device.promise<@__device_6>
%15369 = torch_c.from_builtin_tensor %15368 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15370 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12958 = arith.constant 1 : index
%dim_12959 = tensor.dim %15370, %c1_12958 : tensor<4x?x4096xf16>
%15371 = flow.tensor.transfer %15370 : tensor<4x?x4096xf16>{%dim_12959} to #hal.device.promise<@__device_6>
%15372 = torch_c.from_builtin_tensor %15371 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15373 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12960 = arith.constant 1 : index
%dim_12961 = tensor.dim %15373, %c1_12960 : tensor<4x?x4096xf16>
%15374 = flow.tensor.transfer %15373 : tensor<4x?x4096xf16>{%dim_12961} to #hal.device.promise<@__device_6>
%15375 = torch_c.from_builtin_tensor %15374 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15376 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12962 = arith.constant 1 : index
%dim_12963 = tensor.dim %15376, %c1_12962 : tensor<4x?x4096xf16>
%15377 = flow.tensor.transfer %15376 : tensor<4x?x4096xf16>{%dim_12963} to #hal.device.promise<@__device_6>
%15378 = torch_c.from_builtin_tensor %15377 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15379 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12964 = arith.constant 1 : index
%dim_12965 = tensor.dim %15379, %c1_12964 : tensor<4x?x4096xf16>
%15380 = flow.tensor.transfer %15379 : tensor<4x?x4096xf16>{%dim_12965} to #hal.device.promise<@__device_6>
%15381 = torch_c.from_builtin_tensor %15380 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15382 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12966 = arith.constant 1 : index
%dim_12967 = tensor.dim %15382, %c1_12966 : tensor<4x?x4096xf16>
%15383 = flow.tensor.transfer %15382 : tensor<4x?x4096xf16>{%dim_12967} to #hal.device.promise<@__device_6>
%15384 = torch_c.from_builtin_tensor %15383 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15385 = torch_c.to_builtin_tensor %15198 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12968 = arith.constant 1 : index
%dim_12969 = tensor.dim %15385, %c1_12968 : tensor<4x?x4096xf16>
%15386 = flow.tensor.transfer %15385 : tensor<4x?x4096xf16>{%dim_12969} to #hal.device.promise<@__device_6>
%15387 = torch_c.from_builtin_tensor %15386 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12970 = torch.constant.int 1
%15388 = torch.aten.add.Tensor %15369, %15372, %int1_12970 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12971 = torch.constant.int 1
%15389 = torch.aten.add.Tensor %15388, %15375, %int1_12971 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12972 = torch.constant.int 1
%15390 = torch.aten.add.Tensor %15389, %15378, %int1_12972 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12973 = torch.constant.int 1
%15391 = torch.aten.add.Tensor %15390, %15381, %int1_12973 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12974 = torch.constant.int 1
%15392 = torch.aten.add.Tensor %15391, %15384, %int1_12974 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12975 = torch.constant.int 1
%15393 = torch.aten.add.Tensor %15392, %15191, %int1_12975 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12976 = torch.constant.int 1
%15394 = torch.aten.add.Tensor %15393, %15387, %int1_12976 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15395 = torch_c.to_builtin_tensor %15149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12977 = arith.constant 1 : index
%dim_12978 = tensor.dim %15395, %c1_12977 : tensor<4x?x4096xf16>
%15396 = flow.tensor.transfer %15395 : tensor<4x?x4096xf16>{%dim_12978} to #hal.device.promise<@__device_7>
%15397 = torch_c.from_builtin_tensor %15396 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15398 = torch_c.to_builtin_tensor %15156 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12979 = arith.constant 1 : index
%dim_12980 = tensor.dim %15398, %c1_12979 : tensor<4x?x4096xf16>
%15399 = flow.tensor.transfer %15398 : tensor<4x?x4096xf16>{%dim_12980} to #hal.device.promise<@__device_7>
%15400 = torch_c.from_builtin_tensor %15399 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15401 = torch_c.to_builtin_tensor %15163 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12981 = arith.constant 1 : index
%dim_12982 = tensor.dim %15401, %c1_12981 : tensor<4x?x4096xf16>
%15402 = flow.tensor.transfer %15401 : tensor<4x?x4096xf16>{%dim_12982} to #hal.device.promise<@__device_7>
%15403 = torch_c.from_builtin_tensor %15402 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15404 = torch_c.to_builtin_tensor %15170 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12983 = arith.constant 1 : index
%dim_12984 = tensor.dim %15404, %c1_12983 : tensor<4x?x4096xf16>
%15405 = flow.tensor.transfer %15404 : tensor<4x?x4096xf16>{%dim_12984} to #hal.device.promise<@__device_7>
%15406 = torch_c.from_builtin_tensor %15405 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15407 = torch_c.to_builtin_tensor %15177 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12985 = arith.constant 1 : index
%dim_12986 = tensor.dim %15407, %c1_12985 : tensor<4x?x4096xf16>
%15408 = flow.tensor.transfer %15407 : tensor<4x?x4096xf16>{%dim_12986} to #hal.device.promise<@__device_7>
%15409 = torch_c.from_builtin_tensor %15408 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15410 = torch_c.to_builtin_tensor %15184 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12987 = arith.constant 1 : index
%dim_12988 = tensor.dim %15410, %c1_12987 : tensor<4x?x4096xf16>
%15411 = flow.tensor.transfer %15410 : tensor<4x?x4096xf16>{%dim_12988} to #hal.device.promise<@__device_7>
%15412 = torch_c.from_builtin_tensor %15411 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%15413 = torch_c.to_builtin_tensor %15191 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_12989 = arith.constant 1 : index
%dim_12990 = tensor.dim %15413, %c1_12989 : tensor<4x?x4096xf16>
%15414 = flow.tensor.transfer %15413 : tensor<4x?x4096xf16>{%dim_12990} to #hal.device.promise<@__device_7>
%15415 = torch_c.from_builtin_tensor %15414 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12991 = torch.constant.int 1
%15416 = torch.aten.add.Tensor %15397, %15400, %int1_12991 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12992 = torch.constant.int 1
%15417 = torch.aten.add.Tensor %15416, %15403, %int1_12992 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12993 = torch.constant.int 1
%15418 = torch.aten.add.Tensor %15417, %15406, %int1_12993 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12994 = torch.constant.int 1
%15419 = torch.aten.add.Tensor %15418, %15409, %int1_12994 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12995 = torch.constant.int 1
%15420 = torch.aten.add.Tensor %15419, %15412, %int1_12995 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12996 = torch.constant.int 1
%15421 = torch.aten.add.Tensor %15420, %15415, %int1_12996 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12997 = torch.constant.int 1
%15422 = torch.aten.add.Tensor %15421, %15198, %int1_12997 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12998 = torch.constant.int 1
%15423 = torch.aten.add.Tensor %14903, %15226, %int1_12998 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_12999 = torch.constant.int 1
%15424 = torch.aten.add.Tensor %14904, %15254, %int1_12999 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13000 = torch.constant.int 1
%15425 = torch.aten.add.Tensor %14905, %15282, %int1_13000 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13001 = torch.constant.int 1
%15426 = torch.aten.add.Tensor %14906, %15310, %int1_13001 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13002 = torch.constant.int 1
%15427 = torch.aten.add.Tensor %14907, %15338, %int1_13002 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13003 = torch.constant.int 1
%15428 = torch.aten.add.Tensor %14908, %15366, %int1_13003 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13004 = torch.constant.int 1
%15429 = torch.aten.add.Tensor %14909, %15394, %int1_13004 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13005 = torch.constant.int 1
%15430 = torch.aten.add.Tensor %14910, %15422, %int1_13005 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_13006 = torch.constant.int 6
%15431 = torch.prims.convert_element_type %15423, %int6_13006 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13007 = torch.constant.int 6
%15432 = torch.prims.convert_element_type %15424, %int6_13007 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13008 = torch.constant.int 6
%15433 = torch.prims.convert_element_type %15425, %int6_13008 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13009 = torch.constant.int 6
%15434 = torch.prims.convert_element_type %15426, %int6_13009 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13010 = torch.constant.int 6
%15435 = torch.prims.convert_element_type %15427, %int6_13010 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13011 = torch.constant.int 6
%15436 = torch.prims.convert_element_type %15428, %int6_13011 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13012 = torch.constant.int 6
%15437 = torch.prims.convert_element_type %15429, %int6_13012 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_13013 = torch.constant.int 6
%15438 = torch.prims.convert_element_type %15430, %int6_13013 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13014 = torch.constant.int 2
%15439 = torch.aten.pow.Tensor_Scalar %15431, %int2_13014 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13015 = torch.constant.int 2
%15440 = torch.aten.pow.Tensor_Scalar %15432, %int2_13015 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13016 = torch.constant.int 2
%15441 = torch.aten.pow.Tensor_Scalar %15433, %int2_13016 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13017 = torch.constant.int 2
%15442 = torch.aten.pow.Tensor_Scalar %15434, %int2_13017 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13018 = torch.constant.int 2
%15443 = torch.aten.pow.Tensor_Scalar %15435, %int2_13018 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13019 = torch.constant.int 2
%15444 = torch.aten.pow.Tensor_Scalar %15436, %int2_13019 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13020 = torch.constant.int 2
%15445 = torch.aten.pow.Tensor_Scalar %15437, %int2_13020 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_13021 = torch.constant.int 2
%15446 = torch.aten.pow.Tensor_Scalar %15438, %int2_13021 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_13022 = torch.constant.int -1
%15447 = torch.prim.ListConstruct %int-1_13022 : (!torch.int) -> !torch.list<int>
%true_13023 = torch.constant.bool true
%none_13024 = torch.constant.none
%15448 = torch.aten.mean.dim %15439, %15447, %true_13023, %none_13024 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13025 = torch.constant.int -1
%15449 = torch.prim.ListConstruct %int-1_13025 : (!torch.int) -> !torch.list<int>
%true_13026 = torch.constant.bool true
%none_13027 = torch.constant.none
%15450 = torch.aten.mean.dim %15440, %15449, %true_13026, %none_13027 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13028 = torch.constant.int -1
%15451 = torch.prim.ListConstruct %int-1_13028 : (!torch.int) -> !torch.list<int>
%true_13029 = torch.constant.bool true
%none_13030 = torch.constant.none
%15452 = torch.aten.mean.dim %15441, %15451, %true_13029, %none_13030 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13031 = torch.constant.int -1
%15453 = torch.prim.ListConstruct %int-1_13031 : (!torch.int) -> !torch.list<int>
%true_13032 = torch.constant.bool true
%none_13033 = torch.constant.none
%15454 = torch.aten.mean.dim %15442, %15453, %true_13032, %none_13033 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13034 = torch.constant.int -1
%15455 = torch.prim.ListConstruct %int-1_13034 : (!torch.int) -> !torch.list<int>
%true_13035 = torch.constant.bool true
%none_13036 = torch.constant.none
%15456 = torch.aten.mean.dim %15443, %15455, %true_13035, %none_13036 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13037 = torch.constant.int -1
%15457 = torch.prim.ListConstruct %int-1_13037 : (!torch.int) -> !torch.list<int>
%true_13038 = torch.constant.bool true
%none_13039 = torch.constant.none
%15458 = torch.aten.mean.dim %15444, %15457, %true_13038, %none_13039 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13040 = torch.constant.int -1
%15459 = torch.prim.ListConstruct %int-1_13040 : (!torch.int) -> !torch.list<int>
%true_13041 = torch.constant.bool true
%none_13042 = torch.constant.none
%15460 = torch.aten.mean.dim %15445, %15459, %true_13041, %none_13042 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_13043 = torch.constant.int -1
%15461 = torch.prim.ListConstruct %int-1_13043 : (!torch.int) -> !torch.list<int>
%true_13044 = torch.constant.bool true
%none_13045 = torch.constant.none
%15462 = torch.aten.mean.dim %15446, %15461, %true_13044, %none_13045 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13046 = torch.constant.float 9.9999997473787516E-6
%int1_13047 = torch.constant.int 1
%15463 = torch.aten.add.Scalar %15448, %float9.999990e-06_13046, %int1_13047 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13048 = torch.constant.float 9.9999997473787516E-6
%int1_13049 = torch.constant.int 1
%15464 = torch.aten.add.Scalar %15450, %float9.999990e-06_13048, %int1_13049 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13050 = torch.constant.float 9.9999997473787516E-6
%int1_13051 = torch.constant.int 1
%15465 = torch.aten.add.Scalar %15452, %float9.999990e-06_13050, %int1_13051 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13052 = torch.constant.float 9.9999997473787516E-6
%int1_13053 = torch.constant.int 1
%15466 = torch.aten.add.Scalar %15454, %float9.999990e-06_13052, %int1_13053 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13054 = torch.constant.float 9.9999997473787516E-6
%int1_13055 = torch.constant.int 1
%15467 = torch.aten.add.Scalar %15456, %float9.999990e-06_13054, %int1_13055 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13056 = torch.constant.float 9.9999997473787516E-6
%int1_13057 = torch.constant.int 1
%15468 = torch.aten.add.Scalar %15458, %float9.999990e-06_13056, %int1_13057 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13058 = torch.constant.float 9.9999997473787516E-6
%int1_13059 = torch.constant.int 1
%15469 = torch.aten.add.Scalar %15460, %float9.999990e-06_13058, %int1_13059 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_13060 = torch.constant.float 9.9999997473787516E-6
%int1_13061 = torch.constant.int 1
%15470 = torch.aten.add.Scalar %15462, %float9.999990e-06_13060, %int1_13061 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15471 = torch.aten.rsqrt %15463 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15472 = torch.aten.rsqrt %15464 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15473 = torch.aten.rsqrt %15465 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15474 = torch.aten.rsqrt %15466 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15475 = torch.aten.rsqrt %15467 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15476 = torch.aten.rsqrt %15468 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15477 = torch.aten.rsqrt %15469 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15478 = torch.aten.rsqrt %15470 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %15478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%15479 = torch.aten.mul.Tensor %15431, %15471 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15480 = torch.aten.mul.Tensor %15432, %15472 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15481 = torch.aten.mul.Tensor %15433, %15473 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15482 = torch.aten.mul.Tensor %15434, %15474 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15483 = torch.aten.mul.Tensor %15435, %15475 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15484 = torch.aten.mul.Tensor %15436, %15476 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15485 = torch.aten.mul.Tensor %15437, %15477 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15486 = torch.aten.mul.Tensor %15438, %15478 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15487 = torch.aten.mul.Tensor %512, %15479 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15488 = torch.aten.mul.Tensor %513, %15480 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15489 = torch.aten.mul.Tensor %514, %15481 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15490 = torch.aten.mul.Tensor %515, %15482 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15491 = torch.aten.mul.Tensor %516, %15483 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15492 = torch.aten.mul.Tensor %517, %15484 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15493 = torch.aten.mul.Tensor %518, %15485 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%15494 = torch.aten.mul.Tensor %519, %15486 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %15494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_13062 = torch.constant.int 5
%15495 = torch.prims.convert_element_type %15487, %int5_13062 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13063 = torch.constant.int 5
%15496 = torch.prims.convert_element_type %15488, %int5_13063 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13064 = torch.constant.int 5
%15497 = torch.prims.convert_element_type %15489, %int5_13064 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13065 = torch.constant.int 5
%15498 = torch.prims.convert_element_type %15490, %int5_13065 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13066 = torch.constant.int 5
%15499 = torch.prims.convert_element_type %15491, %int5_13066 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13067 = torch.constant.int 5
%15500 = torch.prims.convert_element_type %15492, %int5_13067 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13068 = torch.constant.int 5
%15501 = torch.prims.convert_element_type %15493, %int5_13068 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_13069 = torch.constant.int 5
%15502 = torch.prims.convert_element_type %15494, %int5_13069 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %15502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_13070 = torch.constant.int 1
%int0_13071 = torch.constant.int 0
%15503 = torch.prim.ListConstruct %int1_13070, %int0_13071 : (!torch.int, !torch.int) -> !torch.list<int>
%15504 = torch.aten.permute %520, %15503 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13072 = torch.constant.int 1
%int0_13073 = torch.constant.int 0
%15505 = torch.prim.ListConstruct %int1_13072, %int0_13073 : (!torch.int, !torch.int) -> !torch.list<int>
%15506 = torch.aten.permute %521, %15505 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13074 = torch.constant.int 1
%int0_13075 = torch.constant.int 0
%15507 = torch.prim.ListConstruct %int1_13074, %int0_13075 : (!torch.int, !torch.int) -> !torch.list<int>
%15508 = torch.aten.permute %522, %15507 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13076 = torch.constant.int 1
%int0_13077 = torch.constant.int 0
%15509 = torch.prim.ListConstruct %int1_13076, %int0_13077 : (!torch.int, !torch.int) -> !torch.list<int>
%15510 = torch.aten.permute %523, %15509 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13078 = torch.constant.int 1
%int0_13079 = torch.constant.int 0
%15511 = torch.prim.ListConstruct %int1_13078, %int0_13079 : (!torch.int, !torch.int) -> !torch.list<int>
%15512 = torch.aten.permute %524, %15511 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13080 = torch.constant.int 1
%int0_13081 = torch.constant.int 0
%15513 = torch.prim.ListConstruct %int1_13080, %int0_13081 : (!torch.int, !torch.int) -> !torch.list<int>
%15514 = torch.aten.permute %525, %15513 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13082 = torch.constant.int 1
%int0_13083 = torch.constant.int 0
%15515 = torch.prim.ListConstruct %int1_13082, %int0_13083 : (!torch.int, !torch.int) -> !torch.list<int>
%15516 = torch.aten.permute %526, %15515 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_13084 = torch.constant.int 1
%int0_13085 = torch.constant.int 0
%15517 = torch.prim.ListConstruct %int1_13084, %int0_13085 : (!torch.int, !torch.int) -> !torch.list<int>
%15518 = torch.aten.permute %527, %15517 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_13086 = torch.constant.int 4
%15519 = torch.aten.mul.int %int4_13086, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13087 = torch.constant.int 4096
%15520 = torch.prim.ListConstruct %15519, %int4096_13087 : (!torch.int, !torch.int) -> !torch.list<int>
%15521 = torch.aten.view %15495, %15520 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15521, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15522 = torch.aten.mm %15521, %15504 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15522, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13088 = torch.constant.int 4
%int512_13089 = torch.constant.int 512
%15523 = torch.prim.ListConstruct %int4_13088, %2482, %int512_13089 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15524 = torch.aten.view %15522, %15523 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13090 = torch.constant.int 4
%15525 = torch.aten.mul.int %int4_13090, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13091 = torch.constant.int 4096
%15526 = torch.prim.ListConstruct %15525, %int4096_13091 : (!torch.int, !torch.int) -> !torch.list<int>
%15527 = torch.aten.view %15496, %15526 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15527, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15528 = torch.aten.mm %15527, %15506 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15528, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13092 = torch.constant.int 4
%int512_13093 = torch.constant.int 512
%15529 = torch.prim.ListConstruct %int4_13092, %2482, %int512_13093 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15530 = torch.aten.view %15528, %15529 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13094 = torch.constant.int 4
%15531 = torch.aten.mul.int %int4_13094, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13095 = torch.constant.int 4096
%15532 = torch.prim.ListConstruct %15531, %int4096_13095 : (!torch.int, !torch.int) -> !torch.list<int>
%15533 = torch.aten.view %15497, %15532 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15533, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15534 = torch.aten.mm %15533, %15508 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15534, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13096 = torch.constant.int 4
%int512_13097 = torch.constant.int 512
%15535 = torch.prim.ListConstruct %int4_13096, %2482, %int512_13097 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15536 = torch.aten.view %15534, %15535 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13098 = torch.constant.int 4
%15537 = torch.aten.mul.int %int4_13098, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13099 = torch.constant.int 4096
%15538 = torch.prim.ListConstruct %15537, %int4096_13099 : (!torch.int, !torch.int) -> !torch.list<int>
%15539 = torch.aten.view %15498, %15538 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15539, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15540 = torch.aten.mm %15539, %15510 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15540, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13100 = torch.constant.int 4
%int512_13101 = torch.constant.int 512
%15541 = torch.prim.ListConstruct %int4_13100, %2482, %int512_13101 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15542 = torch.aten.view %15540, %15541 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13102 = torch.constant.int 4
%15543 = torch.aten.mul.int %int4_13102, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13103 = torch.constant.int 4096
%15544 = torch.prim.ListConstruct %15543, %int4096_13103 : (!torch.int, !torch.int) -> !torch.list<int>
%15545 = torch.aten.view %15499, %15544 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15545, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15546 = torch.aten.mm %15545, %15512 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15546, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13104 = torch.constant.int 4
%int512_13105 = torch.constant.int 512
%15547 = torch.prim.ListConstruct %int4_13104, %2482, %int512_13105 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15548 = torch.aten.view %15546, %15547 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13106 = torch.constant.int 4
%15549 = torch.aten.mul.int %int4_13106, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13107 = torch.constant.int 4096
%15550 = torch.prim.ListConstruct %15549, %int4096_13107 : (!torch.int, !torch.int) -> !torch.list<int>
%15551 = torch.aten.view %15500, %15550 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15551, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15552 = torch.aten.mm %15551, %15514 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15552, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13108 = torch.constant.int 4
%int512_13109 = torch.constant.int 512
%15553 = torch.prim.ListConstruct %int4_13108, %2482, %int512_13109 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15554 = torch.aten.view %15552, %15553 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13110 = torch.constant.int 4
%15555 = torch.aten.mul.int %int4_13110, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13111 = torch.constant.int 4096
%15556 = torch.prim.ListConstruct %15555, %int4096_13111 : (!torch.int, !torch.int) -> !torch.list<int>
%15557 = torch.aten.view %15501, %15556 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15557, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15558 = torch.aten.mm %15557, %15516 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15558, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13112 = torch.constant.int 4
%int512_13113 = torch.constant.int 512
%15559 = torch.prim.ListConstruct %int4_13112, %2482, %int512_13113 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15560 = torch.aten.view %15558, %15559 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_13114 = torch.constant.int 4
%15561 = torch.aten.mul.int %int4_13114, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13115 = torch.constant.int 4096
%15562 = torch.prim.ListConstruct %15561, %int4096_13115 : (!torch.int, !torch.int) -> !torch.list<int>
%15563 = torch.aten.view %15502, %15562 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15563, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15564 = torch.aten.mm %15563, %15518 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %15564, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_13116 = torch.constant.int 4
%int512_13117 = torch.constant.int 512
%15565 = torch.prim.ListConstruct %int4_13116, %2482, %int512_13117 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15566 = torch.aten.view %15564, %15565 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %15566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_13118 = torch.constant.int 1
%int0_13119 = torch.constant.int 0
%15567 = torch.prim.ListConstruct %int1_13118, %int0_13119 : (!torch.int, !torch.int) -> !torch.list<int>
%15568 = torch.aten.permute %528, %15567 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13120 = torch.constant.int 1
%int0_13121 = torch.constant.int 0
%15569 = torch.prim.ListConstruct %int1_13120, %int0_13121 : (!torch.int, !torch.int) -> !torch.list<int>
%15570 = torch.aten.permute %529, %15569 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13122 = torch.constant.int 1
%int0_13123 = torch.constant.int 0
%15571 = torch.prim.ListConstruct %int1_13122, %int0_13123 : (!torch.int, !torch.int) -> !torch.list<int>
%15572 = torch.aten.permute %530, %15571 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13124 = torch.constant.int 1
%int0_13125 = torch.constant.int 0
%15573 = torch.prim.ListConstruct %int1_13124, %int0_13125 : (!torch.int, !torch.int) -> !torch.list<int>
%15574 = torch.aten.permute %531, %15573 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13126 = torch.constant.int 1
%int0_13127 = torch.constant.int 0
%15575 = torch.prim.ListConstruct %int1_13126, %int0_13127 : (!torch.int, !torch.int) -> !torch.list<int>
%15576 = torch.aten.permute %532, %15575 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13128 = torch.constant.int 1
%int0_13129 = torch.constant.int 0
%15577 = torch.prim.ListConstruct %int1_13128, %int0_13129 : (!torch.int, !torch.int) -> !torch.list<int>
%15578 = torch.aten.permute %533, %15577 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13130 = torch.constant.int 1
%int0_13131 = torch.constant.int 0
%15579 = torch.prim.ListConstruct %int1_13130, %int0_13131 : (!torch.int, !torch.int) -> !torch.list<int>
%15580 = torch.aten.permute %534, %15579 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13132 = torch.constant.int 1
%int0_13133 = torch.constant.int 0
%15581 = torch.prim.ListConstruct %int1_13132, %int0_13133 : (!torch.int, !torch.int) -> !torch.list<int>
%15582 = torch.aten.permute %535, %15581 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_13134 = torch.constant.int 4
%15583 = torch.aten.mul.int %int4_13134, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13135 = torch.constant.int 4096
%15584 = torch.prim.ListConstruct %15583, %int4096_13135 : (!torch.int, !torch.int) -> !torch.list<int>
%15585 = torch.aten.view %15495, %15584 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15585, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15586 = torch.aten.mm %15585, %15568 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15586, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13136 = torch.constant.int 4
%int128_13137 = torch.constant.int 128
%15587 = torch.prim.ListConstruct %int4_13136, %2482, %int128_13137 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15588 = torch.aten.view %15586, %15587 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13138 = torch.constant.int 4
%15589 = torch.aten.mul.int %int4_13138, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13139 = torch.constant.int 4096
%15590 = torch.prim.ListConstruct %15589, %int4096_13139 : (!torch.int, !torch.int) -> !torch.list<int>
%15591 = torch.aten.view %15496, %15590 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15591, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15592 = torch.aten.mm %15591, %15570 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15592, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13140 = torch.constant.int 4
%int128_13141 = torch.constant.int 128
%15593 = torch.prim.ListConstruct %int4_13140, %2482, %int128_13141 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15594 = torch.aten.view %15592, %15593 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13142 = torch.constant.int 4
%15595 = torch.aten.mul.int %int4_13142, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13143 = torch.constant.int 4096
%15596 = torch.prim.ListConstruct %15595, %int4096_13143 : (!torch.int, !torch.int) -> !torch.list<int>
%15597 = torch.aten.view %15497, %15596 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15597, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15598 = torch.aten.mm %15597, %15572 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15598, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13144 = torch.constant.int 4
%int128_13145 = torch.constant.int 128
%15599 = torch.prim.ListConstruct %int4_13144, %2482, %int128_13145 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15600 = torch.aten.view %15598, %15599 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13146 = torch.constant.int 4
%15601 = torch.aten.mul.int %int4_13146, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13147 = torch.constant.int 4096
%15602 = torch.prim.ListConstruct %15601, %int4096_13147 : (!torch.int, !torch.int) -> !torch.list<int>
%15603 = torch.aten.view %15498, %15602 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15603, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15604 = torch.aten.mm %15603, %15574 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15604, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13148 = torch.constant.int 4
%int128_13149 = torch.constant.int 128
%15605 = torch.prim.ListConstruct %int4_13148, %2482, %int128_13149 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15606 = torch.aten.view %15604, %15605 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13150 = torch.constant.int 4
%15607 = torch.aten.mul.int %int4_13150, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13151 = torch.constant.int 4096
%15608 = torch.prim.ListConstruct %15607, %int4096_13151 : (!torch.int, !torch.int) -> !torch.list<int>
%15609 = torch.aten.view %15499, %15608 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15609, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15610 = torch.aten.mm %15609, %15576 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15610, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13152 = torch.constant.int 4
%int128_13153 = torch.constant.int 128
%15611 = torch.prim.ListConstruct %int4_13152, %2482, %int128_13153 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15612 = torch.aten.view %15610, %15611 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13154 = torch.constant.int 4
%15613 = torch.aten.mul.int %int4_13154, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13155 = torch.constant.int 4096
%15614 = torch.prim.ListConstruct %15613, %int4096_13155 : (!torch.int, !torch.int) -> !torch.list<int>
%15615 = torch.aten.view %15500, %15614 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15615, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15616 = torch.aten.mm %15615, %15578 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15616, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13156 = torch.constant.int 4
%int128_13157 = torch.constant.int 128
%15617 = torch.prim.ListConstruct %int4_13156, %2482, %int128_13157 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15618 = torch.aten.view %15616, %15617 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13158 = torch.constant.int 4
%15619 = torch.aten.mul.int %int4_13158, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13159 = torch.constant.int 4096
%15620 = torch.prim.ListConstruct %15619, %int4096_13159 : (!torch.int, !torch.int) -> !torch.list<int>
%15621 = torch.aten.view %15501, %15620 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15621, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15622 = torch.aten.mm %15621, %15580 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15622, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13160 = torch.constant.int 4
%int128_13161 = torch.constant.int 128
%15623 = torch.prim.ListConstruct %int4_13160, %2482, %int128_13161 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15624 = torch.aten.view %15622, %15623 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13162 = torch.constant.int 4
%15625 = torch.aten.mul.int %int4_13162, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13163 = torch.constant.int 4096
%15626 = torch.prim.ListConstruct %15625, %int4096_13163 : (!torch.int, !torch.int) -> !torch.list<int>
%15627 = torch.aten.view %15502, %15626 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15627, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15628 = torch.aten.mm %15627, %15582 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15628, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13164 = torch.constant.int 4
%int128_13165 = torch.constant.int 128
%15629 = torch.prim.ListConstruct %int4_13164, %2482, %int128_13165 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15630 = torch.aten.view %15628, %15629 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_13166 = torch.constant.int 1
%int0_13167 = torch.constant.int 0
%15631 = torch.prim.ListConstruct %int1_13166, %int0_13167 : (!torch.int, !torch.int) -> !torch.list<int>
%15632 = torch.aten.permute %536, %15631 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13168 = torch.constant.int 1
%int0_13169 = torch.constant.int 0
%15633 = torch.prim.ListConstruct %int1_13168, %int0_13169 : (!torch.int, !torch.int) -> !torch.list<int>
%15634 = torch.aten.permute %537, %15633 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13170 = torch.constant.int 1
%int0_13171 = torch.constant.int 0
%15635 = torch.prim.ListConstruct %int1_13170, %int0_13171 : (!torch.int, !torch.int) -> !torch.list<int>
%15636 = torch.aten.permute %538, %15635 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13172 = torch.constant.int 1
%int0_13173 = torch.constant.int 0
%15637 = torch.prim.ListConstruct %int1_13172, %int0_13173 : (!torch.int, !torch.int) -> !torch.list<int>
%15638 = torch.aten.permute %539, %15637 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13174 = torch.constant.int 1
%int0_13175 = torch.constant.int 0
%15639 = torch.prim.ListConstruct %int1_13174, %int0_13175 : (!torch.int, !torch.int) -> !torch.list<int>
%15640 = torch.aten.permute %540, %15639 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13176 = torch.constant.int 1
%int0_13177 = torch.constant.int 0
%15641 = torch.prim.ListConstruct %int1_13176, %int0_13177 : (!torch.int, !torch.int) -> !torch.list<int>
%15642 = torch.aten.permute %541, %15641 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13178 = torch.constant.int 1
%int0_13179 = torch.constant.int 0
%15643 = torch.prim.ListConstruct %int1_13178, %int0_13179 : (!torch.int, !torch.int) -> !torch.list<int>
%15644 = torch.aten.permute %542, %15643 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_13180 = torch.constant.int 1
%int0_13181 = torch.constant.int 0
%15645 = torch.prim.ListConstruct %int1_13180, %int0_13181 : (!torch.int, !torch.int) -> !torch.list<int>
%15646 = torch.aten.permute %543, %15645 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_13182 = torch.constant.int 4
%15647 = torch.aten.mul.int %int4_13182, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13183 = torch.constant.int 4096
%15648 = torch.prim.ListConstruct %15647, %int4096_13183 : (!torch.int, !torch.int) -> !torch.list<int>
%15649 = torch.aten.view %15495, %15648 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15649, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15650 = torch.aten.mm %15649, %15632 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15650, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13184 = torch.constant.int 4
%int128_13185 = torch.constant.int 128
%15651 = torch.prim.ListConstruct %int4_13184, %2482, %int128_13185 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15652 = torch.aten.view %15650, %15651 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13186 = torch.constant.int 4
%15653 = torch.aten.mul.int %int4_13186, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13187 = torch.constant.int 4096
%15654 = torch.prim.ListConstruct %15653, %int4096_13187 : (!torch.int, !torch.int) -> !torch.list<int>
%15655 = torch.aten.view %15496, %15654 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15655, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15656 = torch.aten.mm %15655, %15634 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15656, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13188 = torch.constant.int 4
%int128_13189 = torch.constant.int 128
%15657 = torch.prim.ListConstruct %int4_13188, %2482, %int128_13189 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15658 = torch.aten.view %15656, %15657 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13190 = torch.constant.int 4
%15659 = torch.aten.mul.int %int4_13190, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13191 = torch.constant.int 4096
%15660 = torch.prim.ListConstruct %15659, %int4096_13191 : (!torch.int, !torch.int) -> !torch.list<int>
%15661 = torch.aten.view %15497, %15660 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15661, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15662 = torch.aten.mm %15661, %15636 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15662, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13192 = torch.constant.int 4
%int128_13193 = torch.constant.int 128
%15663 = torch.prim.ListConstruct %int4_13192, %2482, %int128_13193 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15664 = torch.aten.view %15662, %15663 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13194 = torch.constant.int 4
%15665 = torch.aten.mul.int %int4_13194, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13195 = torch.constant.int 4096
%15666 = torch.prim.ListConstruct %15665, %int4096_13195 : (!torch.int, !torch.int) -> !torch.list<int>
%15667 = torch.aten.view %15498, %15666 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15667, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15668 = torch.aten.mm %15667, %15638 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15668, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13196 = torch.constant.int 4
%int128_13197 = torch.constant.int 128
%15669 = torch.prim.ListConstruct %int4_13196, %2482, %int128_13197 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15670 = torch.aten.view %15668, %15669 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13198 = torch.constant.int 4
%15671 = torch.aten.mul.int %int4_13198, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13199 = torch.constant.int 4096
%15672 = torch.prim.ListConstruct %15671, %int4096_13199 : (!torch.int, !torch.int) -> !torch.list<int>
%15673 = torch.aten.view %15499, %15672 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15673, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15674 = torch.aten.mm %15673, %15640 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15674, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13200 = torch.constant.int 4
%int128_13201 = torch.constant.int 128
%15675 = torch.prim.ListConstruct %int4_13200, %2482, %int128_13201 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15676 = torch.aten.view %15674, %15675 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13202 = torch.constant.int 4
%15677 = torch.aten.mul.int %int4_13202, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13203 = torch.constant.int 4096
%15678 = torch.prim.ListConstruct %15677, %int4096_13203 : (!torch.int, !torch.int) -> !torch.list<int>
%15679 = torch.aten.view %15500, %15678 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15679, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15680 = torch.aten.mm %15679, %15642 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15680, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13204 = torch.constant.int 4
%int128_13205 = torch.constant.int 128
%15681 = torch.prim.ListConstruct %int4_13204, %2482, %int128_13205 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15682 = torch.aten.view %15680, %15681 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13206 = torch.constant.int 4
%15683 = torch.aten.mul.int %int4_13206, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13207 = torch.constant.int 4096
%15684 = torch.prim.ListConstruct %15683, %int4096_13207 : (!torch.int, !torch.int) -> !torch.list<int>
%15685 = torch.aten.view %15501, %15684 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15685, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15686 = torch.aten.mm %15685, %15644 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15686, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13208 = torch.constant.int 4
%int128_13209 = torch.constant.int 128
%15687 = torch.prim.ListConstruct %int4_13208, %2482, %int128_13209 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15688 = torch.aten.view %15686, %15687 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13210 = torch.constant.int 4
%15689 = torch.aten.mul.int %int4_13210, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_13211 = torch.constant.int 4096
%15690 = torch.prim.ListConstruct %15689, %int4096_13211 : (!torch.int, !torch.int) -> !torch.list<int>
%15691 = torch.aten.view %15502, %15690 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %15691, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%15692 = torch.aten.mm %15691, %15646 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %15692, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_13212 = torch.constant.int 4
%int128_13213 = torch.constant.int 128
%15693 = torch.prim.ListConstruct %int4_13212, %2482, %int128_13213 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15694 = torch.aten.view %15692, %15693 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %15694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_13214 = torch.constant.int 4
%int4_13215 = torch.constant.int 4
%int128_13216 = torch.constant.int 128
%15695 = torch.prim.ListConstruct %int4_13214, %2482, %int4_13215, %int128_13216 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15696 = torch.aten.view %15524, %15695 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13217 = torch.constant.int 4
%int4_13218 = torch.constant.int 4
%int128_13219 = torch.constant.int 128
%15697 = torch.prim.ListConstruct %int4_13217, %2482, %int4_13218, %int128_13219 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15698 = torch.aten.view %15530, %15697 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13220 = torch.constant.int 4
%int4_13221 = torch.constant.int 4
%int128_13222 = torch.constant.int 128
%15699 = torch.prim.ListConstruct %int4_13220, %2482, %int4_13221, %int128_13222 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15700 = torch.aten.view %15536, %15699 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13223 = torch.constant.int 4
%int4_13224 = torch.constant.int 4
%int128_13225 = torch.constant.int 128
%15701 = torch.prim.ListConstruct %int4_13223, %2482, %int4_13224, %int128_13225 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15702 = torch.aten.view %15542, %15701 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13226 = torch.constant.int 4
%int4_13227 = torch.constant.int 4
%int128_13228 = torch.constant.int 128
%15703 = torch.prim.ListConstruct %int4_13226, %2482, %int4_13227, %int128_13228 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15704 = torch.aten.view %15548, %15703 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13229 = torch.constant.int 4
%int4_13230 = torch.constant.int 4
%int128_13231 = torch.constant.int 128
%15705 = torch.prim.ListConstruct %int4_13229, %2482, %int4_13230, %int128_13231 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15706 = torch.aten.view %15554, %15705 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13232 = torch.constant.int 4
%int4_13233 = torch.constant.int 4
%int128_13234 = torch.constant.int 128
%15707 = torch.prim.ListConstruct %int4_13232, %2482, %int4_13233, %int128_13234 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15708 = torch.aten.view %15560, %15707 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13235 = torch.constant.int 4
%int4_13236 = torch.constant.int 4
%int128_13237 = torch.constant.int 128
%15709 = torch.prim.ListConstruct %int4_13235, %2482, %int4_13236, %int128_13237 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15710 = torch.aten.view %15566, %15709 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_13238 = torch.constant.int 4
%int1_13239 = torch.constant.int 1
%int128_13240 = torch.constant.int 128
%15711 = torch.prim.ListConstruct %int4_13238, %2482, %int1_13239, %int128_13240 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15712 = torch.aten.view %15588, %15711 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13241 = torch.constant.int 4
%int1_13242 = torch.constant.int 1
%int128_13243 = torch.constant.int 128
%15713 = torch.prim.ListConstruct %int4_13241, %2482, %int1_13242, %int128_13243 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15714 = torch.aten.view %15594, %15713 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13244 = torch.constant.int 4
%int1_13245 = torch.constant.int 1
%int128_13246 = torch.constant.int 128
%15715 = torch.prim.ListConstruct %int4_13244, %2482, %int1_13245, %int128_13246 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15716 = torch.aten.view %15600, %15715 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13247 = torch.constant.int 4
%int1_13248 = torch.constant.int 1
%int128_13249 = torch.constant.int 128
%15717 = torch.prim.ListConstruct %int4_13247, %2482, %int1_13248, %int128_13249 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15718 = torch.aten.view %15606, %15717 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13250 = torch.constant.int 4
%int1_13251 = torch.constant.int 1
%int128_13252 = torch.constant.int 128
%15719 = torch.prim.ListConstruct %int4_13250, %2482, %int1_13251, %int128_13252 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15720 = torch.aten.view %15612, %15719 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13253 = torch.constant.int 4
%int1_13254 = torch.constant.int 1
%int128_13255 = torch.constant.int 128
%15721 = torch.prim.ListConstruct %int4_13253, %2482, %int1_13254, %int128_13255 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15722 = torch.aten.view %15618, %15721 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13256 = torch.constant.int 4
%int1_13257 = torch.constant.int 1
%int128_13258 = torch.constant.int 128
%15723 = torch.prim.ListConstruct %int4_13256, %2482, %int1_13257, %int128_13258 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15724 = torch.aten.view %15624, %15723 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13259 = torch.constant.int 4
%int1_13260 = torch.constant.int 1
%int128_13261 = torch.constant.int 128
%15725 = torch.prim.ListConstruct %int4_13259, %2482, %int1_13260, %int128_13261 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15726 = torch.aten.view %15630, %15725 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13262 = torch.constant.int 4
%int1_13263 = torch.constant.int 1
%int128_13264 = torch.constant.int 128
%15727 = torch.prim.ListConstruct %int4_13262, %2482, %int1_13263, %int128_13264 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15728 = torch.aten.view %15652, %15727 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13265 = torch.constant.int 4
%int1_13266 = torch.constant.int 1
%int128_13267 = torch.constant.int 128
%15729 = torch.prim.ListConstruct %int4_13265, %2482, %int1_13266, %int128_13267 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15730 = torch.aten.view %15658, %15729 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13268 = torch.constant.int 4
%int1_13269 = torch.constant.int 1
%int128_13270 = torch.constant.int 128
%15731 = torch.prim.ListConstruct %int4_13268, %2482, %int1_13269, %int128_13270 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15732 = torch.aten.view %15664, %15731 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13271 = torch.constant.int 4
%int1_13272 = torch.constant.int 1
%int128_13273 = torch.constant.int 128
%15733 = torch.prim.ListConstruct %int4_13271, %2482, %int1_13272, %int128_13273 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15734 = torch.aten.view %15670, %15733 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13274 = torch.constant.int 4
%int1_13275 = torch.constant.int 1
%int128_13276 = torch.constant.int 128
%15735 = torch.prim.ListConstruct %int4_13274, %2482, %int1_13275, %int128_13276 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15736 = torch.aten.view %15676, %15735 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13277 = torch.constant.int 4
%int1_13278 = torch.constant.int 1
%int128_13279 = torch.constant.int 128
%15737 = torch.prim.ListConstruct %int4_13277, %2482, %int1_13278, %int128_13279 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15738 = torch.aten.view %15682, %15737 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13280 = torch.constant.int 4
%int1_13281 = torch.constant.int 1
%int128_13282 = torch.constant.int 128
%15739 = torch.prim.ListConstruct %int4_13280, %2482, %int1_13281, %int128_13282 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15740 = torch.aten.view %15688, %15739 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_13283 = torch.constant.int 4
%int1_13284 = torch.constant.int 1
%int128_13285 = torch.constant.int 128
%15741 = torch.prim.ListConstruct %int4_13283, %2482, %int1_13284, %int128_13285 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%15742 = torch.aten.view %15694, %15741 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_13286 = torch.constant.int 131072
%none_13287 = torch.constant.none
%none_13288 = torch.constant.none
%cpu_13289 = torch.constant.device "cpu"
%false_13290 = torch.constant.bool false
%15743 = torch.aten.arange %int131072_13286, %none_13287, %none_13288, %cpu_13289, %false_13290 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_13291 = torch.constant.int 0
%int128_13292 = torch.constant.int 128
%int2_13293 = torch.constant.int 2
%none_13294 = torch.constant.none
%none_13295 = torch.constant.none
%cpu_13296 = torch.constant.device "cpu"
%false_13297 = torch.constant.bool false
%15744 = torch.aten.arange.start_step %int0_13291, %int128_13292, %int2_13293, %none_13294, %none_13295, %cpu_13296, %false_13297 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_13298 = torch.constant.int 0
%int0_13299 = torch.constant.int 0
%int64_13300 = torch.constant.int 64
%int1_13301 = torch.constant.int 1
%15745 = torch.aten.slice.Tensor %15744, %int0_13298, %int0_13299, %int64_13300, %int1_13301 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_13302 = torch.constant.int 6
%15746 = torch.prims.convert_element_type %15745, %int6_13302 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_13303 = torch.constant.int 128
%15747 = torch.aten.div.Scalar %15746, %int128_13303 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_13304 = torch.constant.float 5.000000e+05
%15748 = torch.aten.pow.Scalar %float5.000000e05_13304, %15747 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%15749 = torch.aten.reciprocal %15748 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_13305 = torch.constant.float 1.000000e+00
%15750 = torch.aten.mul.Scalar %15749, %float1.000000e00_13305 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_13306 = torch.constant.int 131072
%int1_13307 = torch.constant.int 1
%15751 = torch.prim.ListConstruct %int131072_13306, %int1_13307 : (!torch.int, !torch.int) -> !torch.list<int>
%15752 = torch.aten.view %15743, %15751 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%15753 = torch.aten.mul.Tensor %15752, %15750 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%15754 = torch.aten.cos %15753 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%15755 = torch.aten.sin %15753 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%15756 = torch.aten.complex %15754, %15755 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%15757 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15758 = flow.tensor.transfer %15757 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%15759 = torch_c.from_builtin_tensor %15758 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15760 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15761 = flow.tensor.transfer %15760 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%15762 = torch_c.from_builtin_tensor %15761 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15763 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15764 = flow.tensor.transfer %15763 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%15765 = torch_c.from_builtin_tensor %15764 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15766 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15767 = flow.tensor.transfer %15766 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%15768 = torch_c.from_builtin_tensor %15767 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15769 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15770 = flow.tensor.transfer %15769 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%15771 = torch_c.from_builtin_tensor %15770 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15772 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15773 = flow.tensor.transfer %15772 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%15774 = torch_c.from_builtin_tensor %15773 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15775 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15776 = flow.tensor.transfer %15775 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%15777 = torch_c.from_builtin_tensor %15776 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15778 = torch_c.to_builtin_tensor %15756 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15779 = flow.tensor.transfer %15778 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%15780 = torch_c.from_builtin_tensor %15779 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_13308 = torch.constant.int 1
%15781 = torch.aten.size.int %15524, %int1_13308 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13309 = torch.constant.int 0
%15782 = torch.aten.add.int %int0_13309, %15781 : !torch.int, !torch.int -> !torch.int
%int0_13310 = torch.constant.int 0
%int0_13311 = torch.constant.int 0
%int1_13312 = torch.constant.int 1
%15783 = torch.aten.slice.Tensor %15759, %int0_13310, %int0_13311, %15782, %int1_13312 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15783, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13313 = torch.constant.int 1
%int0_13314 = torch.constant.int 0
%int9223372036854775807_13315 = torch.constant.int 9223372036854775807
%int1_13316 = torch.constant.int 1
%15784 = torch.aten.slice.Tensor %15783, %int1_13313, %int0_13314, %int9223372036854775807_13315, %int1_13316 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15784, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13317 = torch.constant.int 0
%15785 = torch.aten.unsqueeze %15784, %int0_13317 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15785, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13318 = torch.constant.int 2
%15786 = torch.aten.unsqueeze %15785, %int2_13318 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15786, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13319 = torch.constant.int 3
%int0_13320 = torch.constant.int 0
%int9223372036854775807_13321 = torch.constant.int 9223372036854775807
%int1_13322 = torch.constant.int 1
%15787 = torch.aten.slice.Tensor %15786, %int3_13319, %int0_13320, %int9223372036854775807_13321, %int1_13322 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15787, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15788 = torch_c.to_builtin_tensor %15696 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13323 = arith.constant 1 : index
%dim_13324 = tensor.dim %15788, %c1_13323 : tensor<4x?x4x128xf16>
%15789 = flow.tensor.bitcast %15788 : tensor<4x?x4x128xf16>{%dim_13324} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13324}
%15790 = torch_c.from_builtin_tensor %15789 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15791 = torch.aten.mul.Tensor %15790, %15787 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15792 = torch_c.to_builtin_tensor %15791 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13325 = arith.constant 1 : index
%dim_13326 = tensor.dim %15792, %c1_13325 : tensor<4x?x4x64xcomplex<f32>>
%15793 = flow.tensor.bitcast %15792 : tensor<4x?x4x64xcomplex<f32>>{%dim_13326} -> tensor<4x?x4x128xf32>{%dim_13326}
%15794 = torch_c.from_builtin_tensor %15793 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13327 = torch.constant.int 5
%15795 = torch.prims.convert_element_type %15794, %int5_13327 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13328 = torch.constant.int 1
%15796 = torch.aten.size.int %15530, %int1_13328 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13329 = torch.constant.int 0
%15797 = torch.aten.add.int %int0_13329, %15796 : !torch.int, !torch.int -> !torch.int
%int0_13330 = torch.constant.int 0
%int0_13331 = torch.constant.int 0
%int1_13332 = torch.constant.int 1
%15798 = torch.aten.slice.Tensor %15762, %int0_13330, %int0_13331, %15797, %int1_13332 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15798, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13333 = torch.constant.int 1
%int0_13334 = torch.constant.int 0
%int9223372036854775807_13335 = torch.constant.int 9223372036854775807
%int1_13336 = torch.constant.int 1
%15799 = torch.aten.slice.Tensor %15798, %int1_13333, %int0_13334, %int9223372036854775807_13335, %int1_13336 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15799, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13337 = torch.constant.int 0
%15800 = torch.aten.unsqueeze %15799, %int0_13337 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15800, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13338 = torch.constant.int 2
%15801 = torch.aten.unsqueeze %15800, %int2_13338 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15801, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13339 = torch.constant.int 3
%int0_13340 = torch.constant.int 0
%int9223372036854775807_13341 = torch.constant.int 9223372036854775807
%int1_13342 = torch.constant.int 1
%15802 = torch.aten.slice.Tensor %15801, %int3_13339, %int0_13340, %int9223372036854775807_13341, %int1_13342 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15802, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15803 = torch_c.to_builtin_tensor %15698 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13343 = arith.constant 1 : index
%dim_13344 = tensor.dim %15803, %c1_13343 : tensor<4x?x4x128xf16>
%15804 = flow.tensor.bitcast %15803 : tensor<4x?x4x128xf16>{%dim_13344} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13344}
%15805 = torch_c.from_builtin_tensor %15804 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15806 = torch.aten.mul.Tensor %15805, %15802 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15807 = torch_c.to_builtin_tensor %15806 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13345 = arith.constant 1 : index
%dim_13346 = tensor.dim %15807, %c1_13345 : tensor<4x?x4x64xcomplex<f32>>
%15808 = flow.tensor.bitcast %15807 : tensor<4x?x4x64xcomplex<f32>>{%dim_13346} -> tensor<4x?x4x128xf32>{%dim_13346}
%15809 = torch_c.from_builtin_tensor %15808 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13347 = torch.constant.int 5
%15810 = torch.prims.convert_element_type %15809, %int5_13347 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13348 = torch.constant.int 1
%15811 = torch.aten.size.int %15536, %int1_13348 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13349 = torch.constant.int 0
%15812 = torch.aten.add.int %int0_13349, %15811 : !torch.int, !torch.int -> !torch.int
%int0_13350 = torch.constant.int 0
%int0_13351 = torch.constant.int 0
%int1_13352 = torch.constant.int 1
%15813 = torch.aten.slice.Tensor %15765, %int0_13350, %int0_13351, %15812, %int1_13352 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15813, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13353 = torch.constant.int 1
%int0_13354 = torch.constant.int 0
%int9223372036854775807_13355 = torch.constant.int 9223372036854775807
%int1_13356 = torch.constant.int 1
%15814 = torch.aten.slice.Tensor %15813, %int1_13353, %int0_13354, %int9223372036854775807_13355, %int1_13356 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15814, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13357 = torch.constant.int 0
%15815 = torch.aten.unsqueeze %15814, %int0_13357 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15815, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13358 = torch.constant.int 2
%15816 = torch.aten.unsqueeze %15815, %int2_13358 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15816, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13359 = torch.constant.int 3
%int0_13360 = torch.constant.int 0
%int9223372036854775807_13361 = torch.constant.int 9223372036854775807
%int1_13362 = torch.constant.int 1
%15817 = torch.aten.slice.Tensor %15816, %int3_13359, %int0_13360, %int9223372036854775807_13361, %int1_13362 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15817, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15818 = torch_c.to_builtin_tensor %15700 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13363 = arith.constant 1 : index
%dim_13364 = tensor.dim %15818, %c1_13363 : tensor<4x?x4x128xf16>
%15819 = flow.tensor.bitcast %15818 : tensor<4x?x4x128xf16>{%dim_13364} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13364}
%15820 = torch_c.from_builtin_tensor %15819 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15821 = torch.aten.mul.Tensor %15820, %15817 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15822 = torch_c.to_builtin_tensor %15821 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13365 = arith.constant 1 : index
%dim_13366 = tensor.dim %15822, %c1_13365 : tensor<4x?x4x64xcomplex<f32>>
%15823 = flow.tensor.bitcast %15822 : tensor<4x?x4x64xcomplex<f32>>{%dim_13366} -> tensor<4x?x4x128xf32>{%dim_13366}
%15824 = torch_c.from_builtin_tensor %15823 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13367 = torch.constant.int 5
%15825 = torch.prims.convert_element_type %15824, %int5_13367 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13368 = torch.constant.int 1
%15826 = torch.aten.size.int %15542, %int1_13368 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13369 = torch.constant.int 0
%15827 = torch.aten.add.int %int0_13369, %15826 : !torch.int, !torch.int -> !torch.int
%int0_13370 = torch.constant.int 0
%int0_13371 = torch.constant.int 0
%int1_13372 = torch.constant.int 1
%15828 = torch.aten.slice.Tensor %15768, %int0_13370, %int0_13371, %15827, %int1_13372 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15828, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13373 = torch.constant.int 1
%int0_13374 = torch.constant.int 0
%int9223372036854775807_13375 = torch.constant.int 9223372036854775807
%int1_13376 = torch.constant.int 1
%15829 = torch.aten.slice.Tensor %15828, %int1_13373, %int0_13374, %int9223372036854775807_13375, %int1_13376 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15829, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13377 = torch.constant.int 0
%15830 = torch.aten.unsqueeze %15829, %int0_13377 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15830, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13378 = torch.constant.int 2
%15831 = torch.aten.unsqueeze %15830, %int2_13378 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15831, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13379 = torch.constant.int 3
%int0_13380 = torch.constant.int 0
%int9223372036854775807_13381 = torch.constant.int 9223372036854775807
%int1_13382 = torch.constant.int 1
%15832 = torch.aten.slice.Tensor %15831, %int3_13379, %int0_13380, %int9223372036854775807_13381, %int1_13382 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15832, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15833 = torch_c.to_builtin_tensor %15702 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13383 = arith.constant 1 : index
%dim_13384 = tensor.dim %15833, %c1_13383 : tensor<4x?x4x128xf16>
%15834 = flow.tensor.bitcast %15833 : tensor<4x?x4x128xf16>{%dim_13384} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13384}
%15835 = torch_c.from_builtin_tensor %15834 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15836 = torch.aten.mul.Tensor %15835, %15832 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15837 = torch_c.to_builtin_tensor %15836 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13385 = arith.constant 1 : index
%dim_13386 = tensor.dim %15837, %c1_13385 : tensor<4x?x4x64xcomplex<f32>>
%15838 = flow.tensor.bitcast %15837 : tensor<4x?x4x64xcomplex<f32>>{%dim_13386} -> tensor<4x?x4x128xf32>{%dim_13386}
%15839 = torch_c.from_builtin_tensor %15838 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13387 = torch.constant.int 5
%15840 = torch.prims.convert_element_type %15839, %int5_13387 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13388 = torch.constant.int 1
%15841 = torch.aten.size.int %15548, %int1_13388 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13389 = torch.constant.int 0
%15842 = torch.aten.add.int %int0_13389, %15841 : !torch.int, !torch.int -> !torch.int
%int0_13390 = torch.constant.int 0
%int0_13391 = torch.constant.int 0
%int1_13392 = torch.constant.int 1
%15843 = torch.aten.slice.Tensor %15771, %int0_13390, %int0_13391, %15842, %int1_13392 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15843, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13393 = torch.constant.int 1
%int0_13394 = torch.constant.int 0
%int9223372036854775807_13395 = torch.constant.int 9223372036854775807
%int1_13396 = torch.constant.int 1
%15844 = torch.aten.slice.Tensor %15843, %int1_13393, %int0_13394, %int9223372036854775807_13395, %int1_13396 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15844, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13397 = torch.constant.int 0
%15845 = torch.aten.unsqueeze %15844, %int0_13397 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15845, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13398 = torch.constant.int 2
%15846 = torch.aten.unsqueeze %15845, %int2_13398 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15846, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13399 = torch.constant.int 3
%int0_13400 = torch.constant.int 0
%int9223372036854775807_13401 = torch.constant.int 9223372036854775807
%int1_13402 = torch.constant.int 1
%15847 = torch.aten.slice.Tensor %15846, %int3_13399, %int0_13400, %int9223372036854775807_13401, %int1_13402 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15847, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15848 = torch_c.to_builtin_tensor %15704 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13403 = arith.constant 1 : index
%dim_13404 = tensor.dim %15848, %c1_13403 : tensor<4x?x4x128xf16>
%15849 = flow.tensor.bitcast %15848 : tensor<4x?x4x128xf16>{%dim_13404} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13404}
%15850 = torch_c.from_builtin_tensor %15849 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15851 = torch.aten.mul.Tensor %15850, %15847 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15852 = torch_c.to_builtin_tensor %15851 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13405 = arith.constant 1 : index
%dim_13406 = tensor.dim %15852, %c1_13405 : tensor<4x?x4x64xcomplex<f32>>
%15853 = flow.tensor.bitcast %15852 : tensor<4x?x4x64xcomplex<f32>>{%dim_13406} -> tensor<4x?x4x128xf32>{%dim_13406}
%15854 = torch_c.from_builtin_tensor %15853 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13407 = torch.constant.int 5
%15855 = torch.prims.convert_element_type %15854, %int5_13407 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13408 = torch.constant.int 1
%15856 = torch.aten.size.int %15554, %int1_13408 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13409 = torch.constant.int 0
%15857 = torch.aten.add.int %int0_13409, %15856 : !torch.int, !torch.int -> !torch.int
%int0_13410 = torch.constant.int 0
%int0_13411 = torch.constant.int 0
%int1_13412 = torch.constant.int 1
%15858 = torch.aten.slice.Tensor %15774, %int0_13410, %int0_13411, %15857, %int1_13412 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15858, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13413 = torch.constant.int 1
%int0_13414 = torch.constant.int 0
%int9223372036854775807_13415 = torch.constant.int 9223372036854775807
%int1_13416 = torch.constant.int 1
%15859 = torch.aten.slice.Tensor %15858, %int1_13413, %int0_13414, %int9223372036854775807_13415, %int1_13416 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15859, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13417 = torch.constant.int 0
%15860 = torch.aten.unsqueeze %15859, %int0_13417 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15860, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13418 = torch.constant.int 2
%15861 = torch.aten.unsqueeze %15860, %int2_13418 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15861, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13419 = torch.constant.int 3
%int0_13420 = torch.constant.int 0
%int9223372036854775807_13421 = torch.constant.int 9223372036854775807
%int1_13422 = torch.constant.int 1
%15862 = torch.aten.slice.Tensor %15861, %int3_13419, %int0_13420, %int9223372036854775807_13421, %int1_13422 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15862, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15863 = torch_c.to_builtin_tensor %15706 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13423 = arith.constant 1 : index
%dim_13424 = tensor.dim %15863, %c1_13423 : tensor<4x?x4x128xf16>
%15864 = flow.tensor.bitcast %15863 : tensor<4x?x4x128xf16>{%dim_13424} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13424}
%15865 = torch_c.from_builtin_tensor %15864 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15866 = torch.aten.mul.Tensor %15865, %15862 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15867 = torch_c.to_builtin_tensor %15866 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13425 = arith.constant 1 : index
%dim_13426 = tensor.dim %15867, %c1_13425 : tensor<4x?x4x64xcomplex<f32>>
%15868 = flow.tensor.bitcast %15867 : tensor<4x?x4x64xcomplex<f32>>{%dim_13426} -> tensor<4x?x4x128xf32>{%dim_13426}
%15869 = torch_c.from_builtin_tensor %15868 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13427 = torch.constant.int 5
%15870 = torch.prims.convert_element_type %15869, %int5_13427 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13428 = torch.constant.int 1
%15871 = torch.aten.size.int %15560, %int1_13428 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13429 = torch.constant.int 0
%15872 = torch.aten.add.int %int0_13429, %15871 : !torch.int, !torch.int -> !torch.int
%int0_13430 = torch.constant.int 0
%int0_13431 = torch.constant.int 0
%int1_13432 = torch.constant.int 1
%15873 = torch.aten.slice.Tensor %15777, %int0_13430, %int0_13431, %15872, %int1_13432 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15873, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13433 = torch.constant.int 1
%int0_13434 = torch.constant.int 0
%int9223372036854775807_13435 = torch.constant.int 9223372036854775807
%int1_13436 = torch.constant.int 1
%15874 = torch.aten.slice.Tensor %15873, %int1_13433, %int0_13434, %int9223372036854775807_13435, %int1_13436 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15874, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13437 = torch.constant.int 0
%15875 = torch.aten.unsqueeze %15874, %int0_13437 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15875, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13438 = torch.constant.int 2
%15876 = torch.aten.unsqueeze %15875, %int2_13438 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15876, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13439 = torch.constant.int 3
%int0_13440 = torch.constant.int 0
%int9223372036854775807_13441 = torch.constant.int 9223372036854775807
%int1_13442 = torch.constant.int 1
%15877 = torch.aten.slice.Tensor %15876, %int3_13439, %int0_13440, %int9223372036854775807_13441, %int1_13442 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15877, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15878 = torch_c.to_builtin_tensor %15708 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13443 = arith.constant 1 : index
%dim_13444 = tensor.dim %15878, %c1_13443 : tensor<4x?x4x128xf16>
%15879 = flow.tensor.bitcast %15878 : tensor<4x?x4x128xf16>{%dim_13444} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13444}
%15880 = torch_c.from_builtin_tensor %15879 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15881 = torch.aten.mul.Tensor %15880, %15877 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15882 = torch_c.to_builtin_tensor %15881 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13445 = arith.constant 1 : index
%dim_13446 = tensor.dim %15882, %c1_13445 : tensor<4x?x4x64xcomplex<f32>>
%15883 = flow.tensor.bitcast %15882 : tensor<4x?x4x64xcomplex<f32>>{%dim_13446} -> tensor<4x?x4x128xf32>{%dim_13446}
%15884 = torch_c.from_builtin_tensor %15883 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13447 = torch.constant.int 5
%15885 = torch.prims.convert_element_type %15884, %int5_13447 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_13448 = torch.constant.int 1
%15886 = torch.aten.size.int %15566, %int1_13448 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_13449 = torch.constant.int 0
%15887 = torch.aten.add.int %int0_13449, %15886 : !torch.int, !torch.int -> !torch.int
%int0_13450 = torch.constant.int 0
%int0_13451 = torch.constant.int 0
%int1_13452 = torch.constant.int 1
%15888 = torch.aten.slice.Tensor %15780, %int0_13450, %int0_13451, %15887, %int1_13452 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15888, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13453 = torch.constant.int 1
%int0_13454 = torch.constant.int 0
%int9223372036854775807_13455 = torch.constant.int 9223372036854775807
%int1_13456 = torch.constant.int 1
%15889 = torch.aten.slice.Tensor %15888, %int1_13453, %int0_13454, %int9223372036854775807_13455, %int1_13456 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15889, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13457 = torch.constant.int 0
%15890 = torch.aten.unsqueeze %15889, %int0_13457 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15890, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13458 = torch.constant.int 2
%15891 = torch.aten.unsqueeze %15890, %int2_13458 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15891, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13459 = torch.constant.int 3
%int0_13460 = torch.constant.int 0
%int9223372036854775807_13461 = torch.constant.int 9223372036854775807
%int1_13462 = torch.constant.int 1
%15892 = torch.aten.slice.Tensor %15891, %int3_13459, %int0_13460, %int9223372036854775807_13461, %int1_13462 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15892, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15893 = torch_c.to_builtin_tensor %15710 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_13463 = arith.constant 1 : index
%dim_13464 = tensor.dim %15893, %c1_13463 : tensor<4x?x4x128xf16>
%15894 = flow.tensor.bitcast %15893 : tensor<4x?x4x128xf16>{%dim_13464} -> tensor<4x?x4x64xcomplex<f16>>{%dim_13464}
%15895 = torch_c.from_builtin_tensor %15894 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %15895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%15896 = torch.aten.mul.Tensor %15895, %15892 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %15896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%15897 = torch_c.to_builtin_tensor %15896 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_13465 = arith.constant 1 : index
%dim_13466 = tensor.dim %15897, %c1_13465 : tensor<4x?x4x64xcomplex<f32>>
%15898 = flow.tensor.bitcast %15897 : tensor<4x?x4x64xcomplex<f32>>{%dim_13466} -> tensor<4x?x4x128xf32>{%dim_13466}
%15899 = torch_c.from_builtin_tensor %15898 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %15899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_13467 = torch.constant.int 5
%15900 = torch.prims.convert_element_type %15899, %int5_13467 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %15900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_13468 = torch.constant.int 131072
%none_13469 = torch.constant.none
%none_13470 = torch.constant.none
%cpu_13471 = torch.constant.device "cpu"
%false_13472 = torch.constant.bool false
%15901 = torch.aten.arange %int131072_13468, %none_13469, %none_13470, %cpu_13471, %false_13472 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_13473 = torch.constant.int 0
%int128_13474 = torch.constant.int 128
%int2_13475 = torch.constant.int 2
%none_13476 = torch.constant.none
%none_13477 = torch.constant.none
%cpu_13478 = torch.constant.device "cpu"
%false_13479 = torch.constant.bool false
%15902 = torch.aten.arange.start_step %int0_13473, %int128_13474, %int2_13475, %none_13476, %none_13477, %cpu_13478, %false_13479 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_13480 = torch.constant.int 0
%int0_13481 = torch.constant.int 0
%int64_13482 = torch.constant.int 64
%int1_13483 = torch.constant.int 1
%15903 = torch.aten.slice.Tensor %15902, %int0_13480, %int0_13481, %int64_13482, %int1_13483 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_13484 = torch.constant.int 6
%15904 = torch.prims.convert_element_type %15903, %int6_13484 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_13485 = torch.constant.int 128
%15905 = torch.aten.div.Scalar %15904, %int128_13485 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_13486 = torch.constant.float 5.000000e+05
%15906 = torch.aten.pow.Scalar %float5.000000e05_13486, %15905 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%15907 = torch.aten.reciprocal %15906 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_13487 = torch.constant.float 1.000000e+00
%15908 = torch.aten.mul.Scalar %15907, %float1.000000e00_13487 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_13488 = torch.constant.int 131072
%int1_13489 = torch.constant.int 1
%15909 = torch.prim.ListConstruct %int131072_13488, %int1_13489 : (!torch.int, !torch.int) -> !torch.list<int>
%15910 = torch.aten.view %15901, %15909 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%15911 = torch.aten.mul.Tensor %15910, %15908 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%15912 = torch.aten.cos %15911 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%15913 = torch.aten.sin %15911 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%15914 = torch.aten.complex %15912, %15913 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%15915 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15916 = flow.tensor.transfer %15915 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%15917 = torch_c.from_builtin_tensor %15916 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15918 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15919 = flow.tensor.transfer %15918 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%15920 = torch_c.from_builtin_tensor %15919 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15921 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15922 = flow.tensor.transfer %15921 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%15923 = torch_c.from_builtin_tensor %15922 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15924 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15925 = flow.tensor.transfer %15924 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%15926 = torch_c.from_builtin_tensor %15925 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15927 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15928 = flow.tensor.transfer %15927 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%15929 = torch_c.from_builtin_tensor %15928 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15930 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15931 = flow.tensor.transfer %15930 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%15932 = torch_c.from_builtin_tensor %15931 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15933 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15934 = flow.tensor.transfer %15933 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%15935 = torch_c.from_builtin_tensor %15934 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%15936 = torch_c.to_builtin_tensor %15914 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%15937 = flow.tensor.transfer %15936 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%15938 = torch_c.from_builtin_tensor %15937 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_13490 = torch.constant.int 1
%15939 = torch.aten.size.int %15588, %int1_13490 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13491 = torch.constant.int 0
%15940 = torch.aten.add.int %int0_13491, %15939 : !torch.int, !torch.int -> !torch.int
%int0_13492 = torch.constant.int 0
%int0_13493 = torch.constant.int 0
%int1_13494 = torch.constant.int 1
%15941 = torch.aten.slice.Tensor %15917, %int0_13492, %int0_13493, %15940, %int1_13494 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15941, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13495 = torch.constant.int 1
%int0_13496 = torch.constant.int 0
%int9223372036854775807_13497 = torch.constant.int 9223372036854775807
%int1_13498 = torch.constant.int 1
%15942 = torch.aten.slice.Tensor %15941, %int1_13495, %int0_13496, %int9223372036854775807_13497, %int1_13498 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15942, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13499 = torch.constant.int 0
%15943 = torch.aten.unsqueeze %15942, %int0_13499 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15943, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13500 = torch.constant.int 2
%15944 = torch.aten.unsqueeze %15943, %int2_13500 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15944, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13501 = torch.constant.int 3
%int0_13502 = torch.constant.int 0
%int9223372036854775807_13503 = torch.constant.int 9223372036854775807
%int1_13504 = torch.constant.int 1
%15945 = torch.aten.slice.Tensor %15944, %int3_13501, %int0_13502, %int9223372036854775807_13503, %int1_13504 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15945, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15946 = torch_c.to_builtin_tensor %15712 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13505 = arith.constant 1 : index
%dim_13506 = tensor.dim %15946, %c1_13505 : tensor<4x?x1x128xf16>
%15947 = flow.tensor.bitcast %15946 : tensor<4x?x1x128xf16>{%dim_13506} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13506}
%15948 = torch_c.from_builtin_tensor %15947 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %15948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%15949 = torch.aten.mul.Tensor %15948, %15945 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%15950 = torch_c.to_builtin_tensor %15949 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13507 = arith.constant 1 : index
%dim_13508 = tensor.dim %15950, %c1_13507 : tensor<4x?x1x64xcomplex<f32>>
%15951 = flow.tensor.bitcast %15950 : tensor<4x?x1x64xcomplex<f32>>{%dim_13508} -> tensor<4x?x1x128xf32>{%dim_13508}
%15952 = torch_c.from_builtin_tensor %15951 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %15952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13509 = torch.constant.int 5
%15953 = torch.prims.convert_element_type %15952, %int5_13509 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13510 = torch.constant.int 1
%15954 = torch.aten.size.int %15594, %int1_13510 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13511 = torch.constant.int 0
%15955 = torch.aten.add.int %int0_13511, %15954 : !torch.int, !torch.int -> !torch.int
%int0_13512 = torch.constant.int 0
%int0_13513 = torch.constant.int 0
%int1_13514 = torch.constant.int 1
%15956 = torch.aten.slice.Tensor %15920, %int0_13512, %int0_13513, %15955, %int1_13514 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15956, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13515 = torch.constant.int 1
%int0_13516 = torch.constant.int 0
%int9223372036854775807_13517 = torch.constant.int 9223372036854775807
%int1_13518 = torch.constant.int 1
%15957 = torch.aten.slice.Tensor %15956, %int1_13515, %int0_13516, %int9223372036854775807_13517, %int1_13518 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15957, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13519 = torch.constant.int 0
%15958 = torch.aten.unsqueeze %15957, %int0_13519 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15958, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13520 = torch.constant.int 2
%15959 = torch.aten.unsqueeze %15958, %int2_13520 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15959, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13521 = torch.constant.int 3
%int0_13522 = torch.constant.int 0
%int9223372036854775807_13523 = torch.constant.int 9223372036854775807
%int1_13524 = torch.constant.int 1
%15960 = torch.aten.slice.Tensor %15959, %int3_13521, %int0_13522, %int9223372036854775807_13523, %int1_13524 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15960, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15961 = torch_c.to_builtin_tensor %15714 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13525 = arith.constant 1 : index
%dim_13526 = tensor.dim %15961, %c1_13525 : tensor<4x?x1x128xf16>
%15962 = flow.tensor.bitcast %15961 : tensor<4x?x1x128xf16>{%dim_13526} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13526}
%15963 = torch_c.from_builtin_tensor %15962 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %15963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%15964 = torch.aten.mul.Tensor %15963, %15960 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%15965 = torch_c.to_builtin_tensor %15964 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13527 = arith.constant 1 : index
%dim_13528 = tensor.dim %15965, %c1_13527 : tensor<4x?x1x64xcomplex<f32>>
%15966 = flow.tensor.bitcast %15965 : tensor<4x?x1x64xcomplex<f32>>{%dim_13528} -> tensor<4x?x1x128xf32>{%dim_13528}
%15967 = torch_c.from_builtin_tensor %15966 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %15967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13529 = torch.constant.int 5
%15968 = torch.prims.convert_element_type %15967, %int5_13529 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13530 = torch.constant.int 1
%15969 = torch.aten.size.int %15600, %int1_13530 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13531 = torch.constant.int 0
%15970 = torch.aten.add.int %int0_13531, %15969 : !torch.int, !torch.int -> !torch.int
%int0_13532 = torch.constant.int 0
%int0_13533 = torch.constant.int 0
%int1_13534 = torch.constant.int 1
%15971 = torch.aten.slice.Tensor %15923, %int0_13532, %int0_13533, %15970, %int1_13534 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15971, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13535 = torch.constant.int 1
%int0_13536 = torch.constant.int 0
%int9223372036854775807_13537 = torch.constant.int 9223372036854775807
%int1_13538 = torch.constant.int 1
%15972 = torch.aten.slice.Tensor %15971, %int1_13535, %int0_13536, %int9223372036854775807_13537, %int1_13538 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15972, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13539 = torch.constant.int 0
%15973 = torch.aten.unsqueeze %15972, %int0_13539 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15973, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13540 = torch.constant.int 2
%15974 = torch.aten.unsqueeze %15973, %int2_13540 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15974, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13541 = torch.constant.int 3
%int0_13542 = torch.constant.int 0
%int9223372036854775807_13543 = torch.constant.int 9223372036854775807
%int1_13544 = torch.constant.int 1
%15975 = torch.aten.slice.Tensor %15974, %int3_13541, %int0_13542, %int9223372036854775807_13543, %int1_13544 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15975, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15976 = torch_c.to_builtin_tensor %15716 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13545 = arith.constant 1 : index
%dim_13546 = tensor.dim %15976, %c1_13545 : tensor<4x?x1x128xf16>
%15977 = flow.tensor.bitcast %15976 : tensor<4x?x1x128xf16>{%dim_13546} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13546}
%15978 = torch_c.from_builtin_tensor %15977 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %15978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%15979 = torch.aten.mul.Tensor %15978, %15975 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%15980 = torch_c.to_builtin_tensor %15979 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13547 = arith.constant 1 : index
%dim_13548 = tensor.dim %15980, %c1_13547 : tensor<4x?x1x64xcomplex<f32>>
%15981 = flow.tensor.bitcast %15980 : tensor<4x?x1x64xcomplex<f32>>{%dim_13548} -> tensor<4x?x1x128xf32>{%dim_13548}
%15982 = torch_c.from_builtin_tensor %15981 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %15982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13549 = torch.constant.int 5
%15983 = torch.prims.convert_element_type %15982, %int5_13549 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13550 = torch.constant.int 1
%15984 = torch.aten.size.int %15606, %int1_13550 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13551 = torch.constant.int 0
%15985 = torch.aten.add.int %int0_13551, %15984 : !torch.int, !torch.int -> !torch.int
%int0_13552 = torch.constant.int 0
%int0_13553 = torch.constant.int 0
%int1_13554 = torch.constant.int 1
%15986 = torch.aten.slice.Tensor %15926, %int0_13552, %int0_13553, %15985, %int1_13554 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15986, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13555 = torch.constant.int 1
%int0_13556 = torch.constant.int 0
%int9223372036854775807_13557 = torch.constant.int 9223372036854775807
%int1_13558 = torch.constant.int 1
%15987 = torch.aten.slice.Tensor %15986, %int1_13555, %int0_13556, %int9223372036854775807_13557, %int1_13558 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %15987, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13559 = torch.constant.int 0
%15988 = torch.aten.unsqueeze %15987, %int0_13559 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %15988, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13560 = torch.constant.int 2
%15989 = torch.aten.unsqueeze %15988, %int2_13560 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15989, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13561 = torch.constant.int 3
%int0_13562 = torch.constant.int 0
%int9223372036854775807_13563 = torch.constant.int 9223372036854775807
%int1_13564 = torch.constant.int 1
%15990 = torch.aten.slice.Tensor %15989, %int3_13561, %int0_13562, %int9223372036854775807_13563, %int1_13564 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15990, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%15991 = torch_c.to_builtin_tensor %15718 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13565 = arith.constant 1 : index
%dim_13566 = tensor.dim %15991, %c1_13565 : tensor<4x?x1x128xf16>
%15992 = flow.tensor.bitcast %15991 : tensor<4x?x1x128xf16>{%dim_13566} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13566}
%15993 = torch_c.from_builtin_tensor %15992 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %15993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%15994 = torch.aten.mul.Tensor %15993, %15990 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %15994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%15995 = torch_c.to_builtin_tensor %15994 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13567 = arith.constant 1 : index
%dim_13568 = tensor.dim %15995, %c1_13567 : tensor<4x?x1x64xcomplex<f32>>
%15996 = flow.tensor.bitcast %15995 : tensor<4x?x1x64xcomplex<f32>>{%dim_13568} -> tensor<4x?x1x128xf32>{%dim_13568}
%15997 = torch_c.from_builtin_tensor %15996 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %15997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13569 = torch.constant.int 5
%15998 = torch.prims.convert_element_type %15997, %int5_13569 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %15998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13570 = torch.constant.int 1
%15999 = torch.aten.size.int %15612, %int1_13570 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13571 = torch.constant.int 0
%16000 = torch.aten.add.int %int0_13571, %15999 : !torch.int, !torch.int -> !torch.int
%int0_13572 = torch.constant.int 0
%int0_13573 = torch.constant.int 0
%int1_13574 = torch.constant.int 1
%16001 = torch.aten.slice.Tensor %15929, %int0_13572, %int0_13573, %16000, %int1_13574 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16001, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13575 = torch.constant.int 1
%int0_13576 = torch.constant.int 0
%int9223372036854775807_13577 = torch.constant.int 9223372036854775807
%int1_13578 = torch.constant.int 1
%16002 = torch.aten.slice.Tensor %16001, %int1_13575, %int0_13576, %int9223372036854775807_13577, %int1_13578 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16002, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13579 = torch.constant.int 0
%16003 = torch.aten.unsqueeze %16002, %int0_13579 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %16003, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13580 = torch.constant.int 2
%16004 = torch.aten.unsqueeze %16003, %int2_13580 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16004, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13581 = torch.constant.int 3
%int0_13582 = torch.constant.int 0
%int9223372036854775807_13583 = torch.constant.int 9223372036854775807
%int1_13584 = torch.constant.int 1
%16005 = torch.aten.slice.Tensor %16004, %int3_13581, %int0_13582, %int9223372036854775807_13583, %int1_13584 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16005, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%16006 = torch_c.to_builtin_tensor %15720 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13585 = arith.constant 1 : index
%dim_13586 = tensor.dim %16006, %c1_13585 : tensor<4x?x1x128xf16>
%16007 = flow.tensor.bitcast %16006 : tensor<4x?x1x128xf16>{%dim_13586} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13586}
%16008 = torch_c.from_builtin_tensor %16007 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %16008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%16009 = torch.aten.mul.Tensor %16008, %16005 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%16010 = torch_c.to_builtin_tensor %16009 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13587 = arith.constant 1 : index
%dim_13588 = tensor.dim %16010, %c1_13587 : tensor<4x?x1x64xcomplex<f32>>
%16011 = flow.tensor.bitcast %16010 : tensor<4x?x1x64xcomplex<f32>>{%dim_13588} -> tensor<4x?x1x128xf32>{%dim_13588}
%16012 = torch_c.from_builtin_tensor %16011 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %16012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13589 = torch.constant.int 5
%16013 = torch.prims.convert_element_type %16012, %int5_13589 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %16013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13590 = torch.constant.int 1
%16014 = torch.aten.size.int %15618, %int1_13590 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13591 = torch.constant.int 0
%16015 = torch.aten.add.int %int0_13591, %16014 : !torch.int, !torch.int -> !torch.int
%int0_13592 = torch.constant.int 0
%int0_13593 = torch.constant.int 0
%int1_13594 = torch.constant.int 1
%16016 = torch.aten.slice.Tensor %15932, %int0_13592, %int0_13593, %16015, %int1_13594 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16016, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13595 = torch.constant.int 1
%int0_13596 = torch.constant.int 0
%int9223372036854775807_13597 = torch.constant.int 9223372036854775807
%int1_13598 = torch.constant.int 1
%16017 = torch.aten.slice.Tensor %16016, %int1_13595, %int0_13596, %int9223372036854775807_13597, %int1_13598 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16017, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13599 = torch.constant.int 0
%16018 = torch.aten.unsqueeze %16017, %int0_13599 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %16018, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13600 = torch.constant.int 2
%16019 = torch.aten.unsqueeze %16018, %int2_13600 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16019, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13601 = torch.constant.int 3
%int0_13602 = torch.constant.int 0
%int9223372036854775807_13603 = torch.constant.int 9223372036854775807
%int1_13604 = torch.constant.int 1
%16020 = torch.aten.slice.Tensor %16019, %int3_13601, %int0_13602, %int9223372036854775807_13603, %int1_13604 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16020, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%16021 = torch_c.to_builtin_tensor %15722 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13605 = arith.constant 1 : index
%dim_13606 = tensor.dim %16021, %c1_13605 : tensor<4x?x1x128xf16>
%16022 = flow.tensor.bitcast %16021 : tensor<4x?x1x128xf16>{%dim_13606} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13606}
%16023 = torch_c.from_builtin_tensor %16022 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %16023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%16024 = torch.aten.mul.Tensor %16023, %16020 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%16025 = torch_c.to_builtin_tensor %16024 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13607 = arith.constant 1 : index
%dim_13608 = tensor.dim %16025, %c1_13607 : tensor<4x?x1x64xcomplex<f32>>
%16026 = flow.tensor.bitcast %16025 : tensor<4x?x1x64xcomplex<f32>>{%dim_13608} -> tensor<4x?x1x128xf32>{%dim_13608}
%16027 = torch_c.from_builtin_tensor %16026 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %16027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13609 = torch.constant.int 5
%16028 = torch.prims.convert_element_type %16027, %int5_13609 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %16028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13610 = torch.constant.int 1
%16029 = torch.aten.size.int %15624, %int1_13610 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13611 = torch.constant.int 0
%16030 = torch.aten.add.int %int0_13611, %16029 : !torch.int, !torch.int -> !torch.int
%int0_13612 = torch.constant.int 0
%int0_13613 = torch.constant.int 0
%int1_13614 = torch.constant.int 1
%16031 = torch.aten.slice.Tensor %15935, %int0_13612, %int0_13613, %16030, %int1_13614 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16031, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13615 = torch.constant.int 1
%int0_13616 = torch.constant.int 0
%int9223372036854775807_13617 = torch.constant.int 9223372036854775807
%int1_13618 = torch.constant.int 1
%16032 = torch.aten.slice.Tensor %16031, %int1_13615, %int0_13616, %int9223372036854775807_13617, %int1_13618 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16032, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13619 = torch.constant.int 0
%16033 = torch.aten.unsqueeze %16032, %int0_13619 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %16033, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13620 = torch.constant.int 2
%16034 = torch.aten.unsqueeze %16033, %int2_13620 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16034, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13621 = torch.constant.int 3
%int0_13622 = torch.constant.int 0
%int9223372036854775807_13623 = torch.constant.int 9223372036854775807
%int1_13624 = torch.constant.int 1
%16035 = torch.aten.slice.Tensor %16034, %int3_13621, %int0_13622, %int9223372036854775807_13623, %int1_13624 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16035, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%16036 = torch_c.to_builtin_tensor %15724 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13625 = arith.constant 1 : index
%dim_13626 = tensor.dim %16036, %c1_13625 : tensor<4x?x1x128xf16>
%16037 = flow.tensor.bitcast %16036 : tensor<4x?x1x128xf16>{%dim_13626} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13626}
%16038 = torch_c.from_builtin_tensor %16037 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %16038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%16039 = torch.aten.mul.Tensor %16038, %16035 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%16040 = torch_c.to_builtin_tensor %16039 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13627 = arith.constant 1 : index
%dim_13628 = tensor.dim %16040, %c1_13627 : tensor<4x?x1x64xcomplex<f32>>
%16041 = flow.tensor.bitcast %16040 : tensor<4x?x1x64xcomplex<f32>>{%dim_13628} -> tensor<4x?x1x128xf32>{%dim_13628}
%16042 = torch_c.from_builtin_tensor %16041 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %16042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13629 = torch.constant.int 5
%16043 = torch.prims.convert_element_type %16042, %int5_13629 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %16043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_13630 = torch.constant.int 1
%16044 = torch.aten.size.int %15630, %int1_13630 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_13631 = torch.constant.int 0
%16045 = torch.aten.add.int %int0_13631, %16044 : !torch.int, !torch.int -> !torch.int
%int0_13632 = torch.constant.int 0
%int0_13633 = torch.constant.int 0
%int1_13634 = torch.constant.int 1
%16046 = torch.aten.slice.Tensor %15938, %int0_13632, %int0_13633, %16045, %int1_13634 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16046, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_13635 = torch.constant.int 1
%int0_13636 = torch.constant.int 0
%int9223372036854775807_13637 = torch.constant.int 9223372036854775807
%int1_13638 = torch.constant.int 1
%16047 = torch.aten.slice.Tensor %16046, %int1_13635, %int0_13636, %int9223372036854775807_13637, %int1_13638 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %16047, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_13639 = torch.constant.int 0
%16048 = torch.aten.unsqueeze %16047, %int0_13639 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %16048, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_13640 = torch.constant.int 2
%16049 = torch.aten.unsqueeze %16048, %int2_13640 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16049, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_13641 = torch.constant.int 3
%int0_13642 = torch.constant.int 0
%int9223372036854775807_13643 = torch.constant.int 9223372036854775807
%int1_13644 = torch.constant.int 1
%16050 = torch.aten.slice.Tensor %16049, %int3_13641, %int0_13642, %int9223372036854775807_13643, %int1_13644 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16050, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%16051 = torch_c.to_builtin_tensor %15726 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_13645 = arith.constant 1 : index
%dim_13646 = tensor.dim %16051, %c1_13645 : tensor<4x?x1x128xf16>
%16052 = flow.tensor.bitcast %16051 : tensor<4x?x1x128xf16>{%dim_13646} -> tensor<4x?x1x64xcomplex<f16>>{%dim_13646}
%16053 = torch_c.from_builtin_tensor %16052 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %16053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%16054 = torch.aten.mul.Tensor %16053, %16050 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %16054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%16055 = torch_c.to_builtin_tensor %16054 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_13647 = arith.constant 1 : index
%dim_13648 = tensor.dim %16055, %c1_13647 : tensor<4x?x1x64xcomplex<f32>>
%16056 = flow.tensor.bitcast %16055 : tensor<4x?x1x64xcomplex<f32>>{%dim_13648} -> tensor<4x?x1x128xf32>{%dim_13648}
%16057 = torch_c.from_builtin_tensor %16056 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %16057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_13649 = torch.constant.int 5
%16058 = torch.prims.convert_element_type %16057, %int5_13649 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %16058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_13650 = torch.constant.int 64
%16059 = torch.aten.mul.Scalar %2364, %int64_13650 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16059, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13651 = torch.constant.int 64
%16060 = torch.aten.mul.Scalar %2367, %int64_13651 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16060, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13652 = torch.constant.int 64
%16061 = torch.aten.mul.Scalar %2370, %int64_13652 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16061, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13653 = torch.constant.int 64
%16062 = torch.aten.mul.Scalar %2373, %int64_13653 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16062, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13654 = torch.constant.int 64
%16063 = torch.aten.mul.Scalar %2376, %int64_13654 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16063, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13655 = torch.constant.int 64
%16064 = torch.aten.mul.Scalar %2379, %int64_13655 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16064, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13656 = torch.constant.int 64
%16065 = torch.aten.mul.Scalar %2382, %int64_13656 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16065, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_13657 = torch.constant.int 64
%16066 = torch.aten.mul.Scalar %2385, %int64_13657 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16066, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14 = torch.constant.int 14
%int1_13658 = torch.constant.int 1
%16067 = torch.aten.add.Scalar %16059, %int14, %int1_13658 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16067, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13659 = torch.constant.int 14
%int1_13660 = torch.constant.int 1
%16068 = torch.aten.add.Scalar %16060, %int14_13659, %int1_13660 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16068, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13661 = torch.constant.int 14
%int1_13662 = torch.constant.int 1
%16069 = torch.aten.add.Scalar %16061, %int14_13661, %int1_13662 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16069, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13663 = torch.constant.int 14
%int1_13664 = torch.constant.int 1
%16070 = torch.aten.add.Scalar %16062, %int14_13663, %int1_13664 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16070, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13665 = torch.constant.int 14
%int1_13666 = torch.constant.int 1
%16071 = torch.aten.add.Scalar %16063, %int14_13665, %int1_13666 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16071, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13667 = torch.constant.int 14
%int1_13668 = torch.constant.int 1
%16072 = torch.aten.add.Scalar %16064, %int14_13667, %int1_13668 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16072, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13669 = torch.constant.int 14
%int1_13670 = torch.constant.int 1
%16073 = torch.aten.add.Scalar %16065, %int14_13669, %int1_13670 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16073, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int14_13671 = torch.constant.int 14
%int1_13672 = torch.constant.int 1
%16074 = torch.aten.add.Scalar %16066, %int14_13671, %int1_13672 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16074, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_13673 = torch.constant.int 4
%int16_13674 = torch.constant.int 16
%int1_13675 = torch.constant.int 1
%int128_13676 = torch.constant.int 128
%16075 = torch.prim.ListConstruct %int4_13673, %3095, %int16_13674, %int1_13675, %int128_13676 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16076 = torch.aten.view %15953, %16075 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16076, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13677 = torch.constant.int 4
%int16_13678 = torch.constant.int 16
%int1_13679 = torch.constant.int 1
%int128_13680 = torch.constant.int 128
%16077 = torch.prim.ListConstruct %int4_13677, %3095, %int16_13678, %int1_13679, %int128_13680 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16078 = torch.aten.view %15968, %16077 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16078, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13681 = torch.constant.int 4
%int16_13682 = torch.constant.int 16
%int1_13683 = torch.constant.int 1
%int128_13684 = torch.constant.int 128
%16079 = torch.prim.ListConstruct %int4_13681, %3095, %int16_13682, %int1_13683, %int128_13684 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16080 = torch.aten.view %15983, %16079 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16080, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13685 = torch.constant.int 4
%int16_13686 = torch.constant.int 16
%int1_13687 = torch.constant.int 1
%int128_13688 = torch.constant.int 128
%16081 = torch.prim.ListConstruct %int4_13685, %3095, %int16_13686, %int1_13687, %int128_13688 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16082 = torch.aten.view %15998, %16081 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16082, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13689 = torch.constant.int 4
%int16_13690 = torch.constant.int 16
%int1_13691 = torch.constant.int 1
%int128_13692 = torch.constant.int 128
%16083 = torch.prim.ListConstruct %int4_13689, %3095, %int16_13690, %int1_13691, %int128_13692 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16084 = torch.aten.view %16013, %16083 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16084, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13693 = torch.constant.int 4
%int16_13694 = torch.constant.int 16
%int1_13695 = torch.constant.int 1
%int128_13696 = torch.constant.int 128
%16085 = torch.prim.ListConstruct %int4_13693, %3095, %int16_13694, %int1_13695, %int128_13696 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16086 = torch.aten.view %16028, %16085 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16086, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13697 = torch.constant.int 4
%int16_13698 = torch.constant.int 16
%int1_13699 = torch.constant.int 1
%int128_13700 = torch.constant.int 128
%16087 = torch.prim.ListConstruct %int4_13697, %3095, %int16_13698, %int1_13699, %int128_13700 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16088 = torch.aten.view %16043, %16087 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16088, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13701 = torch.constant.int 4
%int16_13702 = torch.constant.int 16
%int1_13703 = torch.constant.int 1
%int128_13704 = torch.constant.int 128
%16089 = torch.prim.ListConstruct %int4_13701, %3095, %int16_13702, %int1_13703, %int128_13704 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16090 = torch.aten.view %16058, %16089 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16090, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13705 = torch.constant.int 4
%16091 = torch.aten.mul.int %int4_13705, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13706 = torch.constant.int 16
%int1_13707 = torch.constant.int 1
%int128_13708 = torch.constant.int 128
%16092 = torch.prim.ListConstruct %16091, %int16_13706, %int1_13707, %int128_13708 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16093 = torch.aten.view %16076, %16092 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16093, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13709 = torch.constant.int 4
%16094 = torch.aten.mul.int %int4_13709, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13710 = torch.constant.int 16
%int1_13711 = torch.constant.int 1
%int128_13712 = torch.constant.int 128
%16095 = torch.prim.ListConstruct %16094, %int16_13710, %int1_13711, %int128_13712 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16096 = torch.aten.view %16078, %16095 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16096, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13713 = torch.constant.int 4
%16097 = torch.aten.mul.int %int4_13713, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13714 = torch.constant.int 16
%int1_13715 = torch.constant.int 1
%int128_13716 = torch.constant.int 128
%16098 = torch.prim.ListConstruct %16097, %int16_13714, %int1_13715, %int128_13716 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16099 = torch.aten.view %16080, %16098 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16099, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13717 = torch.constant.int 4
%16100 = torch.aten.mul.int %int4_13717, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13718 = torch.constant.int 16
%int1_13719 = torch.constant.int 1
%int128_13720 = torch.constant.int 128
%16101 = torch.prim.ListConstruct %16100, %int16_13718, %int1_13719, %int128_13720 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16102 = torch.aten.view %16082, %16101 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16102, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13721 = torch.constant.int 4
%16103 = torch.aten.mul.int %int4_13721, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13722 = torch.constant.int 16
%int1_13723 = torch.constant.int 1
%int128_13724 = torch.constant.int 128
%16104 = torch.prim.ListConstruct %16103, %int16_13722, %int1_13723, %int128_13724 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16105 = torch.aten.view %16084, %16104 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16105, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13725 = torch.constant.int 4
%16106 = torch.aten.mul.int %int4_13725, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13726 = torch.constant.int 16
%int1_13727 = torch.constant.int 1
%int128_13728 = torch.constant.int 128
%16107 = torch.prim.ListConstruct %16106, %int16_13726, %int1_13727, %int128_13728 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16108 = torch.aten.view %16086, %16107 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16108, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13729 = torch.constant.int 4
%16109 = torch.aten.mul.int %int4_13729, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13730 = torch.constant.int 16
%int1_13731 = torch.constant.int 1
%int128_13732 = torch.constant.int 128
%16110 = torch.prim.ListConstruct %16109, %int16_13730, %int1_13731, %int128_13732 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16111 = torch.aten.view %16088, %16110 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16111, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13733 = torch.constant.int 4
%16112 = torch.aten.mul.int %int4_13733, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13734 = torch.constant.int 16
%int1_13735 = torch.constant.int 1
%int128_13736 = torch.constant.int 128
%16113 = torch.prim.ListConstruct %16112, %int16_13734, %int1_13735, %int128_13736 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16114 = torch.aten.view %16090, %16113 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16114, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13737 = torch.constant.int 4
%16115 = torch.aten.mul.int %int4_13737, %3095 : !torch.int, !torch.int -> !torch.int
%16116 = torch.prim.ListConstruct %16115 : (!torch.int) -> !torch.list<int>
%16117 = torch.aten.view %16067, %16116 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16117, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13738 = torch.constant.int 4
%16118 = torch.aten.mul.int %int4_13738, %3095 : !torch.int, !torch.int -> !torch.int
%16119 = torch.prim.ListConstruct %16118 : (!torch.int) -> !torch.list<int>
%16120 = torch.aten.view %16068, %16119 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16120, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13739 = torch.constant.int 4
%16121 = torch.aten.mul.int %int4_13739, %3095 : !torch.int, !torch.int -> !torch.int
%16122 = torch.prim.ListConstruct %16121 : (!torch.int) -> !torch.list<int>
%16123 = torch.aten.view %16069, %16122 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16123, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13740 = torch.constant.int 4
%16124 = torch.aten.mul.int %int4_13740, %3095 : !torch.int, !torch.int -> !torch.int
%16125 = torch.prim.ListConstruct %16124 : (!torch.int) -> !torch.list<int>
%16126 = torch.aten.view %16070, %16125 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16126, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13741 = torch.constant.int 4
%16127 = torch.aten.mul.int %int4_13741, %3095 : !torch.int, !torch.int -> !torch.int
%16128 = torch.prim.ListConstruct %16127 : (!torch.int) -> !torch.list<int>
%16129 = torch.aten.view %16071, %16128 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16129, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13742 = torch.constant.int 4
%16130 = torch.aten.mul.int %int4_13742, %3095 : !torch.int, !torch.int -> !torch.int
%16131 = torch.prim.ListConstruct %16130 : (!torch.int) -> !torch.list<int>
%16132 = torch.aten.view %16072, %16131 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16132, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13743 = torch.constant.int 4
%16133 = torch.aten.mul.int %int4_13743, %3095 : !torch.int, !torch.int -> !torch.int
%16134 = torch.prim.ListConstruct %16133 : (!torch.int) -> !torch.list<int>
%16135 = torch.aten.view %16073, %16134 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16135, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13744 = torch.constant.int 4
%16136 = torch.aten.mul.int %int4_13744, %3095 : !torch.int, !torch.int -> !torch.int
%16137 = torch.prim.ListConstruct %16136 : (!torch.int) -> !torch.list<int>
%16138 = torch.aten.view %16074, %16137 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16138, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13745 = torch.constant.int 4
%int16_13746 = torch.constant.int 16
%int1_13747 = torch.constant.int 1
%int128_13748 = torch.constant.int 128
%16139 = torch.prim.ListConstruct %int4_13745, %3095, %int16_13746, %int1_13747, %int128_13748 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16140 = torch.aten.view %15728, %16139 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16140, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13749 = torch.constant.int 4
%int16_13750 = torch.constant.int 16
%int1_13751 = torch.constant.int 1
%int128_13752 = torch.constant.int 128
%16141 = torch.prim.ListConstruct %int4_13749, %3095, %int16_13750, %int1_13751, %int128_13752 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16142 = torch.aten.view %15730, %16141 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16142, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13753 = torch.constant.int 4
%int16_13754 = torch.constant.int 16
%int1_13755 = torch.constant.int 1
%int128_13756 = torch.constant.int 128
%16143 = torch.prim.ListConstruct %int4_13753, %3095, %int16_13754, %int1_13755, %int128_13756 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16144 = torch.aten.view %15732, %16143 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16144, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13757 = torch.constant.int 4
%int16_13758 = torch.constant.int 16
%int1_13759 = torch.constant.int 1
%int128_13760 = torch.constant.int 128
%16145 = torch.prim.ListConstruct %int4_13757, %3095, %int16_13758, %int1_13759, %int128_13760 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16146 = torch.aten.view %15734, %16145 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16146, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13761 = torch.constant.int 4
%int16_13762 = torch.constant.int 16
%int1_13763 = torch.constant.int 1
%int128_13764 = torch.constant.int 128
%16147 = torch.prim.ListConstruct %int4_13761, %3095, %int16_13762, %int1_13763, %int128_13764 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16148 = torch.aten.view %15736, %16147 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16148, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13765 = torch.constant.int 4
%int16_13766 = torch.constant.int 16
%int1_13767 = torch.constant.int 1
%int128_13768 = torch.constant.int 128
%16149 = torch.prim.ListConstruct %int4_13765, %3095, %int16_13766, %int1_13767, %int128_13768 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16150 = torch.aten.view %15738, %16149 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16150, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13769 = torch.constant.int 4
%int16_13770 = torch.constant.int 16
%int1_13771 = torch.constant.int 1
%int128_13772 = torch.constant.int 128
%16151 = torch.prim.ListConstruct %int4_13769, %3095, %int16_13770, %int1_13771, %int128_13772 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16152 = torch.aten.view %15740, %16151 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16152, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13773 = torch.constant.int 4
%int16_13774 = torch.constant.int 16
%int1_13775 = torch.constant.int 1
%int128_13776 = torch.constant.int 128
%16153 = torch.prim.ListConstruct %int4_13773, %3095, %int16_13774, %int1_13775, %int128_13776 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16154 = torch.aten.view %15742, %16153 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %16154, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_13777 = torch.constant.int 4
%16155 = torch.aten.mul.int %int4_13777, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13778 = torch.constant.int 16
%int1_13779 = torch.constant.int 1
%int128_13780 = torch.constant.int 128
%16156 = torch.prim.ListConstruct %16155, %int16_13778, %int1_13779, %int128_13780 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16157 = torch.aten.view %16140, %16156 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16157, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13781 = torch.constant.int 4
%16158 = torch.aten.mul.int %int4_13781, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13782 = torch.constant.int 16
%int1_13783 = torch.constant.int 1
%int128_13784 = torch.constant.int 128
%16159 = torch.prim.ListConstruct %16158, %int16_13782, %int1_13783, %int128_13784 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16160 = torch.aten.view %16142, %16159 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16160, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13785 = torch.constant.int 4
%16161 = torch.aten.mul.int %int4_13785, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13786 = torch.constant.int 16
%int1_13787 = torch.constant.int 1
%int128_13788 = torch.constant.int 128
%16162 = torch.prim.ListConstruct %16161, %int16_13786, %int1_13787, %int128_13788 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16163 = torch.aten.view %16144, %16162 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16163, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13789 = torch.constant.int 4
%16164 = torch.aten.mul.int %int4_13789, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13790 = torch.constant.int 16
%int1_13791 = torch.constant.int 1
%int128_13792 = torch.constant.int 128
%16165 = torch.prim.ListConstruct %16164, %int16_13790, %int1_13791, %int128_13792 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16166 = torch.aten.view %16146, %16165 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16166, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13793 = torch.constant.int 4
%16167 = torch.aten.mul.int %int4_13793, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13794 = torch.constant.int 16
%int1_13795 = torch.constant.int 1
%int128_13796 = torch.constant.int 128
%16168 = torch.prim.ListConstruct %16167, %int16_13794, %int1_13795, %int128_13796 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16169 = torch.aten.view %16148, %16168 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16169, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13797 = torch.constant.int 4
%16170 = torch.aten.mul.int %int4_13797, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13798 = torch.constant.int 16
%int1_13799 = torch.constant.int 1
%int128_13800 = torch.constant.int 128
%16171 = torch.prim.ListConstruct %16170, %int16_13798, %int1_13799, %int128_13800 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16172 = torch.aten.view %16150, %16171 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16172, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13801 = torch.constant.int 4
%16173 = torch.aten.mul.int %int4_13801, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13802 = torch.constant.int 16
%int1_13803 = torch.constant.int 1
%int128_13804 = torch.constant.int 128
%16174 = torch.prim.ListConstruct %16173, %int16_13802, %int1_13803, %int128_13804 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16175 = torch.aten.view %16152, %16174 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16175, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_13805 = torch.constant.int 4
%16176 = torch.aten.mul.int %int4_13805, %3095 : !torch.int, !torch.int -> !torch.int
%int16_13806 = torch.constant.int 16
%int1_13807 = torch.constant.int 1
%int128_13808 = torch.constant.int 128
%16177 = torch.prim.ListConstruct %16176, %int16_13806, %int1_13807, %int128_13808 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16178 = torch.aten.view %16154, %16177 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16178, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_13809 = torch.constant.int 1
%int1_13810 = torch.constant.int 1
%16179 = torch.aten.add.Scalar %16067, %int1_13809, %int1_13810 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16179, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13811 = torch.constant.int 1
%int1_13812 = torch.constant.int 1
%16180 = torch.aten.add.Scalar %16068, %int1_13811, %int1_13812 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16180, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13813 = torch.constant.int 1
%int1_13814 = torch.constant.int 1
%16181 = torch.aten.add.Scalar %16069, %int1_13813, %int1_13814 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16181, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13815 = torch.constant.int 1
%int1_13816 = torch.constant.int 1
%16182 = torch.aten.add.Scalar %16070, %int1_13815, %int1_13816 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16182, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13817 = torch.constant.int 1
%int1_13818 = torch.constant.int 1
%16183 = torch.aten.add.Scalar %16071, %int1_13817, %int1_13818 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16183, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13819 = torch.constant.int 1
%int1_13820 = torch.constant.int 1
%16184 = torch.aten.add.Scalar %16072, %int1_13819, %int1_13820 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16184, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13821 = torch.constant.int 1
%int1_13822 = torch.constant.int 1
%16185 = torch.aten.add.Scalar %16073, %int1_13821, %int1_13822 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16185, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_13823 = torch.constant.int 1
%int1_13824 = torch.constant.int 1
%16186 = torch.aten.add.Scalar %16074, %int1_13823, %int1_13824 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %16186, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_13825 = torch.constant.int 4
%16187 = torch.aten.mul.int %int4_13825, %3095 : !torch.int, !torch.int -> !torch.int
%16188 = torch.prim.ListConstruct %16187 : (!torch.int) -> !torch.list<int>
%16189 = torch.aten.view %16179, %16188 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16189, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13826 = torch.constant.int 4
%16190 = torch.aten.mul.int %int4_13826, %3095 : !torch.int, !torch.int -> !torch.int
%16191 = torch.prim.ListConstruct %16190 : (!torch.int) -> !torch.list<int>
%16192 = torch.aten.view %16180, %16191 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16192, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13827 = torch.constant.int 4
%16193 = torch.aten.mul.int %int4_13827, %3095 : !torch.int, !torch.int -> !torch.int
%16194 = torch.prim.ListConstruct %16193 : (!torch.int) -> !torch.list<int>
%16195 = torch.aten.view %16181, %16194 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16195, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13828 = torch.constant.int 4
%16196 = torch.aten.mul.int %int4_13828, %3095 : !torch.int, !torch.int -> !torch.int
%16197 = torch.prim.ListConstruct %16196 : (!torch.int) -> !torch.list<int>
%16198 = torch.aten.view %16182, %16197 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16198, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13829 = torch.constant.int 4
%16199 = torch.aten.mul.int %int4_13829, %3095 : !torch.int, !torch.int -> !torch.int
%16200 = torch.prim.ListConstruct %16199 : (!torch.int) -> !torch.list<int>
%16201 = torch.aten.view %16183, %16200 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16201, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13830 = torch.constant.int 4
%16202 = torch.aten.mul.int %int4_13830, %3095 : !torch.int, !torch.int -> !torch.int
%16203 = torch.prim.ListConstruct %16202 : (!torch.int) -> !torch.list<int>
%16204 = torch.aten.view %16184, %16203 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16204, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13831 = torch.constant.int 4
%16205 = torch.aten.mul.int %int4_13831, %3095 : !torch.int, !torch.int -> !torch.int
%16206 = torch.prim.ListConstruct %16205 : (!torch.int) -> !torch.list<int>
%16207 = torch.aten.view %16185, %16206 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16207, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_13832 = torch.constant.int 4
%16208 = torch.aten.mul.int %int4_13832, %3095 : !torch.int, !torch.int -> !torch.int
%16209 = torch.prim.ListConstruct %16208 : (!torch.int) -> !torch.list<int>
%16210 = torch.aten.view %16186, %16209 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16210, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%16211 = torch.prim.ListConstruct %16117, %16189 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13833 = torch.constant.int 0
%16212 = torch.aten.cat %16211, %int0_13833 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16212, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16213 = torch.prim.ListConstruct %16120, %16192 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13834 = torch.constant.int 0
%16214 = torch.aten.cat %16213, %int0_13834 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16214, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16215 = torch.prim.ListConstruct %16123, %16195 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13835 = torch.constant.int 0
%16216 = torch.aten.cat %16215, %int0_13835 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16216, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16217 = torch.prim.ListConstruct %16126, %16198 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13836 = torch.constant.int 0
%16218 = torch.aten.cat %16217, %int0_13836 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16218, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16219 = torch.prim.ListConstruct %16129, %16201 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13837 = torch.constant.int 0
%16220 = torch.aten.cat %16219, %int0_13837 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16220, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16221 = torch.prim.ListConstruct %16132, %16204 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13838 = torch.constant.int 0
%16222 = torch.aten.cat %16221, %int0_13838 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16222, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16223 = torch.prim.ListConstruct %16135, %16207 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13839 = torch.constant.int 0
%16224 = torch.aten.cat %16223, %int0_13839 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16224, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16225 = torch.prim.ListConstruct %16138, %16210 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_13840 = torch.constant.int 0
%16226 = torch.aten.cat %16225, %int0_13840 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %16226, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%16227 = torch.prim.ListConstruct %16093, %16157 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13841 = torch.constant.int 0
%16228 = torch.aten.cat %16227, %int0_13841 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16228, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16229 = torch.prim.ListConstruct %16096, %16160 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13842 = torch.constant.int 0
%16230 = torch.aten.cat %16229, %int0_13842 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16230, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16231 = torch.prim.ListConstruct %16099, %16163 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13843 = torch.constant.int 0
%16232 = torch.aten.cat %16231, %int0_13843 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16232, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16233 = torch.prim.ListConstruct %16102, %16166 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13844 = torch.constant.int 0
%16234 = torch.aten.cat %16233, %int0_13844 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16234, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16235 = torch.prim.ListConstruct %16105, %16169 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13845 = torch.constant.int 0
%16236 = torch.aten.cat %16235, %int0_13845 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16236, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16237 = torch.prim.ListConstruct %16108, %16172 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13846 = torch.constant.int 0
%16238 = torch.aten.cat %16237, %int0_13846 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16238, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16239 = torch.prim.ListConstruct %16111, %16175 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13847 = torch.constant.int 0
%16240 = torch.aten.cat %16239, %int0_13847 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16240, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16241 = torch.prim.ListConstruct %16114, %16178 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_13848 = torch.constant.int 0
%16242 = torch.aten.cat %16241, %int0_13848 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16242, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13849 = torch.constant.int 32
%int2_13850 = torch.constant.int 2
%int16_13851 = torch.constant.int 16
%int1_13852 = torch.constant.int 1
%int128_13853 = torch.constant.int 128
%16243 = torch.prim.ListConstruct %3023, %int32_13849, %int2_13850, %int16_13851, %int1_13852, %int128_13853 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16244 = torch.aten.view %14393, %16243 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16244, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13854 = torch.constant.int 32
%16245 = torch.aten.mul.int %3023, %int32_13854 : !torch.int, !torch.int -> !torch.int
%int2_13855 = torch.constant.int 2
%16246 = torch.aten.mul.int %16245, %int2_13855 : !torch.int, !torch.int -> !torch.int
%int16_13856 = torch.constant.int 16
%int1_13857 = torch.constant.int 1
%int128_13858 = torch.constant.int 128
%16247 = torch.prim.ListConstruct %16246, %int16_13856, %int1_13857, %int128_13858 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16248 = torch.aten.view %16244, %16247 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16248, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16249 = torch.prim.ListConstruct %16212 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13859 = torch.constant.bool false
%16250 = torch.aten.index_put %16248, %16249, %16228, %false_13859 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16250, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13860 = torch.constant.int 32
%int2_13861 = torch.constant.int 2
%int16_13862 = torch.constant.int 16
%int1_13863 = torch.constant.int 1
%int128_13864 = torch.constant.int 128
%16251 = torch.prim.ListConstruct %3023, %int32_13860, %int2_13861, %int16_13862, %int1_13863, %int128_13864 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16252 = torch.aten.view %16250, %16251 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16252, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13865 = torch.constant.int 131072
%16253 = torch.prim.ListConstruct %3023, %int131072_13865 : (!torch.int, !torch.int) -> !torch.list<int>
%16254 = torch.aten.view %16252, %16253 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16254, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13866 = torch.constant.int 32
%int2_13867 = torch.constant.int 2
%int16_13868 = torch.constant.int 16
%int1_13869 = torch.constant.int 1
%int128_13870 = torch.constant.int 128
%16255 = torch.prim.ListConstruct %3026, %int32_13866, %int2_13867, %int16_13868, %int1_13869, %int128_13870 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16256 = torch.aten.view %14405, %16255 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16256, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13871 = torch.constant.int 32
%16257 = torch.aten.mul.int %3026, %int32_13871 : !torch.int, !torch.int -> !torch.int
%int2_13872 = torch.constant.int 2
%16258 = torch.aten.mul.int %16257, %int2_13872 : !torch.int, !torch.int -> !torch.int
%int16_13873 = torch.constant.int 16
%int1_13874 = torch.constant.int 1
%int128_13875 = torch.constant.int 128
%16259 = torch.prim.ListConstruct %16258, %int16_13873, %int1_13874, %int128_13875 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16260 = torch.aten.view %16256, %16259 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16260, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16261 = torch.prim.ListConstruct %16214 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13876 = torch.constant.bool false
%16262 = torch.aten.index_put %16260, %16261, %16230, %false_13876 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16262, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13877 = torch.constant.int 32
%int2_13878 = torch.constant.int 2
%int16_13879 = torch.constant.int 16
%int1_13880 = torch.constant.int 1
%int128_13881 = torch.constant.int 128
%16263 = torch.prim.ListConstruct %3026, %int32_13877, %int2_13878, %int16_13879, %int1_13880, %int128_13881 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16264 = torch.aten.view %16262, %16263 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16264, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13882 = torch.constant.int 131072
%16265 = torch.prim.ListConstruct %3026, %int131072_13882 : (!torch.int, !torch.int) -> !torch.list<int>
%16266 = torch.aten.view %16264, %16265 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16266, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13883 = torch.constant.int 32
%int2_13884 = torch.constant.int 2
%int16_13885 = torch.constant.int 16
%int1_13886 = torch.constant.int 1
%int128_13887 = torch.constant.int 128
%16267 = torch.prim.ListConstruct %3029, %int32_13883, %int2_13884, %int16_13885, %int1_13886, %int128_13887 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16268 = torch.aten.view %14417, %16267 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16268, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13888 = torch.constant.int 32
%16269 = torch.aten.mul.int %3029, %int32_13888 : !torch.int, !torch.int -> !torch.int
%int2_13889 = torch.constant.int 2
%16270 = torch.aten.mul.int %16269, %int2_13889 : !torch.int, !torch.int -> !torch.int
%int16_13890 = torch.constant.int 16
%int1_13891 = torch.constant.int 1
%int128_13892 = torch.constant.int 128
%16271 = torch.prim.ListConstruct %16270, %int16_13890, %int1_13891, %int128_13892 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16272 = torch.aten.view %16268, %16271 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16272, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16273 = torch.prim.ListConstruct %16216 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13893 = torch.constant.bool false
%16274 = torch.aten.index_put %16272, %16273, %16232, %false_13893 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16274, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13894 = torch.constant.int 32
%int2_13895 = torch.constant.int 2
%int16_13896 = torch.constant.int 16
%int1_13897 = torch.constant.int 1
%int128_13898 = torch.constant.int 128
%16275 = torch.prim.ListConstruct %3029, %int32_13894, %int2_13895, %int16_13896, %int1_13897, %int128_13898 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16276 = torch.aten.view %16274, %16275 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16276, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13899 = torch.constant.int 131072
%16277 = torch.prim.ListConstruct %3029, %int131072_13899 : (!torch.int, !torch.int) -> !torch.list<int>
%16278 = torch.aten.view %16276, %16277 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16278, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13900 = torch.constant.int 32
%int2_13901 = torch.constant.int 2
%int16_13902 = torch.constant.int 16
%int1_13903 = torch.constant.int 1
%int128_13904 = torch.constant.int 128
%16279 = torch.prim.ListConstruct %3032, %int32_13900, %int2_13901, %int16_13902, %int1_13903, %int128_13904 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16280 = torch.aten.view %14429, %16279 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16280, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13905 = torch.constant.int 32
%16281 = torch.aten.mul.int %3032, %int32_13905 : !torch.int, !torch.int -> !torch.int
%int2_13906 = torch.constant.int 2
%16282 = torch.aten.mul.int %16281, %int2_13906 : !torch.int, !torch.int -> !torch.int
%int16_13907 = torch.constant.int 16
%int1_13908 = torch.constant.int 1
%int128_13909 = torch.constant.int 128
%16283 = torch.prim.ListConstruct %16282, %int16_13907, %int1_13908, %int128_13909 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16284 = torch.aten.view %16280, %16283 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16284, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16285 = torch.prim.ListConstruct %16218 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13910 = torch.constant.bool false
%16286 = torch.aten.index_put %16284, %16285, %16234, %false_13910 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16286, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13911 = torch.constant.int 32
%int2_13912 = torch.constant.int 2
%int16_13913 = torch.constant.int 16
%int1_13914 = torch.constant.int 1
%int128_13915 = torch.constant.int 128
%16287 = torch.prim.ListConstruct %3032, %int32_13911, %int2_13912, %int16_13913, %int1_13914, %int128_13915 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16288 = torch.aten.view %16286, %16287 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16288, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13916 = torch.constant.int 131072
%16289 = torch.prim.ListConstruct %3032, %int131072_13916 : (!torch.int, !torch.int) -> !torch.list<int>
%16290 = torch.aten.view %16288, %16289 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16290, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13917 = torch.constant.int 32
%int2_13918 = torch.constant.int 2
%int16_13919 = torch.constant.int 16
%int1_13920 = torch.constant.int 1
%int128_13921 = torch.constant.int 128
%16291 = torch.prim.ListConstruct %3035, %int32_13917, %int2_13918, %int16_13919, %int1_13920, %int128_13921 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16292 = torch.aten.view %14441, %16291 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16292, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13922 = torch.constant.int 32
%16293 = torch.aten.mul.int %3035, %int32_13922 : !torch.int, !torch.int -> !torch.int
%int2_13923 = torch.constant.int 2
%16294 = torch.aten.mul.int %16293, %int2_13923 : !torch.int, !torch.int -> !torch.int
%int16_13924 = torch.constant.int 16
%int1_13925 = torch.constant.int 1
%int128_13926 = torch.constant.int 128
%16295 = torch.prim.ListConstruct %16294, %int16_13924, %int1_13925, %int128_13926 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16296 = torch.aten.view %16292, %16295 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16296, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16297 = torch.prim.ListConstruct %16220 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13927 = torch.constant.bool false
%16298 = torch.aten.index_put %16296, %16297, %16236, %false_13927 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16298, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13928 = torch.constant.int 32
%int2_13929 = torch.constant.int 2
%int16_13930 = torch.constant.int 16
%int1_13931 = torch.constant.int 1
%int128_13932 = torch.constant.int 128
%16299 = torch.prim.ListConstruct %3035, %int32_13928, %int2_13929, %int16_13930, %int1_13931, %int128_13932 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16300 = torch.aten.view %16298, %16299 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16300, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13933 = torch.constant.int 131072
%16301 = torch.prim.ListConstruct %3035, %int131072_13933 : (!torch.int, !torch.int) -> !torch.list<int>
%16302 = torch.aten.view %16300, %16301 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16302, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13934 = torch.constant.int 32
%int2_13935 = torch.constant.int 2
%int16_13936 = torch.constant.int 16
%int1_13937 = torch.constant.int 1
%int128_13938 = torch.constant.int 128
%16303 = torch.prim.ListConstruct %3038, %int32_13934, %int2_13935, %int16_13936, %int1_13937, %int128_13938 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16304 = torch.aten.view %14453, %16303 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16304, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13939 = torch.constant.int 32
%16305 = torch.aten.mul.int %3038, %int32_13939 : !torch.int, !torch.int -> !torch.int
%int2_13940 = torch.constant.int 2
%16306 = torch.aten.mul.int %16305, %int2_13940 : !torch.int, !torch.int -> !torch.int
%int16_13941 = torch.constant.int 16
%int1_13942 = torch.constant.int 1
%int128_13943 = torch.constant.int 128
%16307 = torch.prim.ListConstruct %16306, %int16_13941, %int1_13942, %int128_13943 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16308 = torch.aten.view %16304, %16307 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16308, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16309 = torch.prim.ListConstruct %16222 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13944 = torch.constant.bool false
%16310 = torch.aten.index_put %16308, %16309, %16238, %false_13944 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16310, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13945 = torch.constant.int 32
%int2_13946 = torch.constant.int 2
%int16_13947 = torch.constant.int 16
%int1_13948 = torch.constant.int 1
%int128_13949 = torch.constant.int 128
%16311 = torch.prim.ListConstruct %3038, %int32_13945, %int2_13946, %int16_13947, %int1_13948, %int128_13949 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16312 = torch.aten.view %16310, %16311 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16312, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13950 = torch.constant.int 131072
%16313 = torch.prim.ListConstruct %3038, %int131072_13950 : (!torch.int, !torch.int) -> !torch.list<int>
%16314 = torch.aten.view %16312, %16313 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16314, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13951 = torch.constant.int 32
%int2_13952 = torch.constant.int 2
%int16_13953 = torch.constant.int 16
%int1_13954 = torch.constant.int 1
%int128_13955 = torch.constant.int 128
%16315 = torch.prim.ListConstruct %3041, %int32_13951, %int2_13952, %int16_13953, %int1_13954, %int128_13955 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16316 = torch.aten.view %14465, %16315 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16316, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13956 = torch.constant.int 32
%16317 = torch.aten.mul.int %3041, %int32_13956 : !torch.int, !torch.int -> !torch.int
%int2_13957 = torch.constant.int 2
%16318 = torch.aten.mul.int %16317, %int2_13957 : !torch.int, !torch.int -> !torch.int
%int16_13958 = torch.constant.int 16
%int1_13959 = torch.constant.int 1
%int128_13960 = torch.constant.int 128
%16319 = torch.prim.ListConstruct %16318, %int16_13958, %int1_13959, %int128_13960 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16320 = torch.aten.view %16316, %16319 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16320, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16321 = torch.prim.ListConstruct %16224 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13961 = torch.constant.bool false
%16322 = torch.aten.index_put %16320, %16321, %16240, %false_13961 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16322, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13962 = torch.constant.int 32
%int2_13963 = torch.constant.int 2
%int16_13964 = torch.constant.int 16
%int1_13965 = torch.constant.int 1
%int128_13966 = torch.constant.int 128
%16323 = torch.prim.ListConstruct %3041, %int32_13962, %int2_13963, %int16_13964, %int1_13965, %int128_13966 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16324 = torch.aten.view %16322, %16323 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16324, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13967 = torch.constant.int 131072
%16325 = torch.prim.ListConstruct %3041, %int131072_13967 : (!torch.int, !torch.int) -> !torch.list<int>
%16326 = torch.aten.view %16324, %16325 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16326, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_13968 = torch.constant.int 32
%int2_13969 = torch.constant.int 2
%int16_13970 = torch.constant.int 16
%int1_13971 = torch.constant.int 1
%int128_13972 = torch.constant.int 128
%16327 = torch.prim.ListConstruct %3044, %int32_13968, %int2_13969, %int16_13970, %int1_13971, %int128_13972 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16328 = torch.aten.view %14477, %16327 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16328, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_13973 = torch.constant.int 32
%16329 = torch.aten.mul.int %3044, %int32_13973 : !torch.int, !torch.int -> !torch.int
%int2_13974 = torch.constant.int 2
%16330 = torch.aten.mul.int %16329, %int2_13974 : !torch.int, !torch.int -> !torch.int
%int16_13975 = torch.constant.int 16
%int1_13976 = torch.constant.int 1
%int128_13977 = torch.constant.int 128
%16331 = torch.prim.ListConstruct %16330, %int16_13975, %int1_13976, %int128_13977 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16332 = torch.aten.view %16328, %16331 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16332, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%16333 = torch.prim.ListConstruct %16226 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_13978 = torch.constant.bool false
%16334 = torch.aten.index_put %16332, %16333, %16242, %false_13978 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %16334, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_13979 = torch.constant.int 32
%int2_13980 = torch.constant.int 2
%int16_13981 = torch.constant.int 16
%int1_13982 = torch.constant.int 1
%int128_13983 = torch.constant.int 128
%16335 = torch.prim.ListConstruct %3044, %int32_13979, %int2_13980, %int16_13981, %int1_13982, %int128_13983 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16336 = torch.aten.view %16334, %16335 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %16336, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_13984 = torch.constant.int 131072
%16337 = torch.prim.ListConstruct %3044, %int131072_13984 : (!torch.int, !torch.int) -> !torch.list<int>
%16338 = torch.aten.view %16336, %16337 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %16338, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_13985 = torch.constant.int -2
%16339 = torch.aten.unsqueeze %15953, %int-2_13985 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13986 = torch.constant.int -2
%16340 = torch.aten.unsqueeze %15968, %int-2_13986 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13987 = torch.constant.int -2
%16341 = torch.aten.unsqueeze %15983, %int-2_13987 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13988 = torch.constant.int -2
%16342 = torch.aten.unsqueeze %15998, %int-2_13988 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13989 = torch.constant.int -2
%16343 = torch.aten.unsqueeze %16013, %int-2_13989 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13990 = torch.constant.int -2
%16344 = torch.aten.unsqueeze %16028, %int-2_13990 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13991 = torch.constant.int -2
%16345 = torch.aten.unsqueeze %16043, %int-2_13991 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_13992 = torch.constant.int -2
%16346 = torch.aten.unsqueeze %16058, %int-2_13992 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_13993 = torch.constant.int 4
%int1_13994 = torch.constant.int 1
%int4_13995 = torch.constant.int 4
%int128_13996 = torch.constant.int 128
%16347 = torch.prim.ListConstruct %int4_13993, %15939, %int1_13994, %int4_13995, %int128_13996 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_13997 = torch.constant.bool false
%16348 = torch.aten.expand %16339, %16347, %false_13997 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_13998 = torch.constant.int 4
%int1_13999 = torch.constant.int 1
%int4_14000 = torch.constant.int 4
%int128_14001 = torch.constant.int 128
%16349 = torch.prim.ListConstruct %int4_13998, %15939, %int1_13999, %int4_14000, %int128_14001 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14002 = torch.constant.bool false
%16350 = torch.aten.expand %16340, %16349, %false_14002 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14003 = torch.constant.int 4
%int1_14004 = torch.constant.int 1
%int4_14005 = torch.constant.int 4
%int128_14006 = torch.constant.int 128
%16351 = torch.prim.ListConstruct %int4_14003, %15939, %int1_14004, %int4_14005, %int128_14006 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14007 = torch.constant.bool false
%16352 = torch.aten.expand %16341, %16351, %false_14007 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14008 = torch.constant.int 4
%int1_14009 = torch.constant.int 1
%int4_14010 = torch.constant.int 4
%int128_14011 = torch.constant.int 128
%16353 = torch.prim.ListConstruct %int4_14008, %15939, %int1_14009, %int4_14010, %int128_14011 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14012 = torch.constant.bool false
%16354 = torch.aten.expand %16342, %16353, %false_14012 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14013 = torch.constant.int 4
%int1_14014 = torch.constant.int 1
%int4_14015 = torch.constant.int 4
%int128_14016 = torch.constant.int 128
%16355 = torch.prim.ListConstruct %int4_14013, %15939, %int1_14014, %int4_14015, %int128_14016 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14017 = torch.constant.bool false
%16356 = torch.aten.expand %16343, %16355, %false_14017 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14018 = torch.constant.int 4
%int1_14019 = torch.constant.int 1
%int4_14020 = torch.constant.int 4
%int128_14021 = torch.constant.int 128
%16357 = torch.prim.ListConstruct %int4_14018, %15939, %int1_14019, %int4_14020, %int128_14021 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14022 = torch.constant.bool false
%16358 = torch.aten.expand %16344, %16357, %false_14022 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14023 = torch.constant.int 4
%int1_14024 = torch.constant.int 1
%int4_14025 = torch.constant.int 4
%int128_14026 = torch.constant.int 128
%16359 = torch.prim.ListConstruct %int4_14023, %15939, %int1_14024, %int4_14025, %int128_14026 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14027 = torch.constant.bool false
%16360 = torch.aten.expand %16345, %16359, %false_14027 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14028 = torch.constant.int 4
%int1_14029 = torch.constant.int 1
%int4_14030 = torch.constant.int 4
%int128_14031 = torch.constant.int 128
%16361 = torch.prim.ListConstruct %int4_14028, %15939, %int1_14029, %int4_14030, %int128_14031 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14032 = torch.constant.bool false
%16362 = torch.aten.expand %16346, %16361, %false_14032 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14033 = torch.constant.int 4
%int4_14034 = torch.constant.int 4
%int128_14035 = torch.constant.int 128
%16363 = torch.prim.ListConstruct %int4_14033, %15939, %int4_14034, %int128_14035 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16364 = torch.aten.view %16348, %16363 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14036 = torch.constant.int 4
%int4_14037 = torch.constant.int 4
%int128_14038 = torch.constant.int 128
%16365 = torch.prim.ListConstruct %int4_14036, %15939, %int4_14037, %int128_14038 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16366 = torch.aten.view %16350, %16365 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14039 = torch.constant.int 4
%int4_14040 = torch.constant.int 4
%int128_14041 = torch.constant.int 128
%16367 = torch.prim.ListConstruct %int4_14039, %15939, %int4_14040, %int128_14041 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16368 = torch.aten.view %16352, %16367 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14042 = torch.constant.int 4
%int4_14043 = torch.constant.int 4
%int128_14044 = torch.constant.int 128
%16369 = torch.prim.ListConstruct %int4_14042, %15939, %int4_14043, %int128_14044 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16370 = torch.aten.view %16354, %16369 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14045 = torch.constant.int 4
%int4_14046 = torch.constant.int 4
%int128_14047 = torch.constant.int 128
%16371 = torch.prim.ListConstruct %int4_14045, %15939, %int4_14046, %int128_14047 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16372 = torch.aten.view %16356, %16371 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14048 = torch.constant.int 4
%int4_14049 = torch.constant.int 4
%int128_14050 = torch.constant.int 128
%16373 = torch.prim.ListConstruct %int4_14048, %15939, %int4_14049, %int128_14050 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16374 = torch.aten.view %16358, %16373 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14051 = torch.constant.int 4
%int4_14052 = torch.constant.int 4
%int128_14053 = torch.constant.int 128
%16375 = torch.prim.ListConstruct %int4_14051, %15939, %int4_14052, %int128_14053 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16376 = torch.aten.view %16360, %16375 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14054 = torch.constant.int 4
%int4_14055 = torch.constant.int 4
%int128_14056 = torch.constant.int 128
%16377 = torch.prim.ListConstruct %int4_14054, %15939, %int4_14055, %int128_14056 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16378 = torch.aten.view %16362, %16377 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_14057 = torch.constant.int -2
%16379 = torch.aten.unsqueeze %15728, %int-2_14057 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14058 = torch.constant.int -2
%16380 = torch.aten.unsqueeze %15730, %int-2_14058 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14059 = torch.constant.int -2
%16381 = torch.aten.unsqueeze %15732, %int-2_14059 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14060 = torch.constant.int -2
%16382 = torch.aten.unsqueeze %15734, %int-2_14060 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14061 = torch.constant.int -2
%16383 = torch.aten.unsqueeze %15736, %int-2_14061 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14062 = torch.constant.int -2
%16384 = torch.aten.unsqueeze %15738, %int-2_14062 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14063 = torch.constant.int -2
%16385 = torch.aten.unsqueeze %15740, %int-2_14063 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_14064 = torch.constant.int -2
%16386 = torch.aten.unsqueeze %15742, %int-2_14064 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %16386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_14065 = torch.constant.int 1
%16387 = torch.aten.size.int %15652, %int1_14065 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_14066 = torch.constant.int 4
%int1_14067 = torch.constant.int 1
%int4_14068 = torch.constant.int 4
%int128_14069 = torch.constant.int 128
%16388 = torch.prim.ListConstruct %int4_14066, %16387, %int1_14067, %int4_14068, %int128_14069 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14070 = torch.constant.bool false
%16389 = torch.aten.expand %16379, %16388, %false_14070 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14071 = torch.constant.int 4
%int1_14072 = torch.constant.int 1
%int4_14073 = torch.constant.int 4
%int128_14074 = torch.constant.int 128
%16390 = torch.prim.ListConstruct %int4_14071, %16387, %int1_14072, %int4_14073, %int128_14074 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14075 = torch.constant.bool false
%16391 = torch.aten.expand %16380, %16390, %false_14075 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14076 = torch.constant.int 4
%int1_14077 = torch.constant.int 1
%int4_14078 = torch.constant.int 4
%int128_14079 = torch.constant.int 128
%16392 = torch.prim.ListConstruct %int4_14076, %16387, %int1_14077, %int4_14078, %int128_14079 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14080 = torch.constant.bool false
%16393 = torch.aten.expand %16381, %16392, %false_14080 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14081 = torch.constant.int 4
%int1_14082 = torch.constant.int 1
%int4_14083 = torch.constant.int 4
%int128_14084 = torch.constant.int 128
%16394 = torch.prim.ListConstruct %int4_14081, %16387, %int1_14082, %int4_14083, %int128_14084 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14085 = torch.constant.bool false
%16395 = torch.aten.expand %16382, %16394, %false_14085 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14086 = torch.constant.int 4
%int1_14087 = torch.constant.int 1
%int4_14088 = torch.constant.int 4
%int128_14089 = torch.constant.int 128
%16396 = torch.prim.ListConstruct %int4_14086, %16387, %int1_14087, %int4_14088, %int128_14089 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14090 = torch.constant.bool false
%16397 = torch.aten.expand %16383, %16396, %false_14090 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14091 = torch.constant.int 4
%int1_14092 = torch.constant.int 1
%int4_14093 = torch.constant.int 4
%int128_14094 = torch.constant.int 128
%16398 = torch.prim.ListConstruct %int4_14091, %16387, %int1_14092, %int4_14093, %int128_14094 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14095 = torch.constant.bool false
%16399 = torch.aten.expand %16384, %16398, %false_14095 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14096 = torch.constant.int 4
%int1_14097 = torch.constant.int 1
%int4_14098 = torch.constant.int 4
%int128_14099 = torch.constant.int 128
%16400 = torch.prim.ListConstruct %int4_14096, %16387, %int1_14097, %int4_14098, %int128_14099 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14100 = torch.constant.bool false
%16401 = torch.aten.expand %16385, %16400, %false_14100 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14101 = torch.constant.int 4
%int1_14102 = torch.constant.int 1
%int4_14103 = torch.constant.int 4
%int128_14104 = torch.constant.int 128
%16402 = torch.prim.ListConstruct %int4_14101, %16387, %int1_14102, %int4_14103, %int128_14104 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_14105 = torch.constant.bool false
%16403 = torch.aten.expand %16386, %16402, %false_14105 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %16403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_14106 = torch.constant.int 4
%int4_14107 = torch.constant.int 4
%int128_14108 = torch.constant.int 128
%16404 = torch.prim.ListConstruct %int4_14106, %16387, %int4_14107, %int128_14108 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16405 = torch.aten.view %16389, %16404 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14109 = torch.constant.int 4
%int4_14110 = torch.constant.int 4
%int128_14111 = torch.constant.int 128
%16406 = torch.prim.ListConstruct %int4_14109, %16387, %int4_14110, %int128_14111 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16407 = torch.aten.view %16391, %16406 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14112 = torch.constant.int 4
%int4_14113 = torch.constant.int 4
%int128_14114 = torch.constant.int 128
%16408 = torch.prim.ListConstruct %int4_14112, %16387, %int4_14113, %int128_14114 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16409 = torch.aten.view %16393, %16408 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14115 = torch.constant.int 4
%int4_14116 = torch.constant.int 4
%int128_14117 = torch.constant.int 128
%16410 = torch.prim.ListConstruct %int4_14115, %16387, %int4_14116, %int128_14117 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16411 = torch.aten.view %16395, %16410 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14118 = torch.constant.int 4
%int4_14119 = torch.constant.int 4
%int128_14120 = torch.constant.int 128
%16412 = torch.prim.ListConstruct %int4_14118, %16387, %int4_14119, %int128_14120 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16413 = torch.aten.view %16397, %16412 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14121 = torch.constant.int 4
%int4_14122 = torch.constant.int 4
%int128_14123 = torch.constant.int 128
%16414 = torch.prim.ListConstruct %int4_14121, %16387, %int4_14122, %int128_14123 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16415 = torch.aten.view %16399, %16414 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14124 = torch.constant.int 4
%int4_14125 = torch.constant.int 4
%int128_14126 = torch.constant.int 128
%16416 = torch.prim.ListConstruct %int4_14124, %16387, %int4_14125, %int128_14126 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16417 = torch.aten.view %16401, %16416 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14127 = torch.constant.int 4
%int4_14128 = torch.constant.int 4
%int128_14129 = torch.constant.int 128
%16418 = torch.prim.ListConstruct %int4_14127, %16387, %int4_14128, %int128_14129 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16419 = torch.aten.view %16403, %16418 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14130 = torch.constant.int 1
%int2_14131 = torch.constant.int 2
%16420 = torch.aten.transpose.int %15795, %int1_14130, %int2_14131 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16420, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14132 = torch.constant.int 1
%int2_14133 = torch.constant.int 2
%16421 = torch.aten.transpose.int %15810, %int1_14132, %int2_14133 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16421, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14134 = torch.constant.int 1
%int2_14135 = torch.constant.int 2
%16422 = torch.aten.transpose.int %15825, %int1_14134, %int2_14135 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16422, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14136 = torch.constant.int 1
%int2_14137 = torch.constant.int 2
%16423 = torch.aten.transpose.int %15840, %int1_14136, %int2_14137 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16423, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14138 = torch.constant.int 1
%int2_14139 = torch.constant.int 2
%16424 = torch.aten.transpose.int %15855, %int1_14138, %int2_14139 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16424, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14140 = torch.constant.int 1
%int2_14141 = torch.constant.int 2
%16425 = torch.aten.transpose.int %15870, %int1_14140, %int2_14141 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16425, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14142 = torch.constant.int 1
%int2_14143 = torch.constant.int 2
%16426 = torch.aten.transpose.int %15885, %int1_14142, %int2_14143 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16426, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14144 = torch.constant.int 1
%int2_14145 = torch.constant.int 2
%16427 = torch.aten.transpose.int %15900, %int1_14144, %int2_14145 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16427, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14146 = torch.constant.int 1
%int2_14147 = torch.constant.int 2
%16428 = torch.aten.transpose.int %16364, %int1_14146, %int2_14147 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16428, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14148 = torch.constant.int 1
%int2_14149 = torch.constant.int 2
%16429 = torch.aten.transpose.int %16366, %int1_14148, %int2_14149 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16429, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14150 = torch.constant.int 1
%int2_14151 = torch.constant.int 2
%16430 = torch.aten.transpose.int %16368, %int1_14150, %int2_14151 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16430, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14152 = torch.constant.int 1
%int2_14153 = torch.constant.int 2
%16431 = torch.aten.transpose.int %16370, %int1_14152, %int2_14153 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16431, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14154 = torch.constant.int 1
%int2_14155 = torch.constant.int 2
%16432 = torch.aten.transpose.int %16372, %int1_14154, %int2_14155 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16432, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14156 = torch.constant.int 1
%int2_14157 = torch.constant.int 2
%16433 = torch.aten.transpose.int %16374, %int1_14156, %int2_14157 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16433, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14158 = torch.constant.int 1
%int2_14159 = torch.constant.int 2
%16434 = torch.aten.transpose.int %16376, %int1_14158, %int2_14159 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16434, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14160 = torch.constant.int 1
%int2_14161 = torch.constant.int 2
%16435 = torch.aten.transpose.int %16378, %int1_14160, %int2_14161 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16435, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14162 = torch.constant.int 1
%int2_14163 = torch.constant.int 2
%16436 = torch.aten.transpose.int %16405, %int1_14162, %int2_14163 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16436, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14164 = torch.constant.int 1
%int2_14165 = torch.constant.int 2
%16437 = torch.aten.transpose.int %16407, %int1_14164, %int2_14165 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16437, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14166 = torch.constant.int 1
%int2_14167 = torch.constant.int 2
%16438 = torch.aten.transpose.int %16409, %int1_14166, %int2_14167 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16438, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14168 = torch.constant.int 1
%int2_14169 = torch.constant.int 2
%16439 = torch.aten.transpose.int %16411, %int1_14168, %int2_14169 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16439, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14170 = torch.constant.int 1
%int2_14171 = torch.constant.int 2
%16440 = torch.aten.transpose.int %16413, %int1_14170, %int2_14171 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16440, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14172 = torch.constant.int 1
%int2_14173 = torch.constant.int 2
%16441 = torch.aten.transpose.int %16415, %int1_14172, %int2_14173 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16441, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14174 = torch.constant.int 1
%int2_14175 = torch.constant.int 2
%16442 = torch.aten.transpose.int %16417, %int1_14174, %int2_14175 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16442, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14176 = torch.constant.int 1
%int2_14177 = torch.constant.int 2
%16443 = torch.aten.transpose.int %16419, %int1_14176, %int2_14177 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %16443, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14178 = torch.constant.float 0.000000e+00
%true_14179 = torch.constant.bool true
%none_14180 = torch.constant.none
%none_14181 = torch.constant.none
%16444:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16420, %16428, %16436, %float0.000000e00_14178, %true_14179, %none_14180, %none_14181) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16444#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14182 = torch.constant.float 0.000000e+00
%true_14183 = torch.constant.bool true
%none_14184 = torch.constant.none
%none_14185 = torch.constant.none
%16445:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16421, %16429, %16437, %float0.000000e00_14182, %true_14183, %none_14184, %none_14185) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16445#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14186 = torch.constant.float 0.000000e+00
%true_14187 = torch.constant.bool true
%none_14188 = torch.constant.none
%none_14189 = torch.constant.none
%16446:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16422, %16430, %16438, %float0.000000e00_14186, %true_14187, %none_14188, %none_14189) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16446#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14190 = torch.constant.float 0.000000e+00
%true_14191 = torch.constant.bool true
%none_14192 = torch.constant.none
%none_14193 = torch.constant.none
%16447:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16423, %16431, %16439, %float0.000000e00_14190, %true_14191, %none_14192, %none_14193) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16447#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14194 = torch.constant.float 0.000000e+00
%true_14195 = torch.constant.bool true
%none_14196 = torch.constant.none
%none_14197 = torch.constant.none
%16448:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16424, %16432, %16440, %float0.000000e00_14194, %true_14195, %none_14196, %none_14197) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16448#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14198 = torch.constant.float 0.000000e+00
%true_14199 = torch.constant.bool true
%none_14200 = torch.constant.none
%none_14201 = torch.constant.none
%16449:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16425, %16433, %16441, %float0.000000e00_14198, %true_14199, %none_14200, %none_14201) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16449#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14202 = torch.constant.float 0.000000e+00
%true_14203 = torch.constant.bool true
%none_14204 = torch.constant.none
%none_14205 = torch.constant.none
%16450:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16426, %16434, %16442, %float0.000000e00_14202, %true_14203, %none_14204, %none_14205) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16450#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_14206 = torch.constant.float 0.000000e+00
%true_14207 = torch.constant.bool true
%none_14208 = torch.constant.none
%none_14209 = torch.constant.none
%16451:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%16427, %16435, %16443, %float0.000000e00_14206, %true_14207, %none_14208, %none_14209) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %16451#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_14210 = torch.constant.int 1
%int2_14211 = torch.constant.int 2
%16452 = torch.aten.transpose.int %16444#0, %int1_14210, %int2_14211 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14212 = torch.constant.int 1
%int2_14213 = torch.constant.int 2
%16453 = torch.aten.transpose.int %16445#0, %int1_14212, %int2_14213 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14214 = torch.constant.int 1
%int2_14215 = torch.constant.int 2
%16454 = torch.aten.transpose.int %16446#0, %int1_14214, %int2_14215 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14216 = torch.constant.int 1
%int2_14217 = torch.constant.int 2
%16455 = torch.aten.transpose.int %16447#0, %int1_14216, %int2_14217 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14218 = torch.constant.int 1
%int2_14219 = torch.constant.int 2
%16456 = torch.aten.transpose.int %16448#0, %int1_14218, %int2_14219 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14220 = torch.constant.int 1
%int2_14221 = torch.constant.int 2
%16457 = torch.aten.transpose.int %16449#0, %int1_14220, %int2_14221 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14222 = torch.constant.int 1
%int2_14223 = torch.constant.int 2
%16458 = torch.aten.transpose.int %16450#0, %int1_14222, %int2_14223 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_14224 = torch.constant.int 1
%int2_14225 = torch.constant.int 2
%16459 = torch.aten.transpose.int %16451#0, %int1_14224, %int2_14225 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %16459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_14226 = torch.constant.int 4
%int512_14227 = torch.constant.int 512
%16460 = torch.prim.ListConstruct %int4_14226, %15781, %int512_14227 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16461 = torch.aten.view %16452, %16460 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14228 = torch.constant.int 4
%int512_14229 = torch.constant.int 512
%16462 = torch.prim.ListConstruct %int4_14228, %15796, %int512_14229 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16463 = torch.aten.view %16453, %16462 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14230 = torch.constant.int 4
%int512_14231 = torch.constant.int 512
%16464 = torch.prim.ListConstruct %int4_14230, %15811, %int512_14231 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16465 = torch.aten.view %16454, %16464 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14232 = torch.constant.int 4
%int512_14233 = torch.constant.int 512
%16466 = torch.prim.ListConstruct %int4_14232, %15826, %int512_14233 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16467 = torch.aten.view %16455, %16466 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14234 = torch.constant.int 4
%int512_14235 = torch.constant.int 512
%16468 = torch.prim.ListConstruct %int4_14234, %15841, %int512_14235 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16469 = torch.aten.view %16456, %16468 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14236 = torch.constant.int 4
%int512_14237 = torch.constant.int 512
%16470 = torch.prim.ListConstruct %int4_14236, %15856, %int512_14237 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16471 = torch.aten.view %16457, %16470 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14238 = torch.constant.int 4
%int512_14239 = torch.constant.int 512
%16472 = torch.prim.ListConstruct %int4_14238, %15871, %int512_14239 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16473 = torch.aten.view %16458, %16472 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14240 = torch.constant.int 4
%int512_14241 = torch.constant.int 512
%16474 = torch.prim.ListConstruct %int4_14240, %15886, %int512_14241 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16475 = torch.aten.view %16459, %16474 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %16475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_14242 = torch.constant.int 1
%int0_14243 = torch.constant.int 0
%16476 = torch.prim.ListConstruct %int1_14242, %int0_14243 : (!torch.int, !torch.int) -> !torch.list<int>
%16477 = torch.aten.permute %544, %16476 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14244 = torch.constant.int 1
%int0_14245 = torch.constant.int 0
%16478 = torch.prim.ListConstruct %int1_14244, %int0_14245 : (!torch.int, !torch.int) -> !torch.list<int>
%16479 = torch.aten.permute %545, %16478 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14246 = torch.constant.int 1
%int0_14247 = torch.constant.int 0
%16480 = torch.prim.ListConstruct %int1_14246, %int0_14247 : (!torch.int, !torch.int) -> !torch.list<int>
%16481 = torch.aten.permute %546, %16480 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14248 = torch.constant.int 1
%int0_14249 = torch.constant.int 0
%16482 = torch.prim.ListConstruct %int1_14248, %int0_14249 : (!torch.int, !torch.int) -> !torch.list<int>
%16483 = torch.aten.permute %547, %16482 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14250 = torch.constant.int 1
%int0_14251 = torch.constant.int 0
%16484 = torch.prim.ListConstruct %int1_14250, %int0_14251 : (!torch.int, !torch.int) -> !torch.list<int>
%16485 = torch.aten.permute %548, %16484 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14252 = torch.constant.int 1
%int0_14253 = torch.constant.int 0
%16486 = torch.prim.ListConstruct %int1_14252, %int0_14253 : (!torch.int, !torch.int) -> !torch.list<int>
%16487 = torch.aten.permute %549, %16486 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14254 = torch.constant.int 1
%int0_14255 = torch.constant.int 0
%16488 = torch.prim.ListConstruct %int1_14254, %int0_14255 : (!torch.int, !torch.int) -> !torch.list<int>
%16489 = torch.aten.permute %550, %16488 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_14256 = torch.constant.int 1
%int0_14257 = torch.constant.int 0
%16490 = torch.prim.ListConstruct %int1_14256, %int0_14257 : (!torch.int, !torch.int) -> !torch.list<int>
%16491 = torch.aten.permute %551, %16490 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_14258 = torch.constant.int 4
%16492 = torch.aten.mul.int %int4_14258, %15781 : !torch.int, !torch.int -> !torch.int
%int512_14259 = torch.constant.int 512
%16493 = torch.prim.ListConstruct %16492, %int512_14259 : (!torch.int, !torch.int) -> !torch.list<int>
%16494 = torch.aten.view %16461, %16493 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16494, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16495 = torch.aten.mm %16494, %16477 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16495, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14260 = torch.constant.int 4
%int4096_14261 = torch.constant.int 4096
%16496 = torch.prim.ListConstruct %int4_14260, %15781, %int4096_14261 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16497 = torch.aten.view %16495, %16496 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14262 = torch.constant.int 4
%16498 = torch.aten.mul.int %int4_14262, %15796 : !torch.int, !torch.int -> !torch.int
%int512_14263 = torch.constant.int 512
%16499 = torch.prim.ListConstruct %16498, %int512_14263 : (!torch.int, !torch.int) -> !torch.list<int>
%16500 = torch.aten.view %16463, %16499 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16500, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16501 = torch.aten.mm %16500, %16479 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16501, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14264 = torch.constant.int 4
%int4096_14265 = torch.constant.int 4096
%16502 = torch.prim.ListConstruct %int4_14264, %15796, %int4096_14265 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16503 = torch.aten.view %16501, %16502 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14266 = torch.constant.int 4
%16504 = torch.aten.mul.int %int4_14266, %15811 : !torch.int, !torch.int -> !torch.int
%int512_14267 = torch.constant.int 512
%16505 = torch.prim.ListConstruct %16504, %int512_14267 : (!torch.int, !torch.int) -> !torch.list<int>
%16506 = torch.aten.view %16465, %16505 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16506, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16507 = torch.aten.mm %16506, %16481 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16507, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14268 = torch.constant.int 4
%int4096_14269 = torch.constant.int 4096
%16508 = torch.prim.ListConstruct %int4_14268, %15811, %int4096_14269 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16509 = torch.aten.view %16507, %16508 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14270 = torch.constant.int 4
%16510 = torch.aten.mul.int %int4_14270, %15826 : !torch.int, !torch.int -> !torch.int
%int512_14271 = torch.constant.int 512
%16511 = torch.prim.ListConstruct %16510, %int512_14271 : (!torch.int, !torch.int) -> !torch.list<int>
%16512 = torch.aten.view %16467, %16511 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16512, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16513 = torch.aten.mm %16512, %16483 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16513, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14272 = torch.constant.int 4
%int4096_14273 = torch.constant.int 4096
%16514 = torch.prim.ListConstruct %int4_14272, %15826, %int4096_14273 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16515 = torch.aten.view %16513, %16514 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14274 = torch.constant.int 4
%16516 = torch.aten.mul.int %int4_14274, %15841 : !torch.int, !torch.int -> !torch.int
%int512_14275 = torch.constant.int 512
%16517 = torch.prim.ListConstruct %16516, %int512_14275 : (!torch.int, !torch.int) -> !torch.list<int>
%16518 = torch.aten.view %16469, %16517 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16518, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16519 = torch.aten.mm %16518, %16485 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16519, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14276 = torch.constant.int 4
%int4096_14277 = torch.constant.int 4096
%16520 = torch.prim.ListConstruct %int4_14276, %15841, %int4096_14277 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16521 = torch.aten.view %16519, %16520 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14278 = torch.constant.int 4
%16522 = torch.aten.mul.int %int4_14278, %15856 : !torch.int, !torch.int -> !torch.int
%int512_14279 = torch.constant.int 512
%16523 = torch.prim.ListConstruct %16522, %int512_14279 : (!torch.int, !torch.int) -> !torch.list<int>
%16524 = torch.aten.view %16471, %16523 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16524, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16525 = torch.aten.mm %16524, %16487 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16525, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14280 = torch.constant.int 4
%int4096_14281 = torch.constant.int 4096
%16526 = torch.prim.ListConstruct %int4_14280, %15856, %int4096_14281 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16527 = torch.aten.view %16525, %16526 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14282 = torch.constant.int 4
%16528 = torch.aten.mul.int %int4_14282, %15871 : !torch.int, !torch.int -> !torch.int
%int512_14283 = torch.constant.int 512
%16529 = torch.prim.ListConstruct %16528, %int512_14283 : (!torch.int, !torch.int) -> !torch.list<int>
%16530 = torch.aten.view %16473, %16529 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16530, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16531 = torch.aten.mm %16530, %16489 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16531, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14284 = torch.constant.int 4
%int4096_14285 = torch.constant.int 4096
%16532 = torch.prim.ListConstruct %int4_14284, %15871, %int4096_14285 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16533 = torch.aten.view %16531, %16532 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_14286 = torch.constant.int 4
%16534 = torch.aten.mul.int %int4_14286, %15886 : !torch.int, !torch.int -> !torch.int
%int512_14287 = torch.constant.int 512
%16535 = torch.prim.ListConstruct %16534, %int512_14287 : (!torch.int, !torch.int) -> !torch.list<int>
%16536 = torch.aten.view %16475, %16535 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %16536, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%16537 = torch.aten.mm %16536, %16491 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16537, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14288 = torch.constant.int 4
%int4096_14289 = torch.constant.int 4096
%16538 = torch.prim.ListConstruct %int4_14288, %15886, %int4096_14289 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16539 = torch.aten.view %16537, %16538 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16540 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14290 = arith.constant 1 : index
%dim_14291 = tensor.dim %16540, %c1_14290 : tensor<4x?x4096xf16>
%16541 = flow.tensor.transfer %16540 : tensor<4x?x4096xf16>{%dim_14291} to #hal.device.promise<@__device_0>
%16542 = torch_c.from_builtin_tensor %16541 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16543 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14292 = arith.constant 1 : index
%dim_14293 = tensor.dim %16543, %c1_14292 : tensor<4x?x4096xf16>
%16544 = flow.tensor.transfer %16543 : tensor<4x?x4096xf16>{%dim_14293} to #hal.device.promise<@__device_0>
%16545 = torch_c.from_builtin_tensor %16544 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16546 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14294 = arith.constant 1 : index
%dim_14295 = tensor.dim %16546, %c1_14294 : tensor<4x?x4096xf16>
%16547 = flow.tensor.transfer %16546 : tensor<4x?x4096xf16>{%dim_14295} to #hal.device.promise<@__device_0>
%16548 = torch_c.from_builtin_tensor %16547 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16549 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14296 = arith.constant 1 : index
%dim_14297 = tensor.dim %16549, %c1_14296 : tensor<4x?x4096xf16>
%16550 = flow.tensor.transfer %16549 : tensor<4x?x4096xf16>{%dim_14297} to #hal.device.promise<@__device_0>
%16551 = torch_c.from_builtin_tensor %16550 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16552 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14298 = arith.constant 1 : index
%dim_14299 = tensor.dim %16552, %c1_14298 : tensor<4x?x4096xf16>
%16553 = flow.tensor.transfer %16552 : tensor<4x?x4096xf16>{%dim_14299} to #hal.device.promise<@__device_0>
%16554 = torch_c.from_builtin_tensor %16553 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16555 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14300 = arith.constant 1 : index
%dim_14301 = tensor.dim %16555, %c1_14300 : tensor<4x?x4096xf16>
%16556 = flow.tensor.transfer %16555 : tensor<4x?x4096xf16>{%dim_14301} to #hal.device.promise<@__device_0>
%16557 = torch_c.from_builtin_tensor %16556 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16558 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14302 = arith.constant 1 : index
%dim_14303 = tensor.dim %16558, %c1_14302 : tensor<4x?x4096xf16>
%16559 = flow.tensor.transfer %16558 : tensor<4x?x4096xf16>{%dim_14303} to #hal.device.promise<@__device_0>
%16560 = torch_c.from_builtin_tensor %16559 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14304 = torch.constant.int 1
%16561 = torch.aten.add.Tensor %16497, %16542, %int1_14304 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14305 = torch.constant.int 1
%16562 = torch.aten.add.Tensor %16561, %16545, %int1_14305 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14306 = torch.constant.int 1
%16563 = torch.aten.add.Tensor %16562, %16548, %int1_14306 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14307 = torch.constant.int 1
%16564 = torch.aten.add.Tensor %16563, %16551, %int1_14307 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14308 = torch.constant.int 1
%16565 = torch.aten.add.Tensor %16564, %16554, %int1_14308 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14309 = torch.constant.int 1
%16566 = torch.aten.add.Tensor %16565, %16557, %int1_14309 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14310 = torch.constant.int 1
%16567 = torch.aten.add.Tensor %16566, %16560, %int1_14310 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16568 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14311 = arith.constant 1 : index
%dim_14312 = tensor.dim %16568, %c1_14311 : tensor<4x?x4096xf16>
%16569 = flow.tensor.transfer %16568 : tensor<4x?x4096xf16>{%dim_14312} to #hal.device.promise<@__device_1>
%16570 = torch_c.from_builtin_tensor %16569 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16571 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14313 = arith.constant 1 : index
%dim_14314 = tensor.dim %16571, %c1_14313 : tensor<4x?x4096xf16>
%16572 = flow.tensor.transfer %16571 : tensor<4x?x4096xf16>{%dim_14314} to #hal.device.promise<@__device_1>
%16573 = torch_c.from_builtin_tensor %16572 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16574 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14315 = arith.constant 1 : index
%dim_14316 = tensor.dim %16574, %c1_14315 : tensor<4x?x4096xf16>
%16575 = flow.tensor.transfer %16574 : tensor<4x?x4096xf16>{%dim_14316} to #hal.device.promise<@__device_1>
%16576 = torch_c.from_builtin_tensor %16575 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16577 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14317 = arith.constant 1 : index
%dim_14318 = tensor.dim %16577, %c1_14317 : tensor<4x?x4096xf16>
%16578 = flow.tensor.transfer %16577 : tensor<4x?x4096xf16>{%dim_14318} to #hal.device.promise<@__device_1>
%16579 = torch_c.from_builtin_tensor %16578 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16580 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14319 = arith.constant 1 : index
%dim_14320 = tensor.dim %16580, %c1_14319 : tensor<4x?x4096xf16>
%16581 = flow.tensor.transfer %16580 : tensor<4x?x4096xf16>{%dim_14320} to #hal.device.promise<@__device_1>
%16582 = torch_c.from_builtin_tensor %16581 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16583 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14321 = arith.constant 1 : index
%dim_14322 = tensor.dim %16583, %c1_14321 : tensor<4x?x4096xf16>
%16584 = flow.tensor.transfer %16583 : tensor<4x?x4096xf16>{%dim_14322} to #hal.device.promise<@__device_1>
%16585 = torch_c.from_builtin_tensor %16584 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16586 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14323 = arith.constant 1 : index
%dim_14324 = tensor.dim %16586, %c1_14323 : tensor<4x?x4096xf16>
%16587 = flow.tensor.transfer %16586 : tensor<4x?x4096xf16>{%dim_14324} to #hal.device.promise<@__device_1>
%16588 = torch_c.from_builtin_tensor %16587 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14325 = torch.constant.int 1
%16589 = torch.aten.add.Tensor %16570, %16503, %int1_14325 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14326 = torch.constant.int 1
%16590 = torch.aten.add.Tensor %16589, %16573, %int1_14326 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14327 = torch.constant.int 1
%16591 = torch.aten.add.Tensor %16590, %16576, %int1_14327 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14328 = torch.constant.int 1
%16592 = torch.aten.add.Tensor %16591, %16579, %int1_14328 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14329 = torch.constant.int 1
%16593 = torch.aten.add.Tensor %16592, %16582, %int1_14329 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14330 = torch.constant.int 1
%16594 = torch.aten.add.Tensor %16593, %16585, %int1_14330 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14331 = torch.constant.int 1
%16595 = torch.aten.add.Tensor %16594, %16588, %int1_14331 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16596 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14332 = arith.constant 1 : index
%dim_14333 = tensor.dim %16596, %c1_14332 : tensor<4x?x4096xf16>
%16597 = flow.tensor.transfer %16596 : tensor<4x?x4096xf16>{%dim_14333} to #hal.device.promise<@__device_2>
%16598 = torch_c.from_builtin_tensor %16597 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16599 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14334 = arith.constant 1 : index
%dim_14335 = tensor.dim %16599, %c1_14334 : tensor<4x?x4096xf16>
%16600 = flow.tensor.transfer %16599 : tensor<4x?x4096xf16>{%dim_14335} to #hal.device.promise<@__device_2>
%16601 = torch_c.from_builtin_tensor %16600 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16602 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14336 = arith.constant 1 : index
%dim_14337 = tensor.dim %16602, %c1_14336 : tensor<4x?x4096xf16>
%16603 = flow.tensor.transfer %16602 : tensor<4x?x4096xf16>{%dim_14337} to #hal.device.promise<@__device_2>
%16604 = torch_c.from_builtin_tensor %16603 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16605 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14338 = arith.constant 1 : index
%dim_14339 = tensor.dim %16605, %c1_14338 : tensor<4x?x4096xf16>
%16606 = flow.tensor.transfer %16605 : tensor<4x?x4096xf16>{%dim_14339} to #hal.device.promise<@__device_2>
%16607 = torch_c.from_builtin_tensor %16606 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16608 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14340 = arith.constant 1 : index
%dim_14341 = tensor.dim %16608, %c1_14340 : tensor<4x?x4096xf16>
%16609 = flow.tensor.transfer %16608 : tensor<4x?x4096xf16>{%dim_14341} to #hal.device.promise<@__device_2>
%16610 = torch_c.from_builtin_tensor %16609 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16611 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14342 = arith.constant 1 : index
%dim_14343 = tensor.dim %16611, %c1_14342 : tensor<4x?x4096xf16>
%16612 = flow.tensor.transfer %16611 : tensor<4x?x4096xf16>{%dim_14343} to #hal.device.promise<@__device_2>
%16613 = torch_c.from_builtin_tensor %16612 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16614 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14344 = arith.constant 1 : index
%dim_14345 = tensor.dim %16614, %c1_14344 : tensor<4x?x4096xf16>
%16615 = flow.tensor.transfer %16614 : tensor<4x?x4096xf16>{%dim_14345} to #hal.device.promise<@__device_2>
%16616 = torch_c.from_builtin_tensor %16615 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14346 = torch.constant.int 1
%16617 = torch.aten.add.Tensor %16598, %16601, %int1_14346 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14347 = torch.constant.int 1
%16618 = torch.aten.add.Tensor %16617, %16509, %int1_14347 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14348 = torch.constant.int 1
%16619 = torch.aten.add.Tensor %16618, %16604, %int1_14348 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14349 = torch.constant.int 1
%16620 = torch.aten.add.Tensor %16619, %16607, %int1_14349 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14350 = torch.constant.int 1
%16621 = torch.aten.add.Tensor %16620, %16610, %int1_14350 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14351 = torch.constant.int 1
%16622 = torch.aten.add.Tensor %16621, %16613, %int1_14351 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14352 = torch.constant.int 1
%16623 = torch.aten.add.Tensor %16622, %16616, %int1_14352 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16624 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14353 = arith.constant 1 : index
%dim_14354 = tensor.dim %16624, %c1_14353 : tensor<4x?x4096xf16>
%16625 = flow.tensor.transfer %16624 : tensor<4x?x4096xf16>{%dim_14354} to #hal.device.promise<@__device_3>
%16626 = torch_c.from_builtin_tensor %16625 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16627 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14355 = arith.constant 1 : index
%dim_14356 = tensor.dim %16627, %c1_14355 : tensor<4x?x4096xf16>
%16628 = flow.tensor.transfer %16627 : tensor<4x?x4096xf16>{%dim_14356} to #hal.device.promise<@__device_3>
%16629 = torch_c.from_builtin_tensor %16628 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16630 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14357 = arith.constant 1 : index
%dim_14358 = tensor.dim %16630, %c1_14357 : tensor<4x?x4096xf16>
%16631 = flow.tensor.transfer %16630 : tensor<4x?x4096xf16>{%dim_14358} to #hal.device.promise<@__device_3>
%16632 = torch_c.from_builtin_tensor %16631 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16633 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14359 = arith.constant 1 : index
%dim_14360 = tensor.dim %16633, %c1_14359 : tensor<4x?x4096xf16>
%16634 = flow.tensor.transfer %16633 : tensor<4x?x4096xf16>{%dim_14360} to #hal.device.promise<@__device_3>
%16635 = torch_c.from_builtin_tensor %16634 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16636 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14361 = arith.constant 1 : index
%dim_14362 = tensor.dim %16636, %c1_14361 : tensor<4x?x4096xf16>
%16637 = flow.tensor.transfer %16636 : tensor<4x?x4096xf16>{%dim_14362} to #hal.device.promise<@__device_3>
%16638 = torch_c.from_builtin_tensor %16637 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16639 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14363 = arith.constant 1 : index
%dim_14364 = tensor.dim %16639, %c1_14363 : tensor<4x?x4096xf16>
%16640 = flow.tensor.transfer %16639 : tensor<4x?x4096xf16>{%dim_14364} to #hal.device.promise<@__device_3>
%16641 = torch_c.from_builtin_tensor %16640 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16642 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14365 = arith.constant 1 : index
%dim_14366 = tensor.dim %16642, %c1_14365 : tensor<4x?x4096xf16>
%16643 = flow.tensor.transfer %16642 : tensor<4x?x4096xf16>{%dim_14366} to #hal.device.promise<@__device_3>
%16644 = torch_c.from_builtin_tensor %16643 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14367 = torch.constant.int 1
%16645 = torch.aten.add.Tensor %16626, %16629, %int1_14367 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14368 = torch.constant.int 1
%16646 = torch.aten.add.Tensor %16645, %16632, %int1_14368 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14369 = torch.constant.int 1
%16647 = torch.aten.add.Tensor %16646, %16515, %int1_14369 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14370 = torch.constant.int 1
%16648 = torch.aten.add.Tensor %16647, %16635, %int1_14370 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14371 = torch.constant.int 1
%16649 = torch.aten.add.Tensor %16648, %16638, %int1_14371 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14372 = torch.constant.int 1
%16650 = torch.aten.add.Tensor %16649, %16641, %int1_14372 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14373 = torch.constant.int 1
%16651 = torch.aten.add.Tensor %16650, %16644, %int1_14373 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16652 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14374 = arith.constant 1 : index
%dim_14375 = tensor.dim %16652, %c1_14374 : tensor<4x?x4096xf16>
%16653 = flow.tensor.transfer %16652 : tensor<4x?x4096xf16>{%dim_14375} to #hal.device.promise<@__device_4>
%16654 = torch_c.from_builtin_tensor %16653 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16655 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14376 = arith.constant 1 : index
%dim_14377 = tensor.dim %16655, %c1_14376 : tensor<4x?x4096xf16>
%16656 = flow.tensor.transfer %16655 : tensor<4x?x4096xf16>{%dim_14377} to #hal.device.promise<@__device_4>
%16657 = torch_c.from_builtin_tensor %16656 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16658 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14378 = arith.constant 1 : index
%dim_14379 = tensor.dim %16658, %c1_14378 : tensor<4x?x4096xf16>
%16659 = flow.tensor.transfer %16658 : tensor<4x?x4096xf16>{%dim_14379} to #hal.device.promise<@__device_4>
%16660 = torch_c.from_builtin_tensor %16659 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16661 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14380 = arith.constant 1 : index
%dim_14381 = tensor.dim %16661, %c1_14380 : tensor<4x?x4096xf16>
%16662 = flow.tensor.transfer %16661 : tensor<4x?x4096xf16>{%dim_14381} to #hal.device.promise<@__device_4>
%16663 = torch_c.from_builtin_tensor %16662 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16664 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14382 = arith.constant 1 : index
%dim_14383 = tensor.dim %16664, %c1_14382 : tensor<4x?x4096xf16>
%16665 = flow.tensor.transfer %16664 : tensor<4x?x4096xf16>{%dim_14383} to #hal.device.promise<@__device_4>
%16666 = torch_c.from_builtin_tensor %16665 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16667 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14384 = arith.constant 1 : index
%dim_14385 = tensor.dim %16667, %c1_14384 : tensor<4x?x4096xf16>
%16668 = flow.tensor.transfer %16667 : tensor<4x?x4096xf16>{%dim_14385} to #hal.device.promise<@__device_4>
%16669 = torch_c.from_builtin_tensor %16668 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16670 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14386 = arith.constant 1 : index
%dim_14387 = tensor.dim %16670, %c1_14386 : tensor<4x?x4096xf16>
%16671 = flow.tensor.transfer %16670 : tensor<4x?x4096xf16>{%dim_14387} to #hal.device.promise<@__device_4>
%16672 = torch_c.from_builtin_tensor %16671 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14388 = torch.constant.int 1
%16673 = torch.aten.add.Tensor %16654, %16657, %int1_14388 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14389 = torch.constant.int 1
%16674 = torch.aten.add.Tensor %16673, %16660, %int1_14389 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14390 = torch.constant.int 1
%16675 = torch.aten.add.Tensor %16674, %16663, %int1_14390 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14391 = torch.constant.int 1
%16676 = torch.aten.add.Tensor %16675, %16521, %int1_14391 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14392 = torch.constant.int 1
%16677 = torch.aten.add.Tensor %16676, %16666, %int1_14392 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14393 = torch.constant.int 1
%16678 = torch.aten.add.Tensor %16677, %16669, %int1_14393 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14394 = torch.constant.int 1
%16679 = torch.aten.add.Tensor %16678, %16672, %int1_14394 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16680 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14395 = arith.constant 1 : index
%dim_14396 = tensor.dim %16680, %c1_14395 : tensor<4x?x4096xf16>
%16681 = flow.tensor.transfer %16680 : tensor<4x?x4096xf16>{%dim_14396} to #hal.device.promise<@__device_5>
%16682 = torch_c.from_builtin_tensor %16681 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16683 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14397 = arith.constant 1 : index
%dim_14398 = tensor.dim %16683, %c1_14397 : tensor<4x?x4096xf16>
%16684 = flow.tensor.transfer %16683 : tensor<4x?x4096xf16>{%dim_14398} to #hal.device.promise<@__device_5>
%16685 = torch_c.from_builtin_tensor %16684 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16686 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14399 = arith.constant 1 : index
%dim_14400 = tensor.dim %16686, %c1_14399 : tensor<4x?x4096xf16>
%16687 = flow.tensor.transfer %16686 : tensor<4x?x4096xf16>{%dim_14400} to #hal.device.promise<@__device_5>
%16688 = torch_c.from_builtin_tensor %16687 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16689 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14401 = arith.constant 1 : index
%dim_14402 = tensor.dim %16689, %c1_14401 : tensor<4x?x4096xf16>
%16690 = flow.tensor.transfer %16689 : tensor<4x?x4096xf16>{%dim_14402} to #hal.device.promise<@__device_5>
%16691 = torch_c.from_builtin_tensor %16690 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16692 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14403 = arith.constant 1 : index
%dim_14404 = tensor.dim %16692, %c1_14403 : tensor<4x?x4096xf16>
%16693 = flow.tensor.transfer %16692 : tensor<4x?x4096xf16>{%dim_14404} to #hal.device.promise<@__device_5>
%16694 = torch_c.from_builtin_tensor %16693 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16695 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14405 = arith.constant 1 : index
%dim_14406 = tensor.dim %16695, %c1_14405 : tensor<4x?x4096xf16>
%16696 = flow.tensor.transfer %16695 : tensor<4x?x4096xf16>{%dim_14406} to #hal.device.promise<@__device_5>
%16697 = torch_c.from_builtin_tensor %16696 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16698 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14407 = arith.constant 1 : index
%dim_14408 = tensor.dim %16698, %c1_14407 : tensor<4x?x4096xf16>
%16699 = flow.tensor.transfer %16698 : tensor<4x?x4096xf16>{%dim_14408} to #hal.device.promise<@__device_5>
%16700 = torch_c.from_builtin_tensor %16699 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14409 = torch.constant.int 1
%16701 = torch.aten.add.Tensor %16682, %16685, %int1_14409 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14410 = torch.constant.int 1
%16702 = torch.aten.add.Tensor %16701, %16688, %int1_14410 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14411 = torch.constant.int 1
%16703 = torch.aten.add.Tensor %16702, %16691, %int1_14411 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14412 = torch.constant.int 1
%16704 = torch.aten.add.Tensor %16703, %16694, %int1_14412 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14413 = torch.constant.int 1
%16705 = torch.aten.add.Tensor %16704, %16527, %int1_14413 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14414 = torch.constant.int 1
%16706 = torch.aten.add.Tensor %16705, %16697, %int1_14414 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14415 = torch.constant.int 1
%16707 = torch.aten.add.Tensor %16706, %16700, %int1_14415 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16708 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14416 = arith.constant 1 : index
%dim_14417 = tensor.dim %16708, %c1_14416 : tensor<4x?x4096xf16>
%16709 = flow.tensor.transfer %16708 : tensor<4x?x4096xf16>{%dim_14417} to #hal.device.promise<@__device_6>
%16710 = torch_c.from_builtin_tensor %16709 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16711 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14418 = arith.constant 1 : index
%dim_14419 = tensor.dim %16711, %c1_14418 : tensor<4x?x4096xf16>
%16712 = flow.tensor.transfer %16711 : tensor<4x?x4096xf16>{%dim_14419} to #hal.device.promise<@__device_6>
%16713 = torch_c.from_builtin_tensor %16712 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16714 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14420 = arith.constant 1 : index
%dim_14421 = tensor.dim %16714, %c1_14420 : tensor<4x?x4096xf16>
%16715 = flow.tensor.transfer %16714 : tensor<4x?x4096xf16>{%dim_14421} to #hal.device.promise<@__device_6>
%16716 = torch_c.from_builtin_tensor %16715 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16717 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14422 = arith.constant 1 : index
%dim_14423 = tensor.dim %16717, %c1_14422 : tensor<4x?x4096xf16>
%16718 = flow.tensor.transfer %16717 : tensor<4x?x4096xf16>{%dim_14423} to #hal.device.promise<@__device_6>
%16719 = torch_c.from_builtin_tensor %16718 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16720 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14424 = arith.constant 1 : index
%dim_14425 = tensor.dim %16720, %c1_14424 : tensor<4x?x4096xf16>
%16721 = flow.tensor.transfer %16720 : tensor<4x?x4096xf16>{%dim_14425} to #hal.device.promise<@__device_6>
%16722 = torch_c.from_builtin_tensor %16721 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16723 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14426 = arith.constant 1 : index
%dim_14427 = tensor.dim %16723, %c1_14426 : tensor<4x?x4096xf16>
%16724 = flow.tensor.transfer %16723 : tensor<4x?x4096xf16>{%dim_14427} to #hal.device.promise<@__device_6>
%16725 = torch_c.from_builtin_tensor %16724 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16726 = torch_c.to_builtin_tensor %16539 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14428 = arith.constant 1 : index
%dim_14429 = tensor.dim %16726, %c1_14428 : tensor<4x?x4096xf16>
%16727 = flow.tensor.transfer %16726 : tensor<4x?x4096xf16>{%dim_14429} to #hal.device.promise<@__device_6>
%16728 = torch_c.from_builtin_tensor %16727 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14430 = torch.constant.int 1
%16729 = torch.aten.add.Tensor %16710, %16713, %int1_14430 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14431 = torch.constant.int 1
%16730 = torch.aten.add.Tensor %16729, %16716, %int1_14431 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14432 = torch.constant.int 1
%16731 = torch.aten.add.Tensor %16730, %16719, %int1_14432 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14433 = torch.constant.int 1
%16732 = torch.aten.add.Tensor %16731, %16722, %int1_14433 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14434 = torch.constant.int 1
%16733 = torch.aten.add.Tensor %16732, %16725, %int1_14434 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14435 = torch.constant.int 1
%16734 = torch.aten.add.Tensor %16733, %16533, %int1_14435 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14436 = torch.constant.int 1
%16735 = torch.aten.add.Tensor %16734, %16728, %int1_14436 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16736 = torch_c.to_builtin_tensor %16497 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14437 = arith.constant 1 : index
%dim_14438 = tensor.dim %16736, %c1_14437 : tensor<4x?x4096xf16>
%16737 = flow.tensor.transfer %16736 : tensor<4x?x4096xf16>{%dim_14438} to #hal.device.promise<@__device_7>
%16738 = torch_c.from_builtin_tensor %16737 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16739 = torch_c.to_builtin_tensor %16503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14439 = arith.constant 1 : index
%dim_14440 = tensor.dim %16739, %c1_14439 : tensor<4x?x4096xf16>
%16740 = flow.tensor.transfer %16739 : tensor<4x?x4096xf16>{%dim_14440} to #hal.device.promise<@__device_7>
%16741 = torch_c.from_builtin_tensor %16740 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16742 = torch_c.to_builtin_tensor %16509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14441 = arith.constant 1 : index
%dim_14442 = tensor.dim %16742, %c1_14441 : tensor<4x?x4096xf16>
%16743 = flow.tensor.transfer %16742 : tensor<4x?x4096xf16>{%dim_14442} to #hal.device.promise<@__device_7>
%16744 = torch_c.from_builtin_tensor %16743 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16745 = torch_c.to_builtin_tensor %16515 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14443 = arith.constant 1 : index
%dim_14444 = tensor.dim %16745, %c1_14443 : tensor<4x?x4096xf16>
%16746 = flow.tensor.transfer %16745 : tensor<4x?x4096xf16>{%dim_14444} to #hal.device.promise<@__device_7>
%16747 = torch_c.from_builtin_tensor %16746 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16748 = torch_c.to_builtin_tensor %16521 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14445 = arith.constant 1 : index
%dim_14446 = tensor.dim %16748, %c1_14445 : tensor<4x?x4096xf16>
%16749 = flow.tensor.transfer %16748 : tensor<4x?x4096xf16>{%dim_14446} to #hal.device.promise<@__device_7>
%16750 = torch_c.from_builtin_tensor %16749 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16751 = torch_c.to_builtin_tensor %16527 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14447 = arith.constant 1 : index
%dim_14448 = tensor.dim %16751, %c1_14447 : tensor<4x?x4096xf16>
%16752 = flow.tensor.transfer %16751 : tensor<4x?x4096xf16>{%dim_14448} to #hal.device.promise<@__device_7>
%16753 = torch_c.from_builtin_tensor %16752 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%16754 = torch_c.to_builtin_tensor %16533 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14449 = arith.constant 1 : index
%dim_14450 = tensor.dim %16754, %c1_14449 : tensor<4x?x4096xf16>
%16755 = flow.tensor.transfer %16754 : tensor<4x?x4096xf16>{%dim_14450} to #hal.device.promise<@__device_7>
%16756 = torch_c.from_builtin_tensor %16755 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14451 = torch.constant.int 1
%16757 = torch.aten.add.Tensor %16738, %16741, %int1_14451 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14452 = torch.constant.int 1
%16758 = torch.aten.add.Tensor %16757, %16744, %int1_14452 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14453 = torch.constant.int 1
%16759 = torch.aten.add.Tensor %16758, %16747, %int1_14453 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14454 = torch.constant.int 1
%16760 = torch.aten.add.Tensor %16759, %16750, %int1_14454 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14455 = torch.constant.int 1
%16761 = torch.aten.add.Tensor %16760, %16753, %int1_14455 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14456 = torch.constant.int 1
%16762 = torch.aten.add.Tensor %16761, %16756, %int1_14456 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14457 = torch.constant.int 1
%16763 = torch.aten.add.Tensor %16762, %16539, %int1_14457 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14458 = torch.constant.int 1
%16764 = torch.aten.add.Tensor %15423, %16567, %int1_14458 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14459 = torch.constant.int 1
%16765 = torch.aten.add.Tensor %15424, %16595, %int1_14459 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14460 = torch.constant.int 1
%16766 = torch.aten.add.Tensor %15425, %16623, %int1_14460 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14461 = torch.constant.int 1
%16767 = torch.aten.add.Tensor %15426, %16651, %int1_14461 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14462 = torch.constant.int 1
%16768 = torch.aten.add.Tensor %15427, %16679, %int1_14462 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14463 = torch.constant.int 1
%16769 = torch.aten.add.Tensor %15428, %16707, %int1_14463 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14464 = torch.constant.int 1
%16770 = torch.aten.add.Tensor %15429, %16735, %int1_14464 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14465 = torch.constant.int 1
%16771 = torch.aten.add.Tensor %15430, %16763, %int1_14465 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_14466 = torch.constant.int 6
%16772 = torch.prims.convert_element_type %16764, %int6_14466 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14467 = torch.constant.int 6
%16773 = torch.prims.convert_element_type %16765, %int6_14467 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14468 = torch.constant.int 6
%16774 = torch.prims.convert_element_type %16766, %int6_14468 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14469 = torch.constant.int 6
%16775 = torch.prims.convert_element_type %16767, %int6_14469 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14470 = torch.constant.int 6
%16776 = torch.prims.convert_element_type %16768, %int6_14470 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14471 = torch.constant.int 6
%16777 = torch.prims.convert_element_type %16769, %int6_14471 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14472 = torch.constant.int 6
%16778 = torch.prims.convert_element_type %16770, %int6_14472 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14473 = torch.constant.int 6
%16779 = torch.prims.convert_element_type %16771, %int6_14473 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14474 = torch.constant.int 2
%16780 = torch.aten.pow.Tensor_Scalar %16772, %int2_14474 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14475 = torch.constant.int 2
%16781 = torch.aten.pow.Tensor_Scalar %16773, %int2_14475 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14476 = torch.constant.int 2
%16782 = torch.aten.pow.Tensor_Scalar %16774, %int2_14476 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14477 = torch.constant.int 2
%16783 = torch.aten.pow.Tensor_Scalar %16775, %int2_14477 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14478 = torch.constant.int 2
%16784 = torch.aten.pow.Tensor_Scalar %16776, %int2_14478 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14479 = torch.constant.int 2
%16785 = torch.aten.pow.Tensor_Scalar %16777, %int2_14479 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14480 = torch.constant.int 2
%16786 = torch.aten.pow.Tensor_Scalar %16778, %int2_14480 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14481 = torch.constant.int 2
%16787 = torch.aten.pow.Tensor_Scalar %16779, %int2_14481 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_14482 = torch.constant.int -1
%16788 = torch.prim.ListConstruct %int-1_14482 : (!torch.int) -> !torch.list<int>
%true_14483 = torch.constant.bool true
%none_14484 = torch.constant.none
%16789 = torch.aten.mean.dim %16780, %16788, %true_14483, %none_14484 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14485 = torch.constant.int -1
%16790 = torch.prim.ListConstruct %int-1_14485 : (!torch.int) -> !torch.list<int>
%true_14486 = torch.constant.bool true
%none_14487 = torch.constant.none
%16791 = torch.aten.mean.dim %16781, %16790, %true_14486, %none_14487 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14488 = torch.constant.int -1
%16792 = torch.prim.ListConstruct %int-1_14488 : (!torch.int) -> !torch.list<int>
%true_14489 = torch.constant.bool true
%none_14490 = torch.constant.none
%16793 = torch.aten.mean.dim %16782, %16792, %true_14489, %none_14490 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14491 = torch.constant.int -1
%16794 = torch.prim.ListConstruct %int-1_14491 : (!torch.int) -> !torch.list<int>
%true_14492 = torch.constant.bool true
%none_14493 = torch.constant.none
%16795 = torch.aten.mean.dim %16783, %16794, %true_14492, %none_14493 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14494 = torch.constant.int -1
%16796 = torch.prim.ListConstruct %int-1_14494 : (!torch.int) -> !torch.list<int>
%true_14495 = torch.constant.bool true
%none_14496 = torch.constant.none
%16797 = torch.aten.mean.dim %16784, %16796, %true_14495, %none_14496 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14497 = torch.constant.int -1
%16798 = torch.prim.ListConstruct %int-1_14497 : (!torch.int) -> !torch.list<int>
%true_14498 = torch.constant.bool true
%none_14499 = torch.constant.none
%16799 = torch.aten.mean.dim %16785, %16798, %true_14498, %none_14499 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14500 = torch.constant.int -1
%16800 = torch.prim.ListConstruct %int-1_14500 : (!torch.int) -> !torch.list<int>
%true_14501 = torch.constant.bool true
%none_14502 = torch.constant.none
%16801 = torch.aten.mean.dim %16786, %16800, %true_14501, %none_14502 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14503 = torch.constant.int -1
%16802 = torch.prim.ListConstruct %int-1_14503 : (!torch.int) -> !torch.list<int>
%true_14504 = torch.constant.bool true
%none_14505 = torch.constant.none
%16803 = torch.aten.mean.dim %16787, %16802, %true_14504, %none_14505 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14506 = torch.constant.float 9.9999997473787516E-6
%int1_14507 = torch.constant.int 1
%16804 = torch.aten.add.Scalar %16789, %float9.999990e-06_14506, %int1_14507 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14508 = torch.constant.float 9.9999997473787516E-6
%int1_14509 = torch.constant.int 1
%16805 = torch.aten.add.Scalar %16791, %float9.999990e-06_14508, %int1_14509 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14510 = torch.constant.float 9.9999997473787516E-6
%int1_14511 = torch.constant.int 1
%16806 = torch.aten.add.Scalar %16793, %float9.999990e-06_14510, %int1_14511 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14512 = torch.constant.float 9.9999997473787516E-6
%int1_14513 = torch.constant.int 1
%16807 = torch.aten.add.Scalar %16795, %float9.999990e-06_14512, %int1_14513 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14514 = torch.constant.float 9.9999997473787516E-6
%int1_14515 = torch.constant.int 1
%16808 = torch.aten.add.Scalar %16797, %float9.999990e-06_14514, %int1_14515 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14516 = torch.constant.float 9.9999997473787516E-6
%int1_14517 = torch.constant.int 1
%16809 = torch.aten.add.Scalar %16799, %float9.999990e-06_14516, %int1_14517 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14518 = torch.constant.float 9.9999997473787516E-6
%int1_14519 = torch.constant.int 1
%16810 = torch.aten.add.Scalar %16801, %float9.999990e-06_14518, %int1_14519 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14520 = torch.constant.float 9.9999997473787516E-6
%int1_14521 = torch.constant.int 1
%16811 = torch.aten.add.Scalar %16803, %float9.999990e-06_14520, %int1_14521 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16812 = torch.aten.rsqrt %16804 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16813 = torch.aten.rsqrt %16805 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16814 = torch.aten.rsqrt %16806 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16815 = torch.aten.rsqrt %16807 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16816 = torch.aten.rsqrt %16808 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16817 = torch.aten.rsqrt %16809 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16818 = torch.aten.rsqrt %16810 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16819 = torch.aten.rsqrt %16811 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %16819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%16820 = torch.aten.mul.Tensor %16772, %16812 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16821 = torch.aten.mul.Tensor %16773, %16813 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16822 = torch.aten.mul.Tensor %16774, %16814 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16823 = torch.aten.mul.Tensor %16775, %16815 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16824 = torch.aten.mul.Tensor %16776, %16816 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16825 = torch.aten.mul.Tensor %16777, %16817 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16826 = torch.aten.mul.Tensor %16778, %16818 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16827 = torch.aten.mul.Tensor %16779, %16819 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16828 = torch.aten.mul.Tensor %552, %16820 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16829 = torch.aten.mul.Tensor %553, %16821 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16830 = torch.aten.mul.Tensor %554, %16822 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16831 = torch.aten.mul.Tensor %555, %16823 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16832 = torch.aten.mul.Tensor %556, %16824 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16833 = torch.aten.mul.Tensor %557, %16825 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16834 = torch.aten.mul.Tensor %558, %16826 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%16835 = torch.aten.mul.Tensor %559, %16827 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %16835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_14522 = torch.constant.int 5
%16836 = torch.prims.convert_element_type %16828, %int5_14522 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14523 = torch.constant.int 5
%16837 = torch.prims.convert_element_type %16829, %int5_14523 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14524 = torch.constant.int 5
%16838 = torch.prims.convert_element_type %16830, %int5_14524 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14525 = torch.constant.int 5
%16839 = torch.prims.convert_element_type %16831, %int5_14525 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14526 = torch.constant.int 5
%16840 = torch.prims.convert_element_type %16832, %int5_14526 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14527 = torch.constant.int 5
%16841 = torch.prims.convert_element_type %16833, %int5_14527 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14528 = torch.constant.int 5
%16842 = torch.prims.convert_element_type %16834, %int5_14528 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14529 = torch.constant.int 5
%16843 = torch.prims.convert_element_type %16835, %int5_14529 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %16843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14530 = torch.constant.int 1
%int0_14531 = torch.constant.int 0
%16844 = torch.prim.ListConstruct %int1_14530, %int0_14531 : (!torch.int, !torch.int) -> !torch.list<int>
%16845 = torch.aten.permute %560, %16844 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14532 = torch.constant.int 1
%int0_14533 = torch.constant.int 0
%16846 = torch.prim.ListConstruct %int1_14532, %int0_14533 : (!torch.int, !torch.int) -> !torch.list<int>
%16847 = torch.aten.permute %561, %16846 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14534 = torch.constant.int 1
%int0_14535 = torch.constant.int 0
%16848 = torch.prim.ListConstruct %int1_14534, %int0_14535 : (!torch.int, !torch.int) -> !torch.list<int>
%16849 = torch.aten.permute %562, %16848 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14536 = torch.constant.int 1
%int0_14537 = torch.constant.int 0
%16850 = torch.prim.ListConstruct %int1_14536, %int0_14537 : (!torch.int, !torch.int) -> !torch.list<int>
%16851 = torch.aten.permute %563, %16850 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14538 = torch.constant.int 1
%int0_14539 = torch.constant.int 0
%16852 = torch.prim.ListConstruct %int1_14538, %int0_14539 : (!torch.int, !torch.int) -> !torch.list<int>
%16853 = torch.aten.permute %564, %16852 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14540 = torch.constant.int 1
%int0_14541 = torch.constant.int 0
%16854 = torch.prim.ListConstruct %int1_14540, %int0_14541 : (!torch.int, !torch.int) -> !torch.list<int>
%16855 = torch.aten.permute %565, %16854 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14542 = torch.constant.int 1
%int0_14543 = torch.constant.int 0
%16856 = torch.prim.ListConstruct %int1_14542, %int0_14543 : (!torch.int, !torch.int) -> !torch.list<int>
%16857 = torch.aten.permute %566, %16856 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14544 = torch.constant.int 1
%int0_14545 = torch.constant.int 0
%16858 = torch.prim.ListConstruct %int1_14544, %int0_14545 : (!torch.int, !torch.int) -> !torch.list<int>
%16859 = torch.aten.permute %567, %16858 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_14546 = torch.constant.int 4
%16860 = torch.aten.mul.int %int4_14546, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14547 = torch.constant.int 4096
%16861 = torch.prim.ListConstruct %16860, %int4096_14547 : (!torch.int, !torch.int) -> !torch.list<int>
%16862 = torch.aten.view %16836, %16861 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16862, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16863 = torch.aten.mm %16862, %16845 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16863, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14548 = torch.constant.int 4
%int1792_14549 = torch.constant.int 1792
%16864 = torch.prim.ListConstruct %int4_14548, %2482, %int1792_14549 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16865 = torch.aten.view %16863, %16864 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14550 = torch.constant.int 4
%16866 = torch.aten.mul.int %int4_14550, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14551 = torch.constant.int 4096
%16867 = torch.prim.ListConstruct %16866, %int4096_14551 : (!torch.int, !torch.int) -> !torch.list<int>
%16868 = torch.aten.view %16837, %16867 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16868, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16869 = torch.aten.mm %16868, %16847 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16869, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14552 = torch.constant.int 4
%int1792_14553 = torch.constant.int 1792
%16870 = torch.prim.ListConstruct %int4_14552, %2482, %int1792_14553 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16871 = torch.aten.view %16869, %16870 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14554 = torch.constant.int 4
%16872 = torch.aten.mul.int %int4_14554, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14555 = torch.constant.int 4096
%16873 = torch.prim.ListConstruct %16872, %int4096_14555 : (!torch.int, !torch.int) -> !torch.list<int>
%16874 = torch.aten.view %16838, %16873 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16874, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16875 = torch.aten.mm %16874, %16849 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16875, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14556 = torch.constant.int 4
%int1792_14557 = torch.constant.int 1792
%16876 = torch.prim.ListConstruct %int4_14556, %2482, %int1792_14557 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16877 = torch.aten.view %16875, %16876 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14558 = torch.constant.int 4
%16878 = torch.aten.mul.int %int4_14558, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14559 = torch.constant.int 4096
%16879 = torch.prim.ListConstruct %16878, %int4096_14559 : (!torch.int, !torch.int) -> !torch.list<int>
%16880 = torch.aten.view %16839, %16879 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16880, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16881 = torch.aten.mm %16880, %16851 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16881, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14560 = torch.constant.int 4
%int1792_14561 = torch.constant.int 1792
%16882 = torch.prim.ListConstruct %int4_14560, %2482, %int1792_14561 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16883 = torch.aten.view %16881, %16882 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14562 = torch.constant.int 4
%16884 = torch.aten.mul.int %int4_14562, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14563 = torch.constant.int 4096
%16885 = torch.prim.ListConstruct %16884, %int4096_14563 : (!torch.int, !torch.int) -> !torch.list<int>
%16886 = torch.aten.view %16840, %16885 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16886, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16887 = torch.aten.mm %16886, %16853 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16887, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14564 = torch.constant.int 4
%int1792_14565 = torch.constant.int 1792
%16888 = torch.prim.ListConstruct %int4_14564, %2482, %int1792_14565 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16889 = torch.aten.view %16887, %16888 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14566 = torch.constant.int 4
%16890 = torch.aten.mul.int %int4_14566, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14567 = torch.constant.int 4096
%16891 = torch.prim.ListConstruct %16890, %int4096_14567 : (!torch.int, !torch.int) -> !torch.list<int>
%16892 = torch.aten.view %16841, %16891 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16892, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16893 = torch.aten.mm %16892, %16855 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16893, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14568 = torch.constant.int 4
%int1792_14569 = torch.constant.int 1792
%16894 = torch.prim.ListConstruct %int4_14568, %2482, %int1792_14569 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16895 = torch.aten.view %16893, %16894 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14570 = torch.constant.int 4
%16896 = torch.aten.mul.int %int4_14570, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14571 = torch.constant.int 4096
%16897 = torch.prim.ListConstruct %16896, %int4096_14571 : (!torch.int, !torch.int) -> !torch.list<int>
%16898 = torch.aten.view %16842, %16897 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16898, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16899 = torch.aten.mm %16898, %16857 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16899, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14572 = torch.constant.int 4
%int1792_14573 = torch.constant.int 1792
%16900 = torch.prim.ListConstruct %int4_14572, %2482, %int1792_14573 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16901 = torch.aten.view %16899, %16900 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14574 = torch.constant.int 4
%16902 = torch.aten.mul.int %int4_14574, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14575 = torch.constant.int 4096
%16903 = torch.prim.ListConstruct %16902, %int4096_14575 : (!torch.int, !torch.int) -> !torch.list<int>
%16904 = torch.aten.view %16843, %16903 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16904, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16905 = torch.aten.mm %16904, %16859 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16905, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14576 = torch.constant.int 4
%int1792_14577 = torch.constant.int 1792
%16906 = torch.prim.ListConstruct %int4_14576, %2482, %int1792_14577 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16907 = torch.aten.view %16905, %16906 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16908 = torch.aten.silu %16865 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16909 = torch.aten.silu %16871 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16910 = torch.aten.silu %16877 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16911 = torch.aten.silu %16883 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16912 = torch.aten.silu %16889 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16913 = torch.aten.silu %16895 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16914 = torch.aten.silu %16901 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16915 = torch.aten.silu %16907 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_14578 = torch.constant.int 1
%int0_14579 = torch.constant.int 0
%16916 = torch.prim.ListConstruct %int1_14578, %int0_14579 : (!torch.int, !torch.int) -> !torch.list<int>
%16917 = torch.aten.permute %568, %16916 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14580 = torch.constant.int 1
%int0_14581 = torch.constant.int 0
%16918 = torch.prim.ListConstruct %int1_14580, %int0_14581 : (!torch.int, !torch.int) -> !torch.list<int>
%16919 = torch.aten.permute %569, %16918 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14582 = torch.constant.int 1
%int0_14583 = torch.constant.int 0
%16920 = torch.prim.ListConstruct %int1_14582, %int0_14583 : (!torch.int, !torch.int) -> !torch.list<int>
%16921 = torch.aten.permute %570, %16920 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14584 = torch.constant.int 1
%int0_14585 = torch.constant.int 0
%16922 = torch.prim.ListConstruct %int1_14584, %int0_14585 : (!torch.int, !torch.int) -> !torch.list<int>
%16923 = torch.aten.permute %571, %16922 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14586 = torch.constant.int 1
%int0_14587 = torch.constant.int 0
%16924 = torch.prim.ListConstruct %int1_14586, %int0_14587 : (!torch.int, !torch.int) -> !torch.list<int>
%16925 = torch.aten.permute %572, %16924 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14588 = torch.constant.int 1
%int0_14589 = torch.constant.int 0
%16926 = torch.prim.ListConstruct %int1_14588, %int0_14589 : (!torch.int, !torch.int) -> !torch.list<int>
%16927 = torch.aten.permute %573, %16926 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14590 = torch.constant.int 1
%int0_14591 = torch.constant.int 0
%16928 = torch.prim.ListConstruct %int1_14590, %int0_14591 : (!torch.int, !torch.int) -> !torch.list<int>
%16929 = torch.aten.permute %574, %16928 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_14592 = torch.constant.int 1
%int0_14593 = torch.constant.int 0
%16930 = torch.prim.ListConstruct %int1_14592, %int0_14593 : (!torch.int, !torch.int) -> !torch.list<int>
%16931 = torch.aten.permute %575, %16930 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_14594 = torch.constant.int 4
%16932 = torch.aten.mul.int %int4_14594, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14595 = torch.constant.int 4096
%16933 = torch.prim.ListConstruct %16932, %int4096_14595 : (!torch.int, !torch.int) -> !torch.list<int>
%16934 = torch.aten.view %16836, %16933 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16934, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16935 = torch.aten.mm %16934, %16917 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16935, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14596 = torch.constant.int 4
%int1792_14597 = torch.constant.int 1792
%16936 = torch.prim.ListConstruct %int4_14596, %2482, %int1792_14597 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16937 = torch.aten.view %16935, %16936 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14598 = torch.constant.int 4
%16938 = torch.aten.mul.int %int4_14598, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14599 = torch.constant.int 4096
%16939 = torch.prim.ListConstruct %16938, %int4096_14599 : (!torch.int, !torch.int) -> !torch.list<int>
%16940 = torch.aten.view %16837, %16939 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16940, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16941 = torch.aten.mm %16940, %16919 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16941, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14600 = torch.constant.int 4
%int1792_14601 = torch.constant.int 1792
%16942 = torch.prim.ListConstruct %int4_14600, %2482, %int1792_14601 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16943 = torch.aten.view %16941, %16942 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14602 = torch.constant.int 4
%16944 = torch.aten.mul.int %int4_14602, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14603 = torch.constant.int 4096
%16945 = torch.prim.ListConstruct %16944, %int4096_14603 : (!torch.int, !torch.int) -> !torch.list<int>
%16946 = torch.aten.view %16838, %16945 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16946, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16947 = torch.aten.mm %16946, %16921 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16947, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14604 = torch.constant.int 4
%int1792_14605 = torch.constant.int 1792
%16948 = torch.prim.ListConstruct %int4_14604, %2482, %int1792_14605 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16949 = torch.aten.view %16947, %16948 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14606 = torch.constant.int 4
%16950 = torch.aten.mul.int %int4_14606, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14607 = torch.constant.int 4096
%16951 = torch.prim.ListConstruct %16950, %int4096_14607 : (!torch.int, !torch.int) -> !torch.list<int>
%16952 = torch.aten.view %16839, %16951 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16952, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16953 = torch.aten.mm %16952, %16923 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16953, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14608 = torch.constant.int 4
%int1792_14609 = torch.constant.int 1792
%16954 = torch.prim.ListConstruct %int4_14608, %2482, %int1792_14609 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16955 = torch.aten.view %16953, %16954 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14610 = torch.constant.int 4
%16956 = torch.aten.mul.int %int4_14610, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14611 = torch.constant.int 4096
%16957 = torch.prim.ListConstruct %16956, %int4096_14611 : (!torch.int, !torch.int) -> !torch.list<int>
%16958 = torch.aten.view %16840, %16957 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16958, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16959 = torch.aten.mm %16958, %16925 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16959, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14612 = torch.constant.int 4
%int1792_14613 = torch.constant.int 1792
%16960 = torch.prim.ListConstruct %int4_14612, %2482, %int1792_14613 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16961 = torch.aten.view %16959, %16960 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14614 = torch.constant.int 4
%16962 = torch.aten.mul.int %int4_14614, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14615 = torch.constant.int 4096
%16963 = torch.prim.ListConstruct %16962, %int4096_14615 : (!torch.int, !torch.int) -> !torch.list<int>
%16964 = torch.aten.view %16841, %16963 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16964, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16965 = torch.aten.mm %16964, %16927 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16965, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14616 = torch.constant.int 4
%int1792_14617 = torch.constant.int 1792
%16966 = torch.prim.ListConstruct %int4_14616, %2482, %int1792_14617 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16967 = torch.aten.view %16965, %16966 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14618 = torch.constant.int 4
%16968 = torch.aten.mul.int %int4_14618, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14619 = torch.constant.int 4096
%16969 = torch.prim.ListConstruct %16968, %int4096_14619 : (!torch.int, !torch.int) -> !torch.list<int>
%16970 = torch.aten.view %16842, %16969 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16970, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16971 = torch.aten.mm %16970, %16929 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16971, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14620 = torch.constant.int 4
%int1792_14621 = torch.constant.int 1792
%16972 = torch.prim.ListConstruct %int4_14620, %2482, %int1792_14621 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16973 = torch.aten.view %16971, %16972 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_14622 = torch.constant.int 4
%16974 = torch.aten.mul.int %int4_14622, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14623 = torch.constant.int 4096
%16975 = torch.prim.ListConstruct %16974, %int4096_14623 : (!torch.int, !torch.int) -> !torch.list<int>
%16976 = torch.aten.view %16843, %16975 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %16976, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%16977 = torch.aten.mm %16976, %16931 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %16977, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_14624 = torch.constant.int 4
%int1792_14625 = torch.constant.int 1792
%16978 = torch.prim.ListConstruct %int4_14624, %2482, %int1792_14625 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%16979 = torch.aten.view %16977, %16978 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16980 = torch.aten.mul.Tensor %16908, %16937 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16981 = torch.aten.mul.Tensor %16909, %16943 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16982 = torch.aten.mul.Tensor %16910, %16949 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16983 = torch.aten.mul.Tensor %16911, %16955 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16984 = torch.aten.mul.Tensor %16912, %16961 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16985 = torch.aten.mul.Tensor %16913, %16967 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16985, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16986 = torch.aten.mul.Tensor %16914, %16973 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%16987 = torch.aten.mul.Tensor %16915, %16979 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %16987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_14626 = torch.constant.int 1
%int0_14627 = torch.constant.int 0
%16988 = torch.prim.ListConstruct %int1_14626, %int0_14627 : (!torch.int, !torch.int) -> !torch.list<int>
%16989 = torch.aten.permute %576, %16988 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14628 = torch.constant.int 1
%int0_14629 = torch.constant.int 0
%16990 = torch.prim.ListConstruct %int1_14628, %int0_14629 : (!torch.int, !torch.int) -> !torch.list<int>
%16991 = torch.aten.permute %577, %16990 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14630 = torch.constant.int 1
%int0_14631 = torch.constant.int 0
%16992 = torch.prim.ListConstruct %int1_14630, %int0_14631 : (!torch.int, !torch.int) -> !torch.list<int>
%16993 = torch.aten.permute %578, %16992 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14632 = torch.constant.int 1
%int0_14633 = torch.constant.int 0
%16994 = torch.prim.ListConstruct %int1_14632, %int0_14633 : (!torch.int, !torch.int) -> !torch.list<int>
%16995 = torch.aten.permute %579, %16994 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14634 = torch.constant.int 1
%int0_14635 = torch.constant.int 0
%16996 = torch.prim.ListConstruct %int1_14634, %int0_14635 : (!torch.int, !torch.int) -> !torch.list<int>
%16997 = torch.aten.permute %580, %16996 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14636 = torch.constant.int 1
%int0_14637 = torch.constant.int 0
%16998 = torch.prim.ListConstruct %int1_14636, %int0_14637 : (!torch.int, !torch.int) -> !torch.list<int>
%16999 = torch.aten.permute %581, %16998 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14638 = torch.constant.int 1
%int0_14639 = torch.constant.int 0
%17000 = torch.prim.ListConstruct %int1_14638, %int0_14639 : (!torch.int, !torch.int) -> !torch.list<int>
%17001 = torch.aten.permute %582, %17000 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14640 = torch.constant.int 1
%int0_14641 = torch.constant.int 0
%17002 = torch.prim.ListConstruct %int1_14640, %int0_14641 : (!torch.int, !torch.int) -> !torch.list<int>
%17003 = torch.aten.permute %583, %17002 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_14642 = torch.constant.int 1
%17004 = torch.aten.size.int %16865, %int1_14642 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14643 = torch.constant.int 4
%17005 = torch.aten.mul.int %int4_14643, %17004 : !torch.int, !torch.int -> !torch.int
%int1792_14644 = torch.constant.int 1792
%17006 = torch.prim.ListConstruct %17005, %int1792_14644 : (!torch.int, !torch.int) -> !torch.list<int>
%17007 = torch.aten.view %16980, %17006 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17007, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17008 = torch.aten.mm %17007, %16989 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17008, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14645 = torch.constant.int 4
%int4096_14646 = torch.constant.int 4096
%17009 = torch.prim.ListConstruct %int4_14645, %17004, %int4096_14646 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17010 = torch.aten.view %17008, %17009 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14647 = torch.constant.int 1
%17011 = torch.aten.size.int %16871, %int1_14647 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14648 = torch.constant.int 4
%17012 = torch.aten.mul.int %int4_14648, %17011 : !torch.int, !torch.int -> !torch.int
%int1792_14649 = torch.constant.int 1792
%17013 = torch.prim.ListConstruct %17012, %int1792_14649 : (!torch.int, !torch.int) -> !torch.list<int>
%17014 = torch.aten.view %16981, %17013 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17014, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17015 = torch.aten.mm %17014, %16991 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17015, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14650 = torch.constant.int 4
%int4096_14651 = torch.constant.int 4096
%17016 = torch.prim.ListConstruct %int4_14650, %17011, %int4096_14651 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17017 = torch.aten.view %17015, %17016 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14652 = torch.constant.int 1
%17018 = torch.aten.size.int %16877, %int1_14652 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14653 = torch.constant.int 4
%17019 = torch.aten.mul.int %int4_14653, %17018 : !torch.int, !torch.int -> !torch.int
%int1792_14654 = torch.constant.int 1792
%17020 = torch.prim.ListConstruct %17019, %int1792_14654 : (!torch.int, !torch.int) -> !torch.list<int>
%17021 = torch.aten.view %16982, %17020 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17021, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17022 = torch.aten.mm %17021, %16993 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17022, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14655 = torch.constant.int 4
%int4096_14656 = torch.constant.int 4096
%17023 = torch.prim.ListConstruct %int4_14655, %17018, %int4096_14656 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17024 = torch.aten.view %17022, %17023 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14657 = torch.constant.int 1
%17025 = torch.aten.size.int %16883, %int1_14657 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14658 = torch.constant.int 4
%17026 = torch.aten.mul.int %int4_14658, %17025 : !torch.int, !torch.int -> !torch.int
%int1792_14659 = torch.constant.int 1792
%17027 = torch.prim.ListConstruct %17026, %int1792_14659 : (!torch.int, !torch.int) -> !torch.list<int>
%17028 = torch.aten.view %16983, %17027 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17028, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17029 = torch.aten.mm %17028, %16995 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17029, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14660 = torch.constant.int 4
%int4096_14661 = torch.constant.int 4096
%17030 = torch.prim.ListConstruct %int4_14660, %17025, %int4096_14661 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17031 = torch.aten.view %17029, %17030 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14662 = torch.constant.int 1
%17032 = torch.aten.size.int %16889, %int1_14662 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14663 = torch.constant.int 4
%17033 = torch.aten.mul.int %int4_14663, %17032 : !torch.int, !torch.int -> !torch.int
%int1792_14664 = torch.constant.int 1792
%17034 = torch.prim.ListConstruct %17033, %int1792_14664 : (!torch.int, !torch.int) -> !torch.list<int>
%17035 = torch.aten.view %16984, %17034 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17035, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17036 = torch.aten.mm %17035, %16997 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17036, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14665 = torch.constant.int 4
%int4096_14666 = torch.constant.int 4096
%17037 = torch.prim.ListConstruct %int4_14665, %17032, %int4096_14666 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17038 = torch.aten.view %17036, %17037 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14667 = torch.constant.int 1
%17039 = torch.aten.size.int %16895, %int1_14667 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14668 = torch.constant.int 4
%17040 = torch.aten.mul.int %int4_14668, %17039 : !torch.int, !torch.int -> !torch.int
%int1792_14669 = torch.constant.int 1792
%17041 = torch.prim.ListConstruct %17040, %int1792_14669 : (!torch.int, !torch.int) -> !torch.list<int>
%17042 = torch.aten.view %16985, %17041 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17042, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17043 = torch.aten.mm %17042, %16999 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17043, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14670 = torch.constant.int 4
%int4096_14671 = torch.constant.int 4096
%17044 = torch.prim.ListConstruct %int4_14670, %17039, %int4096_14671 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17045 = torch.aten.view %17043, %17044 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14672 = torch.constant.int 1
%17046 = torch.aten.size.int %16901, %int1_14672 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14673 = torch.constant.int 4
%17047 = torch.aten.mul.int %int4_14673, %17046 : !torch.int, !torch.int -> !torch.int
%int1792_14674 = torch.constant.int 1792
%17048 = torch.prim.ListConstruct %17047, %int1792_14674 : (!torch.int, !torch.int) -> !torch.list<int>
%17049 = torch.aten.view %16986, %17048 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17049, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17050 = torch.aten.mm %17049, %17001 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17050, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14675 = torch.constant.int 4
%int4096_14676 = torch.constant.int 4096
%17051 = torch.prim.ListConstruct %int4_14675, %17046, %int4096_14676 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17052 = torch.aten.view %17050, %17051 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14677 = torch.constant.int 1
%17053 = torch.aten.size.int %16907, %int1_14677 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_14678 = torch.constant.int 4
%17054 = torch.aten.mul.int %int4_14678, %17053 : !torch.int, !torch.int -> !torch.int
%int1792_14679 = torch.constant.int 1792
%17055 = torch.prim.ListConstruct %17054, %int1792_14679 : (!torch.int, !torch.int) -> !torch.list<int>
%17056 = torch.aten.view %16987, %17055 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %17056, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%17057 = torch.aten.mm %17056, %17003 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17057, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_14680 = torch.constant.int 4
%int4096_14681 = torch.constant.int 4096
%17058 = torch.prim.ListConstruct %int4_14680, %17053, %int4096_14681 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17059 = torch.aten.view %17057, %17058 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17060 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14682 = arith.constant 1 : index
%dim_14683 = tensor.dim %17060, %c1_14682 : tensor<4x?x4096xf16>
%17061 = flow.tensor.transfer %17060 : tensor<4x?x4096xf16>{%dim_14683} to #hal.device.promise<@__device_0>
%17062 = torch_c.from_builtin_tensor %17061 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17063 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14684 = arith.constant 1 : index
%dim_14685 = tensor.dim %17063, %c1_14684 : tensor<4x?x4096xf16>
%17064 = flow.tensor.transfer %17063 : tensor<4x?x4096xf16>{%dim_14685} to #hal.device.promise<@__device_0>
%17065 = torch_c.from_builtin_tensor %17064 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17066 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14686 = arith.constant 1 : index
%dim_14687 = tensor.dim %17066, %c1_14686 : tensor<4x?x4096xf16>
%17067 = flow.tensor.transfer %17066 : tensor<4x?x4096xf16>{%dim_14687} to #hal.device.promise<@__device_0>
%17068 = torch_c.from_builtin_tensor %17067 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17069 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14688 = arith.constant 1 : index
%dim_14689 = tensor.dim %17069, %c1_14688 : tensor<4x?x4096xf16>
%17070 = flow.tensor.transfer %17069 : tensor<4x?x4096xf16>{%dim_14689} to #hal.device.promise<@__device_0>
%17071 = torch_c.from_builtin_tensor %17070 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17072 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14690 = arith.constant 1 : index
%dim_14691 = tensor.dim %17072, %c1_14690 : tensor<4x?x4096xf16>
%17073 = flow.tensor.transfer %17072 : tensor<4x?x4096xf16>{%dim_14691} to #hal.device.promise<@__device_0>
%17074 = torch_c.from_builtin_tensor %17073 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17075 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14692 = arith.constant 1 : index
%dim_14693 = tensor.dim %17075, %c1_14692 : tensor<4x?x4096xf16>
%17076 = flow.tensor.transfer %17075 : tensor<4x?x4096xf16>{%dim_14693} to #hal.device.promise<@__device_0>
%17077 = torch_c.from_builtin_tensor %17076 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17078 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14694 = arith.constant 1 : index
%dim_14695 = tensor.dim %17078, %c1_14694 : tensor<4x?x4096xf16>
%17079 = flow.tensor.transfer %17078 : tensor<4x?x4096xf16>{%dim_14695} to #hal.device.promise<@__device_0>
%17080 = torch_c.from_builtin_tensor %17079 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14696 = torch.constant.int 1
%17081 = torch.aten.add.Tensor %17010, %17062, %int1_14696 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14697 = torch.constant.int 1
%17082 = torch.aten.add.Tensor %17081, %17065, %int1_14697 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14698 = torch.constant.int 1
%17083 = torch.aten.add.Tensor %17082, %17068, %int1_14698 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14699 = torch.constant.int 1
%17084 = torch.aten.add.Tensor %17083, %17071, %int1_14699 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14700 = torch.constant.int 1
%17085 = torch.aten.add.Tensor %17084, %17074, %int1_14700 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14701 = torch.constant.int 1
%17086 = torch.aten.add.Tensor %17085, %17077, %int1_14701 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14702 = torch.constant.int 1
%17087 = torch.aten.add.Tensor %17086, %17080, %int1_14702 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17088 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14703 = arith.constant 1 : index
%dim_14704 = tensor.dim %17088, %c1_14703 : tensor<4x?x4096xf16>
%17089 = flow.tensor.transfer %17088 : tensor<4x?x4096xf16>{%dim_14704} to #hal.device.promise<@__device_1>
%17090 = torch_c.from_builtin_tensor %17089 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17091 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14705 = arith.constant 1 : index
%dim_14706 = tensor.dim %17091, %c1_14705 : tensor<4x?x4096xf16>
%17092 = flow.tensor.transfer %17091 : tensor<4x?x4096xf16>{%dim_14706} to #hal.device.promise<@__device_1>
%17093 = torch_c.from_builtin_tensor %17092 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17094 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14707 = arith.constant 1 : index
%dim_14708 = tensor.dim %17094, %c1_14707 : tensor<4x?x4096xf16>
%17095 = flow.tensor.transfer %17094 : tensor<4x?x4096xf16>{%dim_14708} to #hal.device.promise<@__device_1>
%17096 = torch_c.from_builtin_tensor %17095 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17097 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14709 = arith.constant 1 : index
%dim_14710 = tensor.dim %17097, %c1_14709 : tensor<4x?x4096xf16>
%17098 = flow.tensor.transfer %17097 : tensor<4x?x4096xf16>{%dim_14710} to #hal.device.promise<@__device_1>
%17099 = torch_c.from_builtin_tensor %17098 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17100 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14711 = arith.constant 1 : index
%dim_14712 = tensor.dim %17100, %c1_14711 : tensor<4x?x4096xf16>
%17101 = flow.tensor.transfer %17100 : tensor<4x?x4096xf16>{%dim_14712} to #hal.device.promise<@__device_1>
%17102 = torch_c.from_builtin_tensor %17101 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17103 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14713 = arith.constant 1 : index
%dim_14714 = tensor.dim %17103, %c1_14713 : tensor<4x?x4096xf16>
%17104 = flow.tensor.transfer %17103 : tensor<4x?x4096xf16>{%dim_14714} to #hal.device.promise<@__device_1>
%17105 = torch_c.from_builtin_tensor %17104 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17106 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14715 = arith.constant 1 : index
%dim_14716 = tensor.dim %17106, %c1_14715 : tensor<4x?x4096xf16>
%17107 = flow.tensor.transfer %17106 : tensor<4x?x4096xf16>{%dim_14716} to #hal.device.promise<@__device_1>
%17108 = torch_c.from_builtin_tensor %17107 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14717 = torch.constant.int 1
%17109 = torch.aten.add.Tensor %17090, %17017, %int1_14717 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14718 = torch.constant.int 1
%17110 = torch.aten.add.Tensor %17109, %17093, %int1_14718 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14719 = torch.constant.int 1
%17111 = torch.aten.add.Tensor %17110, %17096, %int1_14719 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14720 = torch.constant.int 1
%17112 = torch.aten.add.Tensor %17111, %17099, %int1_14720 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14721 = torch.constant.int 1
%17113 = torch.aten.add.Tensor %17112, %17102, %int1_14721 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14722 = torch.constant.int 1
%17114 = torch.aten.add.Tensor %17113, %17105, %int1_14722 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14723 = torch.constant.int 1
%17115 = torch.aten.add.Tensor %17114, %17108, %int1_14723 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17116 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14724 = arith.constant 1 : index
%dim_14725 = tensor.dim %17116, %c1_14724 : tensor<4x?x4096xf16>
%17117 = flow.tensor.transfer %17116 : tensor<4x?x4096xf16>{%dim_14725} to #hal.device.promise<@__device_2>
%17118 = torch_c.from_builtin_tensor %17117 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17119 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14726 = arith.constant 1 : index
%dim_14727 = tensor.dim %17119, %c1_14726 : tensor<4x?x4096xf16>
%17120 = flow.tensor.transfer %17119 : tensor<4x?x4096xf16>{%dim_14727} to #hal.device.promise<@__device_2>
%17121 = torch_c.from_builtin_tensor %17120 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17122 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14728 = arith.constant 1 : index
%dim_14729 = tensor.dim %17122, %c1_14728 : tensor<4x?x4096xf16>
%17123 = flow.tensor.transfer %17122 : tensor<4x?x4096xf16>{%dim_14729} to #hal.device.promise<@__device_2>
%17124 = torch_c.from_builtin_tensor %17123 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17124, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17125 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14730 = arith.constant 1 : index
%dim_14731 = tensor.dim %17125, %c1_14730 : tensor<4x?x4096xf16>
%17126 = flow.tensor.transfer %17125 : tensor<4x?x4096xf16>{%dim_14731} to #hal.device.promise<@__device_2>
%17127 = torch_c.from_builtin_tensor %17126 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17128 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14732 = arith.constant 1 : index
%dim_14733 = tensor.dim %17128, %c1_14732 : tensor<4x?x4096xf16>
%17129 = flow.tensor.transfer %17128 : tensor<4x?x4096xf16>{%dim_14733} to #hal.device.promise<@__device_2>
%17130 = torch_c.from_builtin_tensor %17129 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17131 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14734 = arith.constant 1 : index
%dim_14735 = tensor.dim %17131, %c1_14734 : tensor<4x?x4096xf16>
%17132 = flow.tensor.transfer %17131 : tensor<4x?x4096xf16>{%dim_14735} to #hal.device.promise<@__device_2>
%17133 = torch_c.from_builtin_tensor %17132 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17134 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14736 = arith.constant 1 : index
%dim_14737 = tensor.dim %17134, %c1_14736 : tensor<4x?x4096xf16>
%17135 = flow.tensor.transfer %17134 : tensor<4x?x4096xf16>{%dim_14737} to #hal.device.promise<@__device_2>
%17136 = torch_c.from_builtin_tensor %17135 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14738 = torch.constant.int 1
%17137 = torch.aten.add.Tensor %17118, %17121, %int1_14738 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14739 = torch.constant.int 1
%17138 = torch.aten.add.Tensor %17137, %17024, %int1_14739 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14740 = torch.constant.int 1
%17139 = torch.aten.add.Tensor %17138, %17124, %int1_14740 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14741 = torch.constant.int 1
%17140 = torch.aten.add.Tensor %17139, %17127, %int1_14741 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14742 = torch.constant.int 1
%17141 = torch.aten.add.Tensor %17140, %17130, %int1_14742 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14743 = torch.constant.int 1
%17142 = torch.aten.add.Tensor %17141, %17133, %int1_14743 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14744 = torch.constant.int 1
%17143 = torch.aten.add.Tensor %17142, %17136, %int1_14744 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17144 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14745 = arith.constant 1 : index
%dim_14746 = tensor.dim %17144, %c1_14745 : tensor<4x?x4096xf16>
%17145 = flow.tensor.transfer %17144 : tensor<4x?x4096xf16>{%dim_14746} to #hal.device.promise<@__device_3>
%17146 = torch_c.from_builtin_tensor %17145 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17147 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14747 = arith.constant 1 : index
%dim_14748 = tensor.dim %17147, %c1_14747 : tensor<4x?x4096xf16>
%17148 = flow.tensor.transfer %17147 : tensor<4x?x4096xf16>{%dim_14748} to #hal.device.promise<@__device_3>
%17149 = torch_c.from_builtin_tensor %17148 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17150 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14749 = arith.constant 1 : index
%dim_14750 = tensor.dim %17150, %c1_14749 : tensor<4x?x4096xf16>
%17151 = flow.tensor.transfer %17150 : tensor<4x?x4096xf16>{%dim_14750} to #hal.device.promise<@__device_3>
%17152 = torch_c.from_builtin_tensor %17151 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17153 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14751 = arith.constant 1 : index
%dim_14752 = tensor.dim %17153, %c1_14751 : tensor<4x?x4096xf16>
%17154 = flow.tensor.transfer %17153 : tensor<4x?x4096xf16>{%dim_14752} to #hal.device.promise<@__device_3>
%17155 = torch_c.from_builtin_tensor %17154 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17156 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14753 = arith.constant 1 : index
%dim_14754 = tensor.dim %17156, %c1_14753 : tensor<4x?x4096xf16>
%17157 = flow.tensor.transfer %17156 : tensor<4x?x4096xf16>{%dim_14754} to #hal.device.promise<@__device_3>
%17158 = torch_c.from_builtin_tensor %17157 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17159 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14755 = arith.constant 1 : index
%dim_14756 = tensor.dim %17159, %c1_14755 : tensor<4x?x4096xf16>
%17160 = flow.tensor.transfer %17159 : tensor<4x?x4096xf16>{%dim_14756} to #hal.device.promise<@__device_3>
%17161 = torch_c.from_builtin_tensor %17160 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17162 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14757 = arith.constant 1 : index
%dim_14758 = tensor.dim %17162, %c1_14757 : tensor<4x?x4096xf16>
%17163 = flow.tensor.transfer %17162 : tensor<4x?x4096xf16>{%dim_14758} to #hal.device.promise<@__device_3>
%17164 = torch_c.from_builtin_tensor %17163 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14759 = torch.constant.int 1
%17165 = torch.aten.add.Tensor %17146, %17149, %int1_14759 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14760 = torch.constant.int 1
%17166 = torch.aten.add.Tensor %17165, %17152, %int1_14760 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14761 = torch.constant.int 1
%17167 = torch.aten.add.Tensor %17166, %17031, %int1_14761 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14762 = torch.constant.int 1
%17168 = torch.aten.add.Tensor %17167, %17155, %int1_14762 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14763 = torch.constant.int 1
%17169 = torch.aten.add.Tensor %17168, %17158, %int1_14763 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14764 = torch.constant.int 1
%17170 = torch.aten.add.Tensor %17169, %17161, %int1_14764 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14765 = torch.constant.int 1
%17171 = torch.aten.add.Tensor %17170, %17164, %int1_14765 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17172 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14766 = arith.constant 1 : index
%dim_14767 = tensor.dim %17172, %c1_14766 : tensor<4x?x4096xf16>
%17173 = flow.tensor.transfer %17172 : tensor<4x?x4096xf16>{%dim_14767} to #hal.device.promise<@__device_4>
%17174 = torch_c.from_builtin_tensor %17173 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17175 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14768 = arith.constant 1 : index
%dim_14769 = tensor.dim %17175, %c1_14768 : tensor<4x?x4096xf16>
%17176 = flow.tensor.transfer %17175 : tensor<4x?x4096xf16>{%dim_14769} to #hal.device.promise<@__device_4>
%17177 = torch_c.from_builtin_tensor %17176 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17178 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14770 = arith.constant 1 : index
%dim_14771 = tensor.dim %17178, %c1_14770 : tensor<4x?x4096xf16>
%17179 = flow.tensor.transfer %17178 : tensor<4x?x4096xf16>{%dim_14771} to #hal.device.promise<@__device_4>
%17180 = torch_c.from_builtin_tensor %17179 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17181 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14772 = arith.constant 1 : index
%dim_14773 = tensor.dim %17181, %c1_14772 : tensor<4x?x4096xf16>
%17182 = flow.tensor.transfer %17181 : tensor<4x?x4096xf16>{%dim_14773} to #hal.device.promise<@__device_4>
%17183 = torch_c.from_builtin_tensor %17182 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17184 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14774 = arith.constant 1 : index
%dim_14775 = tensor.dim %17184, %c1_14774 : tensor<4x?x4096xf16>
%17185 = flow.tensor.transfer %17184 : tensor<4x?x4096xf16>{%dim_14775} to #hal.device.promise<@__device_4>
%17186 = torch_c.from_builtin_tensor %17185 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17187 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14776 = arith.constant 1 : index
%dim_14777 = tensor.dim %17187, %c1_14776 : tensor<4x?x4096xf16>
%17188 = flow.tensor.transfer %17187 : tensor<4x?x4096xf16>{%dim_14777} to #hal.device.promise<@__device_4>
%17189 = torch_c.from_builtin_tensor %17188 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17190 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14778 = arith.constant 1 : index
%dim_14779 = tensor.dim %17190, %c1_14778 : tensor<4x?x4096xf16>
%17191 = flow.tensor.transfer %17190 : tensor<4x?x4096xf16>{%dim_14779} to #hal.device.promise<@__device_4>
%17192 = torch_c.from_builtin_tensor %17191 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14780 = torch.constant.int 1
%17193 = torch.aten.add.Tensor %17174, %17177, %int1_14780 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14781 = torch.constant.int 1
%17194 = torch.aten.add.Tensor %17193, %17180, %int1_14781 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14782 = torch.constant.int 1
%17195 = torch.aten.add.Tensor %17194, %17183, %int1_14782 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14783 = torch.constant.int 1
%17196 = torch.aten.add.Tensor %17195, %17038, %int1_14783 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14784 = torch.constant.int 1
%17197 = torch.aten.add.Tensor %17196, %17186, %int1_14784 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14785 = torch.constant.int 1
%17198 = torch.aten.add.Tensor %17197, %17189, %int1_14785 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14786 = torch.constant.int 1
%17199 = torch.aten.add.Tensor %17198, %17192, %int1_14786 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17200 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14787 = arith.constant 1 : index
%dim_14788 = tensor.dim %17200, %c1_14787 : tensor<4x?x4096xf16>
%17201 = flow.tensor.transfer %17200 : tensor<4x?x4096xf16>{%dim_14788} to #hal.device.promise<@__device_5>
%17202 = torch_c.from_builtin_tensor %17201 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17203 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14789 = arith.constant 1 : index
%dim_14790 = tensor.dim %17203, %c1_14789 : tensor<4x?x4096xf16>
%17204 = flow.tensor.transfer %17203 : tensor<4x?x4096xf16>{%dim_14790} to #hal.device.promise<@__device_5>
%17205 = torch_c.from_builtin_tensor %17204 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17206 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14791 = arith.constant 1 : index
%dim_14792 = tensor.dim %17206, %c1_14791 : tensor<4x?x4096xf16>
%17207 = flow.tensor.transfer %17206 : tensor<4x?x4096xf16>{%dim_14792} to #hal.device.promise<@__device_5>
%17208 = torch_c.from_builtin_tensor %17207 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17209 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14793 = arith.constant 1 : index
%dim_14794 = tensor.dim %17209, %c1_14793 : tensor<4x?x4096xf16>
%17210 = flow.tensor.transfer %17209 : tensor<4x?x4096xf16>{%dim_14794} to #hal.device.promise<@__device_5>
%17211 = torch_c.from_builtin_tensor %17210 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17212 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14795 = arith.constant 1 : index
%dim_14796 = tensor.dim %17212, %c1_14795 : tensor<4x?x4096xf16>
%17213 = flow.tensor.transfer %17212 : tensor<4x?x4096xf16>{%dim_14796} to #hal.device.promise<@__device_5>
%17214 = torch_c.from_builtin_tensor %17213 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17215 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14797 = arith.constant 1 : index
%dim_14798 = tensor.dim %17215, %c1_14797 : tensor<4x?x4096xf16>
%17216 = flow.tensor.transfer %17215 : tensor<4x?x4096xf16>{%dim_14798} to #hal.device.promise<@__device_5>
%17217 = torch_c.from_builtin_tensor %17216 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17218 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14799 = arith.constant 1 : index
%dim_14800 = tensor.dim %17218, %c1_14799 : tensor<4x?x4096xf16>
%17219 = flow.tensor.transfer %17218 : tensor<4x?x4096xf16>{%dim_14800} to #hal.device.promise<@__device_5>
%17220 = torch_c.from_builtin_tensor %17219 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14801 = torch.constant.int 1
%17221 = torch.aten.add.Tensor %17202, %17205, %int1_14801 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14802 = torch.constant.int 1
%17222 = torch.aten.add.Tensor %17221, %17208, %int1_14802 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14803 = torch.constant.int 1
%17223 = torch.aten.add.Tensor %17222, %17211, %int1_14803 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14804 = torch.constant.int 1
%17224 = torch.aten.add.Tensor %17223, %17214, %int1_14804 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14805 = torch.constant.int 1
%17225 = torch.aten.add.Tensor %17224, %17045, %int1_14805 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14806 = torch.constant.int 1
%17226 = torch.aten.add.Tensor %17225, %17217, %int1_14806 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14807 = torch.constant.int 1
%17227 = torch.aten.add.Tensor %17226, %17220, %int1_14807 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17228 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14808 = arith.constant 1 : index
%dim_14809 = tensor.dim %17228, %c1_14808 : tensor<4x?x4096xf16>
%17229 = flow.tensor.transfer %17228 : tensor<4x?x4096xf16>{%dim_14809} to #hal.device.promise<@__device_6>
%17230 = torch_c.from_builtin_tensor %17229 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17231 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14810 = arith.constant 1 : index
%dim_14811 = tensor.dim %17231, %c1_14810 : tensor<4x?x4096xf16>
%17232 = flow.tensor.transfer %17231 : tensor<4x?x4096xf16>{%dim_14811} to #hal.device.promise<@__device_6>
%17233 = torch_c.from_builtin_tensor %17232 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17234 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14812 = arith.constant 1 : index
%dim_14813 = tensor.dim %17234, %c1_14812 : tensor<4x?x4096xf16>
%17235 = flow.tensor.transfer %17234 : tensor<4x?x4096xf16>{%dim_14813} to #hal.device.promise<@__device_6>
%17236 = torch_c.from_builtin_tensor %17235 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17236, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17237 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14814 = arith.constant 1 : index
%dim_14815 = tensor.dim %17237, %c1_14814 : tensor<4x?x4096xf16>
%17238 = flow.tensor.transfer %17237 : tensor<4x?x4096xf16>{%dim_14815} to #hal.device.promise<@__device_6>
%17239 = torch_c.from_builtin_tensor %17238 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17240 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14816 = arith.constant 1 : index
%dim_14817 = tensor.dim %17240, %c1_14816 : tensor<4x?x4096xf16>
%17241 = flow.tensor.transfer %17240 : tensor<4x?x4096xf16>{%dim_14817} to #hal.device.promise<@__device_6>
%17242 = torch_c.from_builtin_tensor %17241 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17243 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14818 = arith.constant 1 : index
%dim_14819 = tensor.dim %17243, %c1_14818 : tensor<4x?x4096xf16>
%17244 = flow.tensor.transfer %17243 : tensor<4x?x4096xf16>{%dim_14819} to #hal.device.promise<@__device_6>
%17245 = torch_c.from_builtin_tensor %17244 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17246 = torch_c.to_builtin_tensor %17059 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14820 = arith.constant 1 : index
%dim_14821 = tensor.dim %17246, %c1_14820 : tensor<4x?x4096xf16>
%17247 = flow.tensor.transfer %17246 : tensor<4x?x4096xf16>{%dim_14821} to #hal.device.promise<@__device_6>
%17248 = torch_c.from_builtin_tensor %17247 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14822 = torch.constant.int 1
%17249 = torch.aten.add.Tensor %17230, %17233, %int1_14822 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14823 = torch.constant.int 1
%17250 = torch.aten.add.Tensor %17249, %17236, %int1_14823 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14824 = torch.constant.int 1
%17251 = torch.aten.add.Tensor %17250, %17239, %int1_14824 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14825 = torch.constant.int 1
%17252 = torch.aten.add.Tensor %17251, %17242, %int1_14825 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14826 = torch.constant.int 1
%17253 = torch.aten.add.Tensor %17252, %17245, %int1_14826 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14827 = torch.constant.int 1
%17254 = torch.aten.add.Tensor %17253, %17052, %int1_14827 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14828 = torch.constant.int 1
%17255 = torch.aten.add.Tensor %17254, %17248, %int1_14828 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17256 = torch_c.to_builtin_tensor %17010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14829 = arith.constant 1 : index
%dim_14830 = tensor.dim %17256, %c1_14829 : tensor<4x?x4096xf16>
%17257 = flow.tensor.transfer %17256 : tensor<4x?x4096xf16>{%dim_14830} to #hal.device.promise<@__device_7>
%17258 = torch_c.from_builtin_tensor %17257 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17259 = torch_c.to_builtin_tensor %17017 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14831 = arith.constant 1 : index
%dim_14832 = tensor.dim %17259, %c1_14831 : tensor<4x?x4096xf16>
%17260 = flow.tensor.transfer %17259 : tensor<4x?x4096xf16>{%dim_14832} to #hal.device.promise<@__device_7>
%17261 = torch_c.from_builtin_tensor %17260 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17262 = torch_c.to_builtin_tensor %17024 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14833 = arith.constant 1 : index
%dim_14834 = tensor.dim %17262, %c1_14833 : tensor<4x?x4096xf16>
%17263 = flow.tensor.transfer %17262 : tensor<4x?x4096xf16>{%dim_14834} to #hal.device.promise<@__device_7>
%17264 = torch_c.from_builtin_tensor %17263 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17265 = torch_c.to_builtin_tensor %17031 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14835 = arith.constant 1 : index
%dim_14836 = tensor.dim %17265, %c1_14835 : tensor<4x?x4096xf16>
%17266 = flow.tensor.transfer %17265 : tensor<4x?x4096xf16>{%dim_14836} to #hal.device.promise<@__device_7>
%17267 = torch_c.from_builtin_tensor %17266 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17268 = torch_c.to_builtin_tensor %17038 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14837 = arith.constant 1 : index
%dim_14838 = tensor.dim %17268, %c1_14837 : tensor<4x?x4096xf16>
%17269 = flow.tensor.transfer %17268 : tensor<4x?x4096xf16>{%dim_14838} to #hal.device.promise<@__device_7>
%17270 = torch_c.from_builtin_tensor %17269 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17271 = torch_c.to_builtin_tensor %17045 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14839 = arith.constant 1 : index
%dim_14840 = tensor.dim %17271, %c1_14839 : tensor<4x?x4096xf16>
%17272 = flow.tensor.transfer %17271 : tensor<4x?x4096xf16>{%dim_14840} to #hal.device.promise<@__device_7>
%17273 = torch_c.from_builtin_tensor %17272 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%17274 = torch_c.to_builtin_tensor %17052 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_14841 = arith.constant 1 : index
%dim_14842 = tensor.dim %17274, %c1_14841 : tensor<4x?x4096xf16>
%17275 = flow.tensor.transfer %17274 : tensor<4x?x4096xf16>{%dim_14842} to #hal.device.promise<@__device_7>
%17276 = torch_c.from_builtin_tensor %17275 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14843 = torch.constant.int 1
%17277 = torch.aten.add.Tensor %17258, %17261, %int1_14843 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14844 = torch.constant.int 1
%17278 = torch.aten.add.Tensor %17277, %17264, %int1_14844 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14845 = torch.constant.int 1
%17279 = torch.aten.add.Tensor %17278, %17267, %int1_14845 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14846 = torch.constant.int 1
%17280 = torch.aten.add.Tensor %17279, %17270, %int1_14846 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14847 = torch.constant.int 1
%17281 = torch.aten.add.Tensor %17280, %17273, %int1_14847 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14848 = torch.constant.int 1
%17282 = torch.aten.add.Tensor %17281, %17276, %int1_14848 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14849 = torch.constant.int 1
%17283 = torch.aten.add.Tensor %17282, %17059, %int1_14849 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14850 = torch.constant.int 1
%17284 = torch.aten.add.Tensor %16764, %17087, %int1_14850 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14851 = torch.constant.int 1
%17285 = torch.aten.add.Tensor %16765, %17115, %int1_14851 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14852 = torch.constant.int 1
%17286 = torch.aten.add.Tensor %16766, %17143, %int1_14852 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14853 = torch.constant.int 1
%17287 = torch.aten.add.Tensor %16767, %17171, %int1_14853 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14854 = torch.constant.int 1
%17288 = torch.aten.add.Tensor %16768, %17199, %int1_14854 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14855 = torch.constant.int 1
%17289 = torch.aten.add.Tensor %16769, %17227, %int1_14855 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14856 = torch.constant.int 1
%17290 = torch.aten.add.Tensor %16770, %17255, %int1_14856 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14857 = torch.constant.int 1
%17291 = torch.aten.add.Tensor %16771, %17283, %int1_14857 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_14858 = torch.constant.int 6
%17292 = torch.prims.convert_element_type %17284, %int6_14858 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14859 = torch.constant.int 6
%17293 = torch.prims.convert_element_type %17285, %int6_14859 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14860 = torch.constant.int 6
%17294 = torch.prims.convert_element_type %17286, %int6_14860 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14861 = torch.constant.int 6
%17295 = torch.prims.convert_element_type %17287, %int6_14861 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14862 = torch.constant.int 6
%17296 = torch.prims.convert_element_type %17288, %int6_14862 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14863 = torch.constant.int 6
%17297 = torch.prims.convert_element_type %17289, %int6_14863 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14864 = torch.constant.int 6
%17298 = torch.prims.convert_element_type %17290, %int6_14864 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_14865 = torch.constant.int 6
%17299 = torch.prims.convert_element_type %17291, %int6_14865 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14866 = torch.constant.int 2
%17300 = torch.aten.pow.Tensor_Scalar %17292, %int2_14866 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14867 = torch.constant.int 2
%17301 = torch.aten.pow.Tensor_Scalar %17293, %int2_14867 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14868 = torch.constant.int 2
%17302 = torch.aten.pow.Tensor_Scalar %17294, %int2_14868 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14869 = torch.constant.int 2
%17303 = torch.aten.pow.Tensor_Scalar %17295, %int2_14869 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14870 = torch.constant.int 2
%17304 = torch.aten.pow.Tensor_Scalar %17296, %int2_14870 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14871 = torch.constant.int 2
%17305 = torch.aten.pow.Tensor_Scalar %17297, %int2_14871 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14872 = torch.constant.int 2
%17306 = torch.aten.pow.Tensor_Scalar %17298, %int2_14872 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_14873 = torch.constant.int 2
%17307 = torch.aten.pow.Tensor_Scalar %17299, %int2_14873 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_14874 = torch.constant.int -1
%17308 = torch.prim.ListConstruct %int-1_14874 : (!torch.int) -> !torch.list<int>
%true_14875 = torch.constant.bool true
%none_14876 = torch.constant.none
%17309 = torch.aten.mean.dim %17300, %17308, %true_14875, %none_14876 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14877 = torch.constant.int -1
%17310 = torch.prim.ListConstruct %int-1_14877 : (!torch.int) -> !torch.list<int>
%true_14878 = torch.constant.bool true
%none_14879 = torch.constant.none
%17311 = torch.aten.mean.dim %17301, %17310, %true_14878, %none_14879 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14880 = torch.constant.int -1
%17312 = torch.prim.ListConstruct %int-1_14880 : (!torch.int) -> !torch.list<int>
%true_14881 = torch.constant.bool true
%none_14882 = torch.constant.none
%17313 = torch.aten.mean.dim %17302, %17312, %true_14881, %none_14882 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14883 = torch.constant.int -1
%17314 = torch.prim.ListConstruct %int-1_14883 : (!torch.int) -> !torch.list<int>
%true_14884 = torch.constant.bool true
%none_14885 = torch.constant.none
%17315 = torch.aten.mean.dim %17303, %17314, %true_14884, %none_14885 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14886 = torch.constant.int -1
%17316 = torch.prim.ListConstruct %int-1_14886 : (!torch.int) -> !torch.list<int>
%true_14887 = torch.constant.bool true
%none_14888 = torch.constant.none
%17317 = torch.aten.mean.dim %17304, %17316, %true_14887, %none_14888 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14889 = torch.constant.int -1
%17318 = torch.prim.ListConstruct %int-1_14889 : (!torch.int) -> !torch.list<int>
%true_14890 = torch.constant.bool true
%none_14891 = torch.constant.none
%17319 = torch.aten.mean.dim %17305, %17318, %true_14890, %none_14891 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14892 = torch.constant.int -1
%17320 = torch.prim.ListConstruct %int-1_14892 : (!torch.int) -> !torch.list<int>
%true_14893 = torch.constant.bool true
%none_14894 = torch.constant.none
%17321 = torch.aten.mean.dim %17306, %17320, %true_14893, %none_14894 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_14895 = torch.constant.int -1
%17322 = torch.prim.ListConstruct %int-1_14895 : (!torch.int) -> !torch.list<int>
%true_14896 = torch.constant.bool true
%none_14897 = torch.constant.none
%17323 = torch.aten.mean.dim %17307, %17322, %true_14896, %none_14897 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14898 = torch.constant.float 9.9999997473787516E-6
%int1_14899 = torch.constant.int 1
%17324 = torch.aten.add.Scalar %17309, %float9.999990e-06_14898, %int1_14899 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14900 = torch.constant.float 9.9999997473787516E-6
%int1_14901 = torch.constant.int 1
%17325 = torch.aten.add.Scalar %17311, %float9.999990e-06_14900, %int1_14901 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14902 = torch.constant.float 9.9999997473787516E-6
%int1_14903 = torch.constant.int 1
%17326 = torch.aten.add.Scalar %17313, %float9.999990e-06_14902, %int1_14903 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14904 = torch.constant.float 9.9999997473787516E-6
%int1_14905 = torch.constant.int 1
%17327 = torch.aten.add.Scalar %17315, %float9.999990e-06_14904, %int1_14905 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14906 = torch.constant.float 9.9999997473787516E-6
%int1_14907 = torch.constant.int 1
%17328 = torch.aten.add.Scalar %17317, %float9.999990e-06_14906, %int1_14907 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14908 = torch.constant.float 9.9999997473787516E-6
%int1_14909 = torch.constant.int 1
%17329 = torch.aten.add.Scalar %17319, %float9.999990e-06_14908, %int1_14909 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14910 = torch.constant.float 9.9999997473787516E-6
%int1_14911 = torch.constant.int 1
%17330 = torch.aten.add.Scalar %17321, %float9.999990e-06_14910, %int1_14911 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_14912 = torch.constant.float 9.9999997473787516E-6
%int1_14913 = torch.constant.int 1
%17331 = torch.aten.add.Scalar %17323, %float9.999990e-06_14912, %int1_14913 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17332 = torch.aten.rsqrt %17324 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17333 = torch.aten.rsqrt %17325 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17334 = torch.aten.rsqrt %17326 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17335 = torch.aten.rsqrt %17327 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17336 = torch.aten.rsqrt %17328 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17337 = torch.aten.rsqrt %17329 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17338 = torch.aten.rsqrt %17330 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17339 = torch.aten.rsqrt %17331 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %17339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%17340 = torch.aten.mul.Tensor %17292, %17332 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17341 = torch.aten.mul.Tensor %17293, %17333 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17342 = torch.aten.mul.Tensor %17294, %17334 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17343 = torch.aten.mul.Tensor %17295, %17335 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17344 = torch.aten.mul.Tensor %17296, %17336 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17345 = torch.aten.mul.Tensor %17297, %17337 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17346 = torch.aten.mul.Tensor %17298, %17338 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17347 = torch.aten.mul.Tensor %17299, %17339 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17348 = torch.aten.mul.Tensor %584, %17340 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17349 = torch.aten.mul.Tensor %585, %17341 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17350 = torch.aten.mul.Tensor %586, %17342 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17351 = torch.aten.mul.Tensor %587, %17343 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17352 = torch.aten.mul.Tensor %588, %17344 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17353 = torch.aten.mul.Tensor %589, %17345 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17354 = torch.aten.mul.Tensor %590, %17346 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%17355 = torch.aten.mul.Tensor %591, %17347 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %17355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_14914 = torch.constant.int 5
%17356 = torch.prims.convert_element_type %17348, %int5_14914 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14915 = torch.constant.int 5
%17357 = torch.prims.convert_element_type %17349, %int5_14915 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14916 = torch.constant.int 5
%17358 = torch.prims.convert_element_type %17350, %int5_14916 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14917 = torch.constant.int 5
%17359 = torch.prims.convert_element_type %17351, %int5_14917 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14918 = torch.constant.int 5
%17360 = torch.prims.convert_element_type %17352, %int5_14918 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14919 = torch.constant.int 5
%17361 = torch.prims.convert_element_type %17353, %int5_14919 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14920 = torch.constant.int 5
%17362 = torch.prims.convert_element_type %17354, %int5_14920 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_14921 = torch.constant.int 5
%17363 = torch.prims.convert_element_type %17355, %int5_14921 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %17363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_14922 = torch.constant.int 1
%int0_14923 = torch.constant.int 0
%17364 = torch.prim.ListConstruct %int1_14922, %int0_14923 : (!torch.int, !torch.int) -> !torch.list<int>
%17365 = torch.aten.permute %592, %17364 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14924 = torch.constant.int 1
%int0_14925 = torch.constant.int 0
%17366 = torch.prim.ListConstruct %int1_14924, %int0_14925 : (!torch.int, !torch.int) -> !torch.list<int>
%17367 = torch.aten.permute %593, %17366 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14926 = torch.constant.int 1
%int0_14927 = torch.constant.int 0
%17368 = torch.prim.ListConstruct %int1_14926, %int0_14927 : (!torch.int, !torch.int) -> !torch.list<int>
%17369 = torch.aten.permute %594, %17368 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14928 = torch.constant.int 1
%int0_14929 = torch.constant.int 0
%17370 = torch.prim.ListConstruct %int1_14928, %int0_14929 : (!torch.int, !torch.int) -> !torch.list<int>
%17371 = torch.aten.permute %595, %17370 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14930 = torch.constant.int 1
%int0_14931 = torch.constant.int 0
%17372 = torch.prim.ListConstruct %int1_14930, %int0_14931 : (!torch.int, !torch.int) -> !torch.list<int>
%17373 = torch.aten.permute %596, %17372 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14932 = torch.constant.int 1
%int0_14933 = torch.constant.int 0
%17374 = torch.prim.ListConstruct %int1_14932, %int0_14933 : (!torch.int, !torch.int) -> !torch.list<int>
%17375 = torch.aten.permute %597, %17374 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14934 = torch.constant.int 1
%int0_14935 = torch.constant.int 0
%17376 = torch.prim.ListConstruct %int1_14934, %int0_14935 : (!torch.int, !torch.int) -> !torch.list<int>
%17377 = torch.aten.permute %598, %17376 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_14936 = torch.constant.int 1
%int0_14937 = torch.constant.int 0
%17378 = torch.prim.ListConstruct %int1_14936, %int0_14937 : (!torch.int, !torch.int) -> !torch.list<int>
%17379 = torch.aten.permute %599, %17378 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_14938 = torch.constant.int 4
%17380 = torch.aten.mul.int %int4_14938, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14939 = torch.constant.int 4096
%17381 = torch.prim.ListConstruct %17380, %int4096_14939 : (!torch.int, !torch.int) -> !torch.list<int>
%17382 = torch.aten.view %17356, %17381 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17382, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17383 = torch.aten.mm %17382, %17365 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17383, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14940 = torch.constant.int 4
%int512_14941 = torch.constant.int 512
%17384 = torch.prim.ListConstruct %int4_14940, %2482, %int512_14941 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17385 = torch.aten.view %17383, %17384 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14942 = torch.constant.int 4
%17386 = torch.aten.mul.int %int4_14942, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14943 = torch.constant.int 4096
%17387 = torch.prim.ListConstruct %17386, %int4096_14943 : (!torch.int, !torch.int) -> !torch.list<int>
%17388 = torch.aten.view %17357, %17387 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17388, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17389 = torch.aten.mm %17388, %17367 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17389, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14944 = torch.constant.int 4
%int512_14945 = torch.constant.int 512
%17390 = torch.prim.ListConstruct %int4_14944, %2482, %int512_14945 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17391 = torch.aten.view %17389, %17390 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14946 = torch.constant.int 4
%17392 = torch.aten.mul.int %int4_14946, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14947 = torch.constant.int 4096
%17393 = torch.prim.ListConstruct %17392, %int4096_14947 : (!torch.int, !torch.int) -> !torch.list<int>
%17394 = torch.aten.view %17358, %17393 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17394, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17395 = torch.aten.mm %17394, %17369 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17395, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14948 = torch.constant.int 4
%int512_14949 = torch.constant.int 512
%17396 = torch.prim.ListConstruct %int4_14948, %2482, %int512_14949 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17397 = torch.aten.view %17395, %17396 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14950 = torch.constant.int 4
%17398 = torch.aten.mul.int %int4_14950, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14951 = torch.constant.int 4096
%17399 = torch.prim.ListConstruct %17398, %int4096_14951 : (!torch.int, !torch.int) -> !torch.list<int>
%17400 = torch.aten.view %17359, %17399 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17400, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17401 = torch.aten.mm %17400, %17371 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17401, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14952 = torch.constant.int 4
%int512_14953 = torch.constant.int 512
%17402 = torch.prim.ListConstruct %int4_14952, %2482, %int512_14953 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17403 = torch.aten.view %17401, %17402 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14954 = torch.constant.int 4
%17404 = torch.aten.mul.int %int4_14954, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14955 = torch.constant.int 4096
%17405 = torch.prim.ListConstruct %17404, %int4096_14955 : (!torch.int, !torch.int) -> !torch.list<int>
%17406 = torch.aten.view %17360, %17405 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17406, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17407 = torch.aten.mm %17406, %17373 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17407, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14956 = torch.constant.int 4
%int512_14957 = torch.constant.int 512
%17408 = torch.prim.ListConstruct %int4_14956, %2482, %int512_14957 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17409 = torch.aten.view %17407, %17408 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14958 = torch.constant.int 4
%17410 = torch.aten.mul.int %int4_14958, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14959 = torch.constant.int 4096
%17411 = torch.prim.ListConstruct %17410, %int4096_14959 : (!torch.int, !torch.int) -> !torch.list<int>
%17412 = torch.aten.view %17361, %17411 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17412, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17413 = torch.aten.mm %17412, %17375 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17413, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14960 = torch.constant.int 4
%int512_14961 = torch.constant.int 512
%17414 = torch.prim.ListConstruct %int4_14960, %2482, %int512_14961 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17415 = torch.aten.view %17413, %17414 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14962 = torch.constant.int 4
%17416 = torch.aten.mul.int %int4_14962, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14963 = torch.constant.int 4096
%17417 = torch.prim.ListConstruct %17416, %int4096_14963 : (!torch.int, !torch.int) -> !torch.list<int>
%17418 = torch.aten.view %17362, %17417 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17418, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17419 = torch.aten.mm %17418, %17377 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17419, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14964 = torch.constant.int 4
%int512_14965 = torch.constant.int 512
%17420 = torch.prim.ListConstruct %int4_14964, %2482, %int512_14965 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17421 = torch.aten.view %17419, %17420 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_14966 = torch.constant.int 4
%17422 = torch.aten.mul.int %int4_14966, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14967 = torch.constant.int 4096
%17423 = torch.prim.ListConstruct %17422, %int4096_14967 : (!torch.int, !torch.int) -> !torch.list<int>
%17424 = torch.aten.view %17363, %17423 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17424, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17425 = torch.aten.mm %17424, %17379 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %17425, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_14968 = torch.constant.int 4
%int512_14969 = torch.constant.int 512
%17426 = torch.prim.ListConstruct %int4_14968, %2482, %int512_14969 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17427 = torch.aten.view %17425, %17426 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %17427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_14970 = torch.constant.int 1
%int0_14971 = torch.constant.int 0
%17428 = torch.prim.ListConstruct %int1_14970, %int0_14971 : (!torch.int, !torch.int) -> !torch.list<int>
%17429 = torch.aten.permute %600, %17428 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14972 = torch.constant.int 1
%int0_14973 = torch.constant.int 0
%17430 = torch.prim.ListConstruct %int1_14972, %int0_14973 : (!torch.int, !torch.int) -> !torch.list<int>
%17431 = torch.aten.permute %601, %17430 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14974 = torch.constant.int 1
%int0_14975 = torch.constant.int 0
%17432 = torch.prim.ListConstruct %int1_14974, %int0_14975 : (!torch.int, !torch.int) -> !torch.list<int>
%17433 = torch.aten.permute %602, %17432 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14976 = torch.constant.int 1
%int0_14977 = torch.constant.int 0
%17434 = torch.prim.ListConstruct %int1_14976, %int0_14977 : (!torch.int, !torch.int) -> !torch.list<int>
%17435 = torch.aten.permute %603, %17434 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14978 = torch.constant.int 1
%int0_14979 = torch.constant.int 0
%17436 = torch.prim.ListConstruct %int1_14978, %int0_14979 : (!torch.int, !torch.int) -> !torch.list<int>
%17437 = torch.aten.permute %604, %17436 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14980 = torch.constant.int 1
%int0_14981 = torch.constant.int 0
%17438 = torch.prim.ListConstruct %int1_14980, %int0_14981 : (!torch.int, !torch.int) -> !torch.list<int>
%17439 = torch.aten.permute %605, %17438 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14982 = torch.constant.int 1
%int0_14983 = torch.constant.int 0
%17440 = torch.prim.ListConstruct %int1_14982, %int0_14983 : (!torch.int, !torch.int) -> !torch.list<int>
%17441 = torch.aten.permute %606, %17440 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_14984 = torch.constant.int 1
%int0_14985 = torch.constant.int 0
%17442 = torch.prim.ListConstruct %int1_14984, %int0_14985 : (!torch.int, !torch.int) -> !torch.list<int>
%17443 = torch.aten.permute %607, %17442 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_14986 = torch.constant.int 4
%17444 = torch.aten.mul.int %int4_14986, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14987 = torch.constant.int 4096
%17445 = torch.prim.ListConstruct %17444, %int4096_14987 : (!torch.int, !torch.int) -> !torch.list<int>
%17446 = torch.aten.view %17356, %17445 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17446, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17447 = torch.aten.mm %17446, %17429 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17447, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_14988 = torch.constant.int 4
%int128_14989 = torch.constant.int 128
%17448 = torch.prim.ListConstruct %int4_14988, %2482, %int128_14989 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17449 = torch.aten.view %17447, %17448 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_14990 = torch.constant.int 4
%17450 = torch.aten.mul.int %int4_14990, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14991 = torch.constant.int 4096
%17451 = torch.prim.ListConstruct %17450, %int4096_14991 : (!torch.int, !torch.int) -> !torch.list<int>
%17452 = torch.aten.view %17357, %17451 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17452, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17453 = torch.aten.mm %17452, %17431 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17453, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_14992 = torch.constant.int 4
%int128_14993 = torch.constant.int 128
%17454 = torch.prim.ListConstruct %int4_14992, %2482, %int128_14993 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17455 = torch.aten.view %17453, %17454 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_14994 = torch.constant.int 4
%17456 = torch.aten.mul.int %int4_14994, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14995 = torch.constant.int 4096
%17457 = torch.prim.ListConstruct %17456, %int4096_14995 : (!torch.int, !torch.int) -> !torch.list<int>
%17458 = torch.aten.view %17358, %17457 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17458, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17459 = torch.aten.mm %17458, %17433 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17459, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_14996 = torch.constant.int 4
%int128_14997 = torch.constant.int 128
%17460 = torch.prim.ListConstruct %int4_14996, %2482, %int128_14997 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17461 = torch.aten.view %17459, %17460 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_14998 = torch.constant.int 4
%17462 = torch.aten.mul.int %int4_14998, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_14999 = torch.constant.int 4096
%17463 = torch.prim.ListConstruct %17462, %int4096_14999 : (!torch.int, !torch.int) -> !torch.list<int>
%17464 = torch.aten.view %17359, %17463 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17464, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17465 = torch.aten.mm %17464, %17435 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17465, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15000 = torch.constant.int 4
%int128_15001 = torch.constant.int 128
%17466 = torch.prim.ListConstruct %int4_15000, %2482, %int128_15001 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17467 = torch.aten.view %17465, %17466 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15002 = torch.constant.int 4
%17468 = torch.aten.mul.int %int4_15002, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15003 = torch.constant.int 4096
%17469 = torch.prim.ListConstruct %17468, %int4096_15003 : (!torch.int, !torch.int) -> !torch.list<int>
%17470 = torch.aten.view %17360, %17469 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17470, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17471 = torch.aten.mm %17470, %17437 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17471, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15004 = torch.constant.int 4
%int128_15005 = torch.constant.int 128
%17472 = torch.prim.ListConstruct %int4_15004, %2482, %int128_15005 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17473 = torch.aten.view %17471, %17472 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15006 = torch.constant.int 4
%17474 = torch.aten.mul.int %int4_15006, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15007 = torch.constant.int 4096
%17475 = torch.prim.ListConstruct %17474, %int4096_15007 : (!torch.int, !torch.int) -> !torch.list<int>
%17476 = torch.aten.view %17361, %17475 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17476, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17477 = torch.aten.mm %17476, %17439 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17477, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15008 = torch.constant.int 4
%int128_15009 = torch.constant.int 128
%17478 = torch.prim.ListConstruct %int4_15008, %2482, %int128_15009 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17479 = torch.aten.view %17477, %17478 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15010 = torch.constant.int 4
%17480 = torch.aten.mul.int %int4_15010, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15011 = torch.constant.int 4096
%17481 = torch.prim.ListConstruct %17480, %int4096_15011 : (!torch.int, !torch.int) -> !torch.list<int>
%17482 = torch.aten.view %17362, %17481 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17482, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17483 = torch.aten.mm %17482, %17441 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17483, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15012 = torch.constant.int 4
%int128_15013 = torch.constant.int 128
%17484 = torch.prim.ListConstruct %int4_15012, %2482, %int128_15013 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17485 = torch.aten.view %17483, %17484 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15014 = torch.constant.int 4
%17486 = torch.aten.mul.int %int4_15014, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15015 = torch.constant.int 4096
%17487 = torch.prim.ListConstruct %17486, %int4096_15015 : (!torch.int, !torch.int) -> !torch.list<int>
%17488 = torch.aten.view %17363, %17487 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17488, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17489 = torch.aten.mm %17488, %17443 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17489, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15016 = torch.constant.int 4
%int128_15017 = torch.constant.int 128
%17490 = torch.prim.ListConstruct %int4_15016, %2482, %int128_15017 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17491 = torch.aten.view %17489, %17490 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_15018 = torch.constant.int 1
%int0_15019 = torch.constant.int 0
%17492 = torch.prim.ListConstruct %int1_15018, %int0_15019 : (!torch.int, !torch.int) -> !torch.list<int>
%17493 = torch.aten.permute %608, %17492 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15020 = torch.constant.int 1
%int0_15021 = torch.constant.int 0
%17494 = torch.prim.ListConstruct %int1_15020, %int0_15021 : (!torch.int, !torch.int) -> !torch.list<int>
%17495 = torch.aten.permute %609, %17494 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15022 = torch.constant.int 1
%int0_15023 = torch.constant.int 0
%17496 = torch.prim.ListConstruct %int1_15022, %int0_15023 : (!torch.int, !torch.int) -> !torch.list<int>
%17497 = torch.aten.permute %610, %17496 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15024 = torch.constant.int 1
%int0_15025 = torch.constant.int 0
%17498 = torch.prim.ListConstruct %int1_15024, %int0_15025 : (!torch.int, !torch.int) -> !torch.list<int>
%17499 = torch.aten.permute %611, %17498 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15026 = torch.constant.int 1
%int0_15027 = torch.constant.int 0
%17500 = torch.prim.ListConstruct %int1_15026, %int0_15027 : (!torch.int, !torch.int) -> !torch.list<int>
%17501 = torch.aten.permute %612, %17500 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15028 = torch.constant.int 1
%int0_15029 = torch.constant.int 0
%17502 = torch.prim.ListConstruct %int1_15028, %int0_15029 : (!torch.int, !torch.int) -> !torch.list<int>
%17503 = torch.aten.permute %613, %17502 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15030 = torch.constant.int 1
%int0_15031 = torch.constant.int 0
%17504 = torch.prim.ListConstruct %int1_15030, %int0_15031 : (!torch.int, !torch.int) -> !torch.list<int>
%17505 = torch.aten.permute %614, %17504 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_15032 = torch.constant.int 1
%int0_15033 = torch.constant.int 0
%17506 = torch.prim.ListConstruct %int1_15032, %int0_15033 : (!torch.int, !torch.int) -> !torch.list<int>
%17507 = torch.aten.permute %615, %17506 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_15034 = torch.constant.int 4
%17508 = torch.aten.mul.int %int4_15034, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15035 = torch.constant.int 4096
%17509 = torch.prim.ListConstruct %17508, %int4096_15035 : (!torch.int, !torch.int) -> !torch.list<int>
%17510 = torch.aten.view %17356, %17509 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17510, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17511 = torch.aten.mm %17510, %17493 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17511, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15036 = torch.constant.int 4
%int128_15037 = torch.constant.int 128
%17512 = torch.prim.ListConstruct %int4_15036, %2482, %int128_15037 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17513 = torch.aten.view %17511, %17512 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15038 = torch.constant.int 4
%17514 = torch.aten.mul.int %int4_15038, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15039 = torch.constant.int 4096
%17515 = torch.prim.ListConstruct %17514, %int4096_15039 : (!torch.int, !torch.int) -> !torch.list<int>
%17516 = torch.aten.view %17357, %17515 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17516, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17517 = torch.aten.mm %17516, %17495 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17517, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15040 = torch.constant.int 4
%int128_15041 = torch.constant.int 128
%17518 = torch.prim.ListConstruct %int4_15040, %2482, %int128_15041 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17519 = torch.aten.view %17517, %17518 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15042 = torch.constant.int 4
%17520 = torch.aten.mul.int %int4_15042, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15043 = torch.constant.int 4096
%17521 = torch.prim.ListConstruct %17520, %int4096_15043 : (!torch.int, !torch.int) -> !torch.list<int>
%17522 = torch.aten.view %17358, %17521 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17522, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17523 = torch.aten.mm %17522, %17497 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17523, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15044 = torch.constant.int 4
%int128_15045 = torch.constant.int 128
%17524 = torch.prim.ListConstruct %int4_15044, %2482, %int128_15045 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17525 = torch.aten.view %17523, %17524 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15046 = torch.constant.int 4
%17526 = torch.aten.mul.int %int4_15046, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15047 = torch.constant.int 4096
%17527 = torch.prim.ListConstruct %17526, %int4096_15047 : (!torch.int, !torch.int) -> !torch.list<int>
%17528 = torch.aten.view %17359, %17527 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17528, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17529 = torch.aten.mm %17528, %17499 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17529, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15048 = torch.constant.int 4
%int128_15049 = torch.constant.int 128
%17530 = torch.prim.ListConstruct %int4_15048, %2482, %int128_15049 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17531 = torch.aten.view %17529, %17530 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15050 = torch.constant.int 4
%17532 = torch.aten.mul.int %int4_15050, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15051 = torch.constant.int 4096
%17533 = torch.prim.ListConstruct %17532, %int4096_15051 : (!torch.int, !torch.int) -> !torch.list<int>
%17534 = torch.aten.view %17360, %17533 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17534, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17535 = torch.aten.mm %17534, %17501 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17535, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15052 = torch.constant.int 4
%int128_15053 = torch.constant.int 128
%17536 = torch.prim.ListConstruct %int4_15052, %2482, %int128_15053 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17537 = torch.aten.view %17535, %17536 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15054 = torch.constant.int 4
%17538 = torch.aten.mul.int %int4_15054, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15055 = torch.constant.int 4096
%17539 = torch.prim.ListConstruct %17538, %int4096_15055 : (!torch.int, !torch.int) -> !torch.list<int>
%17540 = torch.aten.view %17361, %17539 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17540, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17541 = torch.aten.mm %17540, %17503 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17541, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15056 = torch.constant.int 4
%int128_15057 = torch.constant.int 128
%17542 = torch.prim.ListConstruct %int4_15056, %2482, %int128_15057 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17543 = torch.aten.view %17541, %17542 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15058 = torch.constant.int 4
%17544 = torch.aten.mul.int %int4_15058, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15059 = torch.constant.int 4096
%17545 = torch.prim.ListConstruct %17544, %int4096_15059 : (!torch.int, !torch.int) -> !torch.list<int>
%17546 = torch.aten.view %17362, %17545 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17546, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17547 = torch.aten.mm %17546, %17505 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17547, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15060 = torch.constant.int 4
%int128_15061 = torch.constant.int 128
%17548 = torch.prim.ListConstruct %int4_15060, %2482, %int128_15061 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17549 = torch.aten.view %17547, %17548 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15062 = torch.constant.int 4
%17550 = torch.aten.mul.int %int4_15062, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_15063 = torch.constant.int 4096
%17551 = torch.prim.ListConstruct %17550, %int4096_15063 : (!torch.int, !torch.int) -> !torch.list<int>
%17552 = torch.aten.view %17363, %17551 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %17552, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%17553 = torch.aten.mm %17552, %17507 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %17553, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_15064 = torch.constant.int 4
%int128_15065 = torch.constant.int 128
%17554 = torch.prim.ListConstruct %int4_15064, %2482, %int128_15065 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17555 = torch.aten.view %17553, %17554 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %17555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_15066 = torch.constant.int 4
%int4_15067 = torch.constant.int 4
%int128_15068 = torch.constant.int 128
%17556 = torch.prim.ListConstruct %int4_15066, %2482, %int4_15067, %int128_15068 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17557 = torch.aten.view %17385, %17556 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15069 = torch.constant.int 4
%int4_15070 = torch.constant.int 4
%int128_15071 = torch.constant.int 128
%17558 = torch.prim.ListConstruct %int4_15069, %2482, %int4_15070, %int128_15071 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17559 = torch.aten.view %17391, %17558 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15072 = torch.constant.int 4
%int4_15073 = torch.constant.int 4
%int128_15074 = torch.constant.int 128
%17560 = torch.prim.ListConstruct %int4_15072, %2482, %int4_15073, %int128_15074 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17561 = torch.aten.view %17397, %17560 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15075 = torch.constant.int 4
%int4_15076 = torch.constant.int 4
%int128_15077 = torch.constant.int 128
%17562 = torch.prim.ListConstruct %int4_15075, %2482, %int4_15076, %int128_15077 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17563 = torch.aten.view %17403, %17562 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15078 = torch.constant.int 4
%int4_15079 = torch.constant.int 4
%int128_15080 = torch.constant.int 128
%17564 = torch.prim.ListConstruct %int4_15078, %2482, %int4_15079, %int128_15080 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17565 = torch.aten.view %17409, %17564 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15081 = torch.constant.int 4
%int4_15082 = torch.constant.int 4
%int128_15083 = torch.constant.int 128
%17566 = torch.prim.ListConstruct %int4_15081, %2482, %int4_15082, %int128_15083 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17567 = torch.aten.view %17415, %17566 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15084 = torch.constant.int 4
%int4_15085 = torch.constant.int 4
%int128_15086 = torch.constant.int 128
%17568 = torch.prim.ListConstruct %int4_15084, %2482, %int4_15085, %int128_15086 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17569 = torch.aten.view %17421, %17568 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15087 = torch.constant.int 4
%int4_15088 = torch.constant.int 4
%int128_15089 = torch.constant.int 128
%17570 = torch.prim.ListConstruct %int4_15087, %2482, %int4_15088, %int128_15089 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17571 = torch.aten.view %17427, %17570 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15090 = torch.constant.int 4
%int1_15091 = torch.constant.int 1
%int128_15092 = torch.constant.int 128
%17572 = torch.prim.ListConstruct %int4_15090, %2482, %int1_15091, %int128_15092 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17573 = torch.aten.view %17449, %17572 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15093 = torch.constant.int 4
%int1_15094 = torch.constant.int 1
%int128_15095 = torch.constant.int 128
%17574 = torch.prim.ListConstruct %int4_15093, %2482, %int1_15094, %int128_15095 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17575 = torch.aten.view %17455, %17574 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15096 = torch.constant.int 4
%int1_15097 = torch.constant.int 1
%int128_15098 = torch.constant.int 128
%17576 = torch.prim.ListConstruct %int4_15096, %2482, %int1_15097, %int128_15098 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17577 = torch.aten.view %17461, %17576 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15099 = torch.constant.int 4
%int1_15100 = torch.constant.int 1
%int128_15101 = torch.constant.int 128
%17578 = torch.prim.ListConstruct %int4_15099, %2482, %int1_15100, %int128_15101 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17579 = torch.aten.view %17467, %17578 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15102 = torch.constant.int 4
%int1_15103 = torch.constant.int 1
%int128_15104 = torch.constant.int 128
%17580 = torch.prim.ListConstruct %int4_15102, %2482, %int1_15103, %int128_15104 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17581 = torch.aten.view %17473, %17580 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15105 = torch.constant.int 4
%int1_15106 = torch.constant.int 1
%int128_15107 = torch.constant.int 128
%17582 = torch.prim.ListConstruct %int4_15105, %2482, %int1_15106, %int128_15107 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17583 = torch.aten.view %17479, %17582 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15108 = torch.constant.int 4
%int1_15109 = torch.constant.int 1
%int128_15110 = torch.constant.int 128
%17584 = torch.prim.ListConstruct %int4_15108, %2482, %int1_15109, %int128_15110 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17585 = torch.aten.view %17485, %17584 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15111 = torch.constant.int 4
%int1_15112 = torch.constant.int 1
%int128_15113 = torch.constant.int 128
%17586 = torch.prim.ListConstruct %int4_15111, %2482, %int1_15112, %int128_15113 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17587 = torch.aten.view %17491, %17586 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15114 = torch.constant.int 4
%int1_15115 = torch.constant.int 1
%int128_15116 = torch.constant.int 128
%17588 = torch.prim.ListConstruct %int4_15114, %2482, %int1_15115, %int128_15116 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17589 = torch.aten.view %17513, %17588 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15117 = torch.constant.int 4
%int1_15118 = torch.constant.int 1
%int128_15119 = torch.constant.int 128
%17590 = torch.prim.ListConstruct %int4_15117, %2482, %int1_15118, %int128_15119 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17591 = torch.aten.view %17519, %17590 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15120 = torch.constant.int 4
%int1_15121 = torch.constant.int 1
%int128_15122 = torch.constant.int 128
%17592 = torch.prim.ListConstruct %int4_15120, %2482, %int1_15121, %int128_15122 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17593 = torch.aten.view %17525, %17592 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15123 = torch.constant.int 4
%int1_15124 = torch.constant.int 1
%int128_15125 = torch.constant.int 128
%17594 = torch.prim.ListConstruct %int4_15123, %2482, %int1_15124, %int128_15125 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17595 = torch.aten.view %17531, %17594 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15126 = torch.constant.int 4
%int1_15127 = torch.constant.int 1
%int128_15128 = torch.constant.int 128
%17596 = torch.prim.ListConstruct %int4_15126, %2482, %int1_15127, %int128_15128 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17597 = torch.aten.view %17537, %17596 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15129 = torch.constant.int 4
%int1_15130 = torch.constant.int 1
%int128_15131 = torch.constant.int 128
%17598 = torch.prim.ListConstruct %int4_15129, %2482, %int1_15130, %int128_15131 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17599 = torch.aten.view %17543, %17598 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15132 = torch.constant.int 4
%int1_15133 = torch.constant.int 1
%int128_15134 = torch.constant.int 128
%17600 = torch.prim.ListConstruct %int4_15132, %2482, %int1_15133, %int128_15134 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17601 = torch.aten.view %17549, %17600 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_15135 = torch.constant.int 4
%int1_15136 = torch.constant.int 1
%int128_15137 = torch.constant.int 128
%17602 = torch.prim.ListConstruct %int4_15135, %2482, %int1_15136, %int128_15137 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17603 = torch.aten.view %17555, %17602 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_15138 = torch.constant.int 131072
%none_15139 = torch.constant.none
%none_15140 = torch.constant.none
%cpu_15141 = torch.constant.device "cpu"
%false_15142 = torch.constant.bool false
%17604 = torch.aten.arange %int131072_15138, %none_15139, %none_15140, %cpu_15141, %false_15142 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_15143 = torch.constant.int 0
%int128_15144 = torch.constant.int 128
%int2_15145 = torch.constant.int 2
%none_15146 = torch.constant.none
%none_15147 = torch.constant.none
%cpu_15148 = torch.constant.device "cpu"
%false_15149 = torch.constant.bool false
%17605 = torch.aten.arange.start_step %int0_15143, %int128_15144, %int2_15145, %none_15146, %none_15147, %cpu_15148, %false_15149 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_15150 = torch.constant.int 0
%int0_15151 = torch.constant.int 0
%int64_15152 = torch.constant.int 64
%int1_15153 = torch.constant.int 1
%17606 = torch.aten.slice.Tensor %17605, %int0_15150, %int0_15151, %int64_15152, %int1_15153 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_15154 = torch.constant.int 6
%17607 = torch.prims.convert_element_type %17606, %int6_15154 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_15155 = torch.constant.int 128
%17608 = torch.aten.div.Scalar %17607, %int128_15155 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_15156 = torch.constant.float 5.000000e+05
%17609 = torch.aten.pow.Scalar %float5.000000e05_15156, %17608 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%17610 = torch.aten.reciprocal %17609 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_15157 = torch.constant.float 1.000000e+00
%17611 = torch.aten.mul.Scalar %17610, %float1.000000e00_15157 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_15158 = torch.constant.int 131072
%int1_15159 = torch.constant.int 1
%17612 = torch.prim.ListConstruct %int131072_15158, %int1_15159 : (!torch.int, !torch.int) -> !torch.list<int>
%17613 = torch.aten.view %17604, %17612 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%17614 = torch.aten.mul.Tensor %17613, %17611 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%17615 = torch.aten.cos %17614 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%17616 = torch.aten.sin %17614 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%17617 = torch.aten.complex %17615, %17616 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%17618 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17619 = flow.tensor.transfer %17618 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%17620 = torch_c.from_builtin_tensor %17619 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17621 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17622 = flow.tensor.transfer %17621 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%17623 = torch_c.from_builtin_tensor %17622 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17624 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17625 = flow.tensor.transfer %17624 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%17626 = torch_c.from_builtin_tensor %17625 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17627 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17628 = flow.tensor.transfer %17627 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%17629 = torch_c.from_builtin_tensor %17628 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17630 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17631 = flow.tensor.transfer %17630 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%17632 = torch_c.from_builtin_tensor %17631 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17633 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17634 = flow.tensor.transfer %17633 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%17635 = torch_c.from_builtin_tensor %17634 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17636 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17637 = flow.tensor.transfer %17636 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%17638 = torch_c.from_builtin_tensor %17637 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17639 = torch_c.to_builtin_tensor %17617 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17640 = flow.tensor.transfer %17639 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%17641 = torch_c.from_builtin_tensor %17640 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_15160 = torch.constant.int 1
%17642 = torch.aten.size.int %17385, %int1_15160 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15161 = torch.constant.int 0
%17643 = torch.aten.add.int %int0_15161, %17642 : !torch.int, !torch.int -> !torch.int
%int0_15162 = torch.constant.int 0
%int0_15163 = torch.constant.int 0
%int1_15164 = torch.constant.int 1
%17644 = torch.aten.slice.Tensor %17620, %int0_15162, %int0_15163, %17643, %int1_15164 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17644, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15165 = torch.constant.int 1
%int0_15166 = torch.constant.int 0
%int9223372036854775807_15167 = torch.constant.int 9223372036854775807
%int1_15168 = torch.constant.int 1
%17645 = torch.aten.slice.Tensor %17644, %int1_15165, %int0_15166, %int9223372036854775807_15167, %int1_15168 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17645, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15169 = torch.constant.int 0
%17646 = torch.aten.unsqueeze %17645, %int0_15169 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17646, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15170 = torch.constant.int 2
%17647 = torch.aten.unsqueeze %17646, %int2_15170 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17647, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15171 = torch.constant.int 3
%int0_15172 = torch.constant.int 0
%int9223372036854775807_15173 = torch.constant.int 9223372036854775807
%int1_15174 = torch.constant.int 1
%17648 = torch.aten.slice.Tensor %17647, %int3_15171, %int0_15172, %int9223372036854775807_15173, %int1_15174 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17648, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17649 = torch_c.to_builtin_tensor %17557 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15175 = arith.constant 1 : index
%dim_15176 = tensor.dim %17649, %c1_15175 : tensor<4x?x4x128xf16>
%17650 = flow.tensor.bitcast %17649 : tensor<4x?x4x128xf16>{%dim_15176} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15176}
%17651 = torch_c.from_builtin_tensor %17650 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17652 = torch.aten.mul.Tensor %17651, %17648 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17653 = torch_c.to_builtin_tensor %17652 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15177 = arith.constant 1 : index
%dim_15178 = tensor.dim %17653, %c1_15177 : tensor<4x?x4x64xcomplex<f32>>
%17654 = flow.tensor.bitcast %17653 : tensor<4x?x4x64xcomplex<f32>>{%dim_15178} -> tensor<4x?x4x128xf32>{%dim_15178}
%17655 = torch_c.from_builtin_tensor %17654 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15179 = torch.constant.int 5
%17656 = torch.prims.convert_element_type %17655, %int5_15179 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15180 = torch.constant.int 1
%17657 = torch.aten.size.int %17391, %int1_15180 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15181 = torch.constant.int 0
%17658 = torch.aten.add.int %int0_15181, %17657 : !torch.int, !torch.int -> !torch.int
%int0_15182 = torch.constant.int 0
%int0_15183 = torch.constant.int 0
%int1_15184 = torch.constant.int 1
%17659 = torch.aten.slice.Tensor %17623, %int0_15182, %int0_15183, %17658, %int1_15184 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17659, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15185 = torch.constant.int 1
%int0_15186 = torch.constant.int 0
%int9223372036854775807_15187 = torch.constant.int 9223372036854775807
%int1_15188 = torch.constant.int 1
%17660 = torch.aten.slice.Tensor %17659, %int1_15185, %int0_15186, %int9223372036854775807_15187, %int1_15188 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17660, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15189 = torch.constant.int 0
%17661 = torch.aten.unsqueeze %17660, %int0_15189 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17661, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15190 = torch.constant.int 2
%17662 = torch.aten.unsqueeze %17661, %int2_15190 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17662, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15191 = torch.constant.int 3
%int0_15192 = torch.constant.int 0
%int9223372036854775807_15193 = torch.constant.int 9223372036854775807
%int1_15194 = torch.constant.int 1
%17663 = torch.aten.slice.Tensor %17662, %int3_15191, %int0_15192, %int9223372036854775807_15193, %int1_15194 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17663, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17664 = torch_c.to_builtin_tensor %17559 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15195 = arith.constant 1 : index
%dim_15196 = tensor.dim %17664, %c1_15195 : tensor<4x?x4x128xf16>
%17665 = flow.tensor.bitcast %17664 : tensor<4x?x4x128xf16>{%dim_15196} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15196}
%17666 = torch_c.from_builtin_tensor %17665 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17667 = torch.aten.mul.Tensor %17666, %17663 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17668 = torch_c.to_builtin_tensor %17667 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15197 = arith.constant 1 : index
%dim_15198 = tensor.dim %17668, %c1_15197 : tensor<4x?x4x64xcomplex<f32>>
%17669 = flow.tensor.bitcast %17668 : tensor<4x?x4x64xcomplex<f32>>{%dim_15198} -> tensor<4x?x4x128xf32>{%dim_15198}
%17670 = torch_c.from_builtin_tensor %17669 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15199 = torch.constant.int 5
%17671 = torch.prims.convert_element_type %17670, %int5_15199 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15200 = torch.constant.int 1
%17672 = torch.aten.size.int %17397, %int1_15200 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15201 = torch.constant.int 0
%17673 = torch.aten.add.int %int0_15201, %17672 : !torch.int, !torch.int -> !torch.int
%int0_15202 = torch.constant.int 0
%int0_15203 = torch.constant.int 0
%int1_15204 = torch.constant.int 1
%17674 = torch.aten.slice.Tensor %17626, %int0_15202, %int0_15203, %17673, %int1_15204 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17674, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15205 = torch.constant.int 1
%int0_15206 = torch.constant.int 0
%int9223372036854775807_15207 = torch.constant.int 9223372036854775807
%int1_15208 = torch.constant.int 1
%17675 = torch.aten.slice.Tensor %17674, %int1_15205, %int0_15206, %int9223372036854775807_15207, %int1_15208 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17675, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15209 = torch.constant.int 0
%17676 = torch.aten.unsqueeze %17675, %int0_15209 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17676, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15210 = torch.constant.int 2
%17677 = torch.aten.unsqueeze %17676, %int2_15210 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17677, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15211 = torch.constant.int 3
%int0_15212 = torch.constant.int 0
%int9223372036854775807_15213 = torch.constant.int 9223372036854775807
%int1_15214 = torch.constant.int 1
%17678 = torch.aten.slice.Tensor %17677, %int3_15211, %int0_15212, %int9223372036854775807_15213, %int1_15214 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17678, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17679 = torch_c.to_builtin_tensor %17561 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15215 = arith.constant 1 : index
%dim_15216 = tensor.dim %17679, %c1_15215 : tensor<4x?x4x128xf16>
%17680 = flow.tensor.bitcast %17679 : tensor<4x?x4x128xf16>{%dim_15216} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15216}
%17681 = torch_c.from_builtin_tensor %17680 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17682 = torch.aten.mul.Tensor %17681, %17678 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17683 = torch_c.to_builtin_tensor %17682 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15217 = arith.constant 1 : index
%dim_15218 = tensor.dim %17683, %c1_15217 : tensor<4x?x4x64xcomplex<f32>>
%17684 = flow.tensor.bitcast %17683 : tensor<4x?x4x64xcomplex<f32>>{%dim_15218} -> tensor<4x?x4x128xf32>{%dim_15218}
%17685 = torch_c.from_builtin_tensor %17684 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15219 = torch.constant.int 5
%17686 = torch.prims.convert_element_type %17685, %int5_15219 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15220 = torch.constant.int 1
%17687 = torch.aten.size.int %17403, %int1_15220 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15221 = torch.constant.int 0
%17688 = torch.aten.add.int %int0_15221, %17687 : !torch.int, !torch.int -> !torch.int
%int0_15222 = torch.constant.int 0
%int0_15223 = torch.constant.int 0
%int1_15224 = torch.constant.int 1
%17689 = torch.aten.slice.Tensor %17629, %int0_15222, %int0_15223, %17688, %int1_15224 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17689, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15225 = torch.constant.int 1
%int0_15226 = torch.constant.int 0
%int9223372036854775807_15227 = torch.constant.int 9223372036854775807
%int1_15228 = torch.constant.int 1
%17690 = torch.aten.slice.Tensor %17689, %int1_15225, %int0_15226, %int9223372036854775807_15227, %int1_15228 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17690, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15229 = torch.constant.int 0
%17691 = torch.aten.unsqueeze %17690, %int0_15229 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17691, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15230 = torch.constant.int 2
%17692 = torch.aten.unsqueeze %17691, %int2_15230 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17692, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15231 = torch.constant.int 3
%int0_15232 = torch.constant.int 0
%int9223372036854775807_15233 = torch.constant.int 9223372036854775807
%int1_15234 = torch.constant.int 1
%17693 = torch.aten.slice.Tensor %17692, %int3_15231, %int0_15232, %int9223372036854775807_15233, %int1_15234 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17693, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17694 = torch_c.to_builtin_tensor %17563 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15235 = arith.constant 1 : index
%dim_15236 = tensor.dim %17694, %c1_15235 : tensor<4x?x4x128xf16>
%17695 = flow.tensor.bitcast %17694 : tensor<4x?x4x128xf16>{%dim_15236} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15236}
%17696 = torch_c.from_builtin_tensor %17695 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17697 = torch.aten.mul.Tensor %17696, %17693 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17698 = torch_c.to_builtin_tensor %17697 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15237 = arith.constant 1 : index
%dim_15238 = tensor.dim %17698, %c1_15237 : tensor<4x?x4x64xcomplex<f32>>
%17699 = flow.tensor.bitcast %17698 : tensor<4x?x4x64xcomplex<f32>>{%dim_15238} -> tensor<4x?x4x128xf32>{%dim_15238}
%17700 = torch_c.from_builtin_tensor %17699 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15239 = torch.constant.int 5
%17701 = torch.prims.convert_element_type %17700, %int5_15239 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15240 = torch.constant.int 1
%17702 = torch.aten.size.int %17409, %int1_15240 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15241 = torch.constant.int 0
%17703 = torch.aten.add.int %int0_15241, %17702 : !torch.int, !torch.int -> !torch.int
%int0_15242 = torch.constant.int 0
%int0_15243 = torch.constant.int 0
%int1_15244 = torch.constant.int 1
%17704 = torch.aten.slice.Tensor %17632, %int0_15242, %int0_15243, %17703, %int1_15244 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17704, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15245 = torch.constant.int 1
%int0_15246 = torch.constant.int 0
%int9223372036854775807_15247 = torch.constant.int 9223372036854775807
%int1_15248 = torch.constant.int 1
%17705 = torch.aten.slice.Tensor %17704, %int1_15245, %int0_15246, %int9223372036854775807_15247, %int1_15248 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17705, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15249 = torch.constant.int 0
%17706 = torch.aten.unsqueeze %17705, %int0_15249 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17706, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15250 = torch.constant.int 2
%17707 = torch.aten.unsqueeze %17706, %int2_15250 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17707, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15251 = torch.constant.int 3
%int0_15252 = torch.constant.int 0
%int9223372036854775807_15253 = torch.constant.int 9223372036854775807
%int1_15254 = torch.constant.int 1
%17708 = torch.aten.slice.Tensor %17707, %int3_15251, %int0_15252, %int9223372036854775807_15253, %int1_15254 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17708, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17709 = torch_c.to_builtin_tensor %17565 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15255 = arith.constant 1 : index
%dim_15256 = tensor.dim %17709, %c1_15255 : tensor<4x?x4x128xf16>
%17710 = flow.tensor.bitcast %17709 : tensor<4x?x4x128xf16>{%dim_15256} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15256}
%17711 = torch_c.from_builtin_tensor %17710 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17712 = torch.aten.mul.Tensor %17711, %17708 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17713 = torch_c.to_builtin_tensor %17712 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15257 = arith.constant 1 : index
%dim_15258 = tensor.dim %17713, %c1_15257 : tensor<4x?x4x64xcomplex<f32>>
%17714 = flow.tensor.bitcast %17713 : tensor<4x?x4x64xcomplex<f32>>{%dim_15258} -> tensor<4x?x4x128xf32>{%dim_15258}
%17715 = torch_c.from_builtin_tensor %17714 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15259 = torch.constant.int 5
%17716 = torch.prims.convert_element_type %17715, %int5_15259 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15260 = torch.constant.int 1
%17717 = torch.aten.size.int %17415, %int1_15260 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15261 = torch.constant.int 0
%17718 = torch.aten.add.int %int0_15261, %17717 : !torch.int, !torch.int -> !torch.int
%int0_15262 = torch.constant.int 0
%int0_15263 = torch.constant.int 0
%int1_15264 = torch.constant.int 1
%17719 = torch.aten.slice.Tensor %17635, %int0_15262, %int0_15263, %17718, %int1_15264 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17719, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15265 = torch.constant.int 1
%int0_15266 = torch.constant.int 0
%int9223372036854775807_15267 = torch.constant.int 9223372036854775807
%int1_15268 = torch.constant.int 1
%17720 = torch.aten.slice.Tensor %17719, %int1_15265, %int0_15266, %int9223372036854775807_15267, %int1_15268 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17720, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15269 = torch.constant.int 0
%17721 = torch.aten.unsqueeze %17720, %int0_15269 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17721, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15270 = torch.constant.int 2
%17722 = torch.aten.unsqueeze %17721, %int2_15270 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17722, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15271 = torch.constant.int 3
%int0_15272 = torch.constant.int 0
%int9223372036854775807_15273 = torch.constant.int 9223372036854775807
%int1_15274 = torch.constant.int 1
%17723 = torch.aten.slice.Tensor %17722, %int3_15271, %int0_15272, %int9223372036854775807_15273, %int1_15274 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17723, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17724 = torch_c.to_builtin_tensor %17567 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15275 = arith.constant 1 : index
%dim_15276 = tensor.dim %17724, %c1_15275 : tensor<4x?x4x128xf16>
%17725 = flow.tensor.bitcast %17724 : tensor<4x?x4x128xf16>{%dim_15276} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15276}
%17726 = torch_c.from_builtin_tensor %17725 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17727 = torch.aten.mul.Tensor %17726, %17723 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17728 = torch_c.to_builtin_tensor %17727 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15277 = arith.constant 1 : index
%dim_15278 = tensor.dim %17728, %c1_15277 : tensor<4x?x4x64xcomplex<f32>>
%17729 = flow.tensor.bitcast %17728 : tensor<4x?x4x64xcomplex<f32>>{%dim_15278} -> tensor<4x?x4x128xf32>{%dim_15278}
%17730 = torch_c.from_builtin_tensor %17729 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15279 = torch.constant.int 5
%17731 = torch.prims.convert_element_type %17730, %int5_15279 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15280 = torch.constant.int 1
%17732 = torch.aten.size.int %17421, %int1_15280 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15281 = torch.constant.int 0
%17733 = torch.aten.add.int %int0_15281, %17732 : !torch.int, !torch.int -> !torch.int
%int0_15282 = torch.constant.int 0
%int0_15283 = torch.constant.int 0
%int1_15284 = torch.constant.int 1
%17734 = torch.aten.slice.Tensor %17638, %int0_15282, %int0_15283, %17733, %int1_15284 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17734, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15285 = torch.constant.int 1
%int0_15286 = torch.constant.int 0
%int9223372036854775807_15287 = torch.constant.int 9223372036854775807
%int1_15288 = torch.constant.int 1
%17735 = torch.aten.slice.Tensor %17734, %int1_15285, %int0_15286, %int9223372036854775807_15287, %int1_15288 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17735, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15289 = torch.constant.int 0
%17736 = torch.aten.unsqueeze %17735, %int0_15289 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17736, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15290 = torch.constant.int 2
%17737 = torch.aten.unsqueeze %17736, %int2_15290 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17737, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15291 = torch.constant.int 3
%int0_15292 = torch.constant.int 0
%int9223372036854775807_15293 = torch.constant.int 9223372036854775807
%int1_15294 = torch.constant.int 1
%17738 = torch.aten.slice.Tensor %17737, %int3_15291, %int0_15292, %int9223372036854775807_15293, %int1_15294 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17738, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17739 = torch_c.to_builtin_tensor %17569 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15295 = arith.constant 1 : index
%dim_15296 = tensor.dim %17739, %c1_15295 : tensor<4x?x4x128xf16>
%17740 = flow.tensor.bitcast %17739 : tensor<4x?x4x128xf16>{%dim_15296} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15296}
%17741 = torch_c.from_builtin_tensor %17740 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17742 = torch.aten.mul.Tensor %17741, %17738 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17743 = torch_c.to_builtin_tensor %17742 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15297 = arith.constant 1 : index
%dim_15298 = tensor.dim %17743, %c1_15297 : tensor<4x?x4x64xcomplex<f32>>
%17744 = flow.tensor.bitcast %17743 : tensor<4x?x4x64xcomplex<f32>>{%dim_15298} -> tensor<4x?x4x128xf32>{%dim_15298}
%17745 = torch_c.from_builtin_tensor %17744 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15299 = torch.constant.int 5
%17746 = torch.prims.convert_element_type %17745, %int5_15299 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15300 = torch.constant.int 1
%17747 = torch.aten.size.int %17427, %int1_15300 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_15301 = torch.constant.int 0
%17748 = torch.aten.add.int %int0_15301, %17747 : !torch.int, !torch.int -> !torch.int
%int0_15302 = torch.constant.int 0
%int0_15303 = torch.constant.int 0
%int1_15304 = torch.constant.int 1
%17749 = torch.aten.slice.Tensor %17641, %int0_15302, %int0_15303, %17748, %int1_15304 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17749, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15305 = torch.constant.int 1
%int0_15306 = torch.constant.int 0
%int9223372036854775807_15307 = torch.constant.int 9223372036854775807
%int1_15308 = torch.constant.int 1
%17750 = torch.aten.slice.Tensor %17749, %int1_15305, %int0_15306, %int9223372036854775807_15307, %int1_15308 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17750, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15309 = torch.constant.int 0
%17751 = torch.aten.unsqueeze %17750, %int0_15309 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17751, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15310 = torch.constant.int 2
%17752 = torch.aten.unsqueeze %17751, %int2_15310 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17752, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15311 = torch.constant.int 3
%int0_15312 = torch.constant.int 0
%int9223372036854775807_15313 = torch.constant.int 9223372036854775807
%int1_15314 = torch.constant.int 1
%17753 = torch.aten.slice.Tensor %17752, %int3_15311, %int0_15312, %int9223372036854775807_15313, %int1_15314 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17753, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17754 = torch_c.to_builtin_tensor %17571 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_15315 = arith.constant 1 : index
%dim_15316 = tensor.dim %17754, %c1_15315 : tensor<4x?x4x128xf16>
%17755 = flow.tensor.bitcast %17754 : tensor<4x?x4x128xf16>{%dim_15316} -> tensor<4x?x4x64xcomplex<f16>>{%dim_15316}
%17756 = torch_c.from_builtin_tensor %17755 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %17756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%17757 = torch.aten.mul.Tensor %17756, %17753 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %17757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%17758 = torch_c.to_builtin_tensor %17757 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_15317 = arith.constant 1 : index
%dim_15318 = tensor.dim %17758, %c1_15317 : tensor<4x?x4x64xcomplex<f32>>
%17759 = flow.tensor.bitcast %17758 : tensor<4x?x4x64xcomplex<f32>>{%dim_15318} -> tensor<4x?x4x128xf32>{%dim_15318}
%17760 = torch_c.from_builtin_tensor %17759 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %17760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_15319 = torch.constant.int 5
%17761 = torch.prims.convert_element_type %17760, %int5_15319 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %17761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_15320 = torch.constant.int 131072
%none_15321 = torch.constant.none
%none_15322 = torch.constant.none
%cpu_15323 = torch.constant.device "cpu"
%false_15324 = torch.constant.bool false
%17762 = torch.aten.arange %int131072_15320, %none_15321, %none_15322, %cpu_15323, %false_15324 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_15325 = torch.constant.int 0
%int128_15326 = torch.constant.int 128
%int2_15327 = torch.constant.int 2
%none_15328 = torch.constant.none
%none_15329 = torch.constant.none
%cpu_15330 = torch.constant.device "cpu"
%false_15331 = torch.constant.bool false
%17763 = torch.aten.arange.start_step %int0_15325, %int128_15326, %int2_15327, %none_15328, %none_15329, %cpu_15330, %false_15331 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_15332 = torch.constant.int 0
%int0_15333 = torch.constant.int 0
%int64_15334 = torch.constant.int 64
%int1_15335 = torch.constant.int 1
%17764 = torch.aten.slice.Tensor %17763, %int0_15332, %int0_15333, %int64_15334, %int1_15335 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_15336 = torch.constant.int 6
%17765 = torch.prims.convert_element_type %17764, %int6_15336 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_15337 = torch.constant.int 128
%17766 = torch.aten.div.Scalar %17765, %int128_15337 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_15338 = torch.constant.float 5.000000e+05
%17767 = torch.aten.pow.Scalar %float5.000000e05_15338, %17766 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%17768 = torch.aten.reciprocal %17767 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_15339 = torch.constant.float 1.000000e+00
%17769 = torch.aten.mul.Scalar %17768, %float1.000000e00_15339 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_15340 = torch.constant.int 131072
%int1_15341 = torch.constant.int 1
%17770 = torch.prim.ListConstruct %int131072_15340, %int1_15341 : (!torch.int, !torch.int) -> !torch.list<int>
%17771 = torch.aten.view %17762, %17770 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%17772 = torch.aten.mul.Tensor %17771, %17769 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%17773 = torch.aten.cos %17772 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%17774 = torch.aten.sin %17772 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%17775 = torch.aten.complex %17773, %17774 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%17776 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17777 = flow.tensor.transfer %17776 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%17778 = torch_c.from_builtin_tensor %17777 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17779 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17780 = flow.tensor.transfer %17779 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%17781 = torch_c.from_builtin_tensor %17780 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17782 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17783 = flow.tensor.transfer %17782 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%17784 = torch_c.from_builtin_tensor %17783 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17785 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17786 = flow.tensor.transfer %17785 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%17787 = torch_c.from_builtin_tensor %17786 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17788 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17789 = flow.tensor.transfer %17788 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%17790 = torch_c.from_builtin_tensor %17789 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17791 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17792 = flow.tensor.transfer %17791 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%17793 = torch_c.from_builtin_tensor %17792 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17794 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17795 = flow.tensor.transfer %17794 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%17796 = torch_c.from_builtin_tensor %17795 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%17797 = torch_c.to_builtin_tensor %17775 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%17798 = flow.tensor.transfer %17797 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%17799 = torch_c.from_builtin_tensor %17798 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_15342 = torch.constant.int 1
%17800 = torch.aten.size.int %17449, %int1_15342 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15343 = torch.constant.int 0
%17801 = torch.aten.add.int %int0_15343, %17800 : !torch.int, !torch.int -> !torch.int
%int0_15344 = torch.constant.int 0
%int0_15345 = torch.constant.int 0
%int1_15346 = torch.constant.int 1
%17802 = torch.aten.slice.Tensor %17778, %int0_15344, %int0_15345, %17801, %int1_15346 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17802, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15347 = torch.constant.int 1
%int0_15348 = torch.constant.int 0
%int9223372036854775807_15349 = torch.constant.int 9223372036854775807
%int1_15350 = torch.constant.int 1
%17803 = torch.aten.slice.Tensor %17802, %int1_15347, %int0_15348, %int9223372036854775807_15349, %int1_15350 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17803, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15351 = torch.constant.int 0
%17804 = torch.aten.unsqueeze %17803, %int0_15351 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17804, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15352 = torch.constant.int 2
%17805 = torch.aten.unsqueeze %17804, %int2_15352 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17805, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15353 = torch.constant.int 3
%int0_15354 = torch.constant.int 0
%int9223372036854775807_15355 = torch.constant.int 9223372036854775807
%int1_15356 = torch.constant.int 1
%17806 = torch.aten.slice.Tensor %17805, %int3_15353, %int0_15354, %int9223372036854775807_15355, %int1_15356 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17806, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17807 = torch_c.to_builtin_tensor %17573 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15357 = arith.constant 1 : index
%dim_15358 = tensor.dim %17807, %c1_15357 : tensor<4x?x1x128xf16>
%17808 = flow.tensor.bitcast %17807 : tensor<4x?x1x128xf16>{%dim_15358} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15358}
%17809 = torch_c.from_builtin_tensor %17808 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17810 = torch.aten.mul.Tensor %17809, %17806 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17811 = torch_c.to_builtin_tensor %17810 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15359 = arith.constant 1 : index
%dim_15360 = tensor.dim %17811, %c1_15359 : tensor<4x?x1x64xcomplex<f32>>
%17812 = flow.tensor.bitcast %17811 : tensor<4x?x1x64xcomplex<f32>>{%dim_15360} -> tensor<4x?x1x128xf32>{%dim_15360}
%17813 = torch_c.from_builtin_tensor %17812 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15361 = torch.constant.int 5
%17814 = torch.prims.convert_element_type %17813, %int5_15361 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15362 = torch.constant.int 1
%17815 = torch.aten.size.int %17455, %int1_15362 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15363 = torch.constant.int 0
%17816 = torch.aten.add.int %int0_15363, %17815 : !torch.int, !torch.int -> !torch.int
%int0_15364 = torch.constant.int 0
%int0_15365 = torch.constant.int 0
%int1_15366 = torch.constant.int 1
%17817 = torch.aten.slice.Tensor %17781, %int0_15364, %int0_15365, %17816, %int1_15366 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17817, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15367 = torch.constant.int 1
%int0_15368 = torch.constant.int 0
%int9223372036854775807_15369 = torch.constant.int 9223372036854775807
%int1_15370 = torch.constant.int 1
%17818 = torch.aten.slice.Tensor %17817, %int1_15367, %int0_15368, %int9223372036854775807_15369, %int1_15370 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17818, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15371 = torch.constant.int 0
%17819 = torch.aten.unsqueeze %17818, %int0_15371 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17819, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15372 = torch.constant.int 2
%17820 = torch.aten.unsqueeze %17819, %int2_15372 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17820, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15373 = torch.constant.int 3
%int0_15374 = torch.constant.int 0
%int9223372036854775807_15375 = torch.constant.int 9223372036854775807
%int1_15376 = torch.constant.int 1
%17821 = torch.aten.slice.Tensor %17820, %int3_15373, %int0_15374, %int9223372036854775807_15375, %int1_15376 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17821, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17822 = torch_c.to_builtin_tensor %17575 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15377 = arith.constant 1 : index
%dim_15378 = tensor.dim %17822, %c1_15377 : tensor<4x?x1x128xf16>
%17823 = flow.tensor.bitcast %17822 : tensor<4x?x1x128xf16>{%dim_15378} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15378}
%17824 = torch_c.from_builtin_tensor %17823 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17825 = torch.aten.mul.Tensor %17824, %17821 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17826 = torch_c.to_builtin_tensor %17825 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15379 = arith.constant 1 : index
%dim_15380 = tensor.dim %17826, %c1_15379 : tensor<4x?x1x64xcomplex<f32>>
%17827 = flow.tensor.bitcast %17826 : tensor<4x?x1x64xcomplex<f32>>{%dim_15380} -> tensor<4x?x1x128xf32>{%dim_15380}
%17828 = torch_c.from_builtin_tensor %17827 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15381 = torch.constant.int 5
%17829 = torch.prims.convert_element_type %17828, %int5_15381 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15382 = torch.constant.int 1
%17830 = torch.aten.size.int %17461, %int1_15382 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15383 = torch.constant.int 0
%17831 = torch.aten.add.int %int0_15383, %17830 : !torch.int, !torch.int -> !torch.int
%int0_15384 = torch.constant.int 0
%int0_15385 = torch.constant.int 0
%int1_15386 = torch.constant.int 1
%17832 = torch.aten.slice.Tensor %17784, %int0_15384, %int0_15385, %17831, %int1_15386 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17832, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15387 = torch.constant.int 1
%int0_15388 = torch.constant.int 0
%int9223372036854775807_15389 = torch.constant.int 9223372036854775807
%int1_15390 = torch.constant.int 1
%17833 = torch.aten.slice.Tensor %17832, %int1_15387, %int0_15388, %int9223372036854775807_15389, %int1_15390 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17833, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15391 = torch.constant.int 0
%17834 = torch.aten.unsqueeze %17833, %int0_15391 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17834, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15392 = torch.constant.int 2
%17835 = torch.aten.unsqueeze %17834, %int2_15392 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17835, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15393 = torch.constant.int 3
%int0_15394 = torch.constant.int 0
%int9223372036854775807_15395 = torch.constant.int 9223372036854775807
%int1_15396 = torch.constant.int 1
%17836 = torch.aten.slice.Tensor %17835, %int3_15393, %int0_15394, %int9223372036854775807_15395, %int1_15396 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17836, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17837 = torch_c.to_builtin_tensor %17577 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15397 = arith.constant 1 : index
%dim_15398 = tensor.dim %17837, %c1_15397 : tensor<4x?x1x128xf16>
%17838 = flow.tensor.bitcast %17837 : tensor<4x?x1x128xf16>{%dim_15398} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15398}
%17839 = torch_c.from_builtin_tensor %17838 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17840 = torch.aten.mul.Tensor %17839, %17836 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17841 = torch_c.to_builtin_tensor %17840 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15399 = arith.constant 1 : index
%dim_15400 = tensor.dim %17841, %c1_15399 : tensor<4x?x1x64xcomplex<f32>>
%17842 = flow.tensor.bitcast %17841 : tensor<4x?x1x64xcomplex<f32>>{%dim_15400} -> tensor<4x?x1x128xf32>{%dim_15400}
%17843 = torch_c.from_builtin_tensor %17842 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15401 = torch.constant.int 5
%17844 = torch.prims.convert_element_type %17843, %int5_15401 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15402 = torch.constant.int 1
%17845 = torch.aten.size.int %17467, %int1_15402 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15403 = torch.constant.int 0
%17846 = torch.aten.add.int %int0_15403, %17845 : !torch.int, !torch.int -> !torch.int
%int0_15404 = torch.constant.int 0
%int0_15405 = torch.constant.int 0
%int1_15406 = torch.constant.int 1
%17847 = torch.aten.slice.Tensor %17787, %int0_15404, %int0_15405, %17846, %int1_15406 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17847, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15407 = torch.constant.int 1
%int0_15408 = torch.constant.int 0
%int9223372036854775807_15409 = torch.constant.int 9223372036854775807
%int1_15410 = torch.constant.int 1
%17848 = torch.aten.slice.Tensor %17847, %int1_15407, %int0_15408, %int9223372036854775807_15409, %int1_15410 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17848, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15411 = torch.constant.int 0
%17849 = torch.aten.unsqueeze %17848, %int0_15411 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17849, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15412 = torch.constant.int 2
%17850 = torch.aten.unsqueeze %17849, %int2_15412 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17850, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15413 = torch.constant.int 3
%int0_15414 = torch.constant.int 0
%int9223372036854775807_15415 = torch.constant.int 9223372036854775807
%int1_15416 = torch.constant.int 1
%17851 = torch.aten.slice.Tensor %17850, %int3_15413, %int0_15414, %int9223372036854775807_15415, %int1_15416 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17851, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17852 = torch_c.to_builtin_tensor %17579 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15417 = arith.constant 1 : index
%dim_15418 = tensor.dim %17852, %c1_15417 : tensor<4x?x1x128xf16>
%17853 = flow.tensor.bitcast %17852 : tensor<4x?x1x128xf16>{%dim_15418} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15418}
%17854 = torch_c.from_builtin_tensor %17853 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17855 = torch.aten.mul.Tensor %17854, %17851 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17856 = torch_c.to_builtin_tensor %17855 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15419 = arith.constant 1 : index
%dim_15420 = tensor.dim %17856, %c1_15419 : tensor<4x?x1x64xcomplex<f32>>
%17857 = flow.tensor.bitcast %17856 : tensor<4x?x1x64xcomplex<f32>>{%dim_15420} -> tensor<4x?x1x128xf32>{%dim_15420}
%17858 = torch_c.from_builtin_tensor %17857 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15421 = torch.constant.int 5
%17859 = torch.prims.convert_element_type %17858, %int5_15421 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15422 = torch.constant.int 1
%17860 = torch.aten.size.int %17473, %int1_15422 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15423 = torch.constant.int 0
%17861 = torch.aten.add.int %int0_15423, %17860 : !torch.int, !torch.int -> !torch.int
%int0_15424 = torch.constant.int 0
%int0_15425 = torch.constant.int 0
%int1_15426 = torch.constant.int 1
%17862 = torch.aten.slice.Tensor %17790, %int0_15424, %int0_15425, %17861, %int1_15426 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17862, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15427 = torch.constant.int 1
%int0_15428 = torch.constant.int 0
%int9223372036854775807_15429 = torch.constant.int 9223372036854775807
%int1_15430 = torch.constant.int 1
%17863 = torch.aten.slice.Tensor %17862, %int1_15427, %int0_15428, %int9223372036854775807_15429, %int1_15430 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17863, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15431 = torch.constant.int 0
%17864 = torch.aten.unsqueeze %17863, %int0_15431 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17864, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15432 = torch.constant.int 2
%17865 = torch.aten.unsqueeze %17864, %int2_15432 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17865, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15433 = torch.constant.int 3
%int0_15434 = torch.constant.int 0
%int9223372036854775807_15435 = torch.constant.int 9223372036854775807
%int1_15436 = torch.constant.int 1
%17866 = torch.aten.slice.Tensor %17865, %int3_15433, %int0_15434, %int9223372036854775807_15435, %int1_15436 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17866, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17867 = torch_c.to_builtin_tensor %17581 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15437 = arith.constant 1 : index
%dim_15438 = tensor.dim %17867, %c1_15437 : tensor<4x?x1x128xf16>
%17868 = flow.tensor.bitcast %17867 : tensor<4x?x1x128xf16>{%dim_15438} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15438}
%17869 = torch_c.from_builtin_tensor %17868 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17870 = torch.aten.mul.Tensor %17869, %17866 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17871 = torch_c.to_builtin_tensor %17870 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15439 = arith.constant 1 : index
%dim_15440 = tensor.dim %17871, %c1_15439 : tensor<4x?x1x64xcomplex<f32>>
%17872 = flow.tensor.bitcast %17871 : tensor<4x?x1x64xcomplex<f32>>{%dim_15440} -> tensor<4x?x1x128xf32>{%dim_15440}
%17873 = torch_c.from_builtin_tensor %17872 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15441 = torch.constant.int 5
%17874 = torch.prims.convert_element_type %17873, %int5_15441 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15442 = torch.constant.int 1
%17875 = torch.aten.size.int %17479, %int1_15442 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15443 = torch.constant.int 0
%17876 = torch.aten.add.int %int0_15443, %17875 : !torch.int, !torch.int -> !torch.int
%int0_15444 = torch.constant.int 0
%int0_15445 = torch.constant.int 0
%int1_15446 = torch.constant.int 1
%17877 = torch.aten.slice.Tensor %17793, %int0_15444, %int0_15445, %17876, %int1_15446 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17877, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15447 = torch.constant.int 1
%int0_15448 = torch.constant.int 0
%int9223372036854775807_15449 = torch.constant.int 9223372036854775807
%int1_15450 = torch.constant.int 1
%17878 = torch.aten.slice.Tensor %17877, %int1_15447, %int0_15448, %int9223372036854775807_15449, %int1_15450 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17878, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15451 = torch.constant.int 0
%17879 = torch.aten.unsqueeze %17878, %int0_15451 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17879, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15452 = torch.constant.int 2
%17880 = torch.aten.unsqueeze %17879, %int2_15452 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17880, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15453 = torch.constant.int 3
%int0_15454 = torch.constant.int 0
%int9223372036854775807_15455 = torch.constant.int 9223372036854775807
%int1_15456 = torch.constant.int 1
%17881 = torch.aten.slice.Tensor %17880, %int3_15453, %int0_15454, %int9223372036854775807_15455, %int1_15456 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17881, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17882 = torch_c.to_builtin_tensor %17583 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15457 = arith.constant 1 : index
%dim_15458 = tensor.dim %17882, %c1_15457 : tensor<4x?x1x128xf16>
%17883 = flow.tensor.bitcast %17882 : tensor<4x?x1x128xf16>{%dim_15458} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15458}
%17884 = torch_c.from_builtin_tensor %17883 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17885 = torch.aten.mul.Tensor %17884, %17881 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17886 = torch_c.to_builtin_tensor %17885 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15459 = arith.constant 1 : index
%dim_15460 = tensor.dim %17886, %c1_15459 : tensor<4x?x1x64xcomplex<f32>>
%17887 = flow.tensor.bitcast %17886 : tensor<4x?x1x64xcomplex<f32>>{%dim_15460} -> tensor<4x?x1x128xf32>{%dim_15460}
%17888 = torch_c.from_builtin_tensor %17887 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15461 = torch.constant.int 5
%17889 = torch.prims.convert_element_type %17888, %int5_15461 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15462 = torch.constant.int 1
%17890 = torch.aten.size.int %17485, %int1_15462 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15463 = torch.constant.int 0
%17891 = torch.aten.add.int %int0_15463, %17890 : !torch.int, !torch.int -> !torch.int
%int0_15464 = torch.constant.int 0
%int0_15465 = torch.constant.int 0
%int1_15466 = torch.constant.int 1
%17892 = torch.aten.slice.Tensor %17796, %int0_15464, %int0_15465, %17891, %int1_15466 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17892, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15467 = torch.constant.int 1
%int0_15468 = torch.constant.int 0
%int9223372036854775807_15469 = torch.constant.int 9223372036854775807
%int1_15470 = torch.constant.int 1
%17893 = torch.aten.slice.Tensor %17892, %int1_15467, %int0_15468, %int9223372036854775807_15469, %int1_15470 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17893, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15471 = torch.constant.int 0
%17894 = torch.aten.unsqueeze %17893, %int0_15471 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17894, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15472 = torch.constant.int 2
%17895 = torch.aten.unsqueeze %17894, %int2_15472 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17895, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15473 = torch.constant.int 3
%int0_15474 = torch.constant.int 0
%int9223372036854775807_15475 = torch.constant.int 9223372036854775807
%int1_15476 = torch.constant.int 1
%17896 = torch.aten.slice.Tensor %17895, %int3_15473, %int0_15474, %int9223372036854775807_15475, %int1_15476 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17896, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17897 = torch_c.to_builtin_tensor %17585 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15477 = arith.constant 1 : index
%dim_15478 = tensor.dim %17897, %c1_15477 : tensor<4x?x1x128xf16>
%17898 = flow.tensor.bitcast %17897 : tensor<4x?x1x128xf16>{%dim_15478} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15478}
%17899 = torch_c.from_builtin_tensor %17898 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17900 = torch.aten.mul.Tensor %17899, %17896 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17901 = torch_c.to_builtin_tensor %17900 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15479 = arith.constant 1 : index
%dim_15480 = tensor.dim %17901, %c1_15479 : tensor<4x?x1x64xcomplex<f32>>
%17902 = flow.tensor.bitcast %17901 : tensor<4x?x1x64xcomplex<f32>>{%dim_15480} -> tensor<4x?x1x128xf32>{%dim_15480}
%17903 = torch_c.from_builtin_tensor %17902 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15481 = torch.constant.int 5
%17904 = torch.prims.convert_element_type %17903, %int5_15481 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_15482 = torch.constant.int 1
%17905 = torch.aten.size.int %17491, %int1_15482 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_15483 = torch.constant.int 0
%17906 = torch.aten.add.int %int0_15483, %17905 : !torch.int, !torch.int -> !torch.int
%int0_15484 = torch.constant.int 0
%int0_15485 = torch.constant.int 0
%int1_15486 = torch.constant.int 1
%17907 = torch.aten.slice.Tensor %17799, %int0_15484, %int0_15485, %17906, %int1_15486 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17907, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_15487 = torch.constant.int 1
%int0_15488 = torch.constant.int 0
%int9223372036854775807_15489 = torch.constant.int 9223372036854775807
%int1_15490 = torch.constant.int 1
%17908 = torch.aten.slice.Tensor %17907, %int1_15487, %int0_15488, %int9223372036854775807_15489, %int1_15490 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %17908, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_15491 = torch.constant.int 0
%17909 = torch.aten.unsqueeze %17908, %int0_15491 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %17909, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_15492 = torch.constant.int 2
%17910 = torch.aten.unsqueeze %17909, %int2_15492 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17910, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_15493 = torch.constant.int 3
%int0_15494 = torch.constant.int 0
%int9223372036854775807_15495 = torch.constant.int 9223372036854775807
%int1_15496 = torch.constant.int 1
%17911 = torch.aten.slice.Tensor %17910, %int3_15493, %int0_15494, %int9223372036854775807_15495, %int1_15496 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17911, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%17912 = torch_c.to_builtin_tensor %17587 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_15497 = arith.constant 1 : index
%dim_15498 = tensor.dim %17912, %c1_15497 : tensor<4x?x1x128xf16>
%17913 = flow.tensor.bitcast %17912 : tensor<4x?x1x128xf16>{%dim_15498} -> tensor<4x?x1x64xcomplex<f16>>{%dim_15498}
%17914 = torch_c.from_builtin_tensor %17913 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %17914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%17915 = torch.aten.mul.Tensor %17914, %17911 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %17915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%17916 = torch_c.to_builtin_tensor %17915 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_15499 = arith.constant 1 : index
%dim_15500 = tensor.dim %17916, %c1_15499 : tensor<4x?x1x64xcomplex<f32>>
%17917 = flow.tensor.bitcast %17916 : tensor<4x?x1x64xcomplex<f32>>{%dim_15500} -> tensor<4x?x1x128xf32>{%dim_15500}
%17918 = torch_c.from_builtin_tensor %17917 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %17918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_15501 = torch.constant.int 5
%17919 = torch.prims.convert_element_type %17918, %int5_15501 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %17919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_15502 = torch.constant.int 64
%17920 = torch.aten.mul.Scalar %2364, %int64_15502 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17920, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15503 = torch.constant.int 64
%17921 = torch.aten.mul.Scalar %2367, %int64_15503 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17921, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15504 = torch.constant.int 64
%17922 = torch.aten.mul.Scalar %2370, %int64_15504 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17922, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15505 = torch.constant.int 64
%17923 = torch.aten.mul.Scalar %2373, %int64_15505 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17923, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15506 = torch.constant.int 64
%17924 = torch.aten.mul.Scalar %2376, %int64_15506 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17924, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15507 = torch.constant.int 64
%17925 = torch.aten.mul.Scalar %2379, %int64_15507 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17925, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15508 = torch.constant.int 64
%17926 = torch.aten.mul.Scalar %2382, %int64_15508 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17926, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_15509 = torch.constant.int 64
%17927 = torch.aten.mul.Scalar %2385, %int64_15509 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17927, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15510 = torch.constant.int 16
%int1_15511 = torch.constant.int 1
%17928 = torch.aten.add.Scalar %17920, %int16_15510, %int1_15511 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17928, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15512 = torch.constant.int 16
%int1_15513 = torch.constant.int 1
%17929 = torch.aten.add.Scalar %17921, %int16_15512, %int1_15513 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17929, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15514 = torch.constant.int 16
%int1_15515 = torch.constant.int 1
%17930 = torch.aten.add.Scalar %17922, %int16_15514, %int1_15515 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17930, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15516 = torch.constant.int 16
%int1_15517 = torch.constant.int 1
%17931 = torch.aten.add.Scalar %17923, %int16_15516, %int1_15517 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17931, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15518 = torch.constant.int 16
%int1_15519 = torch.constant.int 1
%17932 = torch.aten.add.Scalar %17924, %int16_15518, %int1_15519 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17932, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15520 = torch.constant.int 16
%int1_15521 = torch.constant.int 1
%17933 = torch.aten.add.Scalar %17925, %int16_15520, %int1_15521 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17933, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15522 = torch.constant.int 16
%int1_15523 = torch.constant.int 1
%17934 = torch.aten.add.Scalar %17926, %int16_15522, %int1_15523 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17934, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int16_15524 = torch.constant.int 16
%int1_15525 = torch.constant.int 1
%17935 = torch.aten.add.Scalar %17927, %int16_15524, %int1_15525 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %17935, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_15526 = torch.constant.int 4
%int16_15527 = torch.constant.int 16
%int1_15528 = torch.constant.int 1
%int128_15529 = torch.constant.int 128
%17936 = torch.prim.ListConstruct %int4_15526, %3095, %int16_15527, %int1_15528, %int128_15529 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17937 = torch.aten.view %17814, %17936 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17937, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15530 = torch.constant.int 4
%int16_15531 = torch.constant.int 16
%int1_15532 = torch.constant.int 1
%int128_15533 = torch.constant.int 128
%17938 = torch.prim.ListConstruct %int4_15530, %3095, %int16_15531, %int1_15532, %int128_15533 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17939 = torch.aten.view %17829, %17938 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17939, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15534 = torch.constant.int 4
%int16_15535 = torch.constant.int 16
%int1_15536 = torch.constant.int 1
%int128_15537 = torch.constant.int 128
%17940 = torch.prim.ListConstruct %int4_15534, %3095, %int16_15535, %int1_15536, %int128_15537 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17941 = torch.aten.view %17844, %17940 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17941, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15538 = torch.constant.int 4
%int16_15539 = torch.constant.int 16
%int1_15540 = torch.constant.int 1
%int128_15541 = torch.constant.int 128
%17942 = torch.prim.ListConstruct %int4_15538, %3095, %int16_15539, %int1_15540, %int128_15541 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17943 = torch.aten.view %17859, %17942 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17943, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15542 = torch.constant.int 4
%int16_15543 = torch.constant.int 16
%int1_15544 = torch.constant.int 1
%int128_15545 = torch.constant.int 128
%17944 = torch.prim.ListConstruct %int4_15542, %3095, %int16_15543, %int1_15544, %int128_15545 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17945 = torch.aten.view %17874, %17944 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17945, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15546 = torch.constant.int 4
%int16_15547 = torch.constant.int 16
%int1_15548 = torch.constant.int 1
%int128_15549 = torch.constant.int 128
%17946 = torch.prim.ListConstruct %int4_15546, %3095, %int16_15547, %int1_15548, %int128_15549 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17947 = torch.aten.view %17889, %17946 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17947, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15550 = torch.constant.int 4
%int16_15551 = torch.constant.int 16
%int1_15552 = torch.constant.int 1
%int128_15553 = torch.constant.int 128
%17948 = torch.prim.ListConstruct %int4_15550, %3095, %int16_15551, %int1_15552, %int128_15553 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17949 = torch.aten.view %17904, %17948 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17949, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15554 = torch.constant.int 4
%int16_15555 = torch.constant.int 16
%int1_15556 = torch.constant.int 1
%int128_15557 = torch.constant.int 128
%17950 = torch.prim.ListConstruct %int4_15554, %3095, %int16_15555, %int1_15556, %int128_15557 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17951 = torch.aten.view %17919, %17950 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %17951, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15558 = torch.constant.int 4
%17952 = torch.aten.mul.int %int4_15558, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15559 = torch.constant.int 16
%int1_15560 = torch.constant.int 1
%int128_15561 = torch.constant.int 128
%17953 = torch.prim.ListConstruct %17952, %int16_15559, %int1_15560, %int128_15561 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17954 = torch.aten.view %17937, %17953 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17954, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15562 = torch.constant.int 4
%17955 = torch.aten.mul.int %int4_15562, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15563 = torch.constant.int 16
%int1_15564 = torch.constant.int 1
%int128_15565 = torch.constant.int 128
%17956 = torch.prim.ListConstruct %17955, %int16_15563, %int1_15564, %int128_15565 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17957 = torch.aten.view %17939, %17956 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17957, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15566 = torch.constant.int 4
%17958 = torch.aten.mul.int %int4_15566, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15567 = torch.constant.int 16
%int1_15568 = torch.constant.int 1
%int128_15569 = torch.constant.int 128
%17959 = torch.prim.ListConstruct %17958, %int16_15567, %int1_15568, %int128_15569 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17960 = torch.aten.view %17941, %17959 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17960, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15570 = torch.constant.int 4
%17961 = torch.aten.mul.int %int4_15570, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15571 = torch.constant.int 16
%int1_15572 = torch.constant.int 1
%int128_15573 = torch.constant.int 128
%17962 = torch.prim.ListConstruct %17961, %int16_15571, %int1_15572, %int128_15573 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17963 = torch.aten.view %17943, %17962 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17963, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15574 = torch.constant.int 4
%17964 = torch.aten.mul.int %int4_15574, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15575 = torch.constant.int 16
%int1_15576 = torch.constant.int 1
%int128_15577 = torch.constant.int 128
%17965 = torch.prim.ListConstruct %17964, %int16_15575, %int1_15576, %int128_15577 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17966 = torch.aten.view %17945, %17965 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17966, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15578 = torch.constant.int 4
%17967 = torch.aten.mul.int %int4_15578, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15579 = torch.constant.int 16
%int1_15580 = torch.constant.int 1
%int128_15581 = torch.constant.int 128
%17968 = torch.prim.ListConstruct %17967, %int16_15579, %int1_15580, %int128_15581 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17969 = torch.aten.view %17947, %17968 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17969, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15582 = torch.constant.int 4
%17970 = torch.aten.mul.int %int4_15582, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15583 = torch.constant.int 16
%int1_15584 = torch.constant.int 1
%int128_15585 = torch.constant.int 128
%17971 = torch.prim.ListConstruct %17970, %int16_15583, %int1_15584, %int128_15585 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17972 = torch.aten.view %17949, %17971 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17972, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15586 = torch.constant.int 4
%17973 = torch.aten.mul.int %int4_15586, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15587 = torch.constant.int 16
%int1_15588 = torch.constant.int 1
%int128_15589 = torch.constant.int 128
%17974 = torch.prim.ListConstruct %17973, %int16_15587, %int1_15588, %int128_15589 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%17975 = torch.aten.view %17951, %17974 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %17975, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15590 = torch.constant.int 4
%17976 = torch.aten.mul.int %int4_15590, %3095 : !torch.int, !torch.int -> !torch.int
%17977 = torch.prim.ListConstruct %17976 : (!torch.int) -> !torch.list<int>
%17978 = torch.aten.view %17928, %17977 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17978, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15591 = torch.constant.int 4
%17979 = torch.aten.mul.int %int4_15591, %3095 : !torch.int, !torch.int -> !torch.int
%17980 = torch.prim.ListConstruct %17979 : (!torch.int) -> !torch.list<int>
%17981 = torch.aten.view %17929, %17980 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17981, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15592 = torch.constant.int 4
%17982 = torch.aten.mul.int %int4_15592, %3095 : !torch.int, !torch.int -> !torch.int
%17983 = torch.prim.ListConstruct %17982 : (!torch.int) -> !torch.list<int>
%17984 = torch.aten.view %17930, %17983 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17984, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15593 = torch.constant.int 4
%17985 = torch.aten.mul.int %int4_15593, %3095 : !torch.int, !torch.int -> !torch.int
%17986 = torch.prim.ListConstruct %17985 : (!torch.int) -> !torch.list<int>
%17987 = torch.aten.view %17931, %17986 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17987, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15594 = torch.constant.int 4
%17988 = torch.aten.mul.int %int4_15594, %3095 : !torch.int, !torch.int -> !torch.int
%17989 = torch.prim.ListConstruct %17988 : (!torch.int) -> !torch.list<int>
%17990 = torch.aten.view %17932, %17989 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17990, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15595 = torch.constant.int 4
%17991 = torch.aten.mul.int %int4_15595, %3095 : !torch.int, !torch.int -> !torch.int
%17992 = torch.prim.ListConstruct %17991 : (!torch.int) -> !torch.list<int>
%17993 = torch.aten.view %17933, %17992 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17993, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15596 = torch.constant.int 4
%17994 = torch.aten.mul.int %int4_15596, %3095 : !torch.int, !torch.int -> !torch.int
%17995 = torch.prim.ListConstruct %17994 : (!torch.int) -> !torch.list<int>
%17996 = torch.aten.view %17934, %17995 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17996, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15597 = torch.constant.int 4
%17997 = torch.aten.mul.int %int4_15597, %3095 : !torch.int, !torch.int -> !torch.int
%17998 = torch.prim.ListConstruct %17997 : (!torch.int) -> !torch.list<int>
%17999 = torch.aten.view %17935, %17998 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %17999, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15598 = torch.constant.int 4
%int16_15599 = torch.constant.int 16
%int1_15600 = torch.constant.int 1
%int128_15601 = torch.constant.int 128
%18000 = torch.prim.ListConstruct %int4_15598, %3095, %int16_15599, %int1_15600, %int128_15601 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18001 = torch.aten.view %17589, %18000 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18001, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15602 = torch.constant.int 4
%int16_15603 = torch.constant.int 16
%int1_15604 = torch.constant.int 1
%int128_15605 = torch.constant.int 128
%18002 = torch.prim.ListConstruct %int4_15602, %3095, %int16_15603, %int1_15604, %int128_15605 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18003 = torch.aten.view %17591, %18002 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18003, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15606 = torch.constant.int 4
%int16_15607 = torch.constant.int 16
%int1_15608 = torch.constant.int 1
%int128_15609 = torch.constant.int 128
%18004 = torch.prim.ListConstruct %int4_15606, %3095, %int16_15607, %int1_15608, %int128_15609 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18005 = torch.aten.view %17593, %18004 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18005, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15610 = torch.constant.int 4
%int16_15611 = torch.constant.int 16
%int1_15612 = torch.constant.int 1
%int128_15613 = torch.constant.int 128
%18006 = torch.prim.ListConstruct %int4_15610, %3095, %int16_15611, %int1_15612, %int128_15613 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18007 = torch.aten.view %17595, %18006 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18007, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15614 = torch.constant.int 4
%int16_15615 = torch.constant.int 16
%int1_15616 = torch.constant.int 1
%int128_15617 = torch.constant.int 128
%18008 = torch.prim.ListConstruct %int4_15614, %3095, %int16_15615, %int1_15616, %int128_15617 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18009 = torch.aten.view %17597, %18008 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18009, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15618 = torch.constant.int 4
%int16_15619 = torch.constant.int 16
%int1_15620 = torch.constant.int 1
%int128_15621 = torch.constant.int 128
%18010 = torch.prim.ListConstruct %int4_15618, %3095, %int16_15619, %int1_15620, %int128_15621 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18011 = torch.aten.view %17599, %18010 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18011, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15622 = torch.constant.int 4
%int16_15623 = torch.constant.int 16
%int1_15624 = torch.constant.int 1
%int128_15625 = torch.constant.int 128
%18012 = torch.prim.ListConstruct %int4_15622, %3095, %int16_15623, %int1_15624, %int128_15625 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18013 = torch.aten.view %17601, %18012 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18013, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15626 = torch.constant.int 4
%int16_15627 = torch.constant.int 16
%int1_15628 = torch.constant.int 1
%int128_15629 = torch.constant.int 128
%18014 = torch.prim.ListConstruct %int4_15626, %3095, %int16_15627, %int1_15628, %int128_15629 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18015 = torch.aten.view %17603, %18014 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %18015, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_15630 = torch.constant.int 4
%18016 = torch.aten.mul.int %int4_15630, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15631 = torch.constant.int 16
%int1_15632 = torch.constant.int 1
%int128_15633 = torch.constant.int 128
%18017 = torch.prim.ListConstruct %18016, %int16_15631, %int1_15632, %int128_15633 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18018 = torch.aten.view %18001, %18017 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18018, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15634 = torch.constant.int 4
%18019 = torch.aten.mul.int %int4_15634, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15635 = torch.constant.int 16
%int1_15636 = torch.constant.int 1
%int128_15637 = torch.constant.int 128
%18020 = torch.prim.ListConstruct %18019, %int16_15635, %int1_15636, %int128_15637 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18021 = torch.aten.view %18003, %18020 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18021, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15638 = torch.constant.int 4
%18022 = torch.aten.mul.int %int4_15638, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15639 = torch.constant.int 16
%int1_15640 = torch.constant.int 1
%int128_15641 = torch.constant.int 128
%18023 = torch.prim.ListConstruct %18022, %int16_15639, %int1_15640, %int128_15641 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18024 = torch.aten.view %18005, %18023 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18024, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15642 = torch.constant.int 4
%18025 = torch.aten.mul.int %int4_15642, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15643 = torch.constant.int 16
%int1_15644 = torch.constant.int 1
%int128_15645 = torch.constant.int 128
%18026 = torch.prim.ListConstruct %18025, %int16_15643, %int1_15644, %int128_15645 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18027 = torch.aten.view %18007, %18026 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18027, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15646 = torch.constant.int 4
%18028 = torch.aten.mul.int %int4_15646, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15647 = torch.constant.int 16
%int1_15648 = torch.constant.int 1
%int128_15649 = torch.constant.int 128
%18029 = torch.prim.ListConstruct %18028, %int16_15647, %int1_15648, %int128_15649 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18030 = torch.aten.view %18009, %18029 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18030, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15650 = torch.constant.int 4
%18031 = torch.aten.mul.int %int4_15650, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15651 = torch.constant.int 16
%int1_15652 = torch.constant.int 1
%int128_15653 = torch.constant.int 128
%18032 = torch.prim.ListConstruct %18031, %int16_15651, %int1_15652, %int128_15653 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18033 = torch.aten.view %18011, %18032 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18033, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15654 = torch.constant.int 4
%18034 = torch.aten.mul.int %int4_15654, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15655 = torch.constant.int 16
%int1_15656 = torch.constant.int 1
%int128_15657 = torch.constant.int 128
%18035 = torch.prim.ListConstruct %18034, %int16_15655, %int1_15656, %int128_15657 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18036 = torch.aten.view %18013, %18035 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18036, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_15658 = torch.constant.int 4
%18037 = torch.aten.mul.int %int4_15658, %3095 : !torch.int, !torch.int -> !torch.int
%int16_15659 = torch.constant.int 16
%int1_15660 = torch.constant.int 1
%int128_15661 = torch.constant.int 128
%18038 = torch.prim.ListConstruct %18037, %int16_15659, %int1_15660, %int128_15661 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18039 = torch.aten.view %18015, %18038 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18039, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_15662 = torch.constant.int 1
%int1_15663 = torch.constant.int 1
%18040 = torch.aten.add.Scalar %17928, %int1_15662, %int1_15663 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18040, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15664 = torch.constant.int 1
%int1_15665 = torch.constant.int 1
%18041 = torch.aten.add.Scalar %17929, %int1_15664, %int1_15665 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18041, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15666 = torch.constant.int 1
%int1_15667 = torch.constant.int 1
%18042 = torch.aten.add.Scalar %17930, %int1_15666, %int1_15667 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18042, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15668 = torch.constant.int 1
%int1_15669 = torch.constant.int 1
%18043 = torch.aten.add.Scalar %17931, %int1_15668, %int1_15669 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18043, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15670 = torch.constant.int 1
%int1_15671 = torch.constant.int 1
%18044 = torch.aten.add.Scalar %17932, %int1_15670, %int1_15671 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18044, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15672 = torch.constant.int 1
%int1_15673 = torch.constant.int 1
%18045 = torch.aten.add.Scalar %17933, %int1_15672, %int1_15673 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18045, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15674 = torch.constant.int 1
%int1_15675 = torch.constant.int 1
%18046 = torch.aten.add.Scalar %17934, %int1_15674, %int1_15675 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18046, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_15676 = torch.constant.int 1
%int1_15677 = torch.constant.int 1
%18047 = torch.aten.add.Scalar %17935, %int1_15676, %int1_15677 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %18047, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_15678 = torch.constant.int 4
%18048 = torch.aten.mul.int %int4_15678, %3095 : !torch.int, !torch.int -> !torch.int
%18049 = torch.prim.ListConstruct %18048 : (!torch.int) -> !torch.list<int>
%18050 = torch.aten.view %18040, %18049 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18050, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15679 = torch.constant.int 4
%18051 = torch.aten.mul.int %int4_15679, %3095 : !torch.int, !torch.int -> !torch.int
%18052 = torch.prim.ListConstruct %18051 : (!torch.int) -> !torch.list<int>
%18053 = torch.aten.view %18041, %18052 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18053, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15680 = torch.constant.int 4
%18054 = torch.aten.mul.int %int4_15680, %3095 : !torch.int, !torch.int -> !torch.int
%18055 = torch.prim.ListConstruct %18054 : (!torch.int) -> !torch.list<int>
%18056 = torch.aten.view %18042, %18055 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18056, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15681 = torch.constant.int 4
%18057 = torch.aten.mul.int %int4_15681, %3095 : !torch.int, !torch.int -> !torch.int
%18058 = torch.prim.ListConstruct %18057 : (!torch.int) -> !torch.list<int>
%18059 = torch.aten.view %18043, %18058 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18059, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15682 = torch.constant.int 4
%18060 = torch.aten.mul.int %int4_15682, %3095 : !torch.int, !torch.int -> !torch.int
%18061 = torch.prim.ListConstruct %18060 : (!torch.int) -> !torch.list<int>
%18062 = torch.aten.view %18044, %18061 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18062, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15683 = torch.constant.int 4
%18063 = torch.aten.mul.int %int4_15683, %3095 : !torch.int, !torch.int -> !torch.int
%18064 = torch.prim.ListConstruct %18063 : (!torch.int) -> !torch.list<int>
%18065 = torch.aten.view %18045, %18064 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18065, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15684 = torch.constant.int 4
%18066 = torch.aten.mul.int %int4_15684, %3095 : !torch.int, !torch.int -> !torch.int
%18067 = torch.prim.ListConstruct %18066 : (!torch.int) -> !torch.list<int>
%18068 = torch.aten.view %18046, %18067 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18068, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_15685 = torch.constant.int 4
%18069 = torch.aten.mul.int %int4_15685, %3095 : !torch.int, !torch.int -> !torch.int
%18070 = torch.prim.ListConstruct %18069 : (!torch.int) -> !torch.list<int>
%18071 = torch.aten.view %18047, %18070 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18071, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%18072 = torch.prim.ListConstruct %17978, %18050 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15686 = torch.constant.int 0
%18073 = torch.aten.cat %18072, %int0_15686 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18073, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18074 = torch.prim.ListConstruct %17981, %18053 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15687 = torch.constant.int 0
%18075 = torch.aten.cat %18074, %int0_15687 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18075, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18076 = torch.prim.ListConstruct %17984, %18056 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15688 = torch.constant.int 0
%18077 = torch.aten.cat %18076, %int0_15688 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18077, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18078 = torch.prim.ListConstruct %17987, %18059 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15689 = torch.constant.int 0
%18079 = torch.aten.cat %18078, %int0_15689 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18079, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18080 = torch.prim.ListConstruct %17990, %18062 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15690 = torch.constant.int 0
%18081 = torch.aten.cat %18080, %int0_15690 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18081, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18082 = torch.prim.ListConstruct %17993, %18065 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15691 = torch.constant.int 0
%18083 = torch.aten.cat %18082, %int0_15691 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18083, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18084 = torch.prim.ListConstruct %17996, %18068 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15692 = torch.constant.int 0
%18085 = torch.aten.cat %18084, %int0_15692 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18085, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18086 = torch.prim.ListConstruct %17999, %18071 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_15693 = torch.constant.int 0
%18087 = torch.aten.cat %18086, %int0_15693 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %18087, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%18088 = torch.prim.ListConstruct %17954, %18018 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15694 = torch.constant.int 0
%18089 = torch.aten.cat %18088, %int0_15694 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18089, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18090 = torch.prim.ListConstruct %17957, %18021 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15695 = torch.constant.int 0
%18091 = torch.aten.cat %18090, %int0_15695 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18091, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18092 = torch.prim.ListConstruct %17960, %18024 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15696 = torch.constant.int 0
%18093 = torch.aten.cat %18092, %int0_15696 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18093, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18094 = torch.prim.ListConstruct %17963, %18027 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15697 = torch.constant.int 0
%18095 = torch.aten.cat %18094, %int0_15697 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18095, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18096 = torch.prim.ListConstruct %17966, %18030 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15698 = torch.constant.int 0
%18097 = torch.aten.cat %18096, %int0_15698 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18097, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18098 = torch.prim.ListConstruct %17969, %18033 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15699 = torch.constant.int 0
%18099 = torch.aten.cat %18098, %int0_15699 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18099, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18100 = torch.prim.ListConstruct %17972, %18036 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15700 = torch.constant.int 0
%18101 = torch.aten.cat %18100, %int0_15700 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18101, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18102 = torch.prim.ListConstruct %17975, %18039 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_15701 = torch.constant.int 0
%18103 = torch.aten.cat %18102, %int0_15701 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18103, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15702 = torch.constant.int 32
%int2_15703 = torch.constant.int 2
%int16_15704 = torch.constant.int 16
%int1_15705 = torch.constant.int 1
%int128_15706 = torch.constant.int 128
%18104 = torch.prim.ListConstruct %3023, %int32_15702, %int2_15703, %int16_15704, %int1_15705, %int128_15706 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18105 = torch.aten.view %16254, %18104 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18105, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15707 = torch.constant.int 32
%18106 = torch.aten.mul.int %3023, %int32_15707 : !torch.int, !torch.int -> !torch.int
%int2_15708 = torch.constant.int 2
%18107 = torch.aten.mul.int %18106, %int2_15708 : !torch.int, !torch.int -> !torch.int
%int16_15709 = torch.constant.int 16
%int1_15710 = torch.constant.int 1
%int128_15711 = torch.constant.int 128
%18108 = torch.prim.ListConstruct %18107, %int16_15709, %int1_15710, %int128_15711 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18109 = torch.aten.view %18105, %18108 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18109, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18110 = torch.prim.ListConstruct %18073 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15712 = torch.constant.bool false
%18111 = torch.aten.index_put %18109, %18110, %18089, %false_15712 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18111, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15713 = torch.constant.int 32
%int2_15714 = torch.constant.int 2
%int16_15715 = torch.constant.int 16
%int1_15716 = torch.constant.int 1
%int128_15717 = torch.constant.int 128
%18112 = torch.prim.ListConstruct %3023, %int32_15713, %int2_15714, %int16_15715, %int1_15716, %int128_15717 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18113 = torch.aten.view %18111, %18112 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18113, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15718 = torch.constant.int 131072
%18114 = torch.prim.ListConstruct %3023, %int131072_15718 : (!torch.int, !torch.int) -> !torch.list<int>
%18115 = torch.aten.view %18113, %18114 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18115, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15719 = torch.constant.int 32
%int2_15720 = torch.constant.int 2
%int16_15721 = torch.constant.int 16
%int1_15722 = torch.constant.int 1
%int128_15723 = torch.constant.int 128
%18116 = torch.prim.ListConstruct %3026, %int32_15719, %int2_15720, %int16_15721, %int1_15722, %int128_15723 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18117 = torch.aten.view %16266, %18116 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18117, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15724 = torch.constant.int 32
%18118 = torch.aten.mul.int %3026, %int32_15724 : !torch.int, !torch.int -> !torch.int
%int2_15725 = torch.constant.int 2
%18119 = torch.aten.mul.int %18118, %int2_15725 : !torch.int, !torch.int -> !torch.int
%int16_15726 = torch.constant.int 16
%int1_15727 = torch.constant.int 1
%int128_15728 = torch.constant.int 128
%18120 = torch.prim.ListConstruct %18119, %int16_15726, %int1_15727, %int128_15728 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18121 = torch.aten.view %18117, %18120 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18121, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18122 = torch.prim.ListConstruct %18075 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15729 = torch.constant.bool false
%18123 = torch.aten.index_put %18121, %18122, %18091, %false_15729 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18123, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15730 = torch.constant.int 32
%int2_15731 = torch.constant.int 2
%int16_15732 = torch.constant.int 16
%int1_15733 = torch.constant.int 1
%int128_15734 = torch.constant.int 128
%18124 = torch.prim.ListConstruct %3026, %int32_15730, %int2_15731, %int16_15732, %int1_15733, %int128_15734 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18125 = torch.aten.view %18123, %18124 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18125, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15735 = torch.constant.int 131072
%18126 = torch.prim.ListConstruct %3026, %int131072_15735 : (!torch.int, !torch.int) -> !torch.list<int>
%18127 = torch.aten.view %18125, %18126 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18127, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15736 = torch.constant.int 32
%int2_15737 = torch.constant.int 2
%int16_15738 = torch.constant.int 16
%int1_15739 = torch.constant.int 1
%int128_15740 = torch.constant.int 128
%18128 = torch.prim.ListConstruct %3029, %int32_15736, %int2_15737, %int16_15738, %int1_15739, %int128_15740 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18129 = torch.aten.view %16278, %18128 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18129, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15741 = torch.constant.int 32
%18130 = torch.aten.mul.int %3029, %int32_15741 : !torch.int, !torch.int -> !torch.int
%int2_15742 = torch.constant.int 2
%18131 = torch.aten.mul.int %18130, %int2_15742 : !torch.int, !torch.int -> !torch.int
%int16_15743 = torch.constant.int 16
%int1_15744 = torch.constant.int 1
%int128_15745 = torch.constant.int 128
%18132 = torch.prim.ListConstruct %18131, %int16_15743, %int1_15744, %int128_15745 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18133 = torch.aten.view %18129, %18132 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18133, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18134 = torch.prim.ListConstruct %18077 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15746 = torch.constant.bool false
%18135 = torch.aten.index_put %18133, %18134, %18093, %false_15746 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18135, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15747 = torch.constant.int 32
%int2_15748 = torch.constant.int 2
%int16_15749 = torch.constant.int 16
%int1_15750 = torch.constant.int 1
%int128_15751 = torch.constant.int 128
%18136 = torch.prim.ListConstruct %3029, %int32_15747, %int2_15748, %int16_15749, %int1_15750, %int128_15751 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18137 = torch.aten.view %18135, %18136 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18137, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15752 = torch.constant.int 131072
%18138 = torch.prim.ListConstruct %3029, %int131072_15752 : (!torch.int, !torch.int) -> !torch.list<int>
%18139 = torch.aten.view %18137, %18138 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18139, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15753 = torch.constant.int 32
%int2_15754 = torch.constant.int 2
%int16_15755 = torch.constant.int 16
%int1_15756 = torch.constant.int 1
%int128_15757 = torch.constant.int 128
%18140 = torch.prim.ListConstruct %3032, %int32_15753, %int2_15754, %int16_15755, %int1_15756, %int128_15757 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18141 = torch.aten.view %16290, %18140 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18141, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15758 = torch.constant.int 32
%18142 = torch.aten.mul.int %3032, %int32_15758 : !torch.int, !torch.int -> !torch.int
%int2_15759 = torch.constant.int 2
%18143 = torch.aten.mul.int %18142, %int2_15759 : !torch.int, !torch.int -> !torch.int
%int16_15760 = torch.constant.int 16
%int1_15761 = torch.constant.int 1
%int128_15762 = torch.constant.int 128
%18144 = torch.prim.ListConstruct %18143, %int16_15760, %int1_15761, %int128_15762 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18145 = torch.aten.view %18141, %18144 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18145, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18146 = torch.prim.ListConstruct %18079 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15763 = torch.constant.bool false
%18147 = torch.aten.index_put %18145, %18146, %18095, %false_15763 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18147, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15764 = torch.constant.int 32
%int2_15765 = torch.constant.int 2
%int16_15766 = torch.constant.int 16
%int1_15767 = torch.constant.int 1
%int128_15768 = torch.constant.int 128
%18148 = torch.prim.ListConstruct %3032, %int32_15764, %int2_15765, %int16_15766, %int1_15767, %int128_15768 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18149 = torch.aten.view %18147, %18148 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18149, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15769 = torch.constant.int 131072
%18150 = torch.prim.ListConstruct %3032, %int131072_15769 : (!torch.int, !torch.int) -> !torch.list<int>
%18151 = torch.aten.view %18149, %18150 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18151, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15770 = torch.constant.int 32
%int2_15771 = torch.constant.int 2
%int16_15772 = torch.constant.int 16
%int1_15773 = torch.constant.int 1
%int128_15774 = torch.constant.int 128
%18152 = torch.prim.ListConstruct %3035, %int32_15770, %int2_15771, %int16_15772, %int1_15773, %int128_15774 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18153 = torch.aten.view %16302, %18152 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18153, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15775 = torch.constant.int 32
%18154 = torch.aten.mul.int %3035, %int32_15775 : !torch.int, !torch.int -> !torch.int
%int2_15776 = torch.constant.int 2
%18155 = torch.aten.mul.int %18154, %int2_15776 : !torch.int, !torch.int -> !torch.int
%int16_15777 = torch.constant.int 16
%int1_15778 = torch.constant.int 1
%int128_15779 = torch.constant.int 128
%18156 = torch.prim.ListConstruct %18155, %int16_15777, %int1_15778, %int128_15779 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18157 = torch.aten.view %18153, %18156 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18157, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18158 = torch.prim.ListConstruct %18081 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15780 = torch.constant.bool false
%18159 = torch.aten.index_put %18157, %18158, %18097, %false_15780 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18159, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15781 = torch.constant.int 32
%int2_15782 = torch.constant.int 2
%int16_15783 = torch.constant.int 16
%int1_15784 = torch.constant.int 1
%int128_15785 = torch.constant.int 128
%18160 = torch.prim.ListConstruct %3035, %int32_15781, %int2_15782, %int16_15783, %int1_15784, %int128_15785 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18161 = torch.aten.view %18159, %18160 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18161, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15786 = torch.constant.int 131072
%18162 = torch.prim.ListConstruct %3035, %int131072_15786 : (!torch.int, !torch.int) -> !torch.list<int>
%18163 = torch.aten.view %18161, %18162 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18163, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15787 = torch.constant.int 32
%int2_15788 = torch.constant.int 2
%int16_15789 = torch.constant.int 16
%int1_15790 = torch.constant.int 1
%int128_15791 = torch.constant.int 128
%18164 = torch.prim.ListConstruct %3038, %int32_15787, %int2_15788, %int16_15789, %int1_15790, %int128_15791 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18165 = torch.aten.view %16314, %18164 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18165, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15792 = torch.constant.int 32
%18166 = torch.aten.mul.int %3038, %int32_15792 : !torch.int, !torch.int -> !torch.int
%int2_15793 = torch.constant.int 2
%18167 = torch.aten.mul.int %18166, %int2_15793 : !torch.int, !torch.int -> !torch.int
%int16_15794 = torch.constant.int 16
%int1_15795 = torch.constant.int 1
%int128_15796 = torch.constant.int 128
%18168 = torch.prim.ListConstruct %18167, %int16_15794, %int1_15795, %int128_15796 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18169 = torch.aten.view %18165, %18168 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18169, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18170 = torch.prim.ListConstruct %18083 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15797 = torch.constant.bool false
%18171 = torch.aten.index_put %18169, %18170, %18099, %false_15797 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18171, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15798 = torch.constant.int 32
%int2_15799 = torch.constant.int 2
%int16_15800 = torch.constant.int 16
%int1_15801 = torch.constant.int 1
%int128_15802 = torch.constant.int 128
%18172 = torch.prim.ListConstruct %3038, %int32_15798, %int2_15799, %int16_15800, %int1_15801, %int128_15802 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18173 = torch.aten.view %18171, %18172 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18173, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15803 = torch.constant.int 131072
%18174 = torch.prim.ListConstruct %3038, %int131072_15803 : (!torch.int, !torch.int) -> !torch.list<int>
%18175 = torch.aten.view %18173, %18174 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18175, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15804 = torch.constant.int 32
%int2_15805 = torch.constant.int 2
%int16_15806 = torch.constant.int 16
%int1_15807 = torch.constant.int 1
%int128_15808 = torch.constant.int 128
%18176 = torch.prim.ListConstruct %3041, %int32_15804, %int2_15805, %int16_15806, %int1_15807, %int128_15808 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18177 = torch.aten.view %16326, %18176 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18177, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15809 = torch.constant.int 32
%18178 = torch.aten.mul.int %3041, %int32_15809 : !torch.int, !torch.int -> !torch.int
%int2_15810 = torch.constant.int 2
%18179 = torch.aten.mul.int %18178, %int2_15810 : !torch.int, !torch.int -> !torch.int
%int16_15811 = torch.constant.int 16
%int1_15812 = torch.constant.int 1
%int128_15813 = torch.constant.int 128
%18180 = torch.prim.ListConstruct %18179, %int16_15811, %int1_15812, %int128_15813 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18181 = torch.aten.view %18177, %18180 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18181, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18182 = torch.prim.ListConstruct %18085 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15814 = torch.constant.bool false
%18183 = torch.aten.index_put %18181, %18182, %18101, %false_15814 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18183, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15815 = torch.constant.int 32
%int2_15816 = torch.constant.int 2
%int16_15817 = torch.constant.int 16
%int1_15818 = torch.constant.int 1
%int128_15819 = torch.constant.int 128
%18184 = torch.prim.ListConstruct %3041, %int32_15815, %int2_15816, %int16_15817, %int1_15818, %int128_15819 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18185 = torch.aten.view %18183, %18184 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18185, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15820 = torch.constant.int 131072
%18186 = torch.prim.ListConstruct %3041, %int131072_15820 : (!torch.int, !torch.int) -> !torch.list<int>
%18187 = torch.aten.view %18185, %18186 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18187, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_15821 = torch.constant.int 32
%int2_15822 = torch.constant.int 2
%int16_15823 = torch.constant.int 16
%int1_15824 = torch.constant.int 1
%int128_15825 = torch.constant.int 128
%18188 = torch.prim.ListConstruct %3044, %int32_15821, %int2_15822, %int16_15823, %int1_15824, %int128_15825 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18189 = torch.aten.view %16338, %18188 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18189, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_15826 = torch.constant.int 32
%18190 = torch.aten.mul.int %3044, %int32_15826 : !torch.int, !torch.int -> !torch.int
%int2_15827 = torch.constant.int 2
%18191 = torch.aten.mul.int %18190, %int2_15827 : !torch.int, !torch.int -> !torch.int
%int16_15828 = torch.constant.int 16
%int1_15829 = torch.constant.int 1
%int128_15830 = torch.constant.int 128
%18192 = torch.prim.ListConstruct %18191, %int16_15828, %int1_15829, %int128_15830 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18193 = torch.aten.view %18189, %18192 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18193, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%18194 = torch.prim.ListConstruct %18087 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_15831 = torch.constant.bool false
%18195 = torch.aten.index_put %18193, %18194, %18103, %false_15831 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %18195, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_15832 = torch.constant.int 32
%int2_15833 = torch.constant.int 2
%int16_15834 = torch.constant.int 16
%int1_15835 = torch.constant.int 1
%int128_15836 = torch.constant.int 128
%18196 = torch.prim.ListConstruct %3044, %int32_15832, %int2_15833, %int16_15834, %int1_15835, %int128_15836 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18197 = torch.aten.view %18195, %18196 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %18197, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_15837 = torch.constant.int 131072
%18198 = torch.prim.ListConstruct %3044, %int131072_15837 : (!torch.int, !torch.int) -> !torch.list<int>
%18199 = torch.aten.view %18197, %18198 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %18199, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_15838 = torch.constant.int -2
%18200 = torch.aten.unsqueeze %17814, %int-2_15838 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15839 = torch.constant.int -2
%18201 = torch.aten.unsqueeze %17829, %int-2_15839 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15840 = torch.constant.int -2
%18202 = torch.aten.unsqueeze %17844, %int-2_15840 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15841 = torch.constant.int -2
%18203 = torch.aten.unsqueeze %17859, %int-2_15841 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15842 = torch.constant.int -2
%18204 = torch.aten.unsqueeze %17874, %int-2_15842 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15843 = torch.constant.int -2
%18205 = torch.aten.unsqueeze %17889, %int-2_15843 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15844 = torch.constant.int -2
%18206 = torch.aten.unsqueeze %17904, %int-2_15844 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15845 = torch.constant.int -2
%18207 = torch.aten.unsqueeze %17919, %int-2_15845 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_15846 = torch.constant.int 4
%int1_15847 = torch.constant.int 1
%int4_15848 = torch.constant.int 4
%int128_15849 = torch.constant.int 128
%18208 = torch.prim.ListConstruct %int4_15846, %17800, %int1_15847, %int4_15848, %int128_15849 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15850 = torch.constant.bool false
%18209 = torch.aten.expand %18200, %18208, %false_15850 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15851 = torch.constant.int 4
%int1_15852 = torch.constant.int 1
%int4_15853 = torch.constant.int 4
%int128_15854 = torch.constant.int 128
%18210 = torch.prim.ListConstruct %int4_15851, %17800, %int1_15852, %int4_15853, %int128_15854 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15855 = torch.constant.bool false
%18211 = torch.aten.expand %18201, %18210, %false_15855 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15856 = torch.constant.int 4
%int1_15857 = torch.constant.int 1
%int4_15858 = torch.constant.int 4
%int128_15859 = torch.constant.int 128
%18212 = torch.prim.ListConstruct %int4_15856, %17800, %int1_15857, %int4_15858, %int128_15859 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15860 = torch.constant.bool false
%18213 = torch.aten.expand %18202, %18212, %false_15860 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15861 = torch.constant.int 4
%int1_15862 = torch.constant.int 1
%int4_15863 = torch.constant.int 4
%int128_15864 = torch.constant.int 128
%18214 = torch.prim.ListConstruct %int4_15861, %17800, %int1_15862, %int4_15863, %int128_15864 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15865 = torch.constant.bool false
%18215 = torch.aten.expand %18203, %18214, %false_15865 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15866 = torch.constant.int 4
%int1_15867 = torch.constant.int 1
%int4_15868 = torch.constant.int 4
%int128_15869 = torch.constant.int 128
%18216 = torch.prim.ListConstruct %int4_15866, %17800, %int1_15867, %int4_15868, %int128_15869 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15870 = torch.constant.bool false
%18217 = torch.aten.expand %18204, %18216, %false_15870 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15871 = torch.constant.int 4
%int1_15872 = torch.constant.int 1
%int4_15873 = torch.constant.int 4
%int128_15874 = torch.constant.int 128
%18218 = torch.prim.ListConstruct %int4_15871, %17800, %int1_15872, %int4_15873, %int128_15874 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15875 = torch.constant.bool false
%18219 = torch.aten.expand %18205, %18218, %false_15875 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15876 = torch.constant.int 4
%int1_15877 = torch.constant.int 1
%int4_15878 = torch.constant.int 4
%int128_15879 = torch.constant.int 128
%18220 = torch.prim.ListConstruct %int4_15876, %17800, %int1_15877, %int4_15878, %int128_15879 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15880 = torch.constant.bool false
%18221 = torch.aten.expand %18206, %18220, %false_15880 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15881 = torch.constant.int 4
%int1_15882 = torch.constant.int 1
%int4_15883 = torch.constant.int 4
%int128_15884 = torch.constant.int 128
%18222 = torch.prim.ListConstruct %int4_15881, %17800, %int1_15882, %int4_15883, %int128_15884 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15885 = torch.constant.bool false
%18223 = torch.aten.expand %18207, %18222, %false_15885 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15886 = torch.constant.int 4
%int4_15887 = torch.constant.int 4
%int128_15888 = torch.constant.int 128
%18224 = torch.prim.ListConstruct %int4_15886, %17800, %int4_15887, %int128_15888 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18225 = torch.aten.view %18209, %18224 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15889 = torch.constant.int 4
%int4_15890 = torch.constant.int 4
%int128_15891 = torch.constant.int 128
%18226 = torch.prim.ListConstruct %int4_15889, %17800, %int4_15890, %int128_15891 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18227 = torch.aten.view %18211, %18226 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15892 = torch.constant.int 4
%int4_15893 = torch.constant.int 4
%int128_15894 = torch.constant.int 128
%18228 = torch.prim.ListConstruct %int4_15892, %17800, %int4_15893, %int128_15894 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18229 = torch.aten.view %18213, %18228 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15895 = torch.constant.int 4
%int4_15896 = torch.constant.int 4
%int128_15897 = torch.constant.int 128
%18230 = torch.prim.ListConstruct %int4_15895, %17800, %int4_15896, %int128_15897 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18231 = torch.aten.view %18215, %18230 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15898 = torch.constant.int 4
%int4_15899 = torch.constant.int 4
%int128_15900 = torch.constant.int 128
%18232 = torch.prim.ListConstruct %int4_15898, %17800, %int4_15899, %int128_15900 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18233 = torch.aten.view %18217, %18232 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15901 = torch.constant.int 4
%int4_15902 = torch.constant.int 4
%int128_15903 = torch.constant.int 128
%18234 = torch.prim.ListConstruct %int4_15901, %17800, %int4_15902, %int128_15903 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18235 = torch.aten.view %18219, %18234 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15904 = torch.constant.int 4
%int4_15905 = torch.constant.int 4
%int128_15906 = torch.constant.int 128
%18236 = torch.prim.ListConstruct %int4_15904, %17800, %int4_15905, %int128_15906 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18237 = torch.aten.view %18221, %18236 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15907 = torch.constant.int 4
%int4_15908 = torch.constant.int 4
%int128_15909 = torch.constant.int 128
%18238 = torch.prim.ListConstruct %int4_15907, %17800, %int4_15908, %int128_15909 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18239 = torch.aten.view %18223, %18238 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_15910 = torch.constant.int -2
%18240 = torch.aten.unsqueeze %17589, %int-2_15910 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15911 = torch.constant.int -2
%18241 = torch.aten.unsqueeze %17591, %int-2_15911 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15912 = torch.constant.int -2
%18242 = torch.aten.unsqueeze %17593, %int-2_15912 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15913 = torch.constant.int -2
%18243 = torch.aten.unsqueeze %17595, %int-2_15913 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15914 = torch.constant.int -2
%18244 = torch.aten.unsqueeze %17597, %int-2_15914 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15915 = torch.constant.int -2
%18245 = torch.aten.unsqueeze %17599, %int-2_15915 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15916 = torch.constant.int -2
%18246 = torch.aten.unsqueeze %17601, %int-2_15916 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_15917 = torch.constant.int -2
%18247 = torch.aten.unsqueeze %17603, %int-2_15917 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %18247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_15918 = torch.constant.int 1
%18248 = torch.aten.size.int %17513, %int1_15918 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_15919 = torch.constant.int 4
%int1_15920 = torch.constant.int 1
%int4_15921 = torch.constant.int 4
%int128_15922 = torch.constant.int 128
%18249 = torch.prim.ListConstruct %int4_15919, %18248, %int1_15920, %int4_15921, %int128_15922 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15923 = torch.constant.bool false
%18250 = torch.aten.expand %18240, %18249, %false_15923 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15924 = torch.constant.int 4
%int1_15925 = torch.constant.int 1
%int4_15926 = torch.constant.int 4
%int128_15927 = torch.constant.int 128
%18251 = torch.prim.ListConstruct %int4_15924, %18248, %int1_15925, %int4_15926, %int128_15927 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15928 = torch.constant.bool false
%18252 = torch.aten.expand %18241, %18251, %false_15928 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15929 = torch.constant.int 4
%int1_15930 = torch.constant.int 1
%int4_15931 = torch.constant.int 4
%int128_15932 = torch.constant.int 128
%18253 = torch.prim.ListConstruct %int4_15929, %18248, %int1_15930, %int4_15931, %int128_15932 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15933 = torch.constant.bool false
%18254 = torch.aten.expand %18242, %18253, %false_15933 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15934 = torch.constant.int 4
%int1_15935 = torch.constant.int 1
%int4_15936 = torch.constant.int 4
%int128_15937 = torch.constant.int 128
%18255 = torch.prim.ListConstruct %int4_15934, %18248, %int1_15935, %int4_15936, %int128_15937 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15938 = torch.constant.bool false
%18256 = torch.aten.expand %18243, %18255, %false_15938 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15939 = torch.constant.int 4
%int1_15940 = torch.constant.int 1
%int4_15941 = torch.constant.int 4
%int128_15942 = torch.constant.int 128
%18257 = torch.prim.ListConstruct %int4_15939, %18248, %int1_15940, %int4_15941, %int128_15942 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15943 = torch.constant.bool false
%18258 = torch.aten.expand %18244, %18257, %false_15943 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15944 = torch.constant.int 4
%int1_15945 = torch.constant.int 1
%int4_15946 = torch.constant.int 4
%int128_15947 = torch.constant.int 128
%18259 = torch.prim.ListConstruct %int4_15944, %18248, %int1_15945, %int4_15946, %int128_15947 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15948 = torch.constant.bool false
%18260 = torch.aten.expand %18245, %18259, %false_15948 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15949 = torch.constant.int 4
%int1_15950 = torch.constant.int 1
%int4_15951 = torch.constant.int 4
%int128_15952 = torch.constant.int 128
%18261 = torch.prim.ListConstruct %int4_15949, %18248, %int1_15950, %int4_15951, %int128_15952 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15953 = torch.constant.bool false
%18262 = torch.aten.expand %18246, %18261, %false_15953 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15954 = torch.constant.int 4
%int1_15955 = torch.constant.int 1
%int4_15956 = torch.constant.int 4
%int128_15957 = torch.constant.int 128
%18263 = torch.prim.ListConstruct %int4_15954, %18248, %int1_15955, %int4_15956, %int128_15957 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_15958 = torch.constant.bool false
%18264 = torch.aten.expand %18247, %18263, %false_15958 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %18264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_15959 = torch.constant.int 4
%int4_15960 = torch.constant.int 4
%int128_15961 = torch.constant.int 128
%18265 = torch.prim.ListConstruct %int4_15959, %18248, %int4_15960, %int128_15961 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18266 = torch.aten.view %18250, %18265 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15962 = torch.constant.int 4
%int4_15963 = torch.constant.int 4
%int128_15964 = torch.constant.int 128
%18267 = torch.prim.ListConstruct %int4_15962, %18248, %int4_15963, %int128_15964 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18268 = torch.aten.view %18252, %18267 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15965 = torch.constant.int 4
%int4_15966 = torch.constant.int 4
%int128_15967 = torch.constant.int 128
%18269 = torch.prim.ListConstruct %int4_15965, %18248, %int4_15966, %int128_15967 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18270 = torch.aten.view %18254, %18269 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15968 = torch.constant.int 4
%int4_15969 = torch.constant.int 4
%int128_15970 = torch.constant.int 128
%18271 = torch.prim.ListConstruct %int4_15968, %18248, %int4_15969, %int128_15970 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18272 = torch.aten.view %18256, %18271 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15971 = torch.constant.int 4
%int4_15972 = torch.constant.int 4
%int128_15973 = torch.constant.int 128
%18273 = torch.prim.ListConstruct %int4_15971, %18248, %int4_15972, %int128_15973 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18274 = torch.aten.view %18258, %18273 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15974 = torch.constant.int 4
%int4_15975 = torch.constant.int 4
%int128_15976 = torch.constant.int 128
%18275 = torch.prim.ListConstruct %int4_15974, %18248, %int4_15975, %int128_15976 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18276 = torch.aten.view %18260, %18275 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15977 = torch.constant.int 4
%int4_15978 = torch.constant.int 4
%int128_15979 = torch.constant.int 128
%18277 = torch.prim.ListConstruct %int4_15977, %18248, %int4_15978, %int128_15979 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18278 = torch.aten.view %18262, %18277 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_15980 = torch.constant.int 4
%int4_15981 = torch.constant.int 4
%int128_15982 = torch.constant.int 128
%18279 = torch.prim.ListConstruct %int4_15980, %18248, %int4_15981, %int128_15982 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18280 = torch.aten.view %18264, %18279 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_15983 = torch.constant.int 1
%int2_15984 = torch.constant.int 2
%18281 = torch.aten.transpose.int %17656, %int1_15983, %int2_15984 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18281, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15985 = torch.constant.int 1
%int2_15986 = torch.constant.int 2
%18282 = torch.aten.transpose.int %17671, %int1_15985, %int2_15986 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18282, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15987 = torch.constant.int 1
%int2_15988 = torch.constant.int 2
%18283 = torch.aten.transpose.int %17686, %int1_15987, %int2_15988 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18283, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15989 = torch.constant.int 1
%int2_15990 = torch.constant.int 2
%18284 = torch.aten.transpose.int %17701, %int1_15989, %int2_15990 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18284, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15991 = torch.constant.int 1
%int2_15992 = torch.constant.int 2
%18285 = torch.aten.transpose.int %17716, %int1_15991, %int2_15992 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18285, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15993 = torch.constant.int 1
%int2_15994 = torch.constant.int 2
%18286 = torch.aten.transpose.int %17731, %int1_15993, %int2_15994 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18286, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15995 = torch.constant.int 1
%int2_15996 = torch.constant.int 2
%18287 = torch.aten.transpose.int %17746, %int1_15995, %int2_15996 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18287, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15997 = torch.constant.int 1
%int2_15998 = torch.constant.int 2
%18288 = torch.aten.transpose.int %17761, %int1_15997, %int2_15998 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18288, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_15999 = torch.constant.int 1
%int2_16000 = torch.constant.int 2
%18289 = torch.aten.transpose.int %18225, %int1_15999, %int2_16000 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18289, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16001 = torch.constant.int 1
%int2_16002 = torch.constant.int 2
%18290 = torch.aten.transpose.int %18227, %int1_16001, %int2_16002 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18290, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16003 = torch.constant.int 1
%int2_16004 = torch.constant.int 2
%18291 = torch.aten.transpose.int %18229, %int1_16003, %int2_16004 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18291, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16005 = torch.constant.int 1
%int2_16006 = torch.constant.int 2
%18292 = torch.aten.transpose.int %18231, %int1_16005, %int2_16006 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18292, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16007 = torch.constant.int 1
%int2_16008 = torch.constant.int 2
%18293 = torch.aten.transpose.int %18233, %int1_16007, %int2_16008 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18293, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16009 = torch.constant.int 1
%int2_16010 = torch.constant.int 2
%18294 = torch.aten.transpose.int %18235, %int1_16009, %int2_16010 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18294, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16011 = torch.constant.int 1
%int2_16012 = torch.constant.int 2
%18295 = torch.aten.transpose.int %18237, %int1_16011, %int2_16012 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18295, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16013 = torch.constant.int 1
%int2_16014 = torch.constant.int 2
%18296 = torch.aten.transpose.int %18239, %int1_16013, %int2_16014 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18296, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16015 = torch.constant.int 1
%int2_16016 = torch.constant.int 2
%18297 = torch.aten.transpose.int %18266, %int1_16015, %int2_16016 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18297, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16017 = torch.constant.int 1
%int2_16018 = torch.constant.int 2
%18298 = torch.aten.transpose.int %18268, %int1_16017, %int2_16018 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18298, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16019 = torch.constant.int 1
%int2_16020 = torch.constant.int 2
%18299 = torch.aten.transpose.int %18270, %int1_16019, %int2_16020 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18299, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16021 = torch.constant.int 1
%int2_16022 = torch.constant.int 2
%18300 = torch.aten.transpose.int %18272, %int1_16021, %int2_16022 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18300, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16023 = torch.constant.int 1
%int2_16024 = torch.constant.int 2
%18301 = torch.aten.transpose.int %18274, %int1_16023, %int2_16024 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18301, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16025 = torch.constant.int 1
%int2_16026 = torch.constant.int 2
%18302 = torch.aten.transpose.int %18276, %int1_16025, %int2_16026 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18302, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16027 = torch.constant.int 1
%int2_16028 = torch.constant.int 2
%18303 = torch.aten.transpose.int %18278, %int1_16027, %int2_16028 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18303, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16029 = torch.constant.int 1
%int2_16030 = torch.constant.int 2
%18304 = torch.aten.transpose.int %18280, %int1_16029, %int2_16030 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %18304, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16031 = torch.constant.float 0.000000e+00
%true_16032 = torch.constant.bool true
%none_16033 = torch.constant.none
%none_16034 = torch.constant.none
%18305:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18281, %18289, %18297, %float0.000000e00_16031, %true_16032, %none_16033, %none_16034) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18305#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16035 = torch.constant.float 0.000000e+00
%true_16036 = torch.constant.bool true
%none_16037 = torch.constant.none
%none_16038 = torch.constant.none
%18306:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18282, %18290, %18298, %float0.000000e00_16035, %true_16036, %none_16037, %none_16038) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18306#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16039 = torch.constant.float 0.000000e+00
%true_16040 = torch.constant.bool true
%none_16041 = torch.constant.none
%none_16042 = torch.constant.none
%18307:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18283, %18291, %18299, %float0.000000e00_16039, %true_16040, %none_16041, %none_16042) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18307#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16043 = torch.constant.float 0.000000e+00
%true_16044 = torch.constant.bool true
%none_16045 = torch.constant.none
%none_16046 = torch.constant.none
%18308:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18284, %18292, %18300, %float0.000000e00_16043, %true_16044, %none_16045, %none_16046) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18308#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16047 = torch.constant.float 0.000000e+00
%true_16048 = torch.constant.bool true
%none_16049 = torch.constant.none
%none_16050 = torch.constant.none
%18309:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18285, %18293, %18301, %float0.000000e00_16047, %true_16048, %none_16049, %none_16050) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18309#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16051 = torch.constant.float 0.000000e+00
%true_16052 = torch.constant.bool true
%none_16053 = torch.constant.none
%none_16054 = torch.constant.none
%18310:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18286, %18294, %18302, %float0.000000e00_16051, %true_16052, %none_16053, %none_16054) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18310#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16055 = torch.constant.float 0.000000e+00
%true_16056 = torch.constant.bool true
%none_16057 = torch.constant.none
%none_16058 = torch.constant.none
%18311:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18287, %18295, %18303, %float0.000000e00_16055, %true_16056, %none_16057, %none_16058) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18311#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_16059 = torch.constant.float 0.000000e+00
%true_16060 = torch.constant.bool true
%none_16061 = torch.constant.none
%none_16062 = torch.constant.none
%18312:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%18288, %18296, %18304, %float0.000000e00_16059, %true_16060, %none_16061, %none_16062) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %18312#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_16063 = torch.constant.int 1
%int2_16064 = torch.constant.int 2
%18313 = torch.aten.transpose.int %18305#0, %int1_16063, %int2_16064 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16065 = torch.constant.int 1
%int2_16066 = torch.constant.int 2
%18314 = torch.aten.transpose.int %18306#0, %int1_16065, %int2_16066 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16067 = torch.constant.int 1
%int2_16068 = torch.constant.int 2
%18315 = torch.aten.transpose.int %18307#0, %int1_16067, %int2_16068 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16069 = torch.constant.int 1
%int2_16070 = torch.constant.int 2
%18316 = torch.aten.transpose.int %18308#0, %int1_16069, %int2_16070 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16071 = torch.constant.int 1
%int2_16072 = torch.constant.int 2
%18317 = torch.aten.transpose.int %18309#0, %int1_16071, %int2_16072 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16073 = torch.constant.int 1
%int2_16074 = torch.constant.int 2
%18318 = torch.aten.transpose.int %18310#0, %int1_16073, %int2_16074 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16075 = torch.constant.int 1
%int2_16076 = torch.constant.int 2
%18319 = torch.aten.transpose.int %18311#0, %int1_16075, %int2_16076 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_16077 = torch.constant.int 1
%int2_16078 = torch.constant.int 2
%18320 = torch.aten.transpose.int %18312#0, %int1_16077, %int2_16078 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %18320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16079 = torch.constant.int 4
%int512_16080 = torch.constant.int 512
%18321 = torch.prim.ListConstruct %int4_16079, %17642, %int512_16080 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18322 = torch.aten.view %18313, %18321 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16081 = torch.constant.int 4
%int512_16082 = torch.constant.int 512
%18323 = torch.prim.ListConstruct %int4_16081, %17657, %int512_16082 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18324 = torch.aten.view %18314, %18323 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16083 = torch.constant.int 4
%int512_16084 = torch.constant.int 512
%18325 = torch.prim.ListConstruct %int4_16083, %17672, %int512_16084 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18326 = torch.aten.view %18315, %18325 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16085 = torch.constant.int 4
%int512_16086 = torch.constant.int 512
%18327 = torch.prim.ListConstruct %int4_16085, %17687, %int512_16086 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18328 = torch.aten.view %18316, %18327 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16087 = torch.constant.int 4
%int512_16088 = torch.constant.int 512
%18329 = torch.prim.ListConstruct %int4_16087, %17702, %int512_16088 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18330 = torch.aten.view %18317, %18329 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16089 = torch.constant.int 4
%int512_16090 = torch.constant.int 512
%18331 = torch.prim.ListConstruct %int4_16089, %17717, %int512_16090 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18332 = torch.aten.view %18318, %18331 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16091 = torch.constant.int 4
%int512_16092 = torch.constant.int 512
%18333 = torch.prim.ListConstruct %int4_16091, %17732, %int512_16092 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18334 = torch.aten.view %18319, %18333 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16093 = torch.constant.int 4
%int512_16094 = torch.constant.int 512
%18335 = torch.prim.ListConstruct %int4_16093, %17747, %int512_16094 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18336 = torch.aten.view %18320, %18335 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %18336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_16095 = torch.constant.int 1
%int0_16096 = torch.constant.int 0
%18337 = torch.prim.ListConstruct %int1_16095, %int0_16096 : (!torch.int, !torch.int) -> !torch.list<int>
%18338 = torch.aten.permute %616, %18337 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16097 = torch.constant.int 1
%int0_16098 = torch.constant.int 0
%18339 = torch.prim.ListConstruct %int1_16097, %int0_16098 : (!torch.int, !torch.int) -> !torch.list<int>
%18340 = torch.aten.permute %617, %18339 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16099 = torch.constant.int 1
%int0_16100 = torch.constant.int 0
%18341 = torch.prim.ListConstruct %int1_16099, %int0_16100 : (!torch.int, !torch.int) -> !torch.list<int>
%18342 = torch.aten.permute %618, %18341 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16101 = torch.constant.int 1
%int0_16102 = torch.constant.int 0
%18343 = torch.prim.ListConstruct %int1_16101, %int0_16102 : (!torch.int, !torch.int) -> !torch.list<int>
%18344 = torch.aten.permute %619, %18343 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16103 = torch.constant.int 1
%int0_16104 = torch.constant.int 0
%18345 = torch.prim.ListConstruct %int1_16103, %int0_16104 : (!torch.int, !torch.int) -> !torch.list<int>
%18346 = torch.aten.permute %620, %18345 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16105 = torch.constant.int 1
%int0_16106 = torch.constant.int 0
%18347 = torch.prim.ListConstruct %int1_16105, %int0_16106 : (!torch.int, !torch.int) -> !torch.list<int>
%18348 = torch.aten.permute %621, %18347 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16107 = torch.constant.int 1
%int0_16108 = torch.constant.int 0
%18349 = torch.prim.ListConstruct %int1_16107, %int0_16108 : (!torch.int, !torch.int) -> !torch.list<int>
%18350 = torch.aten.permute %622, %18349 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_16109 = torch.constant.int 1
%int0_16110 = torch.constant.int 0
%18351 = torch.prim.ListConstruct %int1_16109, %int0_16110 : (!torch.int, !torch.int) -> !torch.list<int>
%18352 = torch.aten.permute %623, %18351 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_16111 = torch.constant.int 4
%18353 = torch.aten.mul.int %int4_16111, %17642 : !torch.int, !torch.int -> !torch.int
%int512_16112 = torch.constant.int 512
%18354 = torch.prim.ListConstruct %18353, %int512_16112 : (!torch.int, !torch.int) -> !torch.list<int>
%18355 = torch.aten.view %18322, %18354 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18355, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18356 = torch.aten.mm %18355, %18338 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18356, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16113 = torch.constant.int 4
%int4096_16114 = torch.constant.int 4096
%18357 = torch.prim.ListConstruct %int4_16113, %17642, %int4096_16114 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18358 = torch.aten.view %18356, %18357 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16115 = torch.constant.int 4
%18359 = torch.aten.mul.int %int4_16115, %17657 : !torch.int, !torch.int -> !torch.int
%int512_16116 = torch.constant.int 512
%18360 = torch.prim.ListConstruct %18359, %int512_16116 : (!torch.int, !torch.int) -> !torch.list<int>
%18361 = torch.aten.view %18324, %18360 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18361, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18362 = torch.aten.mm %18361, %18340 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18362, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16117 = torch.constant.int 4
%int4096_16118 = torch.constant.int 4096
%18363 = torch.prim.ListConstruct %int4_16117, %17657, %int4096_16118 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18364 = torch.aten.view %18362, %18363 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16119 = torch.constant.int 4
%18365 = torch.aten.mul.int %int4_16119, %17672 : !torch.int, !torch.int -> !torch.int
%int512_16120 = torch.constant.int 512
%18366 = torch.prim.ListConstruct %18365, %int512_16120 : (!torch.int, !torch.int) -> !torch.list<int>
%18367 = torch.aten.view %18326, %18366 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18367, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18368 = torch.aten.mm %18367, %18342 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18368, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16121 = torch.constant.int 4
%int4096_16122 = torch.constant.int 4096
%18369 = torch.prim.ListConstruct %int4_16121, %17672, %int4096_16122 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18370 = torch.aten.view %18368, %18369 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16123 = torch.constant.int 4
%18371 = torch.aten.mul.int %int4_16123, %17687 : !torch.int, !torch.int -> !torch.int
%int512_16124 = torch.constant.int 512
%18372 = torch.prim.ListConstruct %18371, %int512_16124 : (!torch.int, !torch.int) -> !torch.list<int>
%18373 = torch.aten.view %18328, %18372 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18373, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18374 = torch.aten.mm %18373, %18344 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18374, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16125 = torch.constant.int 4
%int4096_16126 = torch.constant.int 4096
%18375 = torch.prim.ListConstruct %int4_16125, %17687, %int4096_16126 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18376 = torch.aten.view %18374, %18375 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16127 = torch.constant.int 4
%18377 = torch.aten.mul.int %int4_16127, %17702 : !torch.int, !torch.int -> !torch.int
%int512_16128 = torch.constant.int 512
%18378 = torch.prim.ListConstruct %18377, %int512_16128 : (!torch.int, !torch.int) -> !torch.list<int>
%18379 = torch.aten.view %18330, %18378 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18379, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18380 = torch.aten.mm %18379, %18346 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18380, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16129 = torch.constant.int 4
%int4096_16130 = torch.constant.int 4096
%18381 = torch.prim.ListConstruct %int4_16129, %17702, %int4096_16130 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18382 = torch.aten.view %18380, %18381 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16131 = torch.constant.int 4
%18383 = torch.aten.mul.int %int4_16131, %17717 : !torch.int, !torch.int -> !torch.int
%int512_16132 = torch.constant.int 512
%18384 = torch.prim.ListConstruct %18383, %int512_16132 : (!torch.int, !torch.int) -> !torch.list<int>
%18385 = torch.aten.view %18332, %18384 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18385, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18386 = torch.aten.mm %18385, %18348 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18386, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16133 = torch.constant.int 4
%int4096_16134 = torch.constant.int 4096
%18387 = torch.prim.ListConstruct %int4_16133, %17717, %int4096_16134 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18388 = torch.aten.view %18386, %18387 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16135 = torch.constant.int 4
%18389 = torch.aten.mul.int %int4_16135, %17732 : !torch.int, !torch.int -> !torch.int
%int512_16136 = torch.constant.int 512
%18390 = torch.prim.ListConstruct %18389, %int512_16136 : (!torch.int, !torch.int) -> !torch.list<int>
%18391 = torch.aten.view %18334, %18390 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18391, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18392 = torch.aten.mm %18391, %18350 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18392, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16137 = torch.constant.int 4
%int4096_16138 = torch.constant.int 4096
%18393 = torch.prim.ListConstruct %int4_16137, %17732, %int4096_16138 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18394 = torch.aten.view %18392, %18393 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_16139 = torch.constant.int 4
%18395 = torch.aten.mul.int %int4_16139, %17747 : !torch.int, !torch.int -> !torch.int
%int512_16140 = torch.constant.int 512
%18396 = torch.prim.ListConstruct %18395, %int512_16140 : (!torch.int, !torch.int) -> !torch.list<int>
%18397 = torch.aten.view %18336, %18396 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %18397, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%18398 = torch.aten.mm %18397, %18352 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18398, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16141 = torch.constant.int 4
%int4096_16142 = torch.constant.int 4096
%18399 = torch.prim.ListConstruct %int4_16141, %17747, %int4096_16142 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18400 = torch.aten.view %18398, %18399 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18401 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16143 = arith.constant 1 : index
%dim_16144 = tensor.dim %18401, %c1_16143 : tensor<4x?x4096xf16>
%18402 = flow.tensor.transfer %18401 : tensor<4x?x4096xf16>{%dim_16144} to #hal.device.promise<@__device_0>
%18403 = torch_c.from_builtin_tensor %18402 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18404 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16145 = arith.constant 1 : index
%dim_16146 = tensor.dim %18404, %c1_16145 : tensor<4x?x4096xf16>
%18405 = flow.tensor.transfer %18404 : tensor<4x?x4096xf16>{%dim_16146} to #hal.device.promise<@__device_0>
%18406 = torch_c.from_builtin_tensor %18405 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18407 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16147 = arith.constant 1 : index
%dim_16148 = tensor.dim %18407, %c1_16147 : tensor<4x?x4096xf16>
%18408 = flow.tensor.transfer %18407 : tensor<4x?x4096xf16>{%dim_16148} to #hal.device.promise<@__device_0>
%18409 = torch_c.from_builtin_tensor %18408 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18410 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16149 = arith.constant 1 : index
%dim_16150 = tensor.dim %18410, %c1_16149 : tensor<4x?x4096xf16>
%18411 = flow.tensor.transfer %18410 : tensor<4x?x4096xf16>{%dim_16150} to #hal.device.promise<@__device_0>
%18412 = torch_c.from_builtin_tensor %18411 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18413 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16151 = arith.constant 1 : index
%dim_16152 = tensor.dim %18413, %c1_16151 : tensor<4x?x4096xf16>
%18414 = flow.tensor.transfer %18413 : tensor<4x?x4096xf16>{%dim_16152} to #hal.device.promise<@__device_0>
%18415 = torch_c.from_builtin_tensor %18414 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18416 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16153 = arith.constant 1 : index
%dim_16154 = tensor.dim %18416, %c1_16153 : tensor<4x?x4096xf16>
%18417 = flow.tensor.transfer %18416 : tensor<4x?x4096xf16>{%dim_16154} to #hal.device.promise<@__device_0>
%18418 = torch_c.from_builtin_tensor %18417 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18419 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16155 = arith.constant 1 : index
%dim_16156 = tensor.dim %18419, %c1_16155 : tensor<4x?x4096xf16>
%18420 = flow.tensor.transfer %18419 : tensor<4x?x4096xf16>{%dim_16156} to #hal.device.promise<@__device_0>
%18421 = torch_c.from_builtin_tensor %18420 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16157 = torch.constant.int 1
%18422 = torch.aten.add.Tensor %18358, %18403, %int1_16157 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16158 = torch.constant.int 1
%18423 = torch.aten.add.Tensor %18422, %18406, %int1_16158 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16159 = torch.constant.int 1
%18424 = torch.aten.add.Tensor %18423, %18409, %int1_16159 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16160 = torch.constant.int 1
%18425 = torch.aten.add.Tensor %18424, %18412, %int1_16160 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16161 = torch.constant.int 1
%18426 = torch.aten.add.Tensor %18425, %18415, %int1_16161 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16162 = torch.constant.int 1
%18427 = torch.aten.add.Tensor %18426, %18418, %int1_16162 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16163 = torch.constant.int 1
%18428 = torch.aten.add.Tensor %18427, %18421, %int1_16163 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18429 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16164 = arith.constant 1 : index
%dim_16165 = tensor.dim %18429, %c1_16164 : tensor<4x?x4096xf16>
%18430 = flow.tensor.transfer %18429 : tensor<4x?x4096xf16>{%dim_16165} to #hal.device.promise<@__device_1>
%18431 = torch_c.from_builtin_tensor %18430 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18432 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16166 = arith.constant 1 : index
%dim_16167 = tensor.dim %18432, %c1_16166 : tensor<4x?x4096xf16>
%18433 = flow.tensor.transfer %18432 : tensor<4x?x4096xf16>{%dim_16167} to #hal.device.promise<@__device_1>
%18434 = torch_c.from_builtin_tensor %18433 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18435 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16168 = arith.constant 1 : index
%dim_16169 = tensor.dim %18435, %c1_16168 : tensor<4x?x4096xf16>
%18436 = flow.tensor.transfer %18435 : tensor<4x?x4096xf16>{%dim_16169} to #hal.device.promise<@__device_1>
%18437 = torch_c.from_builtin_tensor %18436 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18438 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16170 = arith.constant 1 : index
%dim_16171 = tensor.dim %18438, %c1_16170 : tensor<4x?x4096xf16>
%18439 = flow.tensor.transfer %18438 : tensor<4x?x4096xf16>{%dim_16171} to #hal.device.promise<@__device_1>
%18440 = torch_c.from_builtin_tensor %18439 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18441 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16172 = arith.constant 1 : index
%dim_16173 = tensor.dim %18441, %c1_16172 : tensor<4x?x4096xf16>
%18442 = flow.tensor.transfer %18441 : tensor<4x?x4096xf16>{%dim_16173} to #hal.device.promise<@__device_1>
%18443 = torch_c.from_builtin_tensor %18442 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18444 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16174 = arith.constant 1 : index
%dim_16175 = tensor.dim %18444, %c1_16174 : tensor<4x?x4096xf16>
%18445 = flow.tensor.transfer %18444 : tensor<4x?x4096xf16>{%dim_16175} to #hal.device.promise<@__device_1>
%18446 = torch_c.from_builtin_tensor %18445 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18447 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16176 = arith.constant 1 : index
%dim_16177 = tensor.dim %18447, %c1_16176 : tensor<4x?x4096xf16>
%18448 = flow.tensor.transfer %18447 : tensor<4x?x4096xf16>{%dim_16177} to #hal.device.promise<@__device_1>
%18449 = torch_c.from_builtin_tensor %18448 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16178 = torch.constant.int 1
%18450 = torch.aten.add.Tensor %18431, %18364, %int1_16178 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16179 = torch.constant.int 1
%18451 = torch.aten.add.Tensor %18450, %18434, %int1_16179 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16180 = torch.constant.int 1
%18452 = torch.aten.add.Tensor %18451, %18437, %int1_16180 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16181 = torch.constant.int 1
%18453 = torch.aten.add.Tensor %18452, %18440, %int1_16181 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16182 = torch.constant.int 1
%18454 = torch.aten.add.Tensor %18453, %18443, %int1_16182 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16183 = torch.constant.int 1
%18455 = torch.aten.add.Tensor %18454, %18446, %int1_16183 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16184 = torch.constant.int 1
%18456 = torch.aten.add.Tensor %18455, %18449, %int1_16184 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18457 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16185 = arith.constant 1 : index
%dim_16186 = tensor.dim %18457, %c1_16185 : tensor<4x?x4096xf16>
%18458 = flow.tensor.transfer %18457 : tensor<4x?x4096xf16>{%dim_16186} to #hal.device.promise<@__device_2>
%18459 = torch_c.from_builtin_tensor %18458 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18460 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16187 = arith.constant 1 : index
%dim_16188 = tensor.dim %18460, %c1_16187 : tensor<4x?x4096xf16>
%18461 = flow.tensor.transfer %18460 : tensor<4x?x4096xf16>{%dim_16188} to #hal.device.promise<@__device_2>
%18462 = torch_c.from_builtin_tensor %18461 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18463 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16189 = arith.constant 1 : index
%dim_16190 = tensor.dim %18463, %c1_16189 : tensor<4x?x4096xf16>
%18464 = flow.tensor.transfer %18463 : tensor<4x?x4096xf16>{%dim_16190} to #hal.device.promise<@__device_2>
%18465 = torch_c.from_builtin_tensor %18464 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18466 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16191 = arith.constant 1 : index
%dim_16192 = tensor.dim %18466, %c1_16191 : tensor<4x?x4096xf16>
%18467 = flow.tensor.transfer %18466 : tensor<4x?x4096xf16>{%dim_16192} to #hal.device.promise<@__device_2>
%18468 = torch_c.from_builtin_tensor %18467 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18469 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16193 = arith.constant 1 : index
%dim_16194 = tensor.dim %18469, %c1_16193 : tensor<4x?x4096xf16>
%18470 = flow.tensor.transfer %18469 : tensor<4x?x4096xf16>{%dim_16194} to #hal.device.promise<@__device_2>
%18471 = torch_c.from_builtin_tensor %18470 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18472 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16195 = arith.constant 1 : index
%dim_16196 = tensor.dim %18472, %c1_16195 : tensor<4x?x4096xf16>
%18473 = flow.tensor.transfer %18472 : tensor<4x?x4096xf16>{%dim_16196} to #hal.device.promise<@__device_2>
%18474 = torch_c.from_builtin_tensor %18473 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18475 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16197 = arith.constant 1 : index
%dim_16198 = tensor.dim %18475, %c1_16197 : tensor<4x?x4096xf16>
%18476 = flow.tensor.transfer %18475 : tensor<4x?x4096xf16>{%dim_16198} to #hal.device.promise<@__device_2>
%18477 = torch_c.from_builtin_tensor %18476 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16199 = torch.constant.int 1
%18478 = torch.aten.add.Tensor %18459, %18462, %int1_16199 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16200 = torch.constant.int 1
%18479 = torch.aten.add.Tensor %18478, %18370, %int1_16200 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16201 = torch.constant.int 1
%18480 = torch.aten.add.Tensor %18479, %18465, %int1_16201 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16202 = torch.constant.int 1
%18481 = torch.aten.add.Tensor %18480, %18468, %int1_16202 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16203 = torch.constant.int 1
%18482 = torch.aten.add.Tensor %18481, %18471, %int1_16203 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16204 = torch.constant.int 1
%18483 = torch.aten.add.Tensor %18482, %18474, %int1_16204 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16205 = torch.constant.int 1
%18484 = torch.aten.add.Tensor %18483, %18477, %int1_16205 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18485 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16206 = arith.constant 1 : index
%dim_16207 = tensor.dim %18485, %c1_16206 : tensor<4x?x4096xf16>
%18486 = flow.tensor.transfer %18485 : tensor<4x?x4096xf16>{%dim_16207} to #hal.device.promise<@__device_3>
%18487 = torch_c.from_builtin_tensor %18486 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18488 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16208 = arith.constant 1 : index
%dim_16209 = tensor.dim %18488, %c1_16208 : tensor<4x?x4096xf16>
%18489 = flow.tensor.transfer %18488 : tensor<4x?x4096xf16>{%dim_16209} to #hal.device.promise<@__device_3>
%18490 = torch_c.from_builtin_tensor %18489 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18491 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16210 = arith.constant 1 : index
%dim_16211 = tensor.dim %18491, %c1_16210 : tensor<4x?x4096xf16>
%18492 = flow.tensor.transfer %18491 : tensor<4x?x4096xf16>{%dim_16211} to #hal.device.promise<@__device_3>
%18493 = torch_c.from_builtin_tensor %18492 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18494 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16212 = arith.constant 1 : index
%dim_16213 = tensor.dim %18494, %c1_16212 : tensor<4x?x4096xf16>
%18495 = flow.tensor.transfer %18494 : tensor<4x?x4096xf16>{%dim_16213} to #hal.device.promise<@__device_3>
%18496 = torch_c.from_builtin_tensor %18495 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18497 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16214 = arith.constant 1 : index
%dim_16215 = tensor.dim %18497, %c1_16214 : tensor<4x?x4096xf16>
%18498 = flow.tensor.transfer %18497 : tensor<4x?x4096xf16>{%dim_16215} to #hal.device.promise<@__device_3>
%18499 = torch_c.from_builtin_tensor %18498 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18500 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16216 = arith.constant 1 : index
%dim_16217 = tensor.dim %18500, %c1_16216 : tensor<4x?x4096xf16>
%18501 = flow.tensor.transfer %18500 : tensor<4x?x4096xf16>{%dim_16217} to #hal.device.promise<@__device_3>
%18502 = torch_c.from_builtin_tensor %18501 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18503 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16218 = arith.constant 1 : index
%dim_16219 = tensor.dim %18503, %c1_16218 : tensor<4x?x4096xf16>
%18504 = flow.tensor.transfer %18503 : tensor<4x?x4096xf16>{%dim_16219} to #hal.device.promise<@__device_3>
%18505 = torch_c.from_builtin_tensor %18504 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16220 = torch.constant.int 1
%18506 = torch.aten.add.Tensor %18487, %18490, %int1_16220 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16221 = torch.constant.int 1
%18507 = torch.aten.add.Tensor %18506, %18493, %int1_16221 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16222 = torch.constant.int 1
%18508 = torch.aten.add.Tensor %18507, %18376, %int1_16222 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16223 = torch.constant.int 1
%18509 = torch.aten.add.Tensor %18508, %18496, %int1_16223 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16224 = torch.constant.int 1
%18510 = torch.aten.add.Tensor %18509, %18499, %int1_16224 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16225 = torch.constant.int 1
%18511 = torch.aten.add.Tensor %18510, %18502, %int1_16225 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16226 = torch.constant.int 1
%18512 = torch.aten.add.Tensor %18511, %18505, %int1_16226 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18513 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16227 = arith.constant 1 : index
%dim_16228 = tensor.dim %18513, %c1_16227 : tensor<4x?x4096xf16>
%18514 = flow.tensor.transfer %18513 : tensor<4x?x4096xf16>{%dim_16228} to #hal.device.promise<@__device_4>
%18515 = torch_c.from_builtin_tensor %18514 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18516 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16229 = arith.constant 1 : index
%dim_16230 = tensor.dim %18516, %c1_16229 : tensor<4x?x4096xf16>
%18517 = flow.tensor.transfer %18516 : tensor<4x?x4096xf16>{%dim_16230} to #hal.device.promise<@__device_4>
%18518 = torch_c.from_builtin_tensor %18517 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18519 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16231 = arith.constant 1 : index
%dim_16232 = tensor.dim %18519, %c1_16231 : tensor<4x?x4096xf16>
%18520 = flow.tensor.transfer %18519 : tensor<4x?x4096xf16>{%dim_16232} to #hal.device.promise<@__device_4>
%18521 = torch_c.from_builtin_tensor %18520 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18522 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16233 = arith.constant 1 : index
%dim_16234 = tensor.dim %18522, %c1_16233 : tensor<4x?x4096xf16>
%18523 = flow.tensor.transfer %18522 : tensor<4x?x4096xf16>{%dim_16234} to #hal.device.promise<@__device_4>
%18524 = torch_c.from_builtin_tensor %18523 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18525 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16235 = arith.constant 1 : index
%dim_16236 = tensor.dim %18525, %c1_16235 : tensor<4x?x4096xf16>
%18526 = flow.tensor.transfer %18525 : tensor<4x?x4096xf16>{%dim_16236} to #hal.device.promise<@__device_4>
%18527 = torch_c.from_builtin_tensor %18526 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18528 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16237 = arith.constant 1 : index
%dim_16238 = tensor.dim %18528, %c1_16237 : tensor<4x?x4096xf16>
%18529 = flow.tensor.transfer %18528 : tensor<4x?x4096xf16>{%dim_16238} to #hal.device.promise<@__device_4>
%18530 = torch_c.from_builtin_tensor %18529 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18531 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16239 = arith.constant 1 : index
%dim_16240 = tensor.dim %18531, %c1_16239 : tensor<4x?x4096xf16>
%18532 = flow.tensor.transfer %18531 : tensor<4x?x4096xf16>{%dim_16240} to #hal.device.promise<@__device_4>
%18533 = torch_c.from_builtin_tensor %18532 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16241 = torch.constant.int 1
%18534 = torch.aten.add.Tensor %18515, %18518, %int1_16241 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16242 = torch.constant.int 1
%18535 = torch.aten.add.Tensor %18534, %18521, %int1_16242 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16243 = torch.constant.int 1
%18536 = torch.aten.add.Tensor %18535, %18524, %int1_16243 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16244 = torch.constant.int 1
%18537 = torch.aten.add.Tensor %18536, %18382, %int1_16244 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16245 = torch.constant.int 1
%18538 = torch.aten.add.Tensor %18537, %18527, %int1_16245 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16246 = torch.constant.int 1
%18539 = torch.aten.add.Tensor %18538, %18530, %int1_16246 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16247 = torch.constant.int 1
%18540 = torch.aten.add.Tensor %18539, %18533, %int1_16247 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18541 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16248 = arith.constant 1 : index
%dim_16249 = tensor.dim %18541, %c1_16248 : tensor<4x?x4096xf16>
%18542 = flow.tensor.transfer %18541 : tensor<4x?x4096xf16>{%dim_16249} to #hal.device.promise<@__device_5>
%18543 = torch_c.from_builtin_tensor %18542 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18544 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16250 = arith.constant 1 : index
%dim_16251 = tensor.dim %18544, %c1_16250 : tensor<4x?x4096xf16>
%18545 = flow.tensor.transfer %18544 : tensor<4x?x4096xf16>{%dim_16251} to #hal.device.promise<@__device_5>
%18546 = torch_c.from_builtin_tensor %18545 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18547 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16252 = arith.constant 1 : index
%dim_16253 = tensor.dim %18547, %c1_16252 : tensor<4x?x4096xf16>
%18548 = flow.tensor.transfer %18547 : tensor<4x?x4096xf16>{%dim_16253} to #hal.device.promise<@__device_5>
%18549 = torch_c.from_builtin_tensor %18548 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18550 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16254 = arith.constant 1 : index
%dim_16255 = tensor.dim %18550, %c1_16254 : tensor<4x?x4096xf16>
%18551 = flow.tensor.transfer %18550 : tensor<4x?x4096xf16>{%dim_16255} to #hal.device.promise<@__device_5>
%18552 = torch_c.from_builtin_tensor %18551 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18553 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16256 = arith.constant 1 : index
%dim_16257 = tensor.dim %18553, %c1_16256 : tensor<4x?x4096xf16>
%18554 = flow.tensor.transfer %18553 : tensor<4x?x4096xf16>{%dim_16257} to #hal.device.promise<@__device_5>
%18555 = torch_c.from_builtin_tensor %18554 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18556 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16258 = arith.constant 1 : index
%dim_16259 = tensor.dim %18556, %c1_16258 : tensor<4x?x4096xf16>
%18557 = flow.tensor.transfer %18556 : tensor<4x?x4096xf16>{%dim_16259} to #hal.device.promise<@__device_5>
%18558 = torch_c.from_builtin_tensor %18557 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18559 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16260 = arith.constant 1 : index
%dim_16261 = tensor.dim %18559, %c1_16260 : tensor<4x?x4096xf16>
%18560 = flow.tensor.transfer %18559 : tensor<4x?x4096xf16>{%dim_16261} to #hal.device.promise<@__device_5>
%18561 = torch_c.from_builtin_tensor %18560 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16262 = torch.constant.int 1
%18562 = torch.aten.add.Tensor %18543, %18546, %int1_16262 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16263 = torch.constant.int 1
%18563 = torch.aten.add.Tensor %18562, %18549, %int1_16263 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16264 = torch.constant.int 1
%18564 = torch.aten.add.Tensor %18563, %18552, %int1_16264 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16265 = torch.constant.int 1
%18565 = torch.aten.add.Tensor %18564, %18555, %int1_16265 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16266 = torch.constant.int 1
%18566 = torch.aten.add.Tensor %18565, %18388, %int1_16266 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16267 = torch.constant.int 1
%18567 = torch.aten.add.Tensor %18566, %18558, %int1_16267 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16268 = torch.constant.int 1
%18568 = torch.aten.add.Tensor %18567, %18561, %int1_16268 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18569 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16269 = arith.constant 1 : index
%dim_16270 = tensor.dim %18569, %c1_16269 : tensor<4x?x4096xf16>
%18570 = flow.tensor.transfer %18569 : tensor<4x?x4096xf16>{%dim_16270} to #hal.device.promise<@__device_6>
%18571 = torch_c.from_builtin_tensor %18570 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18572 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16271 = arith.constant 1 : index
%dim_16272 = tensor.dim %18572, %c1_16271 : tensor<4x?x4096xf16>
%18573 = flow.tensor.transfer %18572 : tensor<4x?x4096xf16>{%dim_16272} to #hal.device.promise<@__device_6>
%18574 = torch_c.from_builtin_tensor %18573 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18575 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16273 = arith.constant 1 : index
%dim_16274 = tensor.dim %18575, %c1_16273 : tensor<4x?x4096xf16>
%18576 = flow.tensor.transfer %18575 : tensor<4x?x4096xf16>{%dim_16274} to #hal.device.promise<@__device_6>
%18577 = torch_c.from_builtin_tensor %18576 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18578 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16275 = arith.constant 1 : index
%dim_16276 = tensor.dim %18578, %c1_16275 : tensor<4x?x4096xf16>
%18579 = flow.tensor.transfer %18578 : tensor<4x?x4096xf16>{%dim_16276} to #hal.device.promise<@__device_6>
%18580 = torch_c.from_builtin_tensor %18579 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18581 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16277 = arith.constant 1 : index
%dim_16278 = tensor.dim %18581, %c1_16277 : tensor<4x?x4096xf16>
%18582 = flow.tensor.transfer %18581 : tensor<4x?x4096xf16>{%dim_16278} to #hal.device.promise<@__device_6>
%18583 = torch_c.from_builtin_tensor %18582 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18584 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16279 = arith.constant 1 : index
%dim_16280 = tensor.dim %18584, %c1_16279 : tensor<4x?x4096xf16>
%18585 = flow.tensor.transfer %18584 : tensor<4x?x4096xf16>{%dim_16280} to #hal.device.promise<@__device_6>
%18586 = torch_c.from_builtin_tensor %18585 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18587 = torch_c.to_builtin_tensor %18400 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16281 = arith.constant 1 : index
%dim_16282 = tensor.dim %18587, %c1_16281 : tensor<4x?x4096xf16>
%18588 = flow.tensor.transfer %18587 : tensor<4x?x4096xf16>{%dim_16282} to #hal.device.promise<@__device_6>
%18589 = torch_c.from_builtin_tensor %18588 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16283 = torch.constant.int 1
%18590 = torch.aten.add.Tensor %18571, %18574, %int1_16283 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16284 = torch.constant.int 1
%18591 = torch.aten.add.Tensor %18590, %18577, %int1_16284 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16285 = torch.constant.int 1
%18592 = torch.aten.add.Tensor %18591, %18580, %int1_16285 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16286 = torch.constant.int 1
%18593 = torch.aten.add.Tensor %18592, %18583, %int1_16286 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16287 = torch.constant.int 1
%18594 = torch.aten.add.Tensor %18593, %18586, %int1_16287 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16288 = torch.constant.int 1
%18595 = torch.aten.add.Tensor %18594, %18394, %int1_16288 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16289 = torch.constant.int 1
%18596 = torch.aten.add.Tensor %18595, %18589, %int1_16289 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18597 = torch_c.to_builtin_tensor %18358 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16290 = arith.constant 1 : index
%dim_16291 = tensor.dim %18597, %c1_16290 : tensor<4x?x4096xf16>
%18598 = flow.tensor.transfer %18597 : tensor<4x?x4096xf16>{%dim_16291} to #hal.device.promise<@__device_7>
%18599 = torch_c.from_builtin_tensor %18598 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18600 = torch_c.to_builtin_tensor %18364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16292 = arith.constant 1 : index
%dim_16293 = tensor.dim %18600, %c1_16292 : tensor<4x?x4096xf16>
%18601 = flow.tensor.transfer %18600 : tensor<4x?x4096xf16>{%dim_16293} to #hal.device.promise<@__device_7>
%18602 = torch_c.from_builtin_tensor %18601 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18603 = torch_c.to_builtin_tensor %18370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16294 = arith.constant 1 : index
%dim_16295 = tensor.dim %18603, %c1_16294 : tensor<4x?x4096xf16>
%18604 = flow.tensor.transfer %18603 : tensor<4x?x4096xf16>{%dim_16295} to #hal.device.promise<@__device_7>
%18605 = torch_c.from_builtin_tensor %18604 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18606 = torch_c.to_builtin_tensor %18376 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16296 = arith.constant 1 : index
%dim_16297 = tensor.dim %18606, %c1_16296 : tensor<4x?x4096xf16>
%18607 = flow.tensor.transfer %18606 : tensor<4x?x4096xf16>{%dim_16297} to #hal.device.promise<@__device_7>
%18608 = torch_c.from_builtin_tensor %18607 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18609 = torch_c.to_builtin_tensor %18382 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16298 = arith.constant 1 : index
%dim_16299 = tensor.dim %18609, %c1_16298 : tensor<4x?x4096xf16>
%18610 = flow.tensor.transfer %18609 : tensor<4x?x4096xf16>{%dim_16299} to #hal.device.promise<@__device_7>
%18611 = torch_c.from_builtin_tensor %18610 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18612 = torch_c.to_builtin_tensor %18388 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16300 = arith.constant 1 : index
%dim_16301 = tensor.dim %18612, %c1_16300 : tensor<4x?x4096xf16>
%18613 = flow.tensor.transfer %18612 : tensor<4x?x4096xf16>{%dim_16301} to #hal.device.promise<@__device_7>
%18614 = torch_c.from_builtin_tensor %18613 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18615 = torch_c.to_builtin_tensor %18394 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16302 = arith.constant 1 : index
%dim_16303 = tensor.dim %18615, %c1_16302 : tensor<4x?x4096xf16>
%18616 = flow.tensor.transfer %18615 : tensor<4x?x4096xf16>{%dim_16303} to #hal.device.promise<@__device_7>
%18617 = torch_c.from_builtin_tensor %18616 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16304 = torch.constant.int 1
%18618 = torch.aten.add.Tensor %18599, %18602, %int1_16304 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16305 = torch.constant.int 1
%18619 = torch.aten.add.Tensor %18618, %18605, %int1_16305 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16306 = torch.constant.int 1
%18620 = torch.aten.add.Tensor %18619, %18608, %int1_16306 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16307 = torch.constant.int 1
%18621 = torch.aten.add.Tensor %18620, %18611, %int1_16307 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16308 = torch.constant.int 1
%18622 = torch.aten.add.Tensor %18621, %18614, %int1_16308 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16309 = torch.constant.int 1
%18623 = torch.aten.add.Tensor %18622, %18617, %int1_16309 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16310 = torch.constant.int 1
%18624 = torch.aten.add.Tensor %18623, %18400, %int1_16310 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16311 = torch.constant.int 1
%18625 = torch.aten.add.Tensor %17284, %18428, %int1_16311 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16312 = torch.constant.int 1
%18626 = torch.aten.add.Tensor %17285, %18456, %int1_16312 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16313 = torch.constant.int 1
%18627 = torch.aten.add.Tensor %17286, %18484, %int1_16313 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16314 = torch.constant.int 1
%18628 = torch.aten.add.Tensor %17287, %18512, %int1_16314 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16315 = torch.constant.int 1
%18629 = torch.aten.add.Tensor %17288, %18540, %int1_16315 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16316 = torch.constant.int 1
%18630 = torch.aten.add.Tensor %17289, %18568, %int1_16316 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16317 = torch.constant.int 1
%18631 = torch.aten.add.Tensor %17290, %18596, %int1_16317 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16318 = torch.constant.int 1
%18632 = torch.aten.add.Tensor %17291, %18624, %int1_16318 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_16319 = torch.constant.int 6
%18633 = torch.prims.convert_element_type %18625, %int6_16319 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16320 = torch.constant.int 6
%18634 = torch.prims.convert_element_type %18626, %int6_16320 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16321 = torch.constant.int 6
%18635 = torch.prims.convert_element_type %18627, %int6_16321 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16322 = torch.constant.int 6
%18636 = torch.prims.convert_element_type %18628, %int6_16322 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16323 = torch.constant.int 6
%18637 = torch.prims.convert_element_type %18629, %int6_16323 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16324 = torch.constant.int 6
%18638 = torch.prims.convert_element_type %18630, %int6_16324 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16325 = torch.constant.int 6
%18639 = torch.prims.convert_element_type %18631, %int6_16325 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16326 = torch.constant.int 6
%18640 = torch.prims.convert_element_type %18632, %int6_16326 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16327 = torch.constant.int 2
%18641 = torch.aten.pow.Tensor_Scalar %18633, %int2_16327 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16328 = torch.constant.int 2
%18642 = torch.aten.pow.Tensor_Scalar %18634, %int2_16328 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16329 = torch.constant.int 2
%18643 = torch.aten.pow.Tensor_Scalar %18635, %int2_16329 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16330 = torch.constant.int 2
%18644 = torch.aten.pow.Tensor_Scalar %18636, %int2_16330 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16331 = torch.constant.int 2
%18645 = torch.aten.pow.Tensor_Scalar %18637, %int2_16331 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16332 = torch.constant.int 2
%18646 = torch.aten.pow.Tensor_Scalar %18638, %int2_16332 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16333 = torch.constant.int 2
%18647 = torch.aten.pow.Tensor_Scalar %18639, %int2_16333 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16334 = torch.constant.int 2
%18648 = torch.aten.pow.Tensor_Scalar %18640, %int2_16334 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_16335 = torch.constant.int -1
%18649 = torch.prim.ListConstruct %int-1_16335 : (!torch.int) -> !torch.list<int>
%true_16336 = torch.constant.bool true
%none_16337 = torch.constant.none
%18650 = torch.aten.mean.dim %18641, %18649, %true_16336, %none_16337 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16338 = torch.constant.int -1
%18651 = torch.prim.ListConstruct %int-1_16338 : (!torch.int) -> !torch.list<int>
%true_16339 = torch.constant.bool true
%none_16340 = torch.constant.none
%18652 = torch.aten.mean.dim %18642, %18651, %true_16339, %none_16340 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16341 = torch.constant.int -1
%18653 = torch.prim.ListConstruct %int-1_16341 : (!torch.int) -> !torch.list<int>
%true_16342 = torch.constant.bool true
%none_16343 = torch.constant.none
%18654 = torch.aten.mean.dim %18643, %18653, %true_16342, %none_16343 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16344 = torch.constant.int -1
%18655 = torch.prim.ListConstruct %int-1_16344 : (!torch.int) -> !torch.list<int>
%true_16345 = torch.constant.bool true
%none_16346 = torch.constant.none
%18656 = torch.aten.mean.dim %18644, %18655, %true_16345, %none_16346 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16347 = torch.constant.int -1
%18657 = torch.prim.ListConstruct %int-1_16347 : (!torch.int) -> !torch.list<int>
%true_16348 = torch.constant.bool true
%none_16349 = torch.constant.none
%18658 = torch.aten.mean.dim %18645, %18657, %true_16348, %none_16349 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16350 = torch.constant.int -1
%18659 = torch.prim.ListConstruct %int-1_16350 : (!torch.int) -> !torch.list<int>
%true_16351 = torch.constant.bool true
%none_16352 = torch.constant.none
%18660 = torch.aten.mean.dim %18646, %18659, %true_16351, %none_16352 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16353 = torch.constant.int -1
%18661 = torch.prim.ListConstruct %int-1_16353 : (!torch.int) -> !torch.list<int>
%true_16354 = torch.constant.bool true
%none_16355 = torch.constant.none
%18662 = torch.aten.mean.dim %18647, %18661, %true_16354, %none_16355 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16356 = torch.constant.int -1
%18663 = torch.prim.ListConstruct %int-1_16356 : (!torch.int) -> !torch.list<int>
%true_16357 = torch.constant.bool true
%none_16358 = torch.constant.none
%18664 = torch.aten.mean.dim %18648, %18663, %true_16357, %none_16358 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16359 = torch.constant.float 9.9999997473787516E-6
%int1_16360 = torch.constant.int 1
%18665 = torch.aten.add.Scalar %18650, %float9.999990e-06_16359, %int1_16360 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16361 = torch.constant.float 9.9999997473787516E-6
%int1_16362 = torch.constant.int 1
%18666 = torch.aten.add.Scalar %18652, %float9.999990e-06_16361, %int1_16362 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16363 = torch.constant.float 9.9999997473787516E-6
%int1_16364 = torch.constant.int 1
%18667 = torch.aten.add.Scalar %18654, %float9.999990e-06_16363, %int1_16364 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16365 = torch.constant.float 9.9999997473787516E-6
%int1_16366 = torch.constant.int 1
%18668 = torch.aten.add.Scalar %18656, %float9.999990e-06_16365, %int1_16366 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16367 = torch.constant.float 9.9999997473787516E-6
%int1_16368 = torch.constant.int 1
%18669 = torch.aten.add.Scalar %18658, %float9.999990e-06_16367, %int1_16368 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16369 = torch.constant.float 9.9999997473787516E-6
%int1_16370 = torch.constant.int 1
%18670 = torch.aten.add.Scalar %18660, %float9.999990e-06_16369, %int1_16370 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16371 = torch.constant.float 9.9999997473787516E-6
%int1_16372 = torch.constant.int 1
%18671 = torch.aten.add.Scalar %18662, %float9.999990e-06_16371, %int1_16372 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16373 = torch.constant.float 9.9999997473787516E-6
%int1_16374 = torch.constant.int 1
%18672 = torch.aten.add.Scalar %18664, %float9.999990e-06_16373, %int1_16374 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18673 = torch.aten.rsqrt %18665 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18674 = torch.aten.rsqrt %18666 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18675 = torch.aten.rsqrt %18667 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18676 = torch.aten.rsqrt %18668 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18677 = torch.aten.rsqrt %18669 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18678 = torch.aten.rsqrt %18670 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18679 = torch.aten.rsqrt %18671 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18680 = torch.aten.rsqrt %18672 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %18680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%18681 = torch.aten.mul.Tensor %18633, %18673 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18682 = torch.aten.mul.Tensor %18634, %18674 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18683 = torch.aten.mul.Tensor %18635, %18675 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18684 = torch.aten.mul.Tensor %18636, %18676 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18685 = torch.aten.mul.Tensor %18637, %18677 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18686 = torch.aten.mul.Tensor %18638, %18678 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18687 = torch.aten.mul.Tensor %18639, %18679 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18688 = torch.aten.mul.Tensor %18640, %18680 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18689 = torch.aten.mul.Tensor %624, %18681 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18690 = torch.aten.mul.Tensor %625, %18682 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18691 = torch.aten.mul.Tensor %626, %18683 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18692 = torch.aten.mul.Tensor %627, %18684 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18693 = torch.aten.mul.Tensor %628, %18685 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18694 = torch.aten.mul.Tensor %629, %18686 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18695 = torch.aten.mul.Tensor %630, %18687 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%18696 = torch.aten.mul.Tensor %631, %18688 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %18696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_16375 = torch.constant.int 5
%18697 = torch.prims.convert_element_type %18689, %int5_16375 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16376 = torch.constant.int 5
%18698 = torch.prims.convert_element_type %18690, %int5_16376 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16377 = torch.constant.int 5
%18699 = torch.prims.convert_element_type %18691, %int5_16377 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16378 = torch.constant.int 5
%18700 = torch.prims.convert_element_type %18692, %int5_16378 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16379 = torch.constant.int 5
%18701 = torch.prims.convert_element_type %18693, %int5_16379 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16380 = torch.constant.int 5
%18702 = torch.prims.convert_element_type %18694, %int5_16380 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16381 = torch.constant.int 5
%18703 = torch.prims.convert_element_type %18695, %int5_16381 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16382 = torch.constant.int 5
%18704 = torch.prims.convert_element_type %18696, %int5_16382 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16383 = torch.constant.int 1
%int0_16384 = torch.constant.int 0
%18705 = torch.prim.ListConstruct %int1_16383, %int0_16384 : (!torch.int, !torch.int) -> !torch.list<int>
%18706 = torch.aten.permute %632, %18705 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16385 = torch.constant.int 1
%int0_16386 = torch.constant.int 0
%18707 = torch.prim.ListConstruct %int1_16385, %int0_16386 : (!torch.int, !torch.int) -> !torch.list<int>
%18708 = torch.aten.permute %633, %18707 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16387 = torch.constant.int 1
%int0_16388 = torch.constant.int 0
%18709 = torch.prim.ListConstruct %int1_16387, %int0_16388 : (!torch.int, !torch.int) -> !torch.list<int>
%18710 = torch.aten.permute %634, %18709 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16389 = torch.constant.int 1
%int0_16390 = torch.constant.int 0
%18711 = torch.prim.ListConstruct %int1_16389, %int0_16390 : (!torch.int, !torch.int) -> !torch.list<int>
%18712 = torch.aten.permute %635, %18711 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16391 = torch.constant.int 1
%int0_16392 = torch.constant.int 0
%18713 = torch.prim.ListConstruct %int1_16391, %int0_16392 : (!torch.int, !torch.int) -> !torch.list<int>
%18714 = torch.aten.permute %636, %18713 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16393 = torch.constant.int 1
%int0_16394 = torch.constant.int 0
%18715 = torch.prim.ListConstruct %int1_16393, %int0_16394 : (!torch.int, !torch.int) -> !torch.list<int>
%18716 = torch.aten.permute %637, %18715 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16395 = torch.constant.int 1
%int0_16396 = torch.constant.int 0
%18717 = torch.prim.ListConstruct %int1_16395, %int0_16396 : (!torch.int, !torch.int) -> !torch.list<int>
%18718 = torch.aten.permute %638, %18717 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16397 = torch.constant.int 1
%int0_16398 = torch.constant.int 0
%18719 = torch.prim.ListConstruct %int1_16397, %int0_16398 : (!torch.int, !torch.int) -> !torch.list<int>
%18720 = torch.aten.permute %639, %18719 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_16399 = torch.constant.int 4
%18721 = torch.aten.mul.int %int4_16399, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16400 = torch.constant.int 4096
%18722 = torch.prim.ListConstruct %18721, %int4096_16400 : (!torch.int, !torch.int) -> !torch.list<int>
%18723 = torch.aten.view %18697, %18722 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18723, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18724 = torch.aten.mm %18723, %18706 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18724, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16401 = torch.constant.int 4
%int1792_16402 = torch.constant.int 1792
%18725 = torch.prim.ListConstruct %int4_16401, %2482, %int1792_16402 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18726 = torch.aten.view %18724, %18725 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16403 = torch.constant.int 4
%18727 = torch.aten.mul.int %int4_16403, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16404 = torch.constant.int 4096
%18728 = torch.prim.ListConstruct %18727, %int4096_16404 : (!torch.int, !torch.int) -> !torch.list<int>
%18729 = torch.aten.view %18698, %18728 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18729, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18730 = torch.aten.mm %18729, %18708 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18730, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16405 = torch.constant.int 4
%int1792_16406 = torch.constant.int 1792
%18731 = torch.prim.ListConstruct %int4_16405, %2482, %int1792_16406 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18732 = torch.aten.view %18730, %18731 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16407 = torch.constant.int 4
%18733 = torch.aten.mul.int %int4_16407, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16408 = torch.constant.int 4096
%18734 = torch.prim.ListConstruct %18733, %int4096_16408 : (!torch.int, !torch.int) -> !torch.list<int>
%18735 = torch.aten.view %18699, %18734 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18735, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18736 = torch.aten.mm %18735, %18710 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18736, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16409 = torch.constant.int 4
%int1792_16410 = torch.constant.int 1792
%18737 = torch.prim.ListConstruct %int4_16409, %2482, %int1792_16410 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18738 = torch.aten.view %18736, %18737 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16411 = torch.constant.int 4
%18739 = torch.aten.mul.int %int4_16411, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16412 = torch.constant.int 4096
%18740 = torch.prim.ListConstruct %18739, %int4096_16412 : (!torch.int, !torch.int) -> !torch.list<int>
%18741 = torch.aten.view %18700, %18740 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18741, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18742 = torch.aten.mm %18741, %18712 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18742, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16413 = torch.constant.int 4
%int1792_16414 = torch.constant.int 1792
%18743 = torch.prim.ListConstruct %int4_16413, %2482, %int1792_16414 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18744 = torch.aten.view %18742, %18743 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16415 = torch.constant.int 4
%18745 = torch.aten.mul.int %int4_16415, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16416 = torch.constant.int 4096
%18746 = torch.prim.ListConstruct %18745, %int4096_16416 : (!torch.int, !torch.int) -> !torch.list<int>
%18747 = torch.aten.view %18701, %18746 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18747, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18748 = torch.aten.mm %18747, %18714 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18748, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16417 = torch.constant.int 4
%int1792_16418 = torch.constant.int 1792
%18749 = torch.prim.ListConstruct %int4_16417, %2482, %int1792_16418 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18750 = torch.aten.view %18748, %18749 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16419 = torch.constant.int 4
%18751 = torch.aten.mul.int %int4_16419, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16420 = torch.constant.int 4096
%18752 = torch.prim.ListConstruct %18751, %int4096_16420 : (!torch.int, !torch.int) -> !torch.list<int>
%18753 = torch.aten.view %18702, %18752 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18753, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18754 = torch.aten.mm %18753, %18716 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18754, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16421 = torch.constant.int 4
%int1792_16422 = torch.constant.int 1792
%18755 = torch.prim.ListConstruct %int4_16421, %2482, %int1792_16422 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18756 = torch.aten.view %18754, %18755 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16423 = torch.constant.int 4
%18757 = torch.aten.mul.int %int4_16423, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16424 = torch.constant.int 4096
%18758 = torch.prim.ListConstruct %18757, %int4096_16424 : (!torch.int, !torch.int) -> !torch.list<int>
%18759 = torch.aten.view %18703, %18758 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18759, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18760 = torch.aten.mm %18759, %18718 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18760, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16425 = torch.constant.int 4
%int1792_16426 = torch.constant.int 1792
%18761 = torch.prim.ListConstruct %int4_16425, %2482, %int1792_16426 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18762 = torch.aten.view %18760, %18761 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16427 = torch.constant.int 4
%18763 = torch.aten.mul.int %int4_16427, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16428 = torch.constant.int 4096
%18764 = torch.prim.ListConstruct %18763, %int4096_16428 : (!torch.int, !torch.int) -> !torch.list<int>
%18765 = torch.aten.view %18704, %18764 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18765, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18766 = torch.aten.mm %18765, %18720 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18766, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16429 = torch.constant.int 4
%int1792_16430 = torch.constant.int 1792
%18767 = torch.prim.ListConstruct %int4_16429, %2482, %int1792_16430 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18768 = torch.aten.view %18766, %18767 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18769 = torch.aten.silu %18726 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18770 = torch.aten.silu %18732 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18771 = torch.aten.silu %18738 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18772 = torch.aten.silu %18744 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18773 = torch.aten.silu %18750 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18774 = torch.aten.silu %18756 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18775 = torch.aten.silu %18762 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18776 = torch.aten.silu %18768 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_16431 = torch.constant.int 1
%int0_16432 = torch.constant.int 0
%18777 = torch.prim.ListConstruct %int1_16431, %int0_16432 : (!torch.int, !torch.int) -> !torch.list<int>
%18778 = torch.aten.permute %640, %18777 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16433 = torch.constant.int 1
%int0_16434 = torch.constant.int 0
%18779 = torch.prim.ListConstruct %int1_16433, %int0_16434 : (!torch.int, !torch.int) -> !torch.list<int>
%18780 = torch.aten.permute %641, %18779 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16435 = torch.constant.int 1
%int0_16436 = torch.constant.int 0
%18781 = torch.prim.ListConstruct %int1_16435, %int0_16436 : (!torch.int, !torch.int) -> !torch.list<int>
%18782 = torch.aten.permute %642, %18781 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16437 = torch.constant.int 1
%int0_16438 = torch.constant.int 0
%18783 = torch.prim.ListConstruct %int1_16437, %int0_16438 : (!torch.int, !torch.int) -> !torch.list<int>
%18784 = torch.aten.permute %643, %18783 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16439 = torch.constant.int 1
%int0_16440 = torch.constant.int 0
%18785 = torch.prim.ListConstruct %int1_16439, %int0_16440 : (!torch.int, !torch.int) -> !torch.list<int>
%18786 = torch.aten.permute %644, %18785 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16441 = torch.constant.int 1
%int0_16442 = torch.constant.int 0
%18787 = torch.prim.ListConstruct %int1_16441, %int0_16442 : (!torch.int, !torch.int) -> !torch.list<int>
%18788 = torch.aten.permute %645, %18787 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16443 = torch.constant.int 1
%int0_16444 = torch.constant.int 0
%18789 = torch.prim.ListConstruct %int1_16443, %int0_16444 : (!torch.int, !torch.int) -> !torch.list<int>
%18790 = torch.aten.permute %646, %18789 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_16445 = torch.constant.int 1
%int0_16446 = torch.constant.int 0
%18791 = torch.prim.ListConstruct %int1_16445, %int0_16446 : (!torch.int, !torch.int) -> !torch.list<int>
%18792 = torch.aten.permute %647, %18791 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_16447 = torch.constant.int 4
%18793 = torch.aten.mul.int %int4_16447, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16448 = torch.constant.int 4096
%18794 = torch.prim.ListConstruct %18793, %int4096_16448 : (!torch.int, !torch.int) -> !torch.list<int>
%18795 = torch.aten.view %18697, %18794 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18795, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18796 = torch.aten.mm %18795, %18778 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18796, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16449 = torch.constant.int 4
%int1792_16450 = torch.constant.int 1792
%18797 = torch.prim.ListConstruct %int4_16449, %2482, %int1792_16450 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18798 = torch.aten.view %18796, %18797 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16451 = torch.constant.int 4
%18799 = torch.aten.mul.int %int4_16451, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16452 = torch.constant.int 4096
%18800 = torch.prim.ListConstruct %18799, %int4096_16452 : (!torch.int, !torch.int) -> !torch.list<int>
%18801 = torch.aten.view %18698, %18800 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18801, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18802 = torch.aten.mm %18801, %18780 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18802, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16453 = torch.constant.int 4
%int1792_16454 = torch.constant.int 1792
%18803 = torch.prim.ListConstruct %int4_16453, %2482, %int1792_16454 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18804 = torch.aten.view %18802, %18803 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16455 = torch.constant.int 4
%18805 = torch.aten.mul.int %int4_16455, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16456 = torch.constant.int 4096
%18806 = torch.prim.ListConstruct %18805, %int4096_16456 : (!torch.int, !torch.int) -> !torch.list<int>
%18807 = torch.aten.view %18699, %18806 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18807, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18808 = torch.aten.mm %18807, %18782 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18808, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16457 = torch.constant.int 4
%int1792_16458 = torch.constant.int 1792
%18809 = torch.prim.ListConstruct %int4_16457, %2482, %int1792_16458 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18810 = torch.aten.view %18808, %18809 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16459 = torch.constant.int 4
%18811 = torch.aten.mul.int %int4_16459, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16460 = torch.constant.int 4096
%18812 = torch.prim.ListConstruct %18811, %int4096_16460 : (!torch.int, !torch.int) -> !torch.list<int>
%18813 = torch.aten.view %18700, %18812 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18813, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18814 = torch.aten.mm %18813, %18784 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18814, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16461 = torch.constant.int 4
%int1792_16462 = torch.constant.int 1792
%18815 = torch.prim.ListConstruct %int4_16461, %2482, %int1792_16462 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18816 = torch.aten.view %18814, %18815 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16463 = torch.constant.int 4
%18817 = torch.aten.mul.int %int4_16463, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16464 = torch.constant.int 4096
%18818 = torch.prim.ListConstruct %18817, %int4096_16464 : (!torch.int, !torch.int) -> !torch.list<int>
%18819 = torch.aten.view %18701, %18818 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18819, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18820 = torch.aten.mm %18819, %18786 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18820, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16465 = torch.constant.int 4
%int1792_16466 = torch.constant.int 1792
%18821 = torch.prim.ListConstruct %int4_16465, %2482, %int1792_16466 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18822 = torch.aten.view %18820, %18821 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16467 = torch.constant.int 4
%18823 = torch.aten.mul.int %int4_16467, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16468 = torch.constant.int 4096
%18824 = torch.prim.ListConstruct %18823, %int4096_16468 : (!torch.int, !torch.int) -> !torch.list<int>
%18825 = torch.aten.view %18702, %18824 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18825, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18826 = torch.aten.mm %18825, %18788 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18826, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16469 = torch.constant.int 4
%int1792_16470 = torch.constant.int 1792
%18827 = torch.prim.ListConstruct %int4_16469, %2482, %int1792_16470 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18828 = torch.aten.view %18826, %18827 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16471 = torch.constant.int 4
%18829 = torch.aten.mul.int %int4_16471, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16472 = torch.constant.int 4096
%18830 = torch.prim.ListConstruct %18829, %int4096_16472 : (!torch.int, !torch.int) -> !torch.list<int>
%18831 = torch.aten.view %18703, %18830 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18831, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18832 = torch.aten.mm %18831, %18790 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18832, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16473 = torch.constant.int 4
%int1792_16474 = torch.constant.int 1792
%18833 = torch.prim.ListConstruct %int4_16473, %2482, %int1792_16474 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18834 = torch.aten.view %18832, %18833 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_16475 = torch.constant.int 4
%18835 = torch.aten.mul.int %int4_16475, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16476 = torch.constant.int 4096
%18836 = torch.prim.ListConstruct %18835, %int4096_16476 : (!torch.int, !torch.int) -> !torch.list<int>
%18837 = torch.aten.view %18704, %18836 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18837, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%18838 = torch.aten.mm %18837, %18792 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18838, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_16477 = torch.constant.int 4
%int1792_16478 = torch.constant.int 1792
%18839 = torch.prim.ListConstruct %int4_16477, %2482, %int1792_16478 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18840 = torch.aten.view %18838, %18839 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18841 = torch.aten.mul.Tensor %18769, %18798 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18842 = torch.aten.mul.Tensor %18770, %18804 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18843 = torch.aten.mul.Tensor %18771, %18810 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18844 = torch.aten.mul.Tensor %18772, %18816 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18845 = torch.aten.mul.Tensor %18773, %18822 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18846 = torch.aten.mul.Tensor %18774, %18828 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18847 = torch.aten.mul.Tensor %18775, %18834 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%18848 = torch.aten.mul.Tensor %18776, %18840 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %18848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_16479 = torch.constant.int 1
%int0_16480 = torch.constant.int 0
%18849 = torch.prim.ListConstruct %int1_16479, %int0_16480 : (!torch.int, !torch.int) -> !torch.list<int>
%18850 = torch.aten.permute %648, %18849 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16481 = torch.constant.int 1
%int0_16482 = torch.constant.int 0
%18851 = torch.prim.ListConstruct %int1_16481, %int0_16482 : (!torch.int, !torch.int) -> !torch.list<int>
%18852 = torch.aten.permute %649, %18851 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16483 = torch.constant.int 1
%int0_16484 = torch.constant.int 0
%18853 = torch.prim.ListConstruct %int1_16483, %int0_16484 : (!torch.int, !torch.int) -> !torch.list<int>
%18854 = torch.aten.permute %650, %18853 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16485 = torch.constant.int 1
%int0_16486 = torch.constant.int 0
%18855 = torch.prim.ListConstruct %int1_16485, %int0_16486 : (!torch.int, !torch.int) -> !torch.list<int>
%18856 = torch.aten.permute %651, %18855 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16487 = torch.constant.int 1
%int0_16488 = torch.constant.int 0
%18857 = torch.prim.ListConstruct %int1_16487, %int0_16488 : (!torch.int, !torch.int) -> !torch.list<int>
%18858 = torch.aten.permute %652, %18857 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16489 = torch.constant.int 1
%int0_16490 = torch.constant.int 0
%18859 = torch.prim.ListConstruct %int1_16489, %int0_16490 : (!torch.int, !torch.int) -> !torch.list<int>
%18860 = torch.aten.permute %653, %18859 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16491 = torch.constant.int 1
%int0_16492 = torch.constant.int 0
%18861 = torch.prim.ListConstruct %int1_16491, %int0_16492 : (!torch.int, !torch.int) -> !torch.list<int>
%18862 = torch.aten.permute %654, %18861 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16493 = torch.constant.int 1
%int0_16494 = torch.constant.int 0
%18863 = torch.prim.ListConstruct %int1_16493, %int0_16494 : (!torch.int, !torch.int) -> !torch.list<int>
%18864 = torch.aten.permute %655, %18863 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_16495 = torch.constant.int 1
%18865 = torch.aten.size.int %18726, %int1_16495 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16496 = torch.constant.int 4
%18866 = torch.aten.mul.int %int4_16496, %18865 : !torch.int, !torch.int -> !torch.int
%int1792_16497 = torch.constant.int 1792
%18867 = torch.prim.ListConstruct %18866, %int1792_16497 : (!torch.int, !torch.int) -> !torch.list<int>
%18868 = torch.aten.view %18841, %18867 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18868, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18869 = torch.aten.mm %18868, %18850 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18869, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16498 = torch.constant.int 4
%int4096_16499 = torch.constant.int 4096
%18870 = torch.prim.ListConstruct %int4_16498, %18865, %int4096_16499 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18871 = torch.aten.view %18869, %18870 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16500 = torch.constant.int 1
%18872 = torch.aten.size.int %18732, %int1_16500 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16501 = torch.constant.int 4
%18873 = torch.aten.mul.int %int4_16501, %18872 : !torch.int, !torch.int -> !torch.int
%int1792_16502 = torch.constant.int 1792
%18874 = torch.prim.ListConstruct %18873, %int1792_16502 : (!torch.int, !torch.int) -> !torch.list<int>
%18875 = torch.aten.view %18842, %18874 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18875, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18876 = torch.aten.mm %18875, %18852 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18876, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16503 = torch.constant.int 4
%int4096_16504 = torch.constant.int 4096
%18877 = torch.prim.ListConstruct %int4_16503, %18872, %int4096_16504 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18878 = torch.aten.view %18876, %18877 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16505 = torch.constant.int 1
%18879 = torch.aten.size.int %18738, %int1_16505 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16506 = torch.constant.int 4
%18880 = torch.aten.mul.int %int4_16506, %18879 : !torch.int, !torch.int -> !torch.int
%int1792_16507 = torch.constant.int 1792
%18881 = torch.prim.ListConstruct %18880, %int1792_16507 : (!torch.int, !torch.int) -> !torch.list<int>
%18882 = torch.aten.view %18843, %18881 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18882, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18883 = torch.aten.mm %18882, %18854 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18883, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16508 = torch.constant.int 4
%int4096_16509 = torch.constant.int 4096
%18884 = torch.prim.ListConstruct %int4_16508, %18879, %int4096_16509 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18885 = torch.aten.view %18883, %18884 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16510 = torch.constant.int 1
%18886 = torch.aten.size.int %18744, %int1_16510 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16511 = torch.constant.int 4
%18887 = torch.aten.mul.int %int4_16511, %18886 : !torch.int, !torch.int -> !torch.int
%int1792_16512 = torch.constant.int 1792
%18888 = torch.prim.ListConstruct %18887, %int1792_16512 : (!torch.int, !torch.int) -> !torch.list<int>
%18889 = torch.aten.view %18844, %18888 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18889, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18890 = torch.aten.mm %18889, %18856 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18890, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16513 = torch.constant.int 4
%int4096_16514 = torch.constant.int 4096
%18891 = torch.prim.ListConstruct %int4_16513, %18886, %int4096_16514 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18892 = torch.aten.view %18890, %18891 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16515 = torch.constant.int 1
%18893 = torch.aten.size.int %18750, %int1_16515 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16516 = torch.constant.int 4
%18894 = torch.aten.mul.int %int4_16516, %18893 : !torch.int, !torch.int -> !torch.int
%int1792_16517 = torch.constant.int 1792
%18895 = torch.prim.ListConstruct %18894, %int1792_16517 : (!torch.int, !torch.int) -> !torch.list<int>
%18896 = torch.aten.view %18845, %18895 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18896, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18897 = torch.aten.mm %18896, %18858 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18897, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16518 = torch.constant.int 4
%int4096_16519 = torch.constant.int 4096
%18898 = torch.prim.ListConstruct %int4_16518, %18893, %int4096_16519 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18899 = torch.aten.view %18897, %18898 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16520 = torch.constant.int 1
%18900 = torch.aten.size.int %18756, %int1_16520 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16521 = torch.constant.int 4
%18901 = torch.aten.mul.int %int4_16521, %18900 : !torch.int, !torch.int -> !torch.int
%int1792_16522 = torch.constant.int 1792
%18902 = torch.prim.ListConstruct %18901, %int1792_16522 : (!torch.int, !torch.int) -> !torch.list<int>
%18903 = torch.aten.view %18846, %18902 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18903, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18904 = torch.aten.mm %18903, %18860 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18904, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16523 = torch.constant.int 4
%int4096_16524 = torch.constant.int 4096
%18905 = torch.prim.ListConstruct %int4_16523, %18900, %int4096_16524 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18906 = torch.aten.view %18904, %18905 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16525 = torch.constant.int 1
%18907 = torch.aten.size.int %18762, %int1_16525 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16526 = torch.constant.int 4
%18908 = torch.aten.mul.int %int4_16526, %18907 : !torch.int, !torch.int -> !torch.int
%int1792_16527 = torch.constant.int 1792
%18909 = torch.prim.ListConstruct %18908, %int1792_16527 : (!torch.int, !torch.int) -> !torch.list<int>
%18910 = torch.aten.view %18847, %18909 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18910, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18911 = torch.aten.mm %18910, %18862 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18911, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16528 = torch.constant.int 4
%int4096_16529 = torch.constant.int 4096
%18912 = torch.prim.ListConstruct %int4_16528, %18907, %int4096_16529 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18913 = torch.aten.view %18911, %18912 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16530 = torch.constant.int 1
%18914 = torch.aten.size.int %18768, %int1_16530 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_16531 = torch.constant.int 4
%18915 = torch.aten.mul.int %int4_16531, %18914 : !torch.int, !torch.int -> !torch.int
%int1792_16532 = torch.constant.int 1792
%18916 = torch.prim.ListConstruct %18915, %int1792_16532 : (!torch.int, !torch.int) -> !torch.list<int>
%18917 = torch.aten.view %18848, %18916 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %18917, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%18918 = torch.aten.mm %18917, %18864 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %18918, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_16533 = torch.constant.int 4
%int4096_16534 = torch.constant.int 4096
%18919 = torch.prim.ListConstruct %int4_16533, %18914, %int4096_16534 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%18920 = torch.aten.view %18918, %18919 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18921 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16535 = arith.constant 1 : index
%dim_16536 = tensor.dim %18921, %c1_16535 : tensor<4x?x4096xf16>
%18922 = flow.tensor.transfer %18921 : tensor<4x?x4096xf16>{%dim_16536} to #hal.device.promise<@__device_0>
%18923 = torch_c.from_builtin_tensor %18922 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18924 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16537 = arith.constant 1 : index
%dim_16538 = tensor.dim %18924, %c1_16537 : tensor<4x?x4096xf16>
%18925 = flow.tensor.transfer %18924 : tensor<4x?x4096xf16>{%dim_16538} to #hal.device.promise<@__device_0>
%18926 = torch_c.from_builtin_tensor %18925 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18927 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16539 = arith.constant 1 : index
%dim_16540 = tensor.dim %18927, %c1_16539 : tensor<4x?x4096xf16>
%18928 = flow.tensor.transfer %18927 : tensor<4x?x4096xf16>{%dim_16540} to #hal.device.promise<@__device_0>
%18929 = torch_c.from_builtin_tensor %18928 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18930 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16541 = arith.constant 1 : index
%dim_16542 = tensor.dim %18930, %c1_16541 : tensor<4x?x4096xf16>
%18931 = flow.tensor.transfer %18930 : tensor<4x?x4096xf16>{%dim_16542} to #hal.device.promise<@__device_0>
%18932 = torch_c.from_builtin_tensor %18931 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18933 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16543 = arith.constant 1 : index
%dim_16544 = tensor.dim %18933, %c1_16543 : tensor<4x?x4096xf16>
%18934 = flow.tensor.transfer %18933 : tensor<4x?x4096xf16>{%dim_16544} to #hal.device.promise<@__device_0>
%18935 = torch_c.from_builtin_tensor %18934 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18936 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16545 = arith.constant 1 : index
%dim_16546 = tensor.dim %18936, %c1_16545 : tensor<4x?x4096xf16>
%18937 = flow.tensor.transfer %18936 : tensor<4x?x4096xf16>{%dim_16546} to #hal.device.promise<@__device_0>
%18938 = torch_c.from_builtin_tensor %18937 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18939 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16547 = arith.constant 1 : index
%dim_16548 = tensor.dim %18939, %c1_16547 : tensor<4x?x4096xf16>
%18940 = flow.tensor.transfer %18939 : tensor<4x?x4096xf16>{%dim_16548} to #hal.device.promise<@__device_0>
%18941 = torch_c.from_builtin_tensor %18940 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16549 = torch.constant.int 1
%18942 = torch.aten.add.Tensor %18871, %18923, %int1_16549 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16550 = torch.constant.int 1
%18943 = torch.aten.add.Tensor %18942, %18926, %int1_16550 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16551 = torch.constant.int 1
%18944 = torch.aten.add.Tensor %18943, %18929, %int1_16551 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16552 = torch.constant.int 1
%18945 = torch.aten.add.Tensor %18944, %18932, %int1_16552 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16553 = torch.constant.int 1
%18946 = torch.aten.add.Tensor %18945, %18935, %int1_16553 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16554 = torch.constant.int 1
%18947 = torch.aten.add.Tensor %18946, %18938, %int1_16554 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16555 = torch.constant.int 1
%18948 = torch.aten.add.Tensor %18947, %18941, %int1_16555 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18949 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16556 = arith.constant 1 : index
%dim_16557 = tensor.dim %18949, %c1_16556 : tensor<4x?x4096xf16>
%18950 = flow.tensor.transfer %18949 : tensor<4x?x4096xf16>{%dim_16557} to #hal.device.promise<@__device_1>
%18951 = torch_c.from_builtin_tensor %18950 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18952 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16558 = arith.constant 1 : index
%dim_16559 = tensor.dim %18952, %c1_16558 : tensor<4x?x4096xf16>
%18953 = flow.tensor.transfer %18952 : tensor<4x?x4096xf16>{%dim_16559} to #hal.device.promise<@__device_1>
%18954 = torch_c.from_builtin_tensor %18953 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18955 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16560 = arith.constant 1 : index
%dim_16561 = tensor.dim %18955, %c1_16560 : tensor<4x?x4096xf16>
%18956 = flow.tensor.transfer %18955 : tensor<4x?x4096xf16>{%dim_16561} to #hal.device.promise<@__device_1>
%18957 = torch_c.from_builtin_tensor %18956 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18958 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16562 = arith.constant 1 : index
%dim_16563 = tensor.dim %18958, %c1_16562 : tensor<4x?x4096xf16>
%18959 = flow.tensor.transfer %18958 : tensor<4x?x4096xf16>{%dim_16563} to #hal.device.promise<@__device_1>
%18960 = torch_c.from_builtin_tensor %18959 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18961 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16564 = arith.constant 1 : index
%dim_16565 = tensor.dim %18961, %c1_16564 : tensor<4x?x4096xf16>
%18962 = flow.tensor.transfer %18961 : tensor<4x?x4096xf16>{%dim_16565} to #hal.device.promise<@__device_1>
%18963 = torch_c.from_builtin_tensor %18962 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18964 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16566 = arith.constant 1 : index
%dim_16567 = tensor.dim %18964, %c1_16566 : tensor<4x?x4096xf16>
%18965 = flow.tensor.transfer %18964 : tensor<4x?x4096xf16>{%dim_16567} to #hal.device.promise<@__device_1>
%18966 = torch_c.from_builtin_tensor %18965 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18967 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16568 = arith.constant 1 : index
%dim_16569 = tensor.dim %18967, %c1_16568 : tensor<4x?x4096xf16>
%18968 = flow.tensor.transfer %18967 : tensor<4x?x4096xf16>{%dim_16569} to #hal.device.promise<@__device_1>
%18969 = torch_c.from_builtin_tensor %18968 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16570 = torch.constant.int 1
%18970 = torch.aten.add.Tensor %18951, %18878, %int1_16570 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16571 = torch.constant.int 1
%18971 = torch.aten.add.Tensor %18970, %18954, %int1_16571 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16572 = torch.constant.int 1
%18972 = torch.aten.add.Tensor %18971, %18957, %int1_16572 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16573 = torch.constant.int 1
%18973 = torch.aten.add.Tensor %18972, %18960, %int1_16573 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16574 = torch.constant.int 1
%18974 = torch.aten.add.Tensor %18973, %18963, %int1_16574 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16575 = torch.constant.int 1
%18975 = torch.aten.add.Tensor %18974, %18966, %int1_16575 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16576 = torch.constant.int 1
%18976 = torch.aten.add.Tensor %18975, %18969, %int1_16576 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18977 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16577 = arith.constant 1 : index
%dim_16578 = tensor.dim %18977, %c1_16577 : tensor<4x?x4096xf16>
%18978 = flow.tensor.transfer %18977 : tensor<4x?x4096xf16>{%dim_16578} to #hal.device.promise<@__device_2>
%18979 = torch_c.from_builtin_tensor %18978 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18980 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16579 = arith.constant 1 : index
%dim_16580 = tensor.dim %18980, %c1_16579 : tensor<4x?x4096xf16>
%18981 = flow.tensor.transfer %18980 : tensor<4x?x4096xf16>{%dim_16580} to #hal.device.promise<@__device_2>
%18982 = torch_c.from_builtin_tensor %18981 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18983 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16581 = arith.constant 1 : index
%dim_16582 = tensor.dim %18983, %c1_16581 : tensor<4x?x4096xf16>
%18984 = flow.tensor.transfer %18983 : tensor<4x?x4096xf16>{%dim_16582} to #hal.device.promise<@__device_2>
%18985 = torch_c.from_builtin_tensor %18984 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18985, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18986 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16583 = arith.constant 1 : index
%dim_16584 = tensor.dim %18986, %c1_16583 : tensor<4x?x4096xf16>
%18987 = flow.tensor.transfer %18986 : tensor<4x?x4096xf16>{%dim_16584} to #hal.device.promise<@__device_2>
%18988 = torch_c.from_builtin_tensor %18987 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18989 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16585 = arith.constant 1 : index
%dim_16586 = tensor.dim %18989, %c1_16585 : tensor<4x?x4096xf16>
%18990 = flow.tensor.transfer %18989 : tensor<4x?x4096xf16>{%dim_16586} to #hal.device.promise<@__device_2>
%18991 = torch_c.from_builtin_tensor %18990 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18992 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16587 = arith.constant 1 : index
%dim_16588 = tensor.dim %18992, %c1_16587 : tensor<4x?x4096xf16>
%18993 = flow.tensor.transfer %18992 : tensor<4x?x4096xf16>{%dim_16588} to #hal.device.promise<@__device_2>
%18994 = torch_c.from_builtin_tensor %18993 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%18995 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16589 = arith.constant 1 : index
%dim_16590 = tensor.dim %18995, %c1_16589 : tensor<4x?x4096xf16>
%18996 = flow.tensor.transfer %18995 : tensor<4x?x4096xf16>{%dim_16590} to #hal.device.promise<@__device_2>
%18997 = torch_c.from_builtin_tensor %18996 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16591 = torch.constant.int 1
%18998 = torch.aten.add.Tensor %18979, %18982, %int1_16591 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16592 = torch.constant.int 1
%18999 = torch.aten.add.Tensor %18998, %18885, %int1_16592 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %18999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16593 = torch.constant.int 1
%19000 = torch.aten.add.Tensor %18999, %18985, %int1_16593 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16594 = torch.constant.int 1
%19001 = torch.aten.add.Tensor %19000, %18988, %int1_16594 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16595 = torch.constant.int 1
%19002 = torch.aten.add.Tensor %19001, %18991, %int1_16595 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16596 = torch.constant.int 1
%19003 = torch.aten.add.Tensor %19002, %18994, %int1_16596 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16597 = torch.constant.int 1
%19004 = torch.aten.add.Tensor %19003, %18997, %int1_16597 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19005 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16598 = arith.constant 1 : index
%dim_16599 = tensor.dim %19005, %c1_16598 : tensor<4x?x4096xf16>
%19006 = flow.tensor.transfer %19005 : tensor<4x?x4096xf16>{%dim_16599} to #hal.device.promise<@__device_3>
%19007 = torch_c.from_builtin_tensor %19006 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19008 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16600 = arith.constant 1 : index
%dim_16601 = tensor.dim %19008, %c1_16600 : tensor<4x?x4096xf16>
%19009 = flow.tensor.transfer %19008 : tensor<4x?x4096xf16>{%dim_16601} to #hal.device.promise<@__device_3>
%19010 = torch_c.from_builtin_tensor %19009 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19011 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16602 = arith.constant 1 : index
%dim_16603 = tensor.dim %19011, %c1_16602 : tensor<4x?x4096xf16>
%19012 = flow.tensor.transfer %19011 : tensor<4x?x4096xf16>{%dim_16603} to #hal.device.promise<@__device_3>
%19013 = torch_c.from_builtin_tensor %19012 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19014 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16604 = arith.constant 1 : index
%dim_16605 = tensor.dim %19014, %c1_16604 : tensor<4x?x4096xf16>
%19015 = flow.tensor.transfer %19014 : tensor<4x?x4096xf16>{%dim_16605} to #hal.device.promise<@__device_3>
%19016 = torch_c.from_builtin_tensor %19015 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19017 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16606 = arith.constant 1 : index
%dim_16607 = tensor.dim %19017, %c1_16606 : tensor<4x?x4096xf16>
%19018 = flow.tensor.transfer %19017 : tensor<4x?x4096xf16>{%dim_16607} to #hal.device.promise<@__device_3>
%19019 = torch_c.from_builtin_tensor %19018 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19020 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16608 = arith.constant 1 : index
%dim_16609 = tensor.dim %19020, %c1_16608 : tensor<4x?x4096xf16>
%19021 = flow.tensor.transfer %19020 : tensor<4x?x4096xf16>{%dim_16609} to #hal.device.promise<@__device_3>
%19022 = torch_c.from_builtin_tensor %19021 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19023 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16610 = arith.constant 1 : index
%dim_16611 = tensor.dim %19023, %c1_16610 : tensor<4x?x4096xf16>
%19024 = flow.tensor.transfer %19023 : tensor<4x?x4096xf16>{%dim_16611} to #hal.device.promise<@__device_3>
%19025 = torch_c.from_builtin_tensor %19024 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16612 = torch.constant.int 1
%19026 = torch.aten.add.Tensor %19007, %19010, %int1_16612 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16613 = torch.constant.int 1
%19027 = torch.aten.add.Tensor %19026, %19013, %int1_16613 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16614 = torch.constant.int 1
%19028 = torch.aten.add.Tensor %19027, %18892, %int1_16614 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16615 = torch.constant.int 1
%19029 = torch.aten.add.Tensor %19028, %19016, %int1_16615 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16616 = torch.constant.int 1
%19030 = torch.aten.add.Tensor %19029, %19019, %int1_16616 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16617 = torch.constant.int 1
%19031 = torch.aten.add.Tensor %19030, %19022, %int1_16617 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16618 = torch.constant.int 1
%19032 = torch.aten.add.Tensor %19031, %19025, %int1_16618 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19033 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16619 = arith.constant 1 : index
%dim_16620 = tensor.dim %19033, %c1_16619 : tensor<4x?x4096xf16>
%19034 = flow.tensor.transfer %19033 : tensor<4x?x4096xf16>{%dim_16620} to #hal.device.promise<@__device_4>
%19035 = torch_c.from_builtin_tensor %19034 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19036 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16621 = arith.constant 1 : index
%dim_16622 = tensor.dim %19036, %c1_16621 : tensor<4x?x4096xf16>
%19037 = flow.tensor.transfer %19036 : tensor<4x?x4096xf16>{%dim_16622} to #hal.device.promise<@__device_4>
%19038 = torch_c.from_builtin_tensor %19037 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19039 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16623 = arith.constant 1 : index
%dim_16624 = tensor.dim %19039, %c1_16623 : tensor<4x?x4096xf16>
%19040 = flow.tensor.transfer %19039 : tensor<4x?x4096xf16>{%dim_16624} to #hal.device.promise<@__device_4>
%19041 = torch_c.from_builtin_tensor %19040 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19042 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16625 = arith.constant 1 : index
%dim_16626 = tensor.dim %19042, %c1_16625 : tensor<4x?x4096xf16>
%19043 = flow.tensor.transfer %19042 : tensor<4x?x4096xf16>{%dim_16626} to #hal.device.promise<@__device_4>
%19044 = torch_c.from_builtin_tensor %19043 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19045 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16627 = arith.constant 1 : index
%dim_16628 = tensor.dim %19045, %c1_16627 : tensor<4x?x4096xf16>
%19046 = flow.tensor.transfer %19045 : tensor<4x?x4096xf16>{%dim_16628} to #hal.device.promise<@__device_4>
%19047 = torch_c.from_builtin_tensor %19046 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19048 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16629 = arith.constant 1 : index
%dim_16630 = tensor.dim %19048, %c1_16629 : tensor<4x?x4096xf16>
%19049 = flow.tensor.transfer %19048 : tensor<4x?x4096xf16>{%dim_16630} to #hal.device.promise<@__device_4>
%19050 = torch_c.from_builtin_tensor %19049 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19051 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16631 = arith.constant 1 : index
%dim_16632 = tensor.dim %19051, %c1_16631 : tensor<4x?x4096xf16>
%19052 = flow.tensor.transfer %19051 : tensor<4x?x4096xf16>{%dim_16632} to #hal.device.promise<@__device_4>
%19053 = torch_c.from_builtin_tensor %19052 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16633 = torch.constant.int 1
%19054 = torch.aten.add.Tensor %19035, %19038, %int1_16633 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16634 = torch.constant.int 1
%19055 = torch.aten.add.Tensor %19054, %19041, %int1_16634 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16635 = torch.constant.int 1
%19056 = torch.aten.add.Tensor %19055, %19044, %int1_16635 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16636 = torch.constant.int 1
%19057 = torch.aten.add.Tensor %19056, %18899, %int1_16636 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16637 = torch.constant.int 1
%19058 = torch.aten.add.Tensor %19057, %19047, %int1_16637 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16638 = torch.constant.int 1
%19059 = torch.aten.add.Tensor %19058, %19050, %int1_16638 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16639 = torch.constant.int 1
%19060 = torch.aten.add.Tensor %19059, %19053, %int1_16639 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19061 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16640 = arith.constant 1 : index
%dim_16641 = tensor.dim %19061, %c1_16640 : tensor<4x?x4096xf16>
%19062 = flow.tensor.transfer %19061 : tensor<4x?x4096xf16>{%dim_16641} to #hal.device.promise<@__device_5>
%19063 = torch_c.from_builtin_tensor %19062 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19064 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16642 = arith.constant 1 : index
%dim_16643 = tensor.dim %19064, %c1_16642 : tensor<4x?x4096xf16>
%19065 = flow.tensor.transfer %19064 : tensor<4x?x4096xf16>{%dim_16643} to #hal.device.promise<@__device_5>
%19066 = torch_c.from_builtin_tensor %19065 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19067 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16644 = arith.constant 1 : index
%dim_16645 = tensor.dim %19067, %c1_16644 : tensor<4x?x4096xf16>
%19068 = flow.tensor.transfer %19067 : tensor<4x?x4096xf16>{%dim_16645} to #hal.device.promise<@__device_5>
%19069 = torch_c.from_builtin_tensor %19068 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19070 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16646 = arith.constant 1 : index
%dim_16647 = tensor.dim %19070, %c1_16646 : tensor<4x?x4096xf16>
%19071 = flow.tensor.transfer %19070 : tensor<4x?x4096xf16>{%dim_16647} to #hal.device.promise<@__device_5>
%19072 = torch_c.from_builtin_tensor %19071 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19073 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16648 = arith.constant 1 : index
%dim_16649 = tensor.dim %19073, %c1_16648 : tensor<4x?x4096xf16>
%19074 = flow.tensor.transfer %19073 : tensor<4x?x4096xf16>{%dim_16649} to #hal.device.promise<@__device_5>
%19075 = torch_c.from_builtin_tensor %19074 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19076 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16650 = arith.constant 1 : index
%dim_16651 = tensor.dim %19076, %c1_16650 : tensor<4x?x4096xf16>
%19077 = flow.tensor.transfer %19076 : tensor<4x?x4096xf16>{%dim_16651} to #hal.device.promise<@__device_5>
%19078 = torch_c.from_builtin_tensor %19077 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19079 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16652 = arith.constant 1 : index
%dim_16653 = tensor.dim %19079, %c1_16652 : tensor<4x?x4096xf16>
%19080 = flow.tensor.transfer %19079 : tensor<4x?x4096xf16>{%dim_16653} to #hal.device.promise<@__device_5>
%19081 = torch_c.from_builtin_tensor %19080 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16654 = torch.constant.int 1
%19082 = torch.aten.add.Tensor %19063, %19066, %int1_16654 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16655 = torch.constant.int 1
%19083 = torch.aten.add.Tensor %19082, %19069, %int1_16655 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16656 = torch.constant.int 1
%19084 = torch.aten.add.Tensor %19083, %19072, %int1_16656 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16657 = torch.constant.int 1
%19085 = torch.aten.add.Tensor %19084, %19075, %int1_16657 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16658 = torch.constant.int 1
%19086 = torch.aten.add.Tensor %19085, %18906, %int1_16658 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16659 = torch.constant.int 1
%19087 = torch.aten.add.Tensor %19086, %19078, %int1_16659 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16660 = torch.constant.int 1
%19088 = torch.aten.add.Tensor %19087, %19081, %int1_16660 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19089 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16661 = arith.constant 1 : index
%dim_16662 = tensor.dim %19089, %c1_16661 : tensor<4x?x4096xf16>
%19090 = flow.tensor.transfer %19089 : tensor<4x?x4096xf16>{%dim_16662} to #hal.device.promise<@__device_6>
%19091 = torch_c.from_builtin_tensor %19090 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19092 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16663 = arith.constant 1 : index
%dim_16664 = tensor.dim %19092, %c1_16663 : tensor<4x?x4096xf16>
%19093 = flow.tensor.transfer %19092 : tensor<4x?x4096xf16>{%dim_16664} to #hal.device.promise<@__device_6>
%19094 = torch_c.from_builtin_tensor %19093 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19095 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16665 = arith.constant 1 : index
%dim_16666 = tensor.dim %19095, %c1_16665 : tensor<4x?x4096xf16>
%19096 = flow.tensor.transfer %19095 : tensor<4x?x4096xf16>{%dim_16666} to #hal.device.promise<@__device_6>
%19097 = torch_c.from_builtin_tensor %19096 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19098 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16667 = arith.constant 1 : index
%dim_16668 = tensor.dim %19098, %c1_16667 : tensor<4x?x4096xf16>
%19099 = flow.tensor.transfer %19098 : tensor<4x?x4096xf16>{%dim_16668} to #hal.device.promise<@__device_6>
%19100 = torch_c.from_builtin_tensor %19099 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19101 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16669 = arith.constant 1 : index
%dim_16670 = tensor.dim %19101, %c1_16669 : tensor<4x?x4096xf16>
%19102 = flow.tensor.transfer %19101 : tensor<4x?x4096xf16>{%dim_16670} to #hal.device.promise<@__device_6>
%19103 = torch_c.from_builtin_tensor %19102 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19104 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16671 = arith.constant 1 : index
%dim_16672 = tensor.dim %19104, %c1_16671 : tensor<4x?x4096xf16>
%19105 = flow.tensor.transfer %19104 : tensor<4x?x4096xf16>{%dim_16672} to #hal.device.promise<@__device_6>
%19106 = torch_c.from_builtin_tensor %19105 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19107 = torch_c.to_builtin_tensor %18920 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16673 = arith.constant 1 : index
%dim_16674 = tensor.dim %19107, %c1_16673 : tensor<4x?x4096xf16>
%19108 = flow.tensor.transfer %19107 : tensor<4x?x4096xf16>{%dim_16674} to #hal.device.promise<@__device_6>
%19109 = torch_c.from_builtin_tensor %19108 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16675 = torch.constant.int 1
%19110 = torch.aten.add.Tensor %19091, %19094, %int1_16675 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16676 = torch.constant.int 1
%19111 = torch.aten.add.Tensor %19110, %19097, %int1_16676 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16677 = torch.constant.int 1
%19112 = torch.aten.add.Tensor %19111, %19100, %int1_16677 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16678 = torch.constant.int 1
%19113 = torch.aten.add.Tensor %19112, %19103, %int1_16678 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16679 = torch.constant.int 1
%19114 = torch.aten.add.Tensor %19113, %19106, %int1_16679 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16680 = torch.constant.int 1
%19115 = torch.aten.add.Tensor %19114, %18913, %int1_16680 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16681 = torch.constant.int 1
%19116 = torch.aten.add.Tensor %19115, %19109, %int1_16681 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19117 = torch_c.to_builtin_tensor %18871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16682 = arith.constant 1 : index
%dim_16683 = tensor.dim %19117, %c1_16682 : tensor<4x?x4096xf16>
%19118 = flow.tensor.transfer %19117 : tensor<4x?x4096xf16>{%dim_16683} to #hal.device.promise<@__device_7>
%19119 = torch_c.from_builtin_tensor %19118 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19120 = torch_c.to_builtin_tensor %18878 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16684 = arith.constant 1 : index
%dim_16685 = tensor.dim %19120, %c1_16684 : tensor<4x?x4096xf16>
%19121 = flow.tensor.transfer %19120 : tensor<4x?x4096xf16>{%dim_16685} to #hal.device.promise<@__device_7>
%19122 = torch_c.from_builtin_tensor %19121 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19123 = torch_c.to_builtin_tensor %18885 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16686 = arith.constant 1 : index
%dim_16687 = tensor.dim %19123, %c1_16686 : tensor<4x?x4096xf16>
%19124 = flow.tensor.transfer %19123 : tensor<4x?x4096xf16>{%dim_16687} to #hal.device.promise<@__device_7>
%19125 = torch_c.from_builtin_tensor %19124 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19126 = torch_c.to_builtin_tensor %18892 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16688 = arith.constant 1 : index
%dim_16689 = tensor.dim %19126, %c1_16688 : tensor<4x?x4096xf16>
%19127 = flow.tensor.transfer %19126 : tensor<4x?x4096xf16>{%dim_16689} to #hal.device.promise<@__device_7>
%19128 = torch_c.from_builtin_tensor %19127 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19129 = torch_c.to_builtin_tensor %18899 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16690 = arith.constant 1 : index
%dim_16691 = tensor.dim %19129, %c1_16690 : tensor<4x?x4096xf16>
%19130 = flow.tensor.transfer %19129 : tensor<4x?x4096xf16>{%dim_16691} to #hal.device.promise<@__device_7>
%19131 = torch_c.from_builtin_tensor %19130 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19132 = torch_c.to_builtin_tensor %18906 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16692 = arith.constant 1 : index
%dim_16693 = tensor.dim %19132, %c1_16692 : tensor<4x?x4096xf16>
%19133 = flow.tensor.transfer %19132 : tensor<4x?x4096xf16>{%dim_16693} to #hal.device.promise<@__device_7>
%19134 = torch_c.from_builtin_tensor %19133 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%19135 = torch_c.to_builtin_tensor %18913 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_16694 = arith.constant 1 : index
%dim_16695 = tensor.dim %19135, %c1_16694 : tensor<4x?x4096xf16>
%19136 = flow.tensor.transfer %19135 : tensor<4x?x4096xf16>{%dim_16695} to #hal.device.promise<@__device_7>
%19137 = torch_c.from_builtin_tensor %19136 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16696 = torch.constant.int 1
%19138 = torch.aten.add.Tensor %19119, %19122, %int1_16696 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16697 = torch.constant.int 1
%19139 = torch.aten.add.Tensor %19138, %19125, %int1_16697 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16698 = torch.constant.int 1
%19140 = torch.aten.add.Tensor %19139, %19128, %int1_16698 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16699 = torch.constant.int 1
%19141 = torch.aten.add.Tensor %19140, %19131, %int1_16699 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16700 = torch.constant.int 1
%19142 = torch.aten.add.Tensor %19141, %19134, %int1_16700 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16701 = torch.constant.int 1
%19143 = torch.aten.add.Tensor %19142, %19137, %int1_16701 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16702 = torch.constant.int 1
%19144 = torch.aten.add.Tensor %19143, %18920, %int1_16702 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16703 = torch.constant.int 1
%19145 = torch.aten.add.Tensor %18625, %18948, %int1_16703 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16704 = torch.constant.int 1
%19146 = torch.aten.add.Tensor %18626, %18976, %int1_16704 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16705 = torch.constant.int 1
%19147 = torch.aten.add.Tensor %18627, %19004, %int1_16705 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16706 = torch.constant.int 1
%19148 = torch.aten.add.Tensor %18628, %19032, %int1_16706 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16707 = torch.constant.int 1
%19149 = torch.aten.add.Tensor %18629, %19060, %int1_16707 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16708 = torch.constant.int 1
%19150 = torch.aten.add.Tensor %18630, %19088, %int1_16708 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16709 = torch.constant.int 1
%19151 = torch.aten.add.Tensor %18631, %19116, %int1_16709 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16710 = torch.constant.int 1
%19152 = torch.aten.add.Tensor %18632, %19144, %int1_16710 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_16711 = torch.constant.int 6
%19153 = torch.prims.convert_element_type %19145, %int6_16711 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16712 = torch.constant.int 6
%19154 = torch.prims.convert_element_type %19146, %int6_16712 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16713 = torch.constant.int 6
%19155 = torch.prims.convert_element_type %19147, %int6_16713 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16714 = torch.constant.int 6
%19156 = torch.prims.convert_element_type %19148, %int6_16714 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16715 = torch.constant.int 6
%19157 = torch.prims.convert_element_type %19149, %int6_16715 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16716 = torch.constant.int 6
%19158 = torch.prims.convert_element_type %19150, %int6_16716 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16717 = torch.constant.int 6
%19159 = torch.prims.convert_element_type %19151, %int6_16717 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_16718 = torch.constant.int 6
%19160 = torch.prims.convert_element_type %19152, %int6_16718 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16719 = torch.constant.int 2
%19161 = torch.aten.pow.Tensor_Scalar %19153, %int2_16719 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16720 = torch.constant.int 2
%19162 = torch.aten.pow.Tensor_Scalar %19154, %int2_16720 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16721 = torch.constant.int 2
%19163 = torch.aten.pow.Tensor_Scalar %19155, %int2_16721 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16722 = torch.constant.int 2
%19164 = torch.aten.pow.Tensor_Scalar %19156, %int2_16722 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16723 = torch.constant.int 2
%19165 = torch.aten.pow.Tensor_Scalar %19157, %int2_16723 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16724 = torch.constant.int 2
%19166 = torch.aten.pow.Tensor_Scalar %19158, %int2_16724 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16725 = torch.constant.int 2
%19167 = torch.aten.pow.Tensor_Scalar %19159, %int2_16725 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_16726 = torch.constant.int 2
%19168 = torch.aten.pow.Tensor_Scalar %19160, %int2_16726 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_16727 = torch.constant.int -1
%19169 = torch.prim.ListConstruct %int-1_16727 : (!torch.int) -> !torch.list<int>
%true_16728 = torch.constant.bool true
%none_16729 = torch.constant.none
%19170 = torch.aten.mean.dim %19161, %19169, %true_16728, %none_16729 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16730 = torch.constant.int -1
%19171 = torch.prim.ListConstruct %int-1_16730 : (!torch.int) -> !torch.list<int>
%true_16731 = torch.constant.bool true
%none_16732 = torch.constant.none
%19172 = torch.aten.mean.dim %19162, %19171, %true_16731, %none_16732 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16733 = torch.constant.int -1
%19173 = torch.prim.ListConstruct %int-1_16733 : (!torch.int) -> !torch.list<int>
%true_16734 = torch.constant.bool true
%none_16735 = torch.constant.none
%19174 = torch.aten.mean.dim %19163, %19173, %true_16734, %none_16735 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16736 = torch.constant.int -1
%19175 = torch.prim.ListConstruct %int-1_16736 : (!torch.int) -> !torch.list<int>
%true_16737 = torch.constant.bool true
%none_16738 = torch.constant.none
%19176 = torch.aten.mean.dim %19164, %19175, %true_16737, %none_16738 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16739 = torch.constant.int -1
%19177 = torch.prim.ListConstruct %int-1_16739 : (!torch.int) -> !torch.list<int>
%true_16740 = torch.constant.bool true
%none_16741 = torch.constant.none
%19178 = torch.aten.mean.dim %19165, %19177, %true_16740, %none_16741 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16742 = torch.constant.int -1
%19179 = torch.prim.ListConstruct %int-1_16742 : (!torch.int) -> !torch.list<int>
%true_16743 = torch.constant.bool true
%none_16744 = torch.constant.none
%19180 = torch.aten.mean.dim %19166, %19179, %true_16743, %none_16744 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16745 = torch.constant.int -1
%19181 = torch.prim.ListConstruct %int-1_16745 : (!torch.int) -> !torch.list<int>
%true_16746 = torch.constant.bool true
%none_16747 = torch.constant.none
%19182 = torch.aten.mean.dim %19167, %19181, %true_16746, %none_16747 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_16748 = torch.constant.int -1
%19183 = torch.prim.ListConstruct %int-1_16748 : (!torch.int) -> !torch.list<int>
%true_16749 = torch.constant.bool true
%none_16750 = torch.constant.none
%19184 = torch.aten.mean.dim %19168, %19183, %true_16749, %none_16750 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16751 = torch.constant.float 9.9999997473787516E-6
%int1_16752 = torch.constant.int 1
%19185 = torch.aten.add.Scalar %19170, %float9.999990e-06_16751, %int1_16752 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16753 = torch.constant.float 9.9999997473787516E-6
%int1_16754 = torch.constant.int 1
%19186 = torch.aten.add.Scalar %19172, %float9.999990e-06_16753, %int1_16754 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16755 = torch.constant.float 9.9999997473787516E-6
%int1_16756 = torch.constant.int 1
%19187 = torch.aten.add.Scalar %19174, %float9.999990e-06_16755, %int1_16756 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16757 = torch.constant.float 9.9999997473787516E-6
%int1_16758 = torch.constant.int 1
%19188 = torch.aten.add.Scalar %19176, %float9.999990e-06_16757, %int1_16758 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16759 = torch.constant.float 9.9999997473787516E-6
%int1_16760 = torch.constant.int 1
%19189 = torch.aten.add.Scalar %19178, %float9.999990e-06_16759, %int1_16760 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16761 = torch.constant.float 9.9999997473787516E-6
%int1_16762 = torch.constant.int 1
%19190 = torch.aten.add.Scalar %19180, %float9.999990e-06_16761, %int1_16762 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16763 = torch.constant.float 9.9999997473787516E-6
%int1_16764 = torch.constant.int 1
%19191 = torch.aten.add.Scalar %19182, %float9.999990e-06_16763, %int1_16764 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_16765 = torch.constant.float 9.9999997473787516E-6
%int1_16766 = torch.constant.int 1
%19192 = torch.aten.add.Scalar %19184, %float9.999990e-06_16765, %int1_16766 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19193 = torch.aten.rsqrt %19185 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19194 = torch.aten.rsqrt %19186 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19195 = torch.aten.rsqrt %19187 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19196 = torch.aten.rsqrt %19188 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19197 = torch.aten.rsqrt %19189 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19198 = torch.aten.rsqrt %19190 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19199 = torch.aten.rsqrt %19191 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19200 = torch.aten.rsqrt %19192 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %19200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%19201 = torch.aten.mul.Tensor %19153, %19193 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19202 = torch.aten.mul.Tensor %19154, %19194 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19203 = torch.aten.mul.Tensor %19155, %19195 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19204 = torch.aten.mul.Tensor %19156, %19196 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19205 = torch.aten.mul.Tensor %19157, %19197 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19206 = torch.aten.mul.Tensor %19158, %19198 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19207 = torch.aten.mul.Tensor %19159, %19199 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19208 = torch.aten.mul.Tensor %19160, %19200 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19209 = torch.aten.mul.Tensor %656, %19201 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19210 = torch.aten.mul.Tensor %657, %19202 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19211 = torch.aten.mul.Tensor %658, %19203 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19212 = torch.aten.mul.Tensor %659, %19204 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19213 = torch.aten.mul.Tensor %660, %19205 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19214 = torch.aten.mul.Tensor %661, %19206 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19215 = torch.aten.mul.Tensor %662, %19207 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%19216 = torch.aten.mul.Tensor %663, %19208 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %19216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_16767 = torch.constant.int 5
%19217 = torch.prims.convert_element_type %19209, %int5_16767 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16768 = torch.constant.int 5
%19218 = torch.prims.convert_element_type %19210, %int5_16768 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16769 = torch.constant.int 5
%19219 = torch.prims.convert_element_type %19211, %int5_16769 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16770 = torch.constant.int 5
%19220 = torch.prims.convert_element_type %19212, %int5_16770 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16771 = torch.constant.int 5
%19221 = torch.prims.convert_element_type %19213, %int5_16771 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16772 = torch.constant.int 5
%19222 = torch.prims.convert_element_type %19214, %int5_16772 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16773 = torch.constant.int 5
%19223 = torch.prims.convert_element_type %19215, %int5_16773 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_16774 = torch.constant.int 5
%19224 = torch.prims.convert_element_type %19216, %int5_16774 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %19224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_16775 = torch.constant.int 1
%int0_16776 = torch.constant.int 0
%19225 = torch.prim.ListConstruct %int1_16775, %int0_16776 : (!torch.int, !torch.int) -> !torch.list<int>
%19226 = torch.aten.permute %664, %19225 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16777 = torch.constant.int 1
%int0_16778 = torch.constant.int 0
%19227 = torch.prim.ListConstruct %int1_16777, %int0_16778 : (!torch.int, !torch.int) -> !torch.list<int>
%19228 = torch.aten.permute %665, %19227 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16779 = torch.constant.int 1
%int0_16780 = torch.constant.int 0
%19229 = torch.prim.ListConstruct %int1_16779, %int0_16780 : (!torch.int, !torch.int) -> !torch.list<int>
%19230 = torch.aten.permute %666, %19229 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16781 = torch.constant.int 1
%int0_16782 = torch.constant.int 0
%19231 = torch.prim.ListConstruct %int1_16781, %int0_16782 : (!torch.int, !torch.int) -> !torch.list<int>
%19232 = torch.aten.permute %667, %19231 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16783 = torch.constant.int 1
%int0_16784 = torch.constant.int 0
%19233 = torch.prim.ListConstruct %int1_16783, %int0_16784 : (!torch.int, !torch.int) -> !torch.list<int>
%19234 = torch.aten.permute %668, %19233 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16785 = torch.constant.int 1
%int0_16786 = torch.constant.int 0
%19235 = torch.prim.ListConstruct %int1_16785, %int0_16786 : (!torch.int, !torch.int) -> !torch.list<int>
%19236 = torch.aten.permute %669, %19235 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16787 = torch.constant.int 1
%int0_16788 = torch.constant.int 0
%19237 = torch.prim.ListConstruct %int1_16787, %int0_16788 : (!torch.int, !torch.int) -> !torch.list<int>
%19238 = torch.aten.permute %670, %19237 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_16789 = torch.constant.int 1
%int0_16790 = torch.constant.int 0
%19239 = torch.prim.ListConstruct %int1_16789, %int0_16790 : (!torch.int, !torch.int) -> !torch.list<int>
%19240 = torch.aten.permute %671, %19239 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_16791 = torch.constant.int 4
%19241 = torch.aten.mul.int %int4_16791, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16792 = torch.constant.int 4096
%19242 = torch.prim.ListConstruct %19241, %int4096_16792 : (!torch.int, !torch.int) -> !torch.list<int>
%19243 = torch.aten.view %19217, %19242 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19243, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19244 = torch.aten.mm %19243, %19226 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19244, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16793 = torch.constant.int 4
%int512_16794 = torch.constant.int 512
%19245 = torch.prim.ListConstruct %int4_16793, %2482, %int512_16794 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19246 = torch.aten.view %19244, %19245 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16795 = torch.constant.int 4
%19247 = torch.aten.mul.int %int4_16795, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16796 = torch.constant.int 4096
%19248 = torch.prim.ListConstruct %19247, %int4096_16796 : (!torch.int, !torch.int) -> !torch.list<int>
%19249 = torch.aten.view %19218, %19248 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19249, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19250 = torch.aten.mm %19249, %19228 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19250, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16797 = torch.constant.int 4
%int512_16798 = torch.constant.int 512
%19251 = torch.prim.ListConstruct %int4_16797, %2482, %int512_16798 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19252 = torch.aten.view %19250, %19251 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16799 = torch.constant.int 4
%19253 = torch.aten.mul.int %int4_16799, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16800 = torch.constant.int 4096
%19254 = torch.prim.ListConstruct %19253, %int4096_16800 : (!torch.int, !torch.int) -> !torch.list<int>
%19255 = torch.aten.view %19219, %19254 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19255, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19256 = torch.aten.mm %19255, %19230 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19256, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16801 = torch.constant.int 4
%int512_16802 = torch.constant.int 512
%19257 = torch.prim.ListConstruct %int4_16801, %2482, %int512_16802 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19258 = torch.aten.view %19256, %19257 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16803 = torch.constant.int 4
%19259 = torch.aten.mul.int %int4_16803, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16804 = torch.constant.int 4096
%19260 = torch.prim.ListConstruct %19259, %int4096_16804 : (!torch.int, !torch.int) -> !torch.list<int>
%19261 = torch.aten.view %19220, %19260 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19261, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19262 = torch.aten.mm %19261, %19232 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19262, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16805 = torch.constant.int 4
%int512_16806 = torch.constant.int 512
%19263 = torch.prim.ListConstruct %int4_16805, %2482, %int512_16806 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19264 = torch.aten.view %19262, %19263 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16807 = torch.constant.int 4
%19265 = torch.aten.mul.int %int4_16807, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16808 = torch.constant.int 4096
%19266 = torch.prim.ListConstruct %19265, %int4096_16808 : (!torch.int, !torch.int) -> !torch.list<int>
%19267 = torch.aten.view %19221, %19266 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19267, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19268 = torch.aten.mm %19267, %19234 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19268, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16809 = torch.constant.int 4
%int512_16810 = torch.constant.int 512
%19269 = torch.prim.ListConstruct %int4_16809, %2482, %int512_16810 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19270 = torch.aten.view %19268, %19269 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16811 = torch.constant.int 4
%19271 = torch.aten.mul.int %int4_16811, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16812 = torch.constant.int 4096
%19272 = torch.prim.ListConstruct %19271, %int4096_16812 : (!torch.int, !torch.int) -> !torch.list<int>
%19273 = torch.aten.view %19222, %19272 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19273, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19274 = torch.aten.mm %19273, %19236 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19274, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16813 = torch.constant.int 4
%int512_16814 = torch.constant.int 512
%19275 = torch.prim.ListConstruct %int4_16813, %2482, %int512_16814 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19276 = torch.aten.view %19274, %19275 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16815 = torch.constant.int 4
%19277 = torch.aten.mul.int %int4_16815, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16816 = torch.constant.int 4096
%19278 = torch.prim.ListConstruct %19277, %int4096_16816 : (!torch.int, !torch.int) -> !torch.list<int>
%19279 = torch.aten.view %19223, %19278 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19279, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19280 = torch.aten.mm %19279, %19238 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19280, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16817 = torch.constant.int 4
%int512_16818 = torch.constant.int 512
%19281 = torch.prim.ListConstruct %int4_16817, %2482, %int512_16818 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19282 = torch.aten.view %19280, %19281 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_16819 = torch.constant.int 4
%19283 = torch.aten.mul.int %int4_16819, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16820 = torch.constant.int 4096
%19284 = torch.prim.ListConstruct %19283, %int4096_16820 : (!torch.int, !torch.int) -> !torch.list<int>
%19285 = torch.aten.view %19224, %19284 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19285, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19286 = torch.aten.mm %19285, %19240 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %19286, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_16821 = torch.constant.int 4
%int512_16822 = torch.constant.int 512
%19287 = torch.prim.ListConstruct %int4_16821, %2482, %int512_16822 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19288 = torch.aten.view %19286, %19287 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %19288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_16823 = torch.constant.int 1
%int0_16824 = torch.constant.int 0
%19289 = torch.prim.ListConstruct %int1_16823, %int0_16824 : (!torch.int, !torch.int) -> !torch.list<int>
%19290 = torch.aten.permute %672, %19289 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16825 = torch.constant.int 1
%int0_16826 = torch.constant.int 0
%19291 = torch.prim.ListConstruct %int1_16825, %int0_16826 : (!torch.int, !torch.int) -> !torch.list<int>
%19292 = torch.aten.permute %673, %19291 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16827 = torch.constant.int 1
%int0_16828 = torch.constant.int 0
%19293 = torch.prim.ListConstruct %int1_16827, %int0_16828 : (!torch.int, !torch.int) -> !torch.list<int>
%19294 = torch.aten.permute %674, %19293 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16829 = torch.constant.int 1
%int0_16830 = torch.constant.int 0
%19295 = torch.prim.ListConstruct %int1_16829, %int0_16830 : (!torch.int, !torch.int) -> !torch.list<int>
%19296 = torch.aten.permute %675, %19295 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16831 = torch.constant.int 1
%int0_16832 = torch.constant.int 0
%19297 = torch.prim.ListConstruct %int1_16831, %int0_16832 : (!torch.int, !torch.int) -> !torch.list<int>
%19298 = torch.aten.permute %676, %19297 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16833 = torch.constant.int 1
%int0_16834 = torch.constant.int 0
%19299 = torch.prim.ListConstruct %int1_16833, %int0_16834 : (!torch.int, !torch.int) -> !torch.list<int>
%19300 = torch.aten.permute %677, %19299 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16835 = torch.constant.int 1
%int0_16836 = torch.constant.int 0
%19301 = torch.prim.ListConstruct %int1_16835, %int0_16836 : (!torch.int, !torch.int) -> !torch.list<int>
%19302 = torch.aten.permute %678, %19301 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16837 = torch.constant.int 1
%int0_16838 = torch.constant.int 0
%19303 = torch.prim.ListConstruct %int1_16837, %int0_16838 : (!torch.int, !torch.int) -> !torch.list<int>
%19304 = torch.aten.permute %679, %19303 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_16839 = torch.constant.int 4
%19305 = torch.aten.mul.int %int4_16839, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16840 = torch.constant.int 4096
%19306 = torch.prim.ListConstruct %19305, %int4096_16840 : (!torch.int, !torch.int) -> !torch.list<int>
%19307 = torch.aten.view %19217, %19306 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19307, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19308 = torch.aten.mm %19307, %19290 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19308, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16841 = torch.constant.int 4
%int128_16842 = torch.constant.int 128
%19309 = torch.prim.ListConstruct %int4_16841, %2482, %int128_16842 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19310 = torch.aten.view %19308, %19309 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16843 = torch.constant.int 4
%19311 = torch.aten.mul.int %int4_16843, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16844 = torch.constant.int 4096
%19312 = torch.prim.ListConstruct %19311, %int4096_16844 : (!torch.int, !torch.int) -> !torch.list<int>
%19313 = torch.aten.view %19218, %19312 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19313, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19314 = torch.aten.mm %19313, %19292 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19314, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16845 = torch.constant.int 4
%int128_16846 = torch.constant.int 128
%19315 = torch.prim.ListConstruct %int4_16845, %2482, %int128_16846 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19316 = torch.aten.view %19314, %19315 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16847 = torch.constant.int 4
%19317 = torch.aten.mul.int %int4_16847, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16848 = torch.constant.int 4096
%19318 = torch.prim.ListConstruct %19317, %int4096_16848 : (!torch.int, !torch.int) -> !torch.list<int>
%19319 = torch.aten.view %19219, %19318 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19319, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19320 = torch.aten.mm %19319, %19294 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19320, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16849 = torch.constant.int 4
%int128_16850 = torch.constant.int 128
%19321 = torch.prim.ListConstruct %int4_16849, %2482, %int128_16850 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19322 = torch.aten.view %19320, %19321 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16851 = torch.constant.int 4
%19323 = torch.aten.mul.int %int4_16851, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16852 = torch.constant.int 4096
%19324 = torch.prim.ListConstruct %19323, %int4096_16852 : (!torch.int, !torch.int) -> !torch.list<int>
%19325 = torch.aten.view %19220, %19324 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19325, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19326 = torch.aten.mm %19325, %19296 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19326, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16853 = torch.constant.int 4
%int128_16854 = torch.constant.int 128
%19327 = torch.prim.ListConstruct %int4_16853, %2482, %int128_16854 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19328 = torch.aten.view %19326, %19327 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16855 = torch.constant.int 4
%19329 = torch.aten.mul.int %int4_16855, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16856 = torch.constant.int 4096
%19330 = torch.prim.ListConstruct %19329, %int4096_16856 : (!torch.int, !torch.int) -> !torch.list<int>
%19331 = torch.aten.view %19221, %19330 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19331, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19332 = torch.aten.mm %19331, %19298 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19332, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16857 = torch.constant.int 4
%int128_16858 = torch.constant.int 128
%19333 = torch.prim.ListConstruct %int4_16857, %2482, %int128_16858 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19334 = torch.aten.view %19332, %19333 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16859 = torch.constant.int 4
%19335 = torch.aten.mul.int %int4_16859, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16860 = torch.constant.int 4096
%19336 = torch.prim.ListConstruct %19335, %int4096_16860 : (!torch.int, !torch.int) -> !torch.list<int>
%19337 = torch.aten.view %19222, %19336 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19337, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19338 = torch.aten.mm %19337, %19300 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19338, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16861 = torch.constant.int 4
%int128_16862 = torch.constant.int 128
%19339 = torch.prim.ListConstruct %int4_16861, %2482, %int128_16862 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19340 = torch.aten.view %19338, %19339 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16863 = torch.constant.int 4
%19341 = torch.aten.mul.int %int4_16863, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16864 = torch.constant.int 4096
%19342 = torch.prim.ListConstruct %19341, %int4096_16864 : (!torch.int, !torch.int) -> !torch.list<int>
%19343 = torch.aten.view %19223, %19342 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19343, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19344 = torch.aten.mm %19343, %19302 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19344, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16865 = torch.constant.int 4
%int128_16866 = torch.constant.int 128
%19345 = torch.prim.ListConstruct %int4_16865, %2482, %int128_16866 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19346 = torch.aten.view %19344, %19345 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16867 = torch.constant.int 4
%19347 = torch.aten.mul.int %int4_16867, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16868 = torch.constant.int 4096
%19348 = torch.prim.ListConstruct %19347, %int4096_16868 : (!torch.int, !torch.int) -> !torch.list<int>
%19349 = torch.aten.view %19224, %19348 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19349, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19350 = torch.aten.mm %19349, %19304 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19350, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16869 = torch.constant.int 4
%int128_16870 = torch.constant.int 128
%19351 = torch.prim.ListConstruct %int4_16869, %2482, %int128_16870 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19352 = torch.aten.view %19350, %19351 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_16871 = torch.constant.int 1
%int0_16872 = torch.constant.int 0
%19353 = torch.prim.ListConstruct %int1_16871, %int0_16872 : (!torch.int, !torch.int) -> !torch.list<int>
%19354 = torch.aten.permute %680, %19353 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16873 = torch.constant.int 1
%int0_16874 = torch.constant.int 0
%19355 = torch.prim.ListConstruct %int1_16873, %int0_16874 : (!torch.int, !torch.int) -> !torch.list<int>
%19356 = torch.aten.permute %681, %19355 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16875 = torch.constant.int 1
%int0_16876 = torch.constant.int 0
%19357 = torch.prim.ListConstruct %int1_16875, %int0_16876 : (!torch.int, !torch.int) -> !torch.list<int>
%19358 = torch.aten.permute %682, %19357 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16877 = torch.constant.int 1
%int0_16878 = torch.constant.int 0
%19359 = torch.prim.ListConstruct %int1_16877, %int0_16878 : (!torch.int, !torch.int) -> !torch.list<int>
%19360 = torch.aten.permute %683, %19359 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16879 = torch.constant.int 1
%int0_16880 = torch.constant.int 0
%19361 = torch.prim.ListConstruct %int1_16879, %int0_16880 : (!torch.int, !torch.int) -> !torch.list<int>
%19362 = torch.aten.permute %684, %19361 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16881 = torch.constant.int 1
%int0_16882 = torch.constant.int 0
%19363 = torch.prim.ListConstruct %int1_16881, %int0_16882 : (!torch.int, !torch.int) -> !torch.list<int>
%19364 = torch.aten.permute %685, %19363 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16883 = torch.constant.int 1
%int0_16884 = torch.constant.int 0
%19365 = torch.prim.ListConstruct %int1_16883, %int0_16884 : (!torch.int, !torch.int) -> !torch.list<int>
%19366 = torch.aten.permute %686, %19365 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_16885 = torch.constant.int 1
%int0_16886 = torch.constant.int 0
%19367 = torch.prim.ListConstruct %int1_16885, %int0_16886 : (!torch.int, !torch.int) -> !torch.list<int>
%19368 = torch.aten.permute %687, %19367 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_16887 = torch.constant.int 4
%19369 = torch.aten.mul.int %int4_16887, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16888 = torch.constant.int 4096
%19370 = torch.prim.ListConstruct %19369, %int4096_16888 : (!torch.int, !torch.int) -> !torch.list<int>
%19371 = torch.aten.view %19217, %19370 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19371, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19372 = torch.aten.mm %19371, %19354 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19372, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16889 = torch.constant.int 4
%int128_16890 = torch.constant.int 128
%19373 = torch.prim.ListConstruct %int4_16889, %2482, %int128_16890 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19374 = torch.aten.view %19372, %19373 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16891 = torch.constant.int 4
%19375 = torch.aten.mul.int %int4_16891, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16892 = torch.constant.int 4096
%19376 = torch.prim.ListConstruct %19375, %int4096_16892 : (!torch.int, !torch.int) -> !torch.list<int>
%19377 = torch.aten.view %19218, %19376 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19377, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19378 = torch.aten.mm %19377, %19356 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19378, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16893 = torch.constant.int 4
%int128_16894 = torch.constant.int 128
%19379 = torch.prim.ListConstruct %int4_16893, %2482, %int128_16894 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19380 = torch.aten.view %19378, %19379 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16895 = torch.constant.int 4
%19381 = torch.aten.mul.int %int4_16895, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16896 = torch.constant.int 4096
%19382 = torch.prim.ListConstruct %19381, %int4096_16896 : (!torch.int, !torch.int) -> !torch.list<int>
%19383 = torch.aten.view %19219, %19382 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19383, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19384 = torch.aten.mm %19383, %19358 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19384, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16897 = torch.constant.int 4
%int128_16898 = torch.constant.int 128
%19385 = torch.prim.ListConstruct %int4_16897, %2482, %int128_16898 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19386 = torch.aten.view %19384, %19385 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16899 = torch.constant.int 4
%19387 = torch.aten.mul.int %int4_16899, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16900 = torch.constant.int 4096
%19388 = torch.prim.ListConstruct %19387, %int4096_16900 : (!torch.int, !torch.int) -> !torch.list<int>
%19389 = torch.aten.view %19220, %19388 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19389, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19390 = torch.aten.mm %19389, %19360 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19390, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16901 = torch.constant.int 4
%int128_16902 = torch.constant.int 128
%19391 = torch.prim.ListConstruct %int4_16901, %2482, %int128_16902 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19392 = torch.aten.view %19390, %19391 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16903 = torch.constant.int 4
%19393 = torch.aten.mul.int %int4_16903, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16904 = torch.constant.int 4096
%19394 = torch.prim.ListConstruct %19393, %int4096_16904 : (!torch.int, !torch.int) -> !torch.list<int>
%19395 = torch.aten.view %19221, %19394 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19395, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19396 = torch.aten.mm %19395, %19362 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19396, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16905 = torch.constant.int 4
%int128_16906 = torch.constant.int 128
%19397 = torch.prim.ListConstruct %int4_16905, %2482, %int128_16906 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19398 = torch.aten.view %19396, %19397 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16907 = torch.constant.int 4
%19399 = torch.aten.mul.int %int4_16907, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16908 = torch.constant.int 4096
%19400 = torch.prim.ListConstruct %19399, %int4096_16908 : (!torch.int, !torch.int) -> !torch.list<int>
%19401 = torch.aten.view %19222, %19400 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19401, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19402 = torch.aten.mm %19401, %19364 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19402, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16909 = torch.constant.int 4
%int128_16910 = torch.constant.int 128
%19403 = torch.prim.ListConstruct %int4_16909, %2482, %int128_16910 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19404 = torch.aten.view %19402, %19403 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16911 = torch.constant.int 4
%19405 = torch.aten.mul.int %int4_16911, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16912 = torch.constant.int 4096
%19406 = torch.prim.ListConstruct %19405, %int4096_16912 : (!torch.int, !torch.int) -> !torch.list<int>
%19407 = torch.aten.view %19223, %19406 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19407, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19408 = torch.aten.mm %19407, %19366 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19408, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16913 = torch.constant.int 4
%int128_16914 = torch.constant.int 128
%19409 = torch.prim.ListConstruct %int4_16913, %2482, %int128_16914 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19410 = torch.aten.view %19408, %19409 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16915 = torch.constant.int 4
%19411 = torch.aten.mul.int %int4_16915, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_16916 = torch.constant.int 4096
%19412 = torch.prim.ListConstruct %19411, %int4096_16916 : (!torch.int, !torch.int) -> !torch.list<int>
%19413 = torch.aten.view %19224, %19412 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %19413, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%19414 = torch.aten.mm %19413, %19368 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %19414, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_16917 = torch.constant.int 4
%int128_16918 = torch.constant.int 128
%19415 = torch.prim.ListConstruct %int4_16917, %2482, %int128_16918 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19416 = torch.aten.view %19414, %19415 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %19416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_16919 = torch.constant.int 4
%int4_16920 = torch.constant.int 4
%int128_16921 = torch.constant.int 128
%19417 = torch.prim.ListConstruct %int4_16919, %2482, %int4_16920, %int128_16921 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19418 = torch.aten.view %19246, %19417 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16922 = torch.constant.int 4
%int4_16923 = torch.constant.int 4
%int128_16924 = torch.constant.int 128
%19419 = torch.prim.ListConstruct %int4_16922, %2482, %int4_16923, %int128_16924 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19420 = torch.aten.view %19252, %19419 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16925 = torch.constant.int 4
%int4_16926 = torch.constant.int 4
%int128_16927 = torch.constant.int 128
%19421 = torch.prim.ListConstruct %int4_16925, %2482, %int4_16926, %int128_16927 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19422 = torch.aten.view %19258, %19421 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16928 = torch.constant.int 4
%int4_16929 = torch.constant.int 4
%int128_16930 = torch.constant.int 128
%19423 = torch.prim.ListConstruct %int4_16928, %2482, %int4_16929, %int128_16930 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19424 = torch.aten.view %19264, %19423 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16931 = torch.constant.int 4
%int4_16932 = torch.constant.int 4
%int128_16933 = torch.constant.int 128
%19425 = torch.prim.ListConstruct %int4_16931, %2482, %int4_16932, %int128_16933 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19426 = torch.aten.view %19270, %19425 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16934 = torch.constant.int 4
%int4_16935 = torch.constant.int 4
%int128_16936 = torch.constant.int 128
%19427 = torch.prim.ListConstruct %int4_16934, %2482, %int4_16935, %int128_16936 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19428 = torch.aten.view %19276, %19427 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16937 = torch.constant.int 4
%int4_16938 = torch.constant.int 4
%int128_16939 = torch.constant.int 128
%19429 = torch.prim.ListConstruct %int4_16937, %2482, %int4_16938, %int128_16939 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19430 = torch.aten.view %19282, %19429 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16940 = torch.constant.int 4
%int4_16941 = torch.constant.int 4
%int128_16942 = torch.constant.int 128
%19431 = torch.prim.ListConstruct %int4_16940, %2482, %int4_16941, %int128_16942 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19432 = torch.aten.view %19288, %19431 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_16943 = torch.constant.int 4
%int1_16944 = torch.constant.int 1
%int128_16945 = torch.constant.int 128
%19433 = torch.prim.ListConstruct %int4_16943, %2482, %int1_16944, %int128_16945 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19434 = torch.aten.view %19310, %19433 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16946 = torch.constant.int 4
%int1_16947 = torch.constant.int 1
%int128_16948 = torch.constant.int 128
%19435 = torch.prim.ListConstruct %int4_16946, %2482, %int1_16947, %int128_16948 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19436 = torch.aten.view %19316, %19435 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16949 = torch.constant.int 4
%int1_16950 = torch.constant.int 1
%int128_16951 = torch.constant.int 128
%19437 = torch.prim.ListConstruct %int4_16949, %2482, %int1_16950, %int128_16951 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19438 = torch.aten.view %19322, %19437 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16952 = torch.constant.int 4
%int1_16953 = torch.constant.int 1
%int128_16954 = torch.constant.int 128
%19439 = torch.prim.ListConstruct %int4_16952, %2482, %int1_16953, %int128_16954 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19440 = torch.aten.view %19328, %19439 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16955 = torch.constant.int 4
%int1_16956 = torch.constant.int 1
%int128_16957 = torch.constant.int 128
%19441 = torch.prim.ListConstruct %int4_16955, %2482, %int1_16956, %int128_16957 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19442 = torch.aten.view %19334, %19441 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16958 = torch.constant.int 4
%int1_16959 = torch.constant.int 1
%int128_16960 = torch.constant.int 128
%19443 = torch.prim.ListConstruct %int4_16958, %2482, %int1_16959, %int128_16960 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19444 = torch.aten.view %19340, %19443 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16961 = torch.constant.int 4
%int1_16962 = torch.constant.int 1
%int128_16963 = torch.constant.int 128
%19445 = torch.prim.ListConstruct %int4_16961, %2482, %int1_16962, %int128_16963 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19446 = torch.aten.view %19346, %19445 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16964 = torch.constant.int 4
%int1_16965 = torch.constant.int 1
%int128_16966 = torch.constant.int 128
%19447 = torch.prim.ListConstruct %int4_16964, %2482, %int1_16965, %int128_16966 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19448 = torch.aten.view %19352, %19447 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16967 = torch.constant.int 4
%int1_16968 = torch.constant.int 1
%int128_16969 = torch.constant.int 128
%19449 = torch.prim.ListConstruct %int4_16967, %2482, %int1_16968, %int128_16969 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19450 = torch.aten.view %19374, %19449 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16970 = torch.constant.int 4
%int1_16971 = torch.constant.int 1
%int128_16972 = torch.constant.int 128
%19451 = torch.prim.ListConstruct %int4_16970, %2482, %int1_16971, %int128_16972 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19452 = torch.aten.view %19380, %19451 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16973 = torch.constant.int 4
%int1_16974 = torch.constant.int 1
%int128_16975 = torch.constant.int 128
%19453 = torch.prim.ListConstruct %int4_16973, %2482, %int1_16974, %int128_16975 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19454 = torch.aten.view %19386, %19453 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16976 = torch.constant.int 4
%int1_16977 = torch.constant.int 1
%int128_16978 = torch.constant.int 128
%19455 = torch.prim.ListConstruct %int4_16976, %2482, %int1_16977, %int128_16978 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19456 = torch.aten.view %19392, %19455 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16979 = torch.constant.int 4
%int1_16980 = torch.constant.int 1
%int128_16981 = torch.constant.int 128
%19457 = torch.prim.ListConstruct %int4_16979, %2482, %int1_16980, %int128_16981 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19458 = torch.aten.view %19398, %19457 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16982 = torch.constant.int 4
%int1_16983 = torch.constant.int 1
%int128_16984 = torch.constant.int 128
%19459 = torch.prim.ListConstruct %int4_16982, %2482, %int1_16983, %int128_16984 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19460 = torch.aten.view %19404, %19459 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16985 = torch.constant.int 4
%int1_16986 = torch.constant.int 1
%int128_16987 = torch.constant.int 128
%19461 = torch.prim.ListConstruct %int4_16985, %2482, %int1_16986, %int128_16987 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19462 = torch.aten.view %19410, %19461 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_16988 = torch.constant.int 4
%int1_16989 = torch.constant.int 1
%int128_16990 = torch.constant.int 128
%19463 = torch.prim.ListConstruct %int4_16988, %2482, %int1_16989, %int128_16990 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19464 = torch.aten.view %19416, %19463 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_16991 = torch.constant.int 131072
%none_16992 = torch.constant.none
%none_16993 = torch.constant.none
%cpu_16994 = torch.constant.device "cpu"
%false_16995 = torch.constant.bool false
%19465 = torch.aten.arange %int131072_16991, %none_16992, %none_16993, %cpu_16994, %false_16995 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_16996 = torch.constant.int 0
%int128_16997 = torch.constant.int 128
%int2_16998 = torch.constant.int 2
%none_16999 = torch.constant.none
%none_17000 = torch.constant.none
%cpu_17001 = torch.constant.device "cpu"
%false_17002 = torch.constant.bool false
%19466 = torch.aten.arange.start_step %int0_16996, %int128_16997, %int2_16998, %none_16999, %none_17000, %cpu_17001, %false_17002 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_17003 = torch.constant.int 0
%int0_17004 = torch.constant.int 0
%int64_17005 = torch.constant.int 64
%int1_17006 = torch.constant.int 1
%19467 = torch.aten.slice.Tensor %19466, %int0_17003, %int0_17004, %int64_17005, %int1_17006 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_17007 = torch.constant.int 6
%19468 = torch.prims.convert_element_type %19467, %int6_17007 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_17008 = torch.constant.int 128
%19469 = torch.aten.div.Scalar %19468, %int128_17008 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_17009 = torch.constant.float 5.000000e+05
%19470 = torch.aten.pow.Scalar %float5.000000e05_17009, %19469 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%19471 = torch.aten.reciprocal %19470 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_17010 = torch.constant.float 1.000000e+00
%19472 = torch.aten.mul.Scalar %19471, %float1.000000e00_17010 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_17011 = torch.constant.int 131072
%int1_17012 = torch.constant.int 1
%19473 = torch.prim.ListConstruct %int131072_17011, %int1_17012 : (!torch.int, !torch.int) -> !torch.list<int>
%19474 = torch.aten.view %19465, %19473 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%19475 = torch.aten.mul.Tensor %19474, %19472 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%19476 = torch.aten.cos %19475 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%19477 = torch.aten.sin %19475 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%19478 = torch.aten.complex %19476, %19477 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%19479 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19480 = flow.tensor.transfer %19479 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%19481 = torch_c.from_builtin_tensor %19480 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19482 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19483 = flow.tensor.transfer %19482 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%19484 = torch_c.from_builtin_tensor %19483 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19485 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19486 = flow.tensor.transfer %19485 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%19487 = torch_c.from_builtin_tensor %19486 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19488 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19489 = flow.tensor.transfer %19488 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%19490 = torch_c.from_builtin_tensor %19489 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19491 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19492 = flow.tensor.transfer %19491 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%19493 = torch_c.from_builtin_tensor %19492 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19494 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19495 = flow.tensor.transfer %19494 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%19496 = torch_c.from_builtin_tensor %19495 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19497 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19498 = flow.tensor.transfer %19497 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%19499 = torch_c.from_builtin_tensor %19498 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19500 = torch_c.to_builtin_tensor %19478 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19501 = flow.tensor.transfer %19500 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%19502 = torch_c.from_builtin_tensor %19501 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_17013 = torch.constant.int 1
%19503 = torch.aten.size.int %19246, %int1_17013 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17014 = torch.constant.int 0
%19504 = torch.aten.add.int %int0_17014, %19503 : !torch.int, !torch.int -> !torch.int
%int0_17015 = torch.constant.int 0
%int0_17016 = torch.constant.int 0
%int1_17017 = torch.constant.int 1
%19505 = torch.aten.slice.Tensor %19481, %int0_17015, %int0_17016, %19504, %int1_17017 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19505, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17018 = torch.constant.int 1
%int0_17019 = torch.constant.int 0
%int9223372036854775807_17020 = torch.constant.int 9223372036854775807
%int1_17021 = torch.constant.int 1
%19506 = torch.aten.slice.Tensor %19505, %int1_17018, %int0_17019, %int9223372036854775807_17020, %int1_17021 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19506, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17022 = torch.constant.int 0
%19507 = torch.aten.unsqueeze %19506, %int0_17022 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19507, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17023 = torch.constant.int 2
%19508 = torch.aten.unsqueeze %19507, %int2_17023 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19508, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17024 = torch.constant.int 3
%int0_17025 = torch.constant.int 0
%int9223372036854775807_17026 = torch.constant.int 9223372036854775807
%int1_17027 = torch.constant.int 1
%19509 = torch.aten.slice.Tensor %19508, %int3_17024, %int0_17025, %int9223372036854775807_17026, %int1_17027 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19509, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19510 = torch_c.to_builtin_tensor %19418 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17028 = arith.constant 1 : index
%dim_17029 = tensor.dim %19510, %c1_17028 : tensor<4x?x4x128xf16>
%19511 = flow.tensor.bitcast %19510 : tensor<4x?x4x128xf16>{%dim_17029} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17029}
%19512 = torch_c.from_builtin_tensor %19511 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19513 = torch.aten.mul.Tensor %19512, %19509 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19514 = torch_c.to_builtin_tensor %19513 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17030 = arith.constant 1 : index
%dim_17031 = tensor.dim %19514, %c1_17030 : tensor<4x?x4x64xcomplex<f32>>
%19515 = flow.tensor.bitcast %19514 : tensor<4x?x4x64xcomplex<f32>>{%dim_17031} -> tensor<4x?x4x128xf32>{%dim_17031}
%19516 = torch_c.from_builtin_tensor %19515 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17032 = torch.constant.int 5
%19517 = torch.prims.convert_element_type %19516, %int5_17032 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17033 = torch.constant.int 1
%19518 = torch.aten.size.int %19252, %int1_17033 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17034 = torch.constant.int 0
%19519 = torch.aten.add.int %int0_17034, %19518 : !torch.int, !torch.int -> !torch.int
%int0_17035 = torch.constant.int 0
%int0_17036 = torch.constant.int 0
%int1_17037 = torch.constant.int 1
%19520 = torch.aten.slice.Tensor %19484, %int0_17035, %int0_17036, %19519, %int1_17037 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19520, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17038 = torch.constant.int 1
%int0_17039 = torch.constant.int 0
%int9223372036854775807_17040 = torch.constant.int 9223372036854775807
%int1_17041 = torch.constant.int 1
%19521 = torch.aten.slice.Tensor %19520, %int1_17038, %int0_17039, %int9223372036854775807_17040, %int1_17041 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19521, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17042 = torch.constant.int 0
%19522 = torch.aten.unsqueeze %19521, %int0_17042 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19522, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17043 = torch.constant.int 2
%19523 = torch.aten.unsqueeze %19522, %int2_17043 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19523, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17044 = torch.constant.int 3
%int0_17045 = torch.constant.int 0
%int9223372036854775807_17046 = torch.constant.int 9223372036854775807
%int1_17047 = torch.constant.int 1
%19524 = torch.aten.slice.Tensor %19523, %int3_17044, %int0_17045, %int9223372036854775807_17046, %int1_17047 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19524, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19525 = torch_c.to_builtin_tensor %19420 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17048 = arith.constant 1 : index
%dim_17049 = tensor.dim %19525, %c1_17048 : tensor<4x?x4x128xf16>
%19526 = flow.tensor.bitcast %19525 : tensor<4x?x4x128xf16>{%dim_17049} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17049}
%19527 = torch_c.from_builtin_tensor %19526 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19528 = torch.aten.mul.Tensor %19527, %19524 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19529 = torch_c.to_builtin_tensor %19528 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17050 = arith.constant 1 : index
%dim_17051 = tensor.dim %19529, %c1_17050 : tensor<4x?x4x64xcomplex<f32>>
%19530 = flow.tensor.bitcast %19529 : tensor<4x?x4x64xcomplex<f32>>{%dim_17051} -> tensor<4x?x4x128xf32>{%dim_17051}
%19531 = torch_c.from_builtin_tensor %19530 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17052 = torch.constant.int 5
%19532 = torch.prims.convert_element_type %19531, %int5_17052 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17053 = torch.constant.int 1
%19533 = torch.aten.size.int %19258, %int1_17053 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17054 = torch.constant.int 0
%19534 = torch.aten.add.int %int0_17054, %19533 : !torch.int, !torch.int -> !torch.int
%int0_17055 = torch.constant.int 0
%int0_17056 = torch.constant.int 0
%int1_17057 = torch.constant.int 1
%19535 = torch.aten.slice.Tensor %19487, %int0_17055, %int0_17056, %19534, %int1_17057 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19535, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17058 = torch.constant.int 1
%int0_17059 = torch.constant.int 0
%int9223372036854775807_17060 = torch.constant.int 9223372036854775807
%int1_17061 = torch.constant.int 1
%19536 = torch.aten.slice.Tensor %19535, %int1_17058, %int0_17059, %int9223372036854775807_17060, %int1_17061 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19536, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17062 = torch.constant.int 0
%19537 = torch.aten.unsqueeze %19536, %int0_17062 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19537, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17063 = torch.constant.int 2
%19538 = torch.aten.unsqueeze %19537, %int2_17063 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19538, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17064 = torch.constant.int 3
%int0_17065 = torch.constant.int 0
%int9223372036854775807_17066 = torch.constant.int 9223372036854775807
%int1_17067 = torch.constant.int 1
%19539 = torch.aten.slice.Tensor %19538, %int3_17064, %int0_17065, %int9223372036854775807_17066, %int1_17067 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19539, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19540 = torch_c.to_builtin_tensor %19422 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17068 = arith.constant 1 : index
%dim_17069 = tensor.dim %19540, %c1_17068 : tensor<4x?x4x128xf16>
%19541 = flow.tensor.bitcast %19540 : tensor<4x?x4x128xf16>{%dim_17069} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17069}
%19542 = torch_c.from_builtin_tensor %19541 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19543 = torch.aten.mul.Tensor %19542, %19539 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19544 = torch_c.to_builtin_tensor %19543 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17070 = arith.constant 1 : index
%dim_17071 = tensor.dim %19544, %c1_17070 : tensor<4x?x4x64xcomplex<f32>>
%19545 = flow.tensor.bitcast %19544 : tensor<4x?x4x64xcomplex<f32>>{%dim_17071} -> tensor<4x?x4x128xf32>{%dim_17071}
%19546 = torch_c.from_builtin_tensor %19545 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17072 = torch.constant.int 5
%19547 = torch.prims.convert_element_type %19546, %int5_17072 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17073 = torch.constant.int 1
%19548 = torch.aten.size.int %19264, %int1_17073 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17074 = torch.constant.int 0
%19549 = torch.aten.add.int %int0_17074, %19548 : !torch.int, !torch.int -> !torch.int
%int0_17075 = torch.constant.int 0
%int0_17076 = torch.constant.int 0
%int1_17077 = torch.constant.int 1
%19550 = torch.aten.slice.Tensor %19490, %int0_17075, %int0_17076, %19549, %int1_17077 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19550, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17078 = torch.constant.int 1
%int0_17079 = torch.constant.int 0
%int9223372036854775807_17080 = torch.constant.int 9223372036854775807
%int1_17081 = torch.constant.int 1
%19551 = torch.aten.slice.Tensor %19550, %int1_17078, %int0_17079, %int9223372036854775807_17080, %int1_17081 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19551, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17082 = torch.constant.int 0
%19552 = torch.aten.unsqueeze %19551, %int0_17082 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19552, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17083 = torch.constant.int 2
%19553 = torch.aten.unsqueeze %19552, %int2_17083 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19553, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17084 = torch.constant.int 3
%int0_17085 = torch.constant.int 0
%int9223372036854775807_17086 = torch.constant.int 9223372036854775807
%int1_17087 = torch.constant.int 1
%19554 = torch.aten.slice.Tensor %19553, %int3_17084, %int0_17085, %int9223372036854775807_17086, %int1_17087 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19554, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19555 = torch_c.to_builtin_tensor %19424 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17088 = arith.constant 1 : index
%dim_17089 = tensor.dim %19555, %c1_17088 : tensor<4x?x4x128xf16>
%19556 = flow.tensor.bitcast %19555 : tensor<4x?x4x128xf16>{%dim_17089} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17089}
%19557 = torch_c.from_builtin_tensor %19556 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19558 = torch.aten.mul.Tensor %19557, %19554 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19559 = torch_c.to_builtin_tensor %19558 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17090 = arith.constant 1 : index
%dim_17091 = tensor.dim %19559, %c1_17090 : tensor<4x?x4x64xcomplex<f32>>
%19560 = flow.tensor.bitcast %19559 : tensor<4x?x4x64xcomplex<f32>>{%dim_17091} -> tensor<4x?x4x128xf32>{%dim_17091}
%19561 = torch_c.from_builtin_tensor %19560 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17092 = torch.constant.int 5
%19562 = torch.prims.convert_element_type %19561, %int5_17092 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17093 = torch.constant.int 1
%19563 = torch.aten.size.int %19270, %int1_17093 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17094 = torch.constant.int 0
%19564 = torch.aten.add.int %int0_17094, %19563 : !torch.int, !torch.int -> !torch.int
%int0_17095 = torch.constant.int 0
%int0_17096 = torch.constant.int 0
%int1_17097 = torch.constant.int 1
%19565 = torch.aten.slice.Tensor %19493, %int0_17095, %int0_17096, %19564, %int1_17097 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19565, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17098 = torch.constant.int 1
%int0_17099 = torch.constant.int 0
%int9223372036854775807_17100 = torch.constant.int 9223372036854775807
%int1_17101 = torch.constant.int 1
%19566 = torch.aten.slice.Tensor %19565, %int1_17098, %int0_17099, %int9223372036854775807_17100, %int1_17101 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19566, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17102 = torch.constant.int 0
%19567 = torch.aten.unsqueeze %19566, %int0_17102 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19567, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17103 = torch.constant.int 2
%19568 = torch.aten.unsqueeze %19567, %int2_17103 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19568, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17104 = torch.constant.int 3
%int0_17105 = torch.constant.int 0
%int9223372036854775807_17106 = torch.constant.int 9223372036854775807
%int1_17107 = torch.constant.int 1
%19569 = torch.aten.slice.Tensor %19568, %int3_17104, %int0_17105, %int9223372036854775807_17106, %int1_17107 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19569, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19570 = torch_c.to_builtin_tensor %19426 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17108 = arith.constant 1 : index
%dim_17109 = tensor.dim %19570, %c1_17108 : tensor<4x?x4x128xf16>
%19571 = flow.tensor.bitcast %19570 : tensor<4x?x4x128xf16>{%dim_17109} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17109}
%19572 = torch_c.from_builtin_tensor %19571 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19573 = torch.aten.mul.Tensor %19572, %19569 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19574 = torch_c.to_builtin_tensor %19573 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17110 = arith.constant 1 : index
%dim_17111 = tensor.dim %19574, %c1_17110 : tensor<4x?x4x64xcomplex<f32>>
%19575 = flow.tensor.bitcast %19574 : tensor<4x?x4x64xcomplex<f32>>{%dim_17111} -> tensor<4x?x4x128xf32>{%dim_17111}
%19576 = torch_c.from_builtin_tensor %19575 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17112 = torch.constant.int 5
%19577 = torch.prims.convert_element_type %19576, %int5_17112 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17113 = torch.constant.int 1
%19578 = torch.aten.size.int %19276, %int1_17113 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17114 = torch.constant.int 0
%19579 = torch.aten.add.int %int0_17114, %19578 : !torch.int, !torch.int -> !torch.int
%int0_17115 = torch.constant.int 0
%int0_17116 = torch.constant.int 0
%int1_17117 = torch.constant.int 1
%19580 = torch.aten.slice.Tensor %19496, %int0_17115, %int0_17116, %19579, %int1_17117 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19580, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17118 = torch.constant.int 1
%int0_17119 = torch.constant.int 0
%int9223372036854775807_17120 = torch.constant.int 9223372036854775807
%int1_17121 = torch.constant.int 1
%19581 = torch.aten.slice.Tensor %19580, %int1_17118, %int0_17119, %int9223372036854775807_17120, %int1_17121 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19581, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17122 = torch.constant.int 0
%19582 = torch.aten.unsqueeze %19581, %int0_17122 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19582, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17123 = torch.constant.int 2
%19583 = torch.aten.unsqueeze %19582, %int2_17123 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19583, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17124 = torch.constant.int 3
%int0_17125 = torch.constant.int 0
%int9223372036854775807_17126 = torch.constant.int 9223372036854775807
%int1_17127 = torch.constant.int 1
%19584 = torch.aten.slice.Tensor %19583, %int3_17124, %int0_17125, %int9223372036854775807_17126, %int1_17127 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19584, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19585 = torch_c.to_builtin_tensor %19428 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17128 = arith.constant 1 : index
%dim_17129 = tensor.dim %19585, %c1_17128 : tensor<4x?x4x128xf16>
%19586 = flow.tensor.bitcast %19585 : tensor<4x?x4x128xf16>{%dim_17129} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17129}
%19587 = torch_c.from_builtin_tensor %19586 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19588 = torch.aten.mul.Tensor %19587, %19584 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19589 = torch_c.to_builtin_tensor %19588 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17130 = arith.constant 1 : index
%dim_17131 = tensor.dim %19589, %c1_17130 : tensor<4x?x4x64xcomplex<f32>>
%19590 = flow.tensor.bitcast %19589 : tensor<4x?x4x64xcomplex<f32>>{%dim_17131} -> tensor<4x?x4x128xf32>{%dim_17131}
%19591 = torch_c.from_builtin_tensor %19590 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17132 = torch.constant.int 5
%19592 = torch.prims.convert_element_type %19591, %int5_17132 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17133 = torch.constant.int 1
%19593 = torch.aten.size.int %19282, %int1_17133 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17134 = torch.constant.int 0
%19594 = torch.aten.add.int %int0_17134, %19593 : !torch.int, !torch.int -> !torch.int
%int0_17135 = torch.constant.int 0
%int0_17136 = torch.constant.int 0
%int1_17137 = torch.constant.int 1
%19595 = torch.aten.slice.Tensor %19499, %int0_17135, %int0_17136, %19594, %int1_17137 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19595, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17138 = torch.constant.int 1
%int0_17139 = torch.constant.int 0
%int9223372036854775807_17140 = torch.constant.int 9223372036854775807
%int1_17141 = torch.constant.int 1
%19596 = torch.aten.slice.Tensor %19595, %int1_17138, %int0_17139, %int9223372036854775807_17140, %int1_17141 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19596, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17142 = torch.constant.int 0
%19597 = torch.aten.unsqueeze %19596, %int0_17142 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19597, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17143 = torch.constant.int 2
%19598 = torch.aten.unsqueeze %19597, %int2_17143 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19598, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17144 = torch.constant.int 3
%int0_17145 = torch.constant.int 0
%int9223372036854775807_17146 = torch.constant.int 9223372036854775807
%int1_17147 = torch.constant.int 1
%19599 = torch.aten.slice.Tensor %19598, %int3_17144, %int0_17145, %int9223372036854775807_17146, %int1_17147 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19599, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19600 = torch_c.to_builtin_tensor %19430 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17148 = arith.constant 1 : index
%dim_17149 = tensor.dim %19600, %c1_17148 : tensor<4x?x4x128xf16>
%19601 = flow.tensor.bitcast %19600 : tensor<4x?x4x128xf16>{%dim_17149} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17149}
%19602 = torch_c.from_builtin_tensor %19601 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19603 = torch.aten.mul.Tensor %19602, %19599 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19604 = torch_c.to_builtin_tensor %19603 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17150 = arith.constant 1 : index
%dim_17151 = tensor.dim %19604, %c1_17150 : tensor<4x?x4x64xcomplex<f32>>
%19605 = flow.tensor.bitcast %19604 : tensor<4x?x4x64xcomplex<f32>>{%dim_17151} -> tensor<4x?x4x128xf32>{%dim_17151}
%19606 = torch_c.from_builtin_tensor %19605 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17152 = torch.constant.int 5
%19607 = torch.prims.convert_element_type %19606, %int5_17152 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17153 = torch.constant.int 1
%19608 = torch.aten.size.int %19288, %int1_17153 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_17154 = torch.constant.int 0
%19609 = torch.aten.add.int %int0_17154, %19608 : !torch.int, !torch.int -> !torch.int
%int0_17155 = torch.constant.int 0
%int0_17156 = torch.constant.int 0
%int1_17157 = torch.constant.int 1
%19610 = torch.aten.slice.Tensor %19502, %int0_17155, %int0_17156, %19609, %int1_17157 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19610, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17158 = torch.constant.int 1
%int0_17159 = torch.constant.int 0
%int9223372036854775807_17160 = torch.constant.int 9223372036854775807
%int1_17161 = torch.constant.int 1
%19611 = torch.aten.slice.Tensor %19610, %int1_17158, %int0_17159, %int9223372036854775807_17160, %int1_17161 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19611, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17162 = torch.constant.int 0
%19612 = torch.aten.unsqueeze %19611, %int0_17162 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19612, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17163 = torch.constant.int 2
%19613 = torch.aten.unsqueeze %19612, %int2_17163 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19613, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17164 = torch.constant.int 3
%int0_17165 = torch.constant.int 0
%int9223372036854775807_17166 = torch.constant.int 9223372036854775807
%int1_17167 = torch.constant.int 1
%19614 = torch.aten.slice.Tensor %19613, %int3_17164, %int0_17165, %int9223372036854775807_17166, %int1_17167 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19614, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19615 = torch_c.to_builtin_tensor %19432 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_17168 = arith.constant 1 : index
%dim_17169 = tensor.dim %19615, %c1_17168 : tensor<4x?x4x128xf16>
%19616 = flow.tensor.bitcast %19615 : tensor<4x?x4x128xf16>{%dim_17169} -> tensor<4x?x4x64xcomplex<f16>>{%dim_17169}
%19617 = torch_c.from_builtin_tensor %19616 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %19617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%19618 = torch.aten.mul.Tensor %19617, %19614 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %19618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%19619 = torch_c.to_builtin_tensor %19618 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_17170 = arith.constant 1 : index
%dim_17171 = tensor.dim %19619, %c1_17170 : tensor<4x?x4x64xcomplex<f32>>
%19620 = flow.tensor.bitcast %19619 : tensor<4x?x4x64xcomplex<f32>>{%dim_17171} -> tensor<4x?x4x128xf32>{%dim_17171}
%19621 = torch_c.from_builtin_tensor %19620 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %19621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_17172 = torch.constant.int 5
%19622 = torch.prims.convert_element_type %19621, %int5_17172 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %19622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_17173 = torch.constant.int 131072
%none_17174 = torch.constant.none
%none_17175 = torch.constant.none
%cpu_17176 = torch.constant.device "cpu"
%false_17177 = torch.constant.bool false
%19623 = torch.aten.arange %int131072_17173, %none_17174, %none_17175, %cpu_17176, %false_17177 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_17178 = torch.constant.int 0
%int128_17179 = torch.constant.int 128
%int2_17180 = torch.constant.int 2
%none_17181 = torch.constant.none
%none_17182 = torch.constant.none
%cpu_17183 = torch.constant.device "cpu"
%false_17184 = torch.constant.bool false
%19624 = torch.aten.arange.start_step %int0_17178, %int128_17179, %int2_17180, %none_17181, %none_17182, %cpu_17183, %false_17184 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_17185 = torch.constant.int 0
%int0_17186 = torch.constant.int 0
%int64_17187 = torch.constant.int 64
%int1_17188 = torch.constant.int 1
%19625 = torch.aten.slice.Tensor %19624, %int0_17185, %int0_17186, %int64_17187, %int1_17188 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_17189 = torch.constant.int 6
%19626 = torch.prims.convert_element_type %19625, %int6_17189 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_17190 = torch.constant.int 128
%19627 = torch.aten.div.Scalar %19626, %int128_17190 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_17191 = torch.constant.float 5.000000e+05
%19628 = torch.aten.pow.Scalar %float5.000000e05_17191, %19627 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%19629 = torch.aten.reciprocal %19628 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_17192 = torch.constant.float 1.000000e+00
%19630 = torch.aten.mul.Scalar %19629, %float1.000000e00_17192 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_17193 = torch.constant.int 131072
%int1_17194 = torch.constant.int 1
%19631 = torch.prim.ListConstruct %int131072_17193, %int1_17194 : (!torch.int, !torch.int) -> !torch.list<int>
%19632 = torch.aten.view %19623, %19631 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%19633 = torch.aten.mul.Tensor %19632, %19630 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%19634 = torch.aten.cos %19633 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%19635 = torch.aten.sin %19633 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%19636 = torch.aten.complex %19634, %19635 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%19637 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19638 = flow.tensor.transfer %19637 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%19639 = torch_c.from_builtin_tensor %19638 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19640 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19641 = flow.tensor.transfer %19640 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%19642 = torch_c.from_builtin_tensor %19641 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19643 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19644 = flow.tensor.transfer %19643 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%19645 = torch_c.from_builtin_tensor %19644 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19646 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19647 = flow.tensor.transfer %19646 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%19648 = torch_c.from_builtin_tensor %19647 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19649 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19650 = flow.tensor.transfer %19649 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%19651 = torch_c.from_builtin_tensor %19650 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19652 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19653 = flow.tensor.transfer %19652 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%19654 = torch_c.from_builtin_tensor %19653 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19655 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19656 = flow.tensor.transfer %19655 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%19657 = torch_c.from_builtin_tensor %19656 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%19658 = torch_c.to_builtin_tensor %19636 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%19659 = flow.tensor.transfer %19658 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%19660 = torch_c.from_builtin_tensor %19659 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_17195 = torch.constant.int 1
%19661 = torch.aten.size.int %19310, %int1_17195 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17196 = torch.constant.int 0
%19662 = torch.aten.add.int %int0_17196, %19661 : !torch.int, !torch.int -> !torch.int
%int0_17197 = torch.constant.int 0
%int0_17198 = torch.constant.int 0
%int1_17199 = torch.constant.int 1
%19663 = torch.aten.slice.Tensor %19639, %int0_17197, %int0_17198, %19662, %int1_17199 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19663, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17200 = torch.constant.int 1
%int0_17201 = torch.constant.int 0
%int9223372036854775807_17202 = torch.constant.int 9223372036854775807
%int1_17203 = torch.constant.int 1
%19664 = torch.aten.slice.Tensor %19663, %int1_17200, %int0_17201, %int9223372036854775807_17202, %int1_17203 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19664, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17204 = torch.constant.int 0
%19665 = torch.aten.unsqueeze %19664, %int0_17204 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19665, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17205 = torch.constant.int 2
%19666 = torch.aten.unsqueeze %19665, %int2_17205 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19666, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17206 = torch.constant.int 3
%int0_17207 = torch.constant.int 0
%int9223372036854775807_17208 = torch.constant.int 9223372036854775807
%int1_17209 = torch.constant.int 1
%19667 = torch.aten.slice.Tensor %19666, %int3_17206, %int0_17207, %int9223372036854775807_17208, %int1_17209 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19667, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19668 = torch_c.to_builtin_tensor %19434 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17210 = arith.constant 1 : index
%dim_17211 = tensor.dim %19668, %c1_17210 : tensor<4x?x1x128xf16>
%19669 = flow.tensor.bitcast %19668 : tensor<4x?x1x128xf16>{%dim_17211} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17211}
%19670 = torch_c.from_builtin_tensor %19669 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19671 = torch.aten.mul.Tensor %19670, %19667 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19672 = torch_c.to_builtin_tensor %19671 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17212 = arith.constant 1 : index
%dim_17213 = tensor.dim %19672, %c1_17212 : tensor<4x?x1x64xcomplex<f32>>
%19673 = flow.tensor.bitcast %19672 : tensor<4x?x1x64xcomplex<f32>>{%dim_17213} -> tensor<4x?x1x128xf32>{%dim_17213}
%19674 = torch_c.from_builtin_tensor %19673 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17214 = torch.constant.int 5
%19675 = torch.prims.convert_element_type %19674, %int5_17214 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17215 = torch.constant.int 1
%19676 = torch.aten.size.int %19316, %int1_17215 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17216 = torch.constant.int 0
%19677 = torch.aten.add.int %int0_17216, %19676 : !torch.int, !torch.int -> !torch.int
%int0_17217 = torch.constant.int 0
%int0_17218 = torch.constant.int 0
%int1_17219 = torch.constant.int 1
%19678 = torch.aten.slice.Tensor %19642, %int0_17217, %int0_17218, %19677, %int1_17219 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19678, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17220 = torch.constant.int 1
%int0_17221 = torch.constant.int 0
%int9223372036854775807_17222 = torch.constant.int 9223372036854775807
%int1_17223 = torch.constant.int 1
%19679 = torch.aten.slice.Tensor %19678, %int1_17220, %int0_17221, %int9223372036854775807_17222, %int1_17223 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19679, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17224 = torch.constant.int 0
%19680 = torch.aten.unsqueeze %19679, %int0_17224 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19680, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17225 = torch.constant.int 2
%19681 = torch.aten.unsqueeze %19680, %int2_17225 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19681, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17226 = torch.constant.int 3
%int0_17227 = torch.constant.int 0
%int9223372036854775807_17228 = torch.constant.int 9223372036854775807
%int1_17229 = torch.constant.int 1
%19682 = torch.aten.slice.Tensor %19681, %int3_17226, %int0_17227, %int9223372036854775807_17228, %int1_17229 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19682, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19683 = torch_c.to_builtin_tensor %19436 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17230 = arith.constant 1 : index
%dim_17231 = tensor.dim %19683, %c1_17230 : tensor<4x?x1x128xf16>
%19684 = flow.tensor.bitcast %19683 : tensor<4x?x1x128xf16>{%dim_17231} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17231}
%19685 = torch_c.from_builtin_tensor %19684 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19686 = torch.aten.mul.Tensor %19685, %19682 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19687 = torch_c.to_builtin_tensor %19686 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17232 = arith.constant 1 : index
%dim_17233 = tensor.dim %19687, %c1_17232 : tensor<4x?x1x64xcomplex<f32>>
%19688 = flow.tensor.bitcast %19687 : tensor<4x?x1x64xcomplex<f32>>{%dim_17233} -> tensor<4x?x1x128xf32>{%dim_17233}
%19689 = torch_c.from_builtin_tensor %19688 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17234 = torch.constant.int 5
%19690 = torch.prims.convert_element_type %19689, %int5_17234 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17235 = torch.constant.int 1
%19691 = torch.aten.size.int %19322, %int1_17235 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17236 = torch.constant.int 0
%19692 = torch.aten.add.int %int0_17236, %19691 : !torch.int, !torch.int -> !torch.int
%int0_17237 = torch.constant.int 0
%int0_17238 = torch.constant.int 0
%int1_17239 = torch.constant.int 1
%19693 = torch.aten.slice.Tensor %19645, %int0_17237, %int0_17238, %19692, %int1_17239 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19693, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17240 = torch.constant.int 1
%int0_17241 = torch.constant.int 0
%int9223372036854775807_17242 = torch.constant.int 9223372036854775807
%int1_17243 = torch.constant.int 1
%19694 = torch.aten.slice.Tensor %19693, %int1_17240, %int0_17241, %int9223372036854775807_17242, %int1_17243 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19694, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17244 = torch.constant.int 0
%19695 = torch.aten.unsqueeze %19694, %int0_17244 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19695, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17245 = torch.constant.int 2
%19696 = torch.aten.unsqueeze %19695, %int2_17245 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19696, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17246 = torch.constant.int 3
%int0_17247 = torch.constant.int 0
%int9223372036854775807_17248 = torch.constant.int 9223372036854775807
%int1_17249 = torch.constant.int 1
%19697 = torch.aten.slice.Tensor %19696, %int3_17246, %int0_17247, %int9223372036854775807_17248, %int1_17249 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19697, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19698 = torch_c.to_builtin_tensor %19438 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17250 = arith.constant 1 : index
%dim_17251 = tensor.dim %19698, %c1_17250 : tensor<4x?x1x128xf16>
%19699 = flow.tensor.bitcast %19698 : tensor<4x?x1x128xf16>{%dim_17251} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17251}
%19700 = torch_c.from_builtin_tensor %19699 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19701 = torch.aten.mul.Tensor %19700, %19697 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19702 = torch_c.to_builtin_tensor %19701 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17252 = arith.constant 1 : index
%dim_17253 = tensor.dim %19702, %c1_17252 : tensor<4x?x1x64xcomplex<f32>>
%19703 = flow.tensor.bitcast %19702 : tensor<4x?x1x64xcomplex<f32>>{%dim_17253} -> tensor<4x?x1x128xf32>{%dim_17253}
%19704 = torch_c.from_builtin_tensor %19703 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17254 = torch.constant.int 5
%19705 = torch.prims.convert_element_type %19704, %int5_17254 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17255 = torch.constant.int 1
%19706 = torch.aten.size.int %19328, %int1_17255 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17256 = torch.constant.int 0
%19707 = torch.aten.add.int %int0_17256, %19706 : !torch.int, !torch.int -> !torch.int
%int0_17257 = torch.constant.int 0
%int0_17258 = torch.constant.int 0
%int1_17259 = torch.constant.int 1
%19708 = torch.aten.slice.Tensor %19648, %int0_17257, %int0_17258, %19707, %int1_17259 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19708, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17260 = torch.constant.int 1
%int0_17261 = torch.constant.int 0
%int9223372036854775807_17262 = torch.constant.int 9223372036854775807
%int1_17263 = torch.constant.int 1
%19709 = torch.aten.slice.Tensor %19708, %int1_17260, %int0_17261, %int9223372036854775807_17262, %int1_17263 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19709, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17264 = torch.constant.int 0
%19710 = torch.aten.unsqueeze %19709, %int0_17264 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19710, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17265 = torch.constant.int 2
%19711 = torch.aten.unsqueeze %19710, %int2_17265 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19711, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17266 = torch.constant.int 3
%int0_17267 = torch.constant.int 0
%int9223372036854775807_17268 = torch.constant.int 9223372036854775807
%int1_17269 = torch.constant.int 1
%19712 = torch.aten.slice.Tensor %19711, %int3_17266, %int0_17267, %int9223372036854775807_17268, %int1_17269 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19712, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19713 = torch_c.to_builtin_tensor %19440 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17270 = arith.constant 1 : index
%dim_17271 = tensor.dim %19713, %c1_17270 : tensor<4x?x1x128xf16>
%19714 = flow.tensor.bitcast %19713 : tensor<4x?x1x128xf16>{%dim_17271} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17271}
%19715 = torch_c.from_builtin_tensor %19714 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19716 = torch.aten.mul.Tensor %19715, %19712 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19717 = torch_c.to_builtin_tensor %19716 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17272 = arith.constant 1 : index
%dim_17273 = tensor.dim %19717, %c1_17272 : tensor<4x?x1x64xcomplex<f32>>
%19718 = flow.tensor.bitcast %19717 : tensor<4x?x1x64xcomplex<f32>>{%dim_17273} -> tensor<4x?x1x128xf32>{%dim_17273}
%19719 = torch_c.from_builtin_tensor %19718 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17274 = torch.constant.int 5
%19720 = torch.prims.convert_element_type %19719, %int5_17274 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17275 = torch.constant.int 1
%19721 = torch.aten.size.int %19334, %int1_17275 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17276 = torch.constant.int 0
%19722 = torch.aten.add.int %int0_17276, %19721 : !torch.int, !torch.int -> !torch.int
%int0_17277 = torch.constant.int 0
%int0_17278 = torch.constant.int 0
%int1_17279 = torch.constant.int 1
%19723 = torch.aten.slice.Tensor %19651, %int0_17277, %int0_17278, %19722, %int1_17279 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19723, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17280 = torch.constant.int 1
%int0_17281 = torch.constant.int 0
%int9223372036854775807_17282 = torch.constant.int 9223372036854775807
%int1_17283 = torch.constant.int 1
%19724 = torch.aten.slice.Tensor %19723, %int1_17280, %int0_17281, %int9223372036854775807_17282, %int1_17283 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19724, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17284 = torch.constant.int 0
%19725 = torch.aten.unsqueeze %19724, %int0_17284 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19725, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17285 = torch.constant.int 2
%19726 = torch.aten.unsqueeze %19725, %int2_17285 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19726, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17286 = torch.constant.int 3
%int0_17287 = torch.constant.int 0
%int9223372036854775807_17288 = torch.constant.int 9223372036854775807
%int1_17289 = torch.constant.int 1
%19727 = torch.aten.slice.Tensor %19726, %int3_17286, %int0_17287, %int9223372036854775807_17288, %int1_17289 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19727, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19728 = torch_c.to_builtin_tensor %19442 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17290 = arith.constant 1 : index
%dim_17291 = tensor.dim %19728, %c1_17290 : tensor<4x?x1x128xf16>
%19729 = flow.tensor.bitcast %19728 : tensor<4x?x1x128xf16>{%dim_17291} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17291}
%19730 = torch_c.from_builtin_tensor %19729 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19731 = torch.aten.mul.Tensor %19730, %19727 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19732 = torch_c.to_builtin_tensor %19731 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17292 = arith.constant 1 : index
%dim_17293 = tensor.dim %19732, %c1_17292 : tensor<4x?x1x64xcomplex<f32>>
%19733 = flow.tensor.bitcast %19732 : tensor<4x?x1x64xcomplex<f32>>{%dim_17293} -> tensor<4x?x1x128xf32>{%dim_17293}
%19734 = torch_c.from_builtin_tensor %19733 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17294 = torch.constant.int 5
%19735 = torch.prims.convert_element_type %19734, %int5_17294 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17295 = torch.constant.int 1
%19736 = torch.aten.size.int %19340, %int1_17295 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17296 = torch.constant.int 0
%19737 = torch.aten.add.int %int0_17296, %19736 : !torch.int, !torch.int -> !torch.int
%int0_17297 = torch.constant.int 0
%int0_17298 = torch.constant.int 0
%int1_17299 = torch.constant.int 1
%19738 = torch.aten.slice.Tensor %19654, %int0_17297, %int0_17298, %19737, %int1_17299 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19738, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17300 = torch.constant.int 1
%int0_17301 = torch.constant.int 0
%int9223372036854775807_17302 = torch.constant.int 9223372036854775807
%int1_17303 = torch.constant.int 1
%19739 = torch.aten.slice.Tensor %19738, %int1_17300, %int0_17301, %int9223372036854775807_17302, %int1_17303 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19739, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17304 = torch.constant.int 0
%19740 = torch.aten.unsqueeze %19739, %int0_17304 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19740, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17305 = torch.constant.int 2
%19741 = torch.aten.unsqueeze %19740, %int2_17305 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19741, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17306 = torch.constant.int 3
%int0_17307 = torch.constant.int 0
%int9223372036854775807_17308 = torch.constant.int 9223372036854775807
%int1_17309 = torch.constant.int 1
%19742 = torch.aten.slice.Tensor %19741, %int3_17306, %int0_17307, %int9223372036854775807_17308, %int1_17309 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19742, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19743 = torch_c.to_builtin_tensor %19444 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17310 = arith.constant 1 : index
%dim_17311 = tensor.dim %19743, %c1_17310 : tensor<4x?x1x128xf16>
%19744 = flow.tensor.bitcast %19743 : tensor<4x?x1x128xf16>{%dim_17311} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17311}
%19745 = torch_c.from_builtin_tensor %19744 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19746 = torch.aten.mul.Tensor %19745, %19742 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19747 = torch_c.to_builtin_tensor %19746 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17312 = arith.constant 1 : index
%dim_17313 = tensor.dim %19747, %c1_17312 : tensor<4x?x1x64xcomplex<f32>>
%19748 = flow.tensor.bitcast %19747 : tensor<4x?x1x64xcomplex<f32>>{%dim_17313} -> tensor<4x?x1x128xf32>{%dim_17313}
%19749 = torch_c.from_builtin_tensor %19748 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17314 = torch.constant.int 5
%19750 = torch.prims.convert_element_type %19749, %int5_17314 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17315 = torch.constant.int 1
%19751 = torch.aten.size.int %19346, %int1_17315 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17316 = torch.constant.int 0
%19752 = torch.aten.add.int %int0_17316, %19751 : !torch.int, !torch.int -> !torch.int
%int0_17317 = torch.constant.int 0
%int0_17318 = torch.constant.int 0
%int1_17319 = torch.constant.int 1
%19753 = torch.aten.slice.Tensor %19657, %int0_17317, %int0_17318, %19752, %int1_17319 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19753, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17320 = torch.constant.int 1
%int0_17321 = torch.constant.int 0
%int9223372036854775807_17322 = torch.constant.int 9223372036854775807
%int1_17323 = torch.constant.int 1
%19754 = torch.aten.slice.Tensor %19753, %int1_17320, %int0_17321, %int9223372036854775807_17322, %int1_17323 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19754, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17324 = torch.constant.int 0
%19755 = torch.aten.unsqueeze %19754, %int0_17324 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19755, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17325 = torch.constant.int 2
%19756 = torch.aten.unsqueeze %19755, %int2_17325 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19756, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17326 = torch.constant.int 3
%int0_17327 = torch.constant.int 0
%int9223372036854775807_17328 = torch.constant.int 9223372036854775807
%int1_17329 = torch.constant.int 1
%19757 = torch.aten.slice.Tensor %19756, %int3_17326, %int0_17327, %int9223372036854775807_17328, %int1_17329 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19757, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19758 = torch_c.to_builtin_tensor %19446 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17330 = arith.constant 1 : index
%dim_17331 = tensor.dim %19758, %c1_17330 : tensor<4x?x1x128xf16>
%19759 = flow.tensor.bitcast %19758 : tensor<4x?x1x128xf16>{%dim_17331} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17331}
%19760 = torch_c.from_builtin_tensor %19759 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19761 = torch.aten.mul.Tensor %19760, %19757 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19762 = torch_c.to_builtin_tensor %19761 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17332 = arith.constant 1 : index
%dim_17333 = tensor.dim %19762, %c1_17332 : tensor<4x?x1x64xcomplex<f32>>
%19763 = flow.tensor.bitcast %19762 : tensor<4x?x1x64xcomplex<f32>>{%dim_17333} -> tensor<4x?x1x128xf32>{%dim_17333}
%19764 = torch_c.from_builtin_tensor %19763 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17334 = torch.constant.int 5
%19765 = torch.prims.convert_element_type %19764, %int5_17334 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_17335 = torch.constant.int 1
%19766 = torch.aten.size.int %19352, %int1_17335 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_17336 = torch.constant.int 0
%19767 = torch.aten.add.int %int0_17336, %19766 : !torch.int, !torch.int -> !torch.int
%int0_17337 = torch.constant.int 0
%int0_17338 = torch.constant.int 0
%int1_17339 = torch.constant.int 1
%19768 = torch.aten.slice.Tensor %19660, %int0_17337, %int0_17338, %19767, %int1_17339 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19768, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_17340 = torch.constant.int 1
%int0_17341 = torch.constant.int 0
%int9223372036854775807_17342 = torch.constant.int 9223372036854775807
%int1_17343 = torch.constant.int 1
%19769 = torch.aten.slice.Tensor %19768, %int1_17340, %int0_17341, %int9223372036854775807_17342, %int1_17343 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %19769, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_17344 = torch.constant.int 0
%19770 = torch.aten.unsqueeze %19769, %int0_17344 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %19770, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_17345 = torch.constant.int 2
%19771 = torch.aten.unsqueeze %19770, %int2_17345 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19771, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_17346 = torch.constant.int 3
%int0_17347 = torch.constant.int 0
%int9223372036854775807_17348 = torch.constant.int 9223372036854775807
%int1_17349 = torch.constant.int 1
%19772 = torch.aten.slice.Tensor %19771, %int3_17346, %int0_17347, %int9223372036854775807_17348, %int1_17349 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19772, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%19773 = torch_c.to_builtin_tensor %19448 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_17350 = arith.constant 1 : index
%dim_17351 = tensor.dim %19773, %c1_17350 : tensor<4x?x1x128xf16>
%19774 = flow.tensor.bitcast %19773 : tensor<4x?x1x128xf16>{%dim_17351} -> tensor<4x?x1x64xcomplex<f16>>{%dim_17351}
%19775 = torch_c.from_builtin_tensor %19774 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %19775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%19776 = torch.aten.mul.Tensor %19775, %19772 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %19776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%19777 = torch_c.to_builtin_tensor %19776 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_17352 = arith.constant 1 : index
%dim_17353 = tensor.dim %19777, %c1_17352 : tensor<4x?x1x64xcomplex<f32>>
%19778 = flow.tensor.bitcast %19777 : tensor<4x?x1x64xcomplex<f32>>{%dim_17353} -> tensor<4x?x1x128xf32>{%dim_17353}
%19779 = torch_c.from_builtin_tensor %19778 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %19779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_17354 = torch.constant.int 5
%19780 = torch.prims.convert_element_type %19779, %int5_17354 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %19780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_17355 = torch.constant.int 64
%19781 = torch.aten.mul.Scalar %2364, %int64_17355 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19781, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17356 = torch.constant.int 64
%19782 = torch.aten.mul.Scalar %2367, %int64_17356 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19782, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17357 = torch.constant.int 64
%19783 = torch.aten.mul.Scalar %2370, %int64_17357 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19783, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17358 = torch.constant.int 64
%19784 = torch.aten.mul.Scalar %2373, %int64_17358 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19784, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17359 = torch.constant.int 64
%19785 = torch.aten.mul.Scalar %2376, %int64_17359 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19785, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17360 = torch.constant.int 64
%19786 = torch.aten.mul.Scalar %2379, %int64_17360 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19786, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17361 = torch.constant.int 64
%19787 = torch.aten.mul.Scalar %2382, %int64_17361 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19787, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_17362 = torch.constant.int 64
%19788 = torch.aten.mul.Scalar %2385, %int64_17362 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19788, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18 = torch.constant.int 18
%int1_17363 = torch.constant.int 1
%19789 = torch.aten.add.Scalar %19781, %int18, %int1_17363 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19789, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17364 = torch.constant.int 18
%int1_17365 = torch.constant.int 1
%19790 = torch.aten.add.Scalar %19782, %int18_17364, %int1_17365 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19790, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17366 = torch.constant.int 18
%int1_17367 = torch.constant.int 1
%19791 = torch.aten.add.Scalar %19783, %int18_17366, %int1_17367 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19791, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17368 = torch.constant.int 18
%int1_17369 = torch.constant.int 1
%19792 = torch.aten.add.Scalar %19784, %int18_17368, %int1_17369 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19792, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17370 = torch.constant.int 18
%int1_17371 = torch.constant.int 1
%19793 = torch.aten.add.Scalar %19785, %int18_17370, %int1_17371 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19793, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17372 = torch.constant.int 18
%int1_17373 = torch.constant.int 1
%19794 = torch.aten.add.Scalar %19786, %int18_17372, %int1_17373 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19794, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17374 = torch.constant.int 18
%int1_17375 = torch.constant.int 1
%19795 = torch.aten.add.Scalar %19787, %int18_17374, %int1_17375 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19795, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int18_17376 = torch.constant.int 18
%int1_17377 = torch.constant.int 1
%19796 = torch.aten.add.Scalar %19788, %int18_17376, %int1_17377 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19796, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_17378 = torch.constant.int 4
%int16_17379 = torch.constant.int 16
%int1_17380 = torch.constant.int 1
%int128_17381 = torch.constant.int 128
%19797 = torch.prim.ListConstruct %int4_17378, %3095, %int16_17379, %int1_17380, %int128_17381 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19798 = torch.aten.view %19675, %19797 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19798, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17382 = torch.constant.int 4
%int16_17383 = torch.constant.int 16
%int1_17384 = torch.constant.int 1
%int128_17385 = torch.constant.int 128
%19799 = torch.prim.ListConstruct %int4_17382, %3095, %int16_17383, %int1_17384, %int128_17385 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19800 = torch.aten.view %19690, %19799 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19800, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17386 = torch.constant.int 4
%int16_17387 = torch.constant.int 16
%int1_17388 = torch.constant.int 1
%int128_17389 = torch.constant.int 128
%19801 = torch.prim.ListConstruct %int4_17386, %3095, %int16_17387, %int1_17388, %int128_17389 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19802 = torch.aten.view %19705, %19801 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19802, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17390 = torch.constant.int 4
%int16_17391 = torch.constant.int 16
%int1_17392 = torch.constant.int 1
%int128_17393 = torch.constant.int 128
%19803 = torch.prim.ListConstruct %int4_17390, %3095, %int16_17391, %int1_17392, %int128_17393 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19804 = torch.aten.view %19720, %19803 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19804, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17394 = torch.constant.int 4
%int16_17395 = torch.constant.int 16
%int1_17396 = torch.constant.int 1
%int128_17397 = torch.constant.int 128
%19805 = torch.prim.ListConstruct %int4_17394, %3095, %int16_17395, %int1_17396, %int128_17397 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19806 = torch.aten.view %19735, %19805 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19806, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17398 = torch.constant.int 4
%int16_17399 = torch.constant.int 16
%int1_17400 = torch.constant.int 1
%int128_17401 = torch.constant.int 128
%19807 = torch.prim.ListConstruct %int4_17398, %3095, %int16_17399, %int1_17400, %int128_17401 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19808 = torch.aten.view %19750, %19807 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19808, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17402 = torch.constant.int 4
%int16_17403 = torch.constant.int 16
%int1_17404 = torch.constant.int 1
%int128_17405 = torch.constant.int 128
%19809 = torch.prim.ListConstruct %int4_17402, %3095, %int16_17403, %int1_17404, %int128_17405 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19810 = torch.aten.view %19765, %19809 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19810, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17406 = torch.constant.int 4
%int16_17407 = torch.constant.int 16
%int1_17408 = torch.constant.int 1
%int128_17409 = torch.constant.int 128
%19811 = torch.prim.ListConstruct %int4_17406, %3095, %int16_17407, %int1_17408, %int128_17409 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19812 = torch.aten.view %19780, %19811 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19812, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17410 = torch.constant.int 4
%19813 = torch.aten.mul.int %int4_17410, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17411 = torch.constant.int 16
%int1_17412 = torch.constant.int 1
%int128_17413 = torch.constant.int 128
%19814 = torch.prim.ListConstruct %19813, %int16_17411, %int1_17412, %int128_17413 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19815 = torch.aten.view %19798, %19814 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19815, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17414 = torch.constant.int 4
%19816 = torch.aten.mul.int %int4_17414, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17415 = torch.constant.int 16
%int1_17416 = torch.constant.int 1
%int128_17417 = torch.constant.int 128
%19817 = torch.prim.ListConstruct %19816, %int16_17415, %int1_17416, %int128_17417 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19818 = torch.aten.view %19800, %19817 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19818, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17418 = torch.constant.int 4
%19819 = torch.aten.mul.int %int4_17418, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17419 = torch.constant.int 16
%int1_17420 = torch.constant.int 1
%int128_17421 = torch.constant.int 128
%19820 = torch.prim.ListConstruct %19819, %int16_17419, %int1_17420, %int128_17421 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19821 = torch.aten.view %19802, %19820 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19821, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17422 = torch.constant.int 4
%19822 = torch.aten.mul.int %int4_17422, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17423 = torch.constant.int 16
%int1_17424 = torch.constant.int 1
%int128_17425 = torch.constant.int 128
%19823 = torch.prim.ListConstruct %19822, %int16_17423, %int1_17424, %int128_17425 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19824 = torch.aten.view %19804, %19823 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19824, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17426 = torch.constant.int 4
%19825 = torch.aten.mul.int %int4_17426, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17427 = torch.constant.int 16
%int1_17428 = torch.constant.int 1
%int128_17429 = torch.constant.int 128
%19826 = torch.prim.ListConstruct %19825, %int16_17427, %int1_17428, %int128_17429 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19827 = torch.aten.view %19806, %19826 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19827, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17430 = torch.constant.int 4
%19828 = torch.aten.mul.int %int4_17430, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17431 = torch.constant.int 16
%int1_17432 = torch.constant.int 1
%int128_17433 = torch.constant.int 128
%19829 = torch.prim.ListConstruct %19828, %int16_17431, %int1_17432, %int128_17433 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19830 = torch.aten.view %19808, %19829 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19830, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17434 = torch.constant.int 4
%19831 = torch.aten.mul.int %int4_17434, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17435 = torch.constant.int 16
%int1_17436 = torch.constant.int 1
%int128_17437 = torch.constant.int 128
%19832 = torch.prim.ListConstruct %19831, %int16_17435, %int1_17436, %int128_17437 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19833 = torch.aten.view %19810, %19832 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19833, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17438 = torch.constant.int 4
%19834 = torch.aten.mul.int %int4_17438, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17439 = torch.constant.int 16
%int1_17440 = torch.constant.int 1
%int128_17441 = torch.constant.int 128
%19835 = torch.prim.ListConstruct %19834, %int16_17439, %int1_17440, %int128_17441 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19836 = torch.aten.view %19812, %19835 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19836, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17442 = torch.constant.int 4
%19837 = torch.aten.mul.int %int4_17442, %3095 : !torch.int, !torch.int -> !torch.int
%19838 = torch.prim.ListConstruct %19837 : (!torch.int) -> !torch.list<int>
%19839 = torch.aten.view %19789, %19838 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19839, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17443 = torch.constant.int 4
%19840 = torch.aten.mul.int %int4_17443, %3095 : !torch.int, !torch.int -> !torch.int
%19841 = torch.prim.ListConstruct %19840 : (!torch.int) -> !torch.list<int>
%19842 = torch.aten.view %19790, %19841 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19842, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17444 = torch.constant.int 4
%19843 = torch.aten.mul.int %int4_17444, %3095 : !torch.int, !torch.int -> !torch.int
%19844 = torch.prim.ListConstruct %19843 : (!torch.int) -> !torch.list<int>
%19845 = torch.aten.view %19791, %19844 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19845, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17445 = torch.constant.int 4
%19846 = torch.aten.mul.int %int4_17445, %3095 : !torch.int, !torch.int -> !torch.int
%19847 = torch.prim.ListConstruct %19846 : (!torch.int) -> !torch.list<int>
%19848 = torch.aten.view %19792, %19847 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19848, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17446 = torch.constant.int 4
%19849 = torch.aten.mul.int %int4_17446, %3095 : !torch.int, !torch.int -> !torch.int
%19850 = torch.prim.ListConstruct %19849 : (!torch.int) -> !torch.list<int>
%19851 = torch.aten.view %19793, %19850 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19851, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17447 = torch.constant.int 4
%19852 = torch.aten.mul.int %int4_17447, %3095 : !torch.int, !torch.int -> !torch.int
%19853 = torch.prim.ListConstruct %19852 : (!torch.int) -> !torch.list<int>
%19854 = torch.aten.view %19794, %19853 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19854, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17448 = torch.constant.int 4
%19855 = torch.aten.mul.int %int4_17448, %3095 : !torch.int, !torch.int -> !torch.int
%19856 = torch.prim.ListConstruct %19855 : (!torch.int) -> !torch.list<int>
%19857 = torch.aten.view %19795, %19856 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19857, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17449 = torch.constant.int 4
%19858 = torch.aten.mul.int %int4_17449, %3095 : !torch.int, !torch.int -> !torch.int
%19859 = torch.prim.ListConstruct %19858 : (!torch.int) -> !torch.list<int>
%19860 = torch.aten.view %19796, %19859 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19860, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17450 = torch.constant.int 4
%int16_17451 = torch.constant.int 16
%int1_17452 = torch.constant.int 1
%int128_17453 = torch.constant.int 128
%19861 = torch.prim.ListConstruct %int4_17450, %3095, %int16_17451, %int1_17452, %int128_17453 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19862 = torch.aten.view %19450, %19861 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19862, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17454 = torch.constant.int 4
%int16_17455 = torch.constant.int 16
%int1_17456 = torch.constant.int 1
%int128_17457 = torch.constant.int 128
%19863 = torch.prim.ListConstruct %int4_17454, %3095, %int16_17455, %int1_17456, %int128_17457 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19864 = torch.aten.view %19452, %19863 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19864, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17458 = torch.constant.int 4
%int16_17459 = torch.constant.int 16
%int1_17460 = torch.constant.int 1
%int128_17461 = torch.constant.int 128
%19865 = torch.prim.ListConstruct %int4_17458, %3095, %int16_17459, %int1_17460, %int128_17461 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19866 = torch.aten.view %19454, %19865 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19866, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17462 = torch.constant.int 4
%int16_17463 = torch.constant.int 16
%int1_17464 = torch.constant.int 1
%int128_17465 = torch.constant.int 128
%19867 = torch.prim.ListConstruct %int4_17462, %3095, %int16_17463, %int1_17464, %int128_17465 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19868 = torch.aten.view %19456, %19867 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19868, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17466 = torch.constant.int 4
%int16_17467 = torch.constant.int 16
%int1_17468 = torch.constant.int 1
%int128_17469 = torch.constant.int 128
%19869 = torch.prim.ListConstruct %int4_17466, %3095, %int16_17467, %int1_17468, %int128_17469 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19870 = torch.aten.view %19458, %19869 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19870, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17470 = torch.constant.int 4
%int16_17471 = torch.constant.int 16
%int1_17472 = torch.constant.int 1
%int128_17473 = torch.constant.int 128
%19871 = torch.prim.ListConstruct %int4_17470, %3095, %int16_17471, %int1_17472, %int128_17473 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19872 = torch.aten.view %19460, %19871 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19872, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17474 = torch.constant.int 4
%int16_17475 = torch.constant.int 16
%int1_17476 = torch.constant.int 1
%int128_17477 = torch.constant.int 128
%19873 = torch.prim.ListConstruct %int4_17474, %3095, %int16_17475, %int1_17476, %int128_17477 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19874 = torch.aten.view %19462, %19873 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19874, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17478 = torch.constant.int 4
%int16_17479 = torch.constant.int 16
%int1_17480 = torch.constant.int 1
%int128_17481 = torch.constant.int 128
%19875 = torch.prim.ListConstruct %int4_17478, %3095, %int16_17479, %int1_17480, %int128_17481 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19876 = torch.aten.view %19464, %19875 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %19876, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_17482 = torch.constant.int 4
%19877 = torch.aten.mul.int %int4_17482, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17483 = torch.constant.int 16
%int1_17484 = torch.constant.int 1
%int128_17485 = torch.constant.int 128
%19878 = torch.prim.ListConstruct %19877, %int16_17483, %int1_17484, %int128_17485 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19879 = torch.aten.view %19862, %19878 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19879, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17486 = torch.constant.int 4
%19880 = torch.aten.mul.int %int4_17486, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17487 = torch.constant.int 16
%int1_17488 = torch.constant.int 1
%int128_17489 = torch.constant.int 128
%19881 = torch.prim.ListConstruct %19880, %int16_17487, %int1_17488, %int128_17489 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19882 = torch.aten.view %19864, %19881 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19882, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17490 = torch.constant.int 4
%19883 = torch.aten.mul.int %int4_17490, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17491 = torch.constant.int 16
%int1_17492 = torch.constant.int 1
%int128_17493 = torch.constant.int 128
%19884 = torch.prim.ListConstruct %19883, %int16_17491, %int1_17492, %int128_17493 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19885 = torch.aten.view %19866, %19884 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19885, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17494 = torch.constant.int 4
%19886 = torch.aten.mul.int %int4_17494, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17495 = torch.constant.int 16
%int1_17496 = torch.constant.int 1
%int128_17497 = torch.constant.int 128
%19887 = torch.prim.ListConstruct %19886, %int16_17495, %int1_17496, %int128_17497 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19888 = torch.aten.view %19868, %19887 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19888, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17498 = torch.constant.int 4
%19889 = torch.aten.mul.int %int4_17498, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17499 = torch.constant.int 16
%int1_17500 = torch.constant.int 1
%int128_17501 = torch.constant.int 128
%19890 = torch.prim.ListConstruct %19889, %int16_17499, %int1_17500, %int128_17501 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19891 = torch.aten.view %19870, %19890 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19891, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17502 = torch.constant.int 4
%19892 = torch.aten.mul.int %int4_17502, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17503 = torch.constant.int 16
%int1_17504 = torch.constant.int 1
%int128_17505 = torch.constant.int 128
%19893 = torch.prim.ListConstruct %19892, %int16_17503, %int1_17504, %int128_17505 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19894 = torch.aten.view %19872, %19893 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19894, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17506 = torch.constant.int 4
%19895 = torch.aten.mul.int %int4_17506, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17507 = torch.constant.int 16
%int1_17508 = torch.constant.int 1
%int128_17509 = torch.constant.int 128
%19896 = torch.prim.ListConstruct %19895, %int16_17507, %int1_17508, %int128_17509 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19897 = torch.aten.view %19874, %19896 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19897, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_17510 = torch.constant.int 4
%19898 = torch.aten.mul.int %int4_17510, %3095 : !torch.int, !torch.int -> !torch.int
%int16_17511 = torch.constant.int 16
%int1_17512 = torch.constant.int 1
%int128_17513 = torch.constant.int 128
%19899 = torch.prim.ListConstruct %19898, %int16_17511, %int1_17512, %int128_17513 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19900 = torch.aten.view %19876, %19899 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19900, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_17514 = torch.constant.int 1
%int1_17515 = torch.constant.int 1
%19901 = torch.aten.add.Scalar %19789, %int1_17514, %int1_17515 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19901, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17516 = torch.constant.int 1
%int1_17517 = torch.constant.int 1
%19902 = torch.aten.add.Scalar %19790, %int1_17516, %int1_17517 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19902, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17518 = torch.constant.int 1
%int1_17519 = torch.constant.int 1
%19903 = torch.aten.add.Scalar %19791, %int1_17518, %int1_17519 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19903, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17520 = torch.constant.int 1
%int1_17521 = torch.constant.int 1
%19904 = torch.aten.add.Scalar %19792, %int1_17520, %int1_17521 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19904, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17522 = torch.constant.int 1
%int1_17523 = torch.constant.int 1
%19905 = torch.aten.add.Scalar %19793, %int1_17522, %int1_17523 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19905, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17524 = torch.constant.int 1
%int1_17525 = torch.constant.int 1
%19906 = torch.aten.add.Scalar %19794, %int1_17524, %int1_17525 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19906, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17526 = torch.constant.int 1
%int1_17527 = torch.constant.int 1
%19907 = torch.aten.add.Scalar %19795, %int1_17526, %int1_17527 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19907, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_17528 = torch.constant.int 1
%int1_17529 = torch.constant.int 1
%19908 = torch.aten.add.Scalar %19796, %int1_17528, %int1_17529 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %19908, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_17530 = torch.constant.int 4
%19909 = torch.aten.mul.int %int4_17530, %3095 : !torch.int, !torch.int -> !torch.int
%19910 = torch.prim.ListConstruct %19909 : (!torch.int) -> !torch.list<int>
%19911 = torch.aten.view %19901, %19910 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19911, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17531 = torch.constant.int 4
%19912 = torch.aten.mul.int %int4_17531, %3095 : !torch.int, !torch.int -> !torch.int
%19913 = torch.prim.ListConstruct %19912 : (!torch.int) -> !torch.list<int>
%19914 = torch.aten.view %19902, %19913 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19914, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17532 = torch.constant.int 4
%19915 = torch.aten.mul.int %int4_17532, %3095 : !torch.int, !torch.int -> !torch.int
%19916 = torch.prim.ListConstruct %19915 : (!torch.int) -> !torch.list<int>
%19917 = torch.aten.view %19903, %19916 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19917, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17533 = torch.constant.int 4
%19918 = torch.aten.mul.int %int4_17533, %3095 : !torch.int, !torch.int -> !torch.int
%19919 = torch.prim.ListConstruct %19918 : (!torch.int) -> !torch.list<int>
%19920 = torch.aten.view %19904, %19919 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19920, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17534 = torch.constant.int 4
%19921 = torch.aten.mul.int %int4_17534, %3095 : !torch.int, !torch.int -> !torch.int
%19922 = torch.prim.ListConstruct %19921 : (!torch.int) -> !torch.list<int>
%19923 = torch.aten.view %19905, %19922 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19923, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17535 = torch.constant.int 4
%19924 = torch.aten.mul.int %int4_17535, %3095 : !torch.int, !torch.int -> !torch.int
%19925 = torch.prim.ListConstruct %19924 : (!torch.int) -> !torch.list<int>
%19926 = torch.aten.view %19906, %19925 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19926, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17536 = torch.constant.int 4
%19927 = torch.aten.mul.int %int4_17536, %3095 : !torch.int, !torch.int -> !torch.int
%19928 = torch.prim.ListConstruct %19927 : (!torch.int) -> !torch.list<int>
%19929 = torch.aten.view %19907, %19928 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19929, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_17537 = torch.constant.int 4
%19930 = torch.aten.mul.int %int4_17537, %3095 : !torch.int, !torch.int -> !torch.int
%19931 = torch.prim.ListConstruct %19930 : (!torch.int) -> !torch.list<int>
%19932 = torch.aten.view %19908, %19931 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19932, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%19933 = torch.prim.ListConstruct %19839, %19911 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17538 = torch.constant.int 0
%19934 = torch.aten.cat %19933, %int0_17538 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19934, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19935 = torch.prim.ListConstruct %19842, %19914 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17539 = torch.constant.int 0
%19936 = torch.aten.cat %19935, %int0_17539 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19936, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19937 = torch.prim.ListConstruct %19845, %19917 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17540 = torch.constant.int 0
%19938 = torch.aten.cat %19937, %int0_17540 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19938, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19939 = torch.prim.ListConstruct %19848, %19920 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17541 = torch.constant.int 0
%19940 = torch.aten.cat %19939, %int0_17541 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19940, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19941 = torch.prim.ListConstruct %19851, %19923 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17542 = torch.constant.int 0
%19942 = torch.aten.cat %19941, %int0_17542 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19942, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19943 = torch.prim.ListConstruct %19854, %19926 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17543 = torch.constant.int 0
%19944 = torch.aten.cat %19943, %int0_17543 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19944, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19945 = torch.prim.ListConstruct %19857, %19929 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17544 = torch.constant.int 0
%19946 = torch.aten.cat %19945, %int0_17544 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19946, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19947 = torch.prim.ListConstruct %19860, %19932 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_17545 = torch.constant.int 0
%19948 = torch.aten.cat %19947, %int0_17545 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %19948, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%19949 = torch.prim.ListConstruct %19815, %19879 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17546 = torch.constant.int 0
%19950 = torch.aten.cat %19949, %int0_17546 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19950, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19951 = torch.prim.ListConstruct %19818, %19882 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17547 = torch.constant.int 0
%19952 = torch.aten.cat %19951, %int0_17547 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19952, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19953 = torch.prim.ListConstruct %19821, %19885 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17548 = torch.constant.int 0
%19954 = torch.aten.cat %19953, %int0_17548 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19954, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19955 = torch.prim.ListConstruct %19824, %19888 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17549 = torch.constant.int 0
%19956 = torch.aten.cat %19955, %int0_17549 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19956, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19957 = torch.prim.ListConstruct %19827, %19891 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17550 = torch.constant.int 0
%19958 = torch.aten.cat %19957, %int0_17550 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19958, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19959 = torch.prim.ListConstruct %19830, %19894 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17551 = torch.constant.int 0
%19960 = torch.aten.cat %19959, %int0_17551 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19960, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19961 = torch.prim.ListConstruct %19833, %19897 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17552 = torch.constant.int 0
%19962 = torch.aten.cat %19961, %int0_17552 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19962, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19963 = torch.prim.ListConstruct %19836, %19900 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_17553 = torch.constant.int 0
%19964 = torch.aten.cat %19963, %int0_17553 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19964, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17554 = torch.constant.int 32
%int2_17555 = torch.constant.int 2
%int16_17556 = torch.constant.int 16
%int1_17557 = torch.constant.int 1
%int128_17558 = torch.constant.int 128
%19965 = torch.prim.ListConstruct %3023, %int32_17554, %int2_17555, %int16_17556, %int1_17557, %int128_17558 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19966 = torch.aten.view %18115, %19965 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %19966, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17559 = torch.constant.int 32
%19967 = torch.aten.mul.int %3023, %int32_17559 : !torch.int, !torch.int -> !torch.int
%int2_17560 = torch.constant.int 2
%19968 = torch.aten.mul.int %19967, %int2_17560 : !torch.int, !torch.int -> !torch.int
%int16_17561 = torch.constant.int 16
%int1_17562 = torch.constant.int 1
%int128_17563 = torch.constant.int 128
%19969 = torch.prim.ListConstruct %19968, %int16_17561, %int1_17562, %int128_17563 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19970 = torch.aten.view %19966, %19969 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19970, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19971 = torch.prim.ListConstruct %19934 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17564 = torch.constant.bool false
%19972 = torch.aten.index_put %19970, %19971, %19950, %false_17564 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19972, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17565 = torch.constant.int 32
%int2_17566 = torch.constant.int 2
%int16_17567 = torch.constant.int 16
%int1_17568 = torch.constant.int 1
%int128_17569 = torch.constant.int 128
%19973 = torch.prim.ListConstruct %3023, %int32_17565, %int2_17566, %int16_17567, %int1_17568, %int128_17569 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19974 = torch.aten.view %19972, %19973 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %19974, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17570 = torch.constant.int 131072
%19975 = torch.prim.ListConstruct %3023, %int131072_17570 : (!torch.int, !torch.int) -> !torch.list<int>
%19976 = torch.aten.view %19974, %19975 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %19976, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17571 = torch.constant.int 32
%int2_17572 = torch.constant.int 2
%int16_17573 = torch.constant.int 16
%int1_17574 = torch.constant.int 1
%int128_17575 = torch.constant.int 128
%19977 = torch.prim.ListConstruct %3026, %int32_17571, %int2_17572, %int16_17573, %int1_17574, %int128_17575 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19978 = torch.aten.view %18127, %19977 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %19978, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17576 = torch.constant.int 32
%19979 = torch.aten.mul.int %3026, %int32_17576 : !torch.int, !torch.int -> !torch.int
%int2_17577 = torch.constant.int 2
%19980 = torch.aten.mul.int %19979, %int2_17577 : !torch.int, !torch.int -> !torch.int
%int16_17578 = torch.constant.int 16
%int1_17579 = torch.constant.int 1
%int128_17580 = torch.constant.int 128
%19981 = torch.prim.ListConstruct %19980, %int16_17578, %int1_17579, %int128_17580 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19982 = torch.aten.view %19978, %19981 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19982, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19983 = torch.prim.ListConstruct %19936 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17581 = torch.constant.bool false
%19984 = torch.aten.index_put %19982, %19983, %19952, %false_17581 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19984, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17582 = torch.constant.int 32
%int2_17583 = torch.constant.int 2
%int16_17584 = torch.constant.int 16
%int1_17585 = torch.constant.int 1
%int128_17586 = torch.constant.int 128
%19985 = torch.prim.ListConstruct %3026, %int32_17582, %int2_17583, %int16_17584, %int1_17585, %int128_17586 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19986 = torch.aten.view %19984, %19985 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %19986, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17587 = torch.constant.int 131072
%19987 = torch.prim.ListConstruct %3026, %int131072_17587 : (!torch.int, !torch.int) -> !torch.list<int>
%19988 = torch.aten.view %19986, %19987 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %19988, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17588 = torch.constant.int 32
%int2_17589 = torch.constant.int 2
%int16_17590 = torch.constant.int 16
%int1_17591 = torch.constant.int 1
%int128_17592 = torch.constant.int 128
%19989 = torch.prim.ListConstruct %3029, %int32_17588, %int2_17589, %int16_17590, %int1_17591, %int128_17592 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19990 = torch.aten.view %18139, %19989 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %19990, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17593 = torch.constant.int 32
%19991 = torch.aten.mul.int %3029, %int32_17593 : !torch.int, !torch.int -> !torch.int
%int2_17594 = torch.constant.int 2
%19992 = torch.aten.mul.int %19991, %int2_17594 : !torch.int, !torch.int -> !torch.int
%int16_17595 = torch.constant.int 16
%int1_17596 = torch.constant.int 1
%int128_17597 = torch.constant.int 128
%19993 = torch.prim.ListConstruct %19992, %int16_17595, %int1_17596, %int128_17597 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19994 = torch.aten.view %19990, %19993 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19994, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%19995 = torch.prim.ListConstruct %19938 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17598 = torch.constant.bool false
%19996 = torch.aten.index_put %19994, %19995, %19954, %false_17598 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %19996, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17599 = torch.constant.int 32
%int2_17600 = torch.constant.int 2
%int16_17601 = torch.constant.int 16
%int1_17602 = torch.constant.int 1
%int128_17603 = torch.constant.int 128
%19997 = torch.prim.ListConstruct %3029, %int32_17599, %int2_17600, %int16_17601, %int1_17602, %int128_17603 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%19998 = torch.aten.view %19996, %19997 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %19998, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17604 = torch.constant.int 131072
%19999 = torch.prim.ListConstruct %3029, %int131072_17604 : (!torch.int, !torch.int) -> !torch.list<int>
%20000 = torch.aten.view %19998, %19999 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %20000, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17605 = torch.constant.int 32
%int2_17606 = torch.constant.int 2
%int16_17607 = torch.constant.int 16
%int1_17608 = torch.constant.int 1
%int128_17609 = torch.constant.int 128
%20001 = torch.prim.ListConstruct %3032, %int32_17605, %int2_17606, %int16_17607, %int1_17608, %int128_17609 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20002 = torch.aten.view %18151, %20001 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20002, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17610 = torch.constant.int 32
%20003 = torch.aten.mul.int %3032, %int32_17610 : !torch.int, !torch.int -> !torch.int
%int2_17611 = torch.constant.int 2
%20004 = torch.aten.mul.int %20003, %int2_17611 : !torch.int, !torch.int -> !torch.int
%int16_17612 = torch.constant.int 16
%int1_17613 = torch.constant.int 1
%int128_17614 = torch.constant.int 128
%20005 = torch.prim.ListConstruct %20004, %int16_17612, %int1_17613, %int128_17614 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20006 = torch.aten.view %20002, %20005 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20006, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%20007 = torch.prim.ListConstruct %19940 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17615 = torch.constant.bool false
%20008 = torch.aten.index_put %20006, %20007, %19956, %false_17615 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20008, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17616 = torch.constant.int 32
%int2_17617 = torch.constant.int 2
%int16_17618 = torch.constant.int 16
%int1_17619 = torch.constant.int 1
%int128_17620 = torch.constant.int 128
%20009 = torch.prim.ListConstruct %3032, %int32_17616, %int2_17617, %int16_17618, %int1_17619, %int128_17620 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20010 = torch.aten.view %20008, %20009 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20010, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17621 = torch.constant.int 131072
%20011 = torch.prim.ListConstruct %3032, %int131072_17621 : (!torch.int, !torch.int) -> !torch.list<int>
%20012 = torch.aten.view %20010, %20011 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %20012, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17622 = torch.constant.int 32
%int2_17623 = torch.constant.int 2
%int16_17624 = torch.constant.int 16
%int1_17625 = torch.constant.int 1
%int128_17626 = torch.constant.int 128
%20013 = torch.prim.ListConstruct %3035, %int32_17622, %int2_17623, %int16_17624, %int1_17625, %int128_17626 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20014 = torch.aten.view %18163, %20013 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20014, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17627 = torch.constant.int 32
%20015 = torch.aten.mul.int %3035, %int32_17627 : !torch.int, !torch.int -> !torch.int
%int2_17628 = torch.constant.int 2
%20016 = torch.aten.mul.int %20015, %int2_17628 : !torch.int, !torch.int -> !torch.int
%int16_17629 = torch.constant.int 16
%int1_17630 = torch.constant.int 1
%int128_17631 = torch.constant.int 128
%20017 = torch.prim.ListConstruct %20016, %int16_17629, %int1_17630, %int128_17631 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20018 = torch.aten.view %20014, %20017 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20018, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%20019 = torch.prim.ListConstruct %19942 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17632 = torch.constant.bool false
%20020 = torch.aten.index_put %20018, %20019, %19958, %false_17632 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20020, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17633 = torch.constant.int 32
%int2_17634 = torch.constant.int 2
%int16_17635 = torch.constant.int 16
%int1_17636 = torch.constant.int 1
%int128_17637 = torch.constant.int 128
%20021 = torch.prim.ListConstruct %3035, %int32_17633, %int2_17634, %int16_17635, %int1_17636, %int128_17637 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20022 = torch.aten.view %20020, %20021 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20022, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17638 = torch.constant.int 131072
%20023 = torch.prim.ListConstruct %3035, %int131072_17638 : (!torch.int, !torch.int) -> !torch.list<int>
%20024 = torch.aten.view %20022, %20023 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %20024, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17639 = torch.constant.int 32
%int2_17640 = torch.constant.int 2
%int16_17641 = torch.constant.int 16
%int1_17642 = torch.constant.int 1
%int128_17643 = torch.constant.int 128
%20025 = torch.prim.ListConstruct %3038, %int32_17639, %int2_17640, %int16_17641, %int1_17642, %int128_17643 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20026 = torch.aten.view %18175, %20025 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20026, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17644 = torch.constant.int 32
%20027 = torch.aten.mul.int %3038, %int32_17644 : !torch.int, !torch.int -> !torch.int
%int2_17645 = torch.constant.int 2
%20028 = torch.aten.mul.int %20027, %int2_17645 : !torch.int, !torch.int -> !torch.int
%int16_17646 = torch.constant.int 16
%int1_17647 = torch.constant.int 1
%int128_17648 = torch.constant.int 128
%20029 = torch.prim.ListConstruct %20028, %int16_17646, %int1_17647, %int128_17648 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20030 = torch.aten.view %20026, %20029 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20030, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%20031 = torch.prim.ListConstruct %19944 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17649 = torch.constant.bool false
%20032 = torch.aten.index_put %20030, %20031, %19960, %false_17649 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20032, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17650 = torch.constant.int 32
%int2_17651 = torch.constant.int 2
%int16_17652 = torch.constant.int 16
%int1_17653 = torch.constant.int 1
%int128_17654 = torch.constant.int 128
%20033 = torch.prim.ListConstruct %3038, %int32_17650, %int2_17651, %int16_17652, %int1_17653, %int128_17654 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20034 = torch.aten.view %20032, %20033 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20034, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17655 = torch.constant.int 131072
%20035 = torch.prim.ListConstruct %3038, %int131072_17655 : (!torch.int, !torch.int) -> !torch.list<int>
%20036 = torch.aten.view %20034, %20035 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %20036, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17656 = torch.constant.int 32
%int2_17657 = torch.constant.int 2
%int16_17658 = torch.constant.int 16
%int1_17659 = torch.constant.int 1
%int128_17660 = torch.constant.int 128
%20037 = torch.prim.ListConstruct %3041, %int32_17656, %int2_17657, %int16_17658, %int1_17659, %int128_17660 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20038 = torch.aten.view %18187, %20037 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20038, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17661 = torch.constant.int 32
%20039 = torch.aten.mul.int %3041, %int32_17661 : !torch.int, !torch.int -> !torch.int
%int2_17662 = torch.constant.int 2
%20040 = torch.aten.mul.int %20039, %int2_17662 : !torch.int, !torch.int -> !torch.int
%int16_17663 = torch.constant.int 16
%int1_17664 = torch.constant.int 1
%int128_17665 = torch.constant.int 128
%20041 = torch.prim.ListConstruct %20040, %int16_17663, %int1_17664, %int128_17665 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20042 = torch.aten.view %20038, %20041 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20042, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%20043 = torch.prim.ListConstruct %19946 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17666 = torch.constant.bool false
%20044 = torch.aten.index_put %20042, %20043, %19962, %false_17666 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20044, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17667 = torch.constant.int 32
%int2_17668 = torch.constant.int 2
%int16_17669 = torch.constant.int 16
%int1_17670 = torch.constant.int 1
%int128_17671 = torch.constant.int 128
%20045 = torch.prim.ListConstruct %3041, %int32_17667, %int2_17668, %int16_17669, %int1_17670, %int128_17671 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20046 = torch.aten.view %20044, %20045 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20046, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17672 = torch.constant.int 131072
%20047 = torch.prim.ListConstruct %3041, %int131072_17672 : (!torch.int, !torch.int) -> !torch.list<int>
%20048 = torch.aten.view %20046, %20047 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %20048, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_17673 = torch.constant.int 32
%int2_17674 = torch.constant.int 2
%int16_17675 = torch.constant.int 16
%int1_17676 = torch.constant.int 1
%int128_17677 = torch.constant.int 128
%20049 = torch.prim.ListConstruct %3044, %int32_17673, %int2_17674, %int16_17675, %int1_17676, %int128_17677 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20050 = torch.aten.view %18199, %20049 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20050, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_17678 = torch.constant.int 32
%20051 = torch.aten.mul.int %3044, %int32_17678 : !torch.int, !torch.int -> !torch.int
%int2_17679 = torch.constant.int 2
%20052 = torch.aten.mul.int %20051, %int2_17679 : !torch.int, !torch.int -> !torch.int
%int16_17680 = torch.constant.int 16
%int1_17681 = torch.constant.int 1
%int128_17682 = torch.constant.int 128
%20053 = torch.prim.ListConstruct %20052, %int16_17680, %int1_17681, %int128_17682 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20054 = torch.aten.view %20050, %20053 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20054, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%20055 = torch.prim.ListConstruct %19948 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_17683 = torch.constant.bool false
%20056 = torch.aten.index_put %20054, %20055, %19964, %false_17683 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %20056, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_17684 = torch.constant.int 32
%int2_17685 = torch.constant.int 2
%int16_17686 = torch.constant.int 16
%int1_17687 = torch.constant.int 1
%int128_17688 = torch.constant.int 128
%20057 = torch.prim.ListConstruct %3044, %int32_17684, %int2_17685, %int16_17686, %int1_17687, %int128_17688 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20058 = torch.aten.view %20056, %20057 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %20058, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_17689 = torch.constant.int 131072
%20059 = torch.prim.ListConstruct %3044, %int131072_17689 : (!torch.int, !torch.int) -> !torch.list<int>
%20060 = torch.aten.view %20058, %20059 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %20060, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_17690 = torch.constant.int -2
%20061 = torch.aten.unsqueeze %19675, %int-2_17690 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17691 = torch.constant.int -2
%20062 = torch.aten.unsqueeze %19690, %int-2_17691 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17692 = torch.constant.int -2
%20063 = torch.aten.unsqueeze %19705, %int-2_17692 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17693 = torch.constant.int -2
%20064 = torch.aten.unsqueeze %19720, %int-2_17693 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17694 = torch.constant.int -2
%20065 = torch.aten.unsqueeze %19735, %int-2_17694 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17695 = torch.constant.int -2
%20066 = torch.aten.unsqueeze %19750, %int-2_17695 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17696 = torch.constant.int -2
%20067 = torch.aten.unsqueeze %19765, %int-2_17696 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17697 = torch.constant.int -2
%20068 = torch.aten.unsqueeze %19780, %int-2_17697 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_17698 = torch.constant.int 4
%int1_17699 = torch.constant.int 1
%int4_17700 = torch.constant.int 4
%int128_17701 = torch.constant.int 128
%20069 = torch.prim.ListConstruct %int4_17698, %19661, %int1_17699, %int4_17700, %int128_17701 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17702 = torch.constant.bool false
%20070 = torch.aten.expand %20061, %20069, %false_17702 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17703 = torch.constant.int 4
%int1_17704 = torch.constant.int 1
%int4_17705 = torch.constant.int 4
%int128_17706 = torch.constant.int 128
%20071 = torch.prim.ListConstruct %int4_17703, %19661, %int1_17704, %int4_17705, %int128_17706 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17707 = torch.constant.bool false
%20072 = torch.aten.expand %20062, %20071, %false_17707 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17708 = torch.constant.int 4
%int1_17709 = torch.constant.int 1
%int4_17710 = torch.constant.int 4
%int128_17711 = torch.constant.int 128
%20073 = torch.prim.ListConstruct %int4_17708, %19661, %int1_17709, %int4_17710, %int128_17711 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17712 = torch.constant.bool false
%20074 = torch.aten.expand %20063, %20073, %false_17712 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17713 = torch.constant.int 4
%int1_17714 = torch.constant.int 1
%int4_17715 = torch.constant.int 4
%int128_17716 = torch.constant.int 128
%20075 = torch.prim.ListConstruct %int4_17713, %19661, %int1_17714, %int4_17715, %int128_17716 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17717 = torch.constant.bool false
%20076 = torch.aten.expand %20064, %20075, %false_17717 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17718 = torch.constant.int 4
%int1_17719 = torch.constant.int 1
%int4_17720 = torch.constant.int 4
%int128_17721 = torch.constant.int 128
%20077 = torch.prim.ListConstruct %int4_17718, %19661, %int1_17719, %int4_17720, %int128_17721 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17722 = torch.constant.bool false
%20078 = torch.aten.expand %20065, %20077, %false_17722 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17723 = torch.constant.int 4
%int1_17724 = torch.constant.int 1
%int4_17725 = torch.constant.int 4
%int128_17726 = torch.constant.int 128
%20079 = torch.prim.ListConstruct %int4_17723, %19661, %int1_17724, %int4_17725, %int128_17726 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17727 = torch.constant.bool false
%20080 = torch.aten.expand %20066, %20079, %false_17727 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17728 = torch.constant.int 4
%int1_17729 = torch.constant.int 1
%int4_17730 = torch.constant.int 4
%int128_17731 = torch.constant.int 128
%20081 = torch.prim.ListConstruct %int4_17728, %19661, %int1_17729, %int4_17730, %int128_17731 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17732 = torch.constant.bool false
%20082 = torch.aten.expand %20067, %20081, %false_17732 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17733 = torch.constant.int 4
%int1_17734 = torch.constant.int 1
%int4_17735 = torch.constant.int 4
%int128_17736 = torch.constant.int 128
%20083 = torch.prim.ListConstruct %int4_17733, %19661, %int1_17734, %int4_17735, %int128_17736 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17737 = torch.constant.bool false
%20084 = torch.aten.expand %20068, %20083, %false_17737 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17738 = torch.constant.int 4
%int4_17739 = torch.constant.int 4
%int128_17740 = torch.constant.int 128
%20085 = torch.prim.ListConstruct %int4_17738, %19661, %int4_17739, %int128_17740 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20086 = torch.aten.view %20070, %20085 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17741 = torch.constant.int 4
%int4_17742 = torch.constant.int 4
%int128_17743 = torch.constant.int 128
%20087 = torch.prim.ListConstruct %int4_17741, %19661, %int4_17742, %int128_17743 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20088 = torch.aten.view %20072, %20087 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17744 = torch.constant.int 4
%int4_17745 = torch.constant.int 4
%int128_17746 = torch.constant.int 128
%20089 = torch.prim.ListConstruct %int4_17744, %19661, %int4_17745, %int128_17746 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20090 = torch.aten.view %20074, %20089 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17747 = torch.constant.int 4
%int4_17748 = torch.constant.int 4
%int128_17749 = torch.constant.int 128
%20091 = torch.prim.ListConstruct %int4_17747, %19661, %int4_17748, %int128_17749 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20092 = torch.aten.view %20076, %20091 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17750 = torch.constant.int 4
%int4_17751 = torch.constant.int 4
%int128_17752 = torch.constant.int 128
%20093 = torch.prim.ListConstruct %int4_17750, %19661, %int4_17751, %int128_17752 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20094 = torch.aten.view %20078, %20093 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17753 = torch.constant.int 4
%int4_17754 = torch.constant.int 4
%int128_17755 = torch.constant.int 128
%20095 = torch.prim.ListConstruct %int4_17753, %19661, %int4_17754, %int128_17755 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20096 = torch.aten.view %20080, %20095 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17756 = torch.constant.int 4
%int4_17757 = torch.constant.int 4
%int128_17758 = torch.constant.int 128
%20097 = torch.prim.ListConstruct %int4_17756, %19661, %int4_17757, %int128_17758 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20098 = torch.aten.view %20082, %20097 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17759 = torch.constant.int 4
%int4_17760 = torch.constant.int 4
%int128_17761 = torch.constant.int 128
%20099 = torch.prim.ListConstruct %int4_17759, %19661, %int4_17760, %int128_17761 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20100 = torch.aten.view %20084, %20099 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_17762 = torch.constant.int -2
%20101 = torch.aten.unsqueeze %19450, %int-2_17762 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17763 = torch.constant.int -2
%20102 = torch.aten.unsqueeze %19452, %int-2_17763 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17764 = torch.constant.int -2
%20103 = torch.aten.unsqueeze %19454, %int-2_17764 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17765 = torch.constant.int -2
%20104 = torch.aten.unsqueeze %19456, %int-2_17765 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17766 = torch.constant.int -2
%20105 = torch.aten.unsqueeze %19458, %int-2_17766 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17767 = torch.constant.int -2
%20106 = torch.aten.unsqueeze %19460, %int-2_17767 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17768 = torch.constant.int -2
%20107 = torch.aten.unsqueeze %19462, %int-2_17768 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_17769 = torch.constant.int -2
%20108 = torch.aten.unsqueeze %19464, %int-2_17769 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %20108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_17770 = torch.constant.int 1
%20109 = torch.aten.size.int %19374, %int1_17770 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_17771 = torch.constant.int 4
%int1_17772 = torch.constant.int 1
%int4_17773 = torch.constant.int 4
%int128_17774 = torch.constant.int 128
%20110 = torch.prim.ListConstruct %int4_17771, %20109, %int1_17772, %int4_17773, %int128_17774 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17775 = torch.constant.bool false
%20111 = torch.aten.expand %20101, %20110, %false_17775 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17776 = torch.constant.int 4
%int1_17777 = torch.constant.int 1
%int4_17778 = torch.constant.int 4
%int128_17779 = torch.constant.int 128
%20112 = torch.prim.ListConstruct %int4_17776, %20109, %int1_17777, %int4_17778, %int128_17779 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17780 = torch.constant.bool false
%20113 = torch.aten.expand %20102, %20112, %false_17780 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17781 = torch.constant.int 4
%int1_17782 = torch.constant.int 1
%int4_17783 = torch.constant.int 4
%int128_17784 = torch.constant.int 128
%20114 = torch.prim.ListConstruct %int4_17781, %20109, %int1_17782, %int4_17783, %int128_17784 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17785 = torch.constant.bool false
%20115 = torch.aten.expand %20103, %20114, %false_17785 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17786 = torch.constant.int 4
%int1_17787 = torch.constant.int 1
%int4_17788 = torch.constant.int 4
%int128_17789 = torch.constant.int 128
%20116 = torch.prim.ListConstruct %int4_17786, %20109, %int1_17787, %int4_17788, %int128_17789 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17790 = torch.constant.bool false
%20117 = torch.aten.expand %20104, %20116, %false_17790 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17791 = torch.constant.int 4
%int1_17792 = torch.constant.int 1
%int4_17793 = torch.constant.int 4
%int128_17794 = torch.constant.int 128
%20118 = torch.prim.ListConstruct %int4_17791, %20109, %int1_17792, %int4_17793, %int128_17794 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17795 = torch.constant.bool false
%20119 = torch.aten.expand %20105, %20118, %false_17795 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17796 = torch.constant.int 4
%int1_17797 = torch.constant.int 1
%int4_17798 = torch.constant.int 4
%int128_17799 = torch.constant.int 128
%20120 = torch.prim.ListConstruct %int4_17796, %20109, %int1_17797, %int4_17798, %int128_17799 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17800 = torch.constant.bool false
%20121 = torch.aten.expand %20106, %20120, %false_17800 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17801 = torch.constant.int 4
%int1_17802 = torch.constant.int 1
%int4_17803 = torch.constant.int 4
%int128_17804 = torch.constant.int 128
%20122 = torch.prim.ListConstruct %int4_17801, %20109, %int1_17802, %int4_17803, %int128_17804 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17805 = torch.constant.bool false
%20123 = torch.aten.expand %20107, %20122, %false_17805 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17806 = torch.constant.int 4
%int1_17807 = torch.constant.int 1
%int4_17808 = torch.constant.int 4
%int128_17809 = torch.constant.int 128
%20124 = torch.prim.ListConstruct %int4_17806, %20109, %int1_17807, %int4_17808, %int128_17809 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_17810 = torch.constant.bool false
%20125 = torch.aten.expand %20108, %20124, %false_17810 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %20125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_17811 = torch.constant.int 4
%int4_17812 = torch.constant.int 4
%int128_17813 = torch.constant.int 128
%20126 = torch.prim.ListConstruct %int4_17811, %20109, %int4_17812, %int128_17813 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20127 = torch.aten.view %20111, %20126 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17814 = torch.constant.int 4
%int4_17815 = torch.constant.int 4
%int128_17816 = torch.constant.int 128
%20128 = torch.prim.ListConstruct %int4_17814, %20109, %int4_17815, %int128_17816 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20129 = torch.aten.view %20113, %20128 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17817 = torch.constant.int 4
%int4_17818 = torch.constant.int 4
%int128_17819 = torch.constant.int 128
%20130 = torch.prim.ListConstruct %int4_17817, %20109, %int4_17818, %int128_17819 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20131 = torch.aten.view %20115, %20130 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17820 = torch.constant.int 4
%int4_17821 = torch.constant.int 4
%int128_17822 = torch.constant.int 128
%20132 = torch.prim.ListConstruct %int4_17820, %20109, %int4_17821, %int128_17822 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20133 = torch.aten.view %20117, %20132 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17823 = torch.constant.int 4
%int4_17824 = torch.constant.int 4
%int128_17825 = torch.constant.int 128
%20134 = torch.prim.ListConstruct %int4_17823, %20109, %int4_17824, %int128_17825 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20135 = torch.aten.view %20119, %20134 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17826 = torch.constant.int 4
%int4_17827 = torch.constant.int 4
%int128_17828 = torch.constant.int 128
%20136 = torch.prim.ListConstruct %int4_17826, %20109, %int4_17827, %int128_17828 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20137 = torch.aten.view %20121, %20136 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17829 = torch.constant.int 4
%int4_17830 = torch.constant.int 4
%int128_17831 = torch.constant.int 128
%20138 = torch.prim.ListConstruct %int4_17829, %20109, %int4_17830, %int128_17831 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20139 = torch.aten.view %20123, %20138 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17832 = torch.constant.int 4
%int4_17833 = torch.constant.int 4
%int128_17834 = torch.constant.int 128
%20140 = torch.prim.ListConstruct %int4_17832, %20109, %int4_17833, %int128_17834 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20141 = torch.aten.view %20125, %20140 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17835 = torch.constant.int 1
%int2_17836 = torch.constant.int 2
%20142 = torch.aten.transpose.int %19517, %int1_17835, %int2_17836 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20142, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17837 = torch.constant.int 1
%int2_17838 = torch.constant.int 2
%20143 = torch.aten.transpose.int %19532, %int1_17837, %int2_17838 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20143, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17839 = torch.constant.int 1
%int2_17840 = torch.constant.int 2
%20144 = torch.aten.transpose.int %19547, %int1_17839, %int2_17840 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20144, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17841 = torch.constant.int 1
%int2_17842 = torch.constant.int 2
%20145 = torch.aten.transpose.int %19562, %int1_17841, %int2_17842 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20145, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17843 = torch.constant.int 1
%int2_17844 = torch.constant.int 2
%20146 = torch.aten.transpose.int %19577, %int1_17843, %int2_17844 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20146, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17845 = torch.constant.int 1
%int2_17846 = torch.constant.int 2
%20147 = torch.aten.transpose.int %19592, %int1_17845, %int2_17846 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20147, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17847 = torch.constant.int 1
%int2_17848 = torch.constant.int 2
%20148 = torch.aten.transpose.int %19607, %int1_17847, %int2_17848 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20148, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17849 = torch.constant.int 1
%int2_17850 = torch.constant.int 2
%20149 = torch.aten.transpose.int %19622, %int1_17849, %int2_17850 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20149, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17851 = torch.constant.int 1
%int2_17852 = torch.constant.int 2
%20150 = torch.aten.transpose.int %20086, %int1_17851, %int2_17852 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20150, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17853 = torch.constant.int 1
%int2_17854 = torch.constant.int 2
%20151 = torch.aten.transpose.int %20088, %int1_17853, %int2_17854 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20151, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17855 = torch.constant.int 1
%int2_17856 = torch.constant.int 2
%20152 = torch.aten.transpose.int %20090, %int1_17855, %int2_17856 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20152, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17857 = torch.constant.int 1
%int2_17858 = torch.constant.int 2
%20153 = torch.aten.transpose.int %20092, %int1_17857, %int2_17858 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20153, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17859 = torch.constant.int 1
%int2_17860 = torch.constant.int 2
%20154 = torch.aten.transpose.int %20094, %int1_17859, %int2_17860 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20154, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17861 = torch.constant.int 1
%int2_17862 = torch.constant.int 2
%20155 = torch.aten.transpose.int %20096, %int1_17861, %int2_17862 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20155, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17863 = torch.constant.int 1
%int2_17864 = torch.constant.int 2
%20156 = torch.aten.transpose.int %20098, %int1_17863, %int2_17864 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20156, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17865 = torch.constant.int 1
%int2_17866 = torch.constant.int 2
%20157 = torch.aten.transpose.int %20100, %int1_17865, %int2_17866 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20157, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17867 = torch.constant.int 1
%int2_17868 = torch.constant.int 2
%20158 = torch.aten.transpose.int %20127, %int1_17867, %int2_17868 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20158, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17869 = torch.constant.int 1
%int2_17870 = torch.constant.int 2
%20159 = torch.aten.transpose.int %20129, %int1_17869, %int2_17870 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20159, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17871 = torch.constant.int 1
%int2_17872 = torch.constant.int 2
%20160 = torch.aten.transpose.int %20131, %int1_17871, %int2_17872 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20160, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17873 = torch.constant.int 1
%int2_17874 = torch.constant.int 2
%20161 = torch.aten.transpose.int %20133, %int1_17873, %int2_17874 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20161, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17875 = torch.constant.int 1
%int2_17876 = torch.constant.int 2
%20162 = torch.aten.transpose.int %20135, %int1_17875, %int2_17876 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20162, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17877 = torch.constant.int 1
%int2_17878 = torch.constant.int 2
%20163 = torch.aten.transpose.int %20137, %int1_17877, %int2_17878 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20163, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17879 = torch.constant.int 1
%int2_17880 = torch.constant.int 2
%20164 = torch.aten.transpose.int %20139, %int1_17879, %int2_17880 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20164, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17881 = torch.constant.int 1
%int2_17882 = torch.constant.int 2
%20165 = torch.aten.transpose.int %20141, %int1_17881, %int2_17882 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %20165, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17883 = torch.constant.float 0.000000e+00
%true_17884 = torch.constant.bool true
%none_17885 = torch.constant.none
%none_17886 = torch.constant.none
%20166:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20142, %20150, %20158, %float0.000000e00_17883, %true_17884, %none_17885, %none_17886) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20166#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17887 = torch.constant.float 0.000000e+00
%true_17888 = torch.constant.bool true
%none_17889 = torch.constant.none
%none_17890 = torch.constant.none
%20167:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20143, %20151, %20159, %float0.000000e00_17887, %true_17888, %none_17889, %none_17890) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20167#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17891 = torch.constant.float 0.000000e+00
%true_17892 = torch.constant.bool true
%none_17893 = torch.constant.none
%none_17894 = torch.constant.none
%20168:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20144, %20152, %20160, %float0.000000e00_17891, %true_17892, %none_17893, %none_17894) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20168#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17895 = torch.constant.float 0.000000e+00
%true_17896 = torch.constant.bool true
%none_17897 = torch.constant.none
%none_17898 = torch.constant.none
%20169:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20145, %20153, %20161, %float0.000000e00_17895, %true_17896, %none_17897, %none_17898) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20169#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17899 = torch.constant.float 0.000000e+00
%true_17900 = torch.constant.bool true
%none_17901 = torch.constant.none
%none_17902 = torch.constant.none
%20170:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20146, %20154, %20162, %float0.000000e00_17899, %true_17900, %none_17901, %none_17902) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20170#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17903 = torch.constant.float 0.000000e+00
%true_17904 = torch.constant.bool true
%none_17905 = torch.constant.none
%none_17906 = torch.constant.none
%20171:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20147, %20155, %20163, %float0.000000e00_17903, %true_17904, %none_17905, %none_17906) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20171#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17907 = torch.constant.float 0.000000e+00
%true_17908 = torch.constant.bool true
%none_17909 = torch.constant.none
%none_17910 = torch.constant.none
%20172:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20148, %20156, %20164, %float0.000000e00_17907, %true_17908, %none_17909, %none_17910) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20172#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_17911 = torch.constant.float 0.000000e+00
%true_17912 = torch.constant.bool true
%none_17913 = torch.constant.none
%none_17914 = torch.constant.none
%20173:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%20149, %20157, %20165, %float0.000000e00_17911, %true_17912, %none_17913, %none_17914) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %20173#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_17915 = torch.constant.int 1
%int2_17916 = torch.constant.int 2
%20174 = torch.aten.transpose.int %20166#0, %int1_17915, %int2_17916 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17917 = torch.constant.int 1
%int2_17918 = torch.constant.int 2
%20175 = torch.aten.transpose.int %20167#0, %int1_17917, %int2_17918 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17919 = torch.constant.int 1
%int2_17920 = torch.constant.int 2
%20176 = torch.aten.transpose.int %20168#0, %int1_17919, %int2_17920 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17921 = torch.constant.int 1
%int2_17922 = torch.constant.int 2
%20177 = torch.aten.transpose.int %20169#0, %int1_17921, %int2_17922 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17923 = torch.constant.int 1
%int2_17924 = torch.constant.int 2
%20178 = torch.aten.transpose.int %20170#0, %int1_17923, %int2_17924 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17925 = torch.constant.int 1
%int2_17926 = torch.constant.int 2
%20179 = torch.aten.transpose.int %20171#0, %int1_17925, %int2_17926 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17927 = torch.constant.int 1
%int2_17928 = torch.constant.int 2
%20180 = torch.aten.transpose.int %20172#0, %int1_17927, %int2_17928 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_17929 = torch.constant.int 1
%int2_17930 = torch.constant.int 2
%20181 = torch.aten.transpose.int %20173#0, %int1_17929, %int2_17930 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %20181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_17931 = torch.constant.int 4
%int512_17932 = torch.constant.int 512
%20182 = torch.prim.ListConstruct %int4_17931, %19503, %int512_17932 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20183 = torch.aten.view %20174, %20182 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17933 = torch.constant.int 4
%int512_17934 = torch.constant.int 512
%20184 = torch.prim.ListConstruct %int4_17933, %19518, %int512_17934 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20185 = torch.aten.view %20175, %20184 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17935 = torch.constant.int 4
%int512_17936 = torch.constant.int 512
%20186 = torch.prim.ListConstruct %int4_17935, %19533, %int512_17936 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20187 = torch.aten.view %20176, %20186 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17937 = torch.constant.int 4
%int512_17938 = torch.constant.int 512
%20188 = torch.prim.ListConstruct %int4_17937, %19548, %int512_17938 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20189 = torch.aten.view %20177, %20188 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17939 = torch.constant.int 4
%int512_17940 = torch.constant.int 512
%20190 = torch.prim.ListConstruct %int4_17939, %19563, %int512_17940 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20191 = torch.aten.view %20178, %20190 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17941 = torch.constant.int 4
%int512_17942 = torch.constant.int 512
%20192 = torch.prim.ListConstruct %int4_17941, %19578, %int512_17942 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20193 = torch.aten.view %20179, %20192 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17943 = torch.constant.int 4
%int512_17944 = torch.constant.int 512
%20194 = torch.prim.ListConstruct %int4_17943, %19593, %int512_17944 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20195 = torch.aten.view %20180, %20194 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_17945 = torch.constant.int 4
%int512_17946 = torch.constant.int 512
%20196 = torch.prim.ListConstruct %int4_17945, %19608, %int512_17946 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20197 = torch.aten.view %20181, %20196 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %20197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_17947 = torch.constant.int 1
%int0_17948 = torch.constant.int 0
%20198 = torch.prim.ListConstruct %int1_17947, %int0_17948 : (!torch.int, !torch.int) -> !torch.list<int>
%20199 = torch.aten.permute %688, %20198 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17949 = torch.constant.int 1
%int0_17950 = torch.constant.int 0
%20200 = torch.prim.ListConstruct %int1_17949, %int0_17950 : (!torch.int, !torch.int) -> !torch.list<int>
%20201 = torch.aten.permute %689, %20200 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17951 = torch.constant.int 1
%int0_17952 = torch.constant.int 0
%20202 = torch.prim.ListConstruct %int1_17951, %int0_17952 : (!torch.int, !torch.int) -> !torch.list<int>
%20203 = torch.aten.permute %690, %20202 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17953 = torch.constant.int 1
%int0_17954 = torch.constant.int 0
%20204 = torch.prim.ListConstruct %int1_17953, %int0_17954 : (!torch.int, !torch.int) -> !torch.list<int>
%20205 = torch.aten.permute %691, %20204 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17955 = torch.constant.int 1
%int0_17956 = torch.constant.int 0
%20206 = torch.prim.ListConstruct %int1_17955, %int0_17956 : (!torch.int, !torch.int) -> !torch.list<int>
%20207 = torch.aten.permute %692, %20206 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17957 = torch.constant.int 1
%int0_17958 = torch.constant.int 0
%20208 = torch.prim.ListConstruct %int1_17957, %int0_17958 : (!torch.int, !torch.int) -> !torch.list<int>
%20209 = torch.aten.permute %693, %20208 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17959 = torch.constant.int 1
%int0_17960 = torch.constant.int 0
%20210 = torch.prim.ListConstruct %int1_17959, %int0_17960 : (!torch.int, !torch.int) -> !torch.list<int>
%20211 = torch.aten.permute %694, %20210 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_17961 = torch.constant.int 1
%int0_17962 = torch.constant.int 0
%20212 = torch.prim.ListConstruct %int1_17961, %int0_17962 : (!torch.int, !torch.int) -> !torch.list<int>
%20213 = torch.aten.permute %695, %20212 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_17963 = torch.constant.int 4
%20214 = torch.aten.mul.int %int4_17963, %19503 : !torch.int, !torch.int -> !torch.int
%int512_17964 = torch.constant.int 512
%20215 = torch.prim.ListConstruct %20214, %int512_17964 : (!torch.int, !torch.int) -> !torch.list<int>
%20216 = torch.aten.view %20183, %20215 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20216, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20217 = torch.aten.mm %20216, %20199 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20217, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17965 = torch.constant.int 4
%int4096_17966 = torch.constant.int 4096
%20218 = torch.prim.ListConstruct %int4_17965, %19503, %int4096_17966 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20219 = torch.aten.view %20217, %20218 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17967 = torch.constant.int 4
%20220 = torch.aten.mul.int %int4_17967, %19518 : !torch.int, !torch.int -> !torch.int
%int512_17968 = torch.constant.int 512
%20221 = torch.prim.ListConstruct %20220, %int512_17968 : (!torch.int, !torch.int) -> !torch.list<int>
%20222 = torch.aten.view %20185, %20221 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20222, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20223 = torch.aten.mm %20222, %20201 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20223, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17969 = torch.constant.int 4
%int4096_17970 = torch.constant.int 4096
%20224 = torch.prim.ListConstruct %int4_17969, %19518, %int4096_17970 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20225 = torch.aten.view %20223, %20224 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17971 = torch.constant.int 4
%20226 = torch.aten.mul.int %int4_17971, %19533 : !torch.int, !torch.int -> !torch.int
%int512_17972 = torch.constant.int 512
%20227 = torch.prim.ListConstruct %20226, %int512_17972 : (!torch.int, !torch.int) -> !torch.list<int>
%20228 = torch.aten.view %20187, %20227 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20228, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20229 = torch.aten.mm %20228, %20203 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20229, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17973 = torch.constant.int 4
%int4096_17974 = torch.constant.int 4096
%20230 = torch.prim.ListConstruct %int4_17973, %19533, %int4096_17974 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20231 = torch.aten.view %20229, %20230 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17975 = torch.constant.int 4
%20232 = torch.aten.mul.int %int4_17975, %19548 : !torch.int, !torch.int -> !torch.int
%int512_17976 = torch.constant.int 512
%20233 = torch.prim.ListConstruct %20232, %int512_17976 : (!torch.int, !torch.int) -> !torch.list<int>
%20234 = torch.aten.view %20189, %20233 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20234, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20235 = torch.aten.mm %20234, %20205 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20235, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17977 = torch.constant.int 4
%int4096_17978 = torch.constant.int 4096
%20236 = torch.prim.ListConstruct %int4_17977, %19548, %int4096_17978 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20237 = torch.aten.view %20235, %20236 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17979 = torch.constant.int 4
%20238 = torch.aten.mul.int %int4_17979, %19563 : !torch.int, !torch.int -> !torch.int
%int512_17980 = torch.constant.int 512
%20239 = torch.prim.ListConstruct %20238, %int512_17980 : (!torch.int, !torch.int) -> !torch.list<int>
%20240 = torch.aten.view %20191, %20239 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20240, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20241 = torch.aten.mm %20240, %20207 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20241, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17981 = torch.constant.int 4
%int4096_17982 = torch.constant.int 4096
%20242 = torch.prim.ListConstruct %int4_17981, %19563, %int4096_17982 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20243 = torch.aten.view %20241, %20242 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17983 = torch.constant.int 4
%20244 = torch.aten.mul.int %int4_17983, %19578 : !torch.int, !torch.int -> !torch.int
%int512_17984 = torch.constant.int 512
%20245 = torch.prim.ListConstruct %20244, %int512_17984 : (!torch.int, !torch.int) -> !torch.list<int>
%20246 = torch.aten.view %20193, %20245 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20246, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20247 = torch.aten.mm %20246, %20209 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20247, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17985 = torch.constant.int 4
%int4096_17986 = torch.constant.int 4096
%20248 = torch.prim.ListConstruct %int4_17985, %19578, %int4096_17986 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20249 = torch.aten.view %20247, %20248 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17987 = torch.constant.int 4
%20250 = torch.aten.mul.int %int4_17987, %19593 : !torch.int, !torch.int -> !torch.int
%int512_17988 = torch.constant.int 512
%20251 = torch.prim.ListConstruct %20250, %int512_17988 : (!torch.int, !torch.int) -> !torch.list<int>
%20252 = torch.aten.view %20195, %20251 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20252, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20253 = torch.aten.mm %20252, %20211 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20253, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17989 = torch.constant.int 4
%int4096_17990 = torch.constant.int 4096
%20254 = torch.prim.ListConstruct %int4_17989, %19593, %int4096_17990 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20255 = torch.aten.view %20253, %20254 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_17991 = torch.constant.int 4
%20256 = torch.aten.mul.int %int4_17991, %19608 : !torch.int, !torch.int -> !torch.int
%int512_17992 = torch.constant.int 512
%20257 = torch.prim.ListConstruct %20256, %int512_17992 : (!torch.int, !torch.int) -> !torch.list<int>
%20258 = torch.aten.view %20197, %20257 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %20258, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%20259 = torch.aten.mm %20258, %20213 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20259, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_17993 = torch.constant.int 4
%int4096_17994 = torch.constant.int 4096
%20260 = torch.prim.ListConstruct %int4_17993, %19608, %int4096_17994 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20261 = torch.aten.view %20259, %20260 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20262 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_17995 = arith.constant 1 : index
%dim_17996 = tensor.dim %20262, %c1_17995 : tensor<4x?x4096xf16>
%20263 = flow.tensor.transfer %20262 : tensor<4x?x4096xf16>{%dim_17996} to #hal.device.promise<@__device_0>
%20264 = torch_c.from_builtin_tensor %20263 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20265 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_17997 = arith.constant 1 : index
%dim_17998 = tensor.dim %20265, %c1_17997 : tensor<4x?x4096xf16>
%20266 = flow.tensor.transfer %20265 : tensor<4x?x4096xf16>{%dim_17998} to #hal.device.promise<@__device_0>
%20267 = torch_c.from_builtin_tensor %20266 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20268 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_17999 = arith.constant 1 : index
%dim_18000 = tensor.dim %20268, %c1_17999 : tensor<4x?x4096xf16>
%20269 = flow.tensor.transfer %20268 : tensor<4x?x4096xf16>{%dim_18000} to #hal.device.promise<@__device_0>
%20270 = torch_c.from_builtin_tensor %20269 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20271 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18001 = arith.constant 1 : index
%dim_18002 = tensor.dim %20271, %c1_18001 : tensor<4x?x4096xf16>
%20272 = flow.tensor.transfer %20271 : tensor<4x?x4096xf16>{%dim_18002} to #hal.device.promise<@__device_0>
%20273 = torch_c.from_builtin_tensor %20272 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20274 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18003 = arith.constant 1 : index
%dim_18004 = tensor.dim %20274, %c1_18003 : tensor<4x?x4096xf16>
%20275 = flow.tensor.transfer %20274 : tensor<4x?x4096xf16>{%dim_18004} to #hal.device.promise<@__device_0>
%20276 = torch_c.from_builtin_tensor %20275 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20277 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18005 = arith.constant 1 : index
%dim_18006 = tensor.dim %20277, %c1_18005 : tensor<4x?x4096xf16>
%20278 = flow.tensor.transfer %20277 : tensor<4x?x4096xf16>{%dim_18006} to #hal.device.promise<@__device_0>
%20279 = torch_c.from_builtin_tensor %20278 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20280 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18007 = arith.constant 1 : index
%dim_18008 = tensor.dim %20280, %c1_18007 : tensor<4x?x4096xf16>
%20281 = flow.tensor.transfer %20280 : tensor<4x?x4096xf16>{%dim_18008} to #hal.device.promise<@__device_0>
%20282 = torch_c.from_builtin_tensor %20281 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18009 = torch.constant.int 1
%20283 = torch.aten.add.Tensor %20219, %20264, %int1_18009 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18010 = torch.constant.int 1
%20284 = torch.aten.add.Tensor %20283, %20267, %int1_18010 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18011 = torch.constant.int 1
%20285 = torch.aten.add.Tensor %20284, %20270, %int1_18011 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18012 = torch.constant.int 1
%20286 = torch.aten.add.Tensor %20285, %20273, %int1_18012 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18013 = torch.constant.int 1
%20287 = torch.aten.add.Tensor %20286, %20276, %int1_18013 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18014 = torch.constant.int 1
%20288 = torch.aten.add.Tensor %20287, %20279, %int1_18014 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18015 = torch.constant.int 1
%20289 = torch.aten.add.Tensor %20288, %20282, %int1_18015 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20290 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18016 = arith.constant 1 : index
%dim_18017 = tensor.dim %20290, %c1_18016 : tensor<4x?x4096xf16>
%20291 = flow.tensor.transfer %20290 : tensor<4x?x4096xf16>{%dim_18017} to #hal.device.promise<@__device_1>
%20292 = torch_c.from_builtin_tensor %20291 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20293 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18018 = arith.constant 1 : index
%dim_18019 = tensor.dim %20293, %c1_18018 : tensor<4x?x4096xf16>
%20294 = flow.tensor.transfer %20293 : tensor<4x?x4096xf16>{%dim_18019} to #hal.device.promise<@__device_1>
%20295 = torch_c.from_builtin_tensor %20294 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20296 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18020 = arith.constant 1 : index
%dim_18021 = tensor.dim %20296, %c1_18020 : tensor<4x?x4096xf16>
%20297 = flow.tensor.transfer %20296 : tensor<4x?x4096xf16>{%dim_18021} to #hal.device.promise<@__device_1>
%20298 = torch_c.from_builtin_tensor %20297 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20299 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18022 = arith.constant 1 : index
%dim_18023 = tensor.dim %20299, %c1_18022 : tensor<4x?x4096xf16>
%20300 = flow.tensor.transfer %20299 : tensor<4x?x4096xf16>{%dim_18023} to #hal.device.promise<@__device_1>
%20301 = torch_c.from_builtin_tensor %20300 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20302 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18024 = arith.constant 1 : index
%dim_18025 = tensor.dim %20302, %c1_18024 : tensor<4x?x4096xf16>
%20303 = flow.tensor.transfer %20302 : tensor<4x?x4096xf16>{%dim_18025} to #hal.device.promise<@__device_1>
%20304 = torch_c.from_builtin_tensor %20303 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20305 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18026 = arith.constant 1 : index
%dim_18027 = tensor.dim %20305, %c1_18026 : tensor<4x?x4096xf16>
%20306 = flow.tensor.transfer %20305 : tensor<4x?x4096xf16>{%dim_18027} to #hal.device.promise<@__device_1>
%20307 = torch_c.from_builtin_tensor %20306 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20308 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18028 = arith.constant 1 : index
%dim_18029 = tensor.dim %20308, %c1_18028 : tensor<4x?x4096xf16>
%20309 = flow.tensor.transfer %20308 : tensor<4x?x4096xf16>{%dim_18029} to #hal.device.promise<@__device_1>
%20310 = torch_c.from_builtin_tensor %20309 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18030 = torch.constant.int 1
%20311 = torch.aten.add.Tensor %20292, %20225, %int1_18030 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18031 = torch.constant.int 1
%20312 = torch.aten.add.Tensor %20311, %20295, %int1_18031 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18032 = torch.constant.int 1
%20313 = torch.aten.add.Tensor %20312, %20298, %int1_18032 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18033 = torch.constant.int 1
%20314 = torch.aten.add.Tensor %20313, %20301, %int1_18033 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18034 = torch.constant.int 1
%20315 = torch.aten.add.Tensor %20314, %20304, %int1_18034 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18035 = torch.constant.int 1
%20316 = torch.aten.add.Tensor %20315, %20307, %int1_18035 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18036 = torch.constant.int 1
%20317 = torch.aten.add.Tensor %20316, %20310, %int1_18036 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20318 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18037 = arith.constant 1 : index
%dim_18038 = tensor.dim %20318, %c1_18037 : tensor<4x?x4096xf16>
%20319 = flow.tensor.transfer %20318 : tensor<4x?x4096xf16>{%dim_18038} to #hal.device.promise<@__device_2>
%20320 = torch_c.from_builtin_tensor %20319 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20321 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18039 = arith.constant 1 : index
%dim_18040 = tensor.dim %20321, %c1_18039 : tensor<4x?x4096xf16>
%20322 = flow.tensor.transfer %20321 : tensor<4x?x4096xf16>{%dim_18040} to #hal.device.promise<@__device_2>
%20323 = torch_c.from_builtin_tensor %20322 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20324 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18041 = arith.constant 1 : index
%dim_18042 = tensor.dim %20324, %c1_18041 : tensor<4x?x4096xf16>
%20325 = flow.tensor.transfer %20324 : tensor<4x?x4096xf16>{%dim_18042} to #hal.device.promise<@__device_2>
%20326 = torch_c.from_builtin_tensor %20325 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20327 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18043 = arith.constant 1 : index
%dim_18044 = tensor.dim %20327, %c1_18043 : tensor<4x?x4096xf16>
%20328 = flow.tensor.transfer %20327 : tensor<4x?x4096xf16>{%dim_18044} to #hal.device.promise<@__device_2>
%20329 = torch_c.from_builtin_tensor %20328 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20330 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18045 = arith.constant 1 : index
%dim_18046 = tensor.dim %20330, %c1_18045 : tensor<4x?x4096xf16>
%20331 = flow.tensor.transfer %20330 : tensor<4x?x4096xf16>{%dim_18046} to #hal.device.promise<@__device_2>
%20332 = torch_c.from_builtin_tensor %20331 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20333 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18047 = arith.constant 1 : index
%dim_18048 = tensor.dim %20333, %c1_18047 : tensor<4x?x4096xf16>
%20334 = flow.tensor.transfer %20333 : tensor<4x?x4096xf16>{%dim_18048} to #hal.device.promise<@__device_2>
%20335 = torch_c.from_builtin_tensor %20334 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20336 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18049 = arith.constant 1 : index
%dim_18050 = tensor.dim %20336, %c1_18049 : tensor<4x?x4096xf16>
%20337 = flow.tensor.transfer %20336 : tensor<4x?x4096xf16>{%dim_18050} to #hal.device.promise<@__device_2>
%20338 = torch_c.from_builtin_tensor %20337 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18051 = torch.constant.int 1
%20339 = torch.aten.add.Tensor %20320, %20323, %int1_18051 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18052 = torch.constant.int 1
%20340 = torch.aten.add.Tensor %20339, %20231, %int1_18052 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18053 = torch.constant.int 1
%20341 = torch.aten.add.Tensor %20340, %20326, %int1_18053 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18054 = torch.constant.int 1
%20342 = torch.aten.add.Tensor %20341, %20329, %int1_18054 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18055 = torch.constant.int 1
%20343 = torch.aten.add.Tensor %20342, %20332, %int1_18055 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18056 = torch.constant.int 1
%20344 = torch.aten.add.Tensor %20343, %20335, %int1_18056 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18057 = torch.constant.int 1
%20345 = torch.aten.add.Tensor %20344, %20338, %int1_18057 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20346 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18058 = arith.constant 1 : index
%dim_18059 = tensor.dim %20346, %c1_18058 : tensor<4x?x4096xf16>
%20347 = flow.tensor.transfer %20346 : tensor<4x?x4096xf16>{%dim_18059} to #hal.device.promise<@__device_3>
%20348 = torch_c.from_builtin_tensor %20347 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20349 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18060 = arith.constant 1 : index
%dim_18061 = tensor.dim %20349, %c1_18060 : tensor<4x?x4096xf16>
%20350 = flow.tensor.transfer %20349 : tensor<4x?x4096xf16>{%dim_18061} to #hal.device.promise<@__device_3>
%20351 = torch_c.from_builtin_tensor %20350 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20352 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18062 = arith.constant 1 : index
%dim_18063 = tensor.dim %20352, %c1_18062 : tensor<4x?x4096xf16>
%20353 = flow.tensor.transfer %20352 : tensor<4x?x4096xf16>{%dim_18063} to #hal.device.promise<@__device_3>
%20354 = torch_c.from_builtin_tensor %20353 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20355 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18064 = arith.constant 1 : index
%dim_18065 = tensor.dim %20355, %c1_18064 : tensor<4x?x4096xf16>
%20356 = flow.tensor.transfer %20355 : tensor<4x?x4096xf16>{%dim_18065} to #hal.device.promise<@__device_3>
%20357 = torch_c.from_builtin_tensor %20356 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20358 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18066 = arith.constant 1 : index
%dim_18067 = tensor.dim %20358, %c1_18066 : tensor<4x?x4096xf16>
%20359 = flow.tensor.transfer %20358 : tensor<4x?x4096xf16>{%dim_18067} to #hal.device.promise<@__device_3>
%20360 = torch_c.from_builtin_tensor %20359 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20361 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18068 = arith.constant 1 : index
%dim_18069 = tensor.dim %20361, %c1_18068 : tensor<4x?x4096xf16>
%20362 = flow.tensor.transfer %20361 : tensor<4x?x4096xf16>{%dim_18069} to #hal.device.promise<@__device_3>
%20363 = torch_c.from_builtin_tensor %20362 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20364 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18070 = arith.constant 1 : index
%dim_18071 = tensor.dim %20364, %c1_18070 : tensor<4x?x4096xf16>
%20365 = flow.tensor.transfer %20364 : tensor<4x?x4096xf16>{%dim_18071} to #hal.device.promise<@__device_3>
%20366 = torch_c.from_builtin_tensor %20365 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18072 = torch.constant.int 1
%20367 = torch.aten.add.Tensor %20348, %20351, %int1_18072 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18073 = torch.constant.int 1
%20368 = torch.aten.add.Tensor %20367, %20354, %int1_18073 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18074 = torch.constant.int 1
%20369 = torch.aten.add.Tensor %20368, %20237, %int1_18074 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18075 = torch.constant.int 1
%20370 = torch.aten.add.Tensor %20369, %20357, %int1_18075 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18076 = torch.constant.int 1
%20371 = torch.aten.add.Tensor %20370, %20360, %int1_18076 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18077 = torch.constant.int 1
%20372 = torch.aten.add.Tensor %20371, %20363, %int1_18077 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18078 = torch.constant.int 1
%20373 = torch.aten.add.Tensor %20372, %20366, %int1_18078 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20374 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18079 = arith.constant 1 : index
%dim_18080 = tensor.dim %20374, %c1_18079 : tensor<4x?x4096xf16>
%20375 = flow.tensor.transfer %20374 : tensor<4x?x4096xf16>{%dim_18080} to #hal.device.promise<@__device_4>
%20376 = torch_c.from_builtin_tensor %20375 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20377 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18081 = arith.constant 1 : index
%dim_18082 = tensor.dim %20377, %c1_18081 : tensor<4x?x4096xf16>
%20378 = flow.tensor.transfer %20377 : tensor<4x?x4096xf16>{%dim_18082} to #hal.device.promise<@__device_4>
%20379 = torch_c.from_builtin_tensor %20378 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20380 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18083 = arith.constant 1 : index
%dim_18084 = tensor.dim %20380, %c1_18083 : tensor<4x?x4096xf16>
%20381 = flow.tensor.transfer %20380 : tensor<4x?x4096xf16>{%dim_18084} to #hal.device.promise<@__device_4>
%20382 = torch_c.from_builtin_tensor %20381 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20383 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18085 = arith.constant 1 : index
%dim_18086 = tensor.dim %20383, %c1_18085 : tensor<4x?x4096xf16>
%20384 = flow.tensor.transfer %20383 : tensor<4x?x4096xf16>{%dim_18086} to #hal.device.promise<@__device_4>
%20385 = torch_c.from_builtin_tensor %20384 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20386 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18087 = arith.constant 1 : index
%dim_18088 = tensor.dim %20386, %c1_18087 : tensor<4x?x4096xf16>
%20387 = flow.tensor.transfer %20386 : tensor<4x?x4096xf16>{%dim_18088} to #hal.device.promise<@__device_4>
%20388 = torch_c.from_builtin_tensor %20387 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20389 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18089 = arith.constant 1 : index
%dim_18090 = tensor.dim %20389, %c1_18089 : tensor<4x?x4096xf16>
%20390 = flow.tensor.transfer %20389 : tensor<4x?x4096xf16>{%dim_18090} to #hal.device.promise<@__device_4>
%20391 = torch_c.from_builtin_tensor %20390 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20392 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18091 = arith.constant 1 : index
%dim_18092 = tensor.dim %20392, %c1_18091 : tensor<4x?x4096xf16>
%20393 = flow.tensor.transfer %20392 : tensor<4x?x4096xf16>{%dim_18092} to #hal.device.promise<@__device_4>
%20394 = torch_c.from_builtin_tensor %20393 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18093 = torch.constant.int 1
%20395 = torch.aten.add.Tensor %20376, %20379, %int1_18093 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18094 = torch.constant.int 1
%20396 = torch.aten.add.Tensor %20395, %20382, %int1_18094 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18095 = torch.constant.int 1
%20397 = torch.aten.add.Tensor %20396, %20385, %int1_18095 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18096 = torch.constant.int 1
%20398 = torch.aten.add.Tensor %20397, %20243, %int1_18096 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18097 = torch.constant.int 1
%20399 = torch.aten.add.Tensor %20398, %20388, %int1_18097 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18098 = torch.constant.int 1
%20400 = torch.aten.add.Tensor %20399, %20391, %int1_18098 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18099 = torch.constant.int 1
%20401 = torch.aten.add.Tensor %20400, %20394, %int1_18099 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20402 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18100 = arith.constant 1 : index
%dim_18101 = tensor.dim %20402, %c1_18100 : tensor<4x?x4096xf16>
%20403 = flow.tensor.transfer %20402 : tensor<4x?x4096xf16>{%dim_18101} to #hal.device.promise<@__device_5>
%20404 = torch_c.from_builtin_tensor %20403 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20405 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18102 = arith.constant 1 : index
%dim_18103 = tensor.dim %20405, %c1_18102 : tensor<4x?x4096xf16>
%20406 = flow.tensor.transfer %20405 : tensor<4x?x4096xf16>{%dim_18103} to #hal.device.promise<@__device_5>
%20407 = torch_c.from_builtin_tensor %20406 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20408 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18104 = arith.constant 1 : index
%dim_18105 = tensor.dim %20408, %c1_18104 : tensor<4x?x4096xf16>
%20409 = flow.tensor.transfer %20408 : tensor<4x?x4096xf16>{%dim_18105} to #hal.device.promise<@__device_5>
%20410 = torch_c.from_builtin_tensor %20409 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20411 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18106 = arith.constant 1 : index
%dim_18107 = tensor.dim %20411, %c1_18106 : tensor<4x?x4096xf16>
%20412 = flow.tensor.transfer %20411 : tensor<4x?x4096xf16>{%dim_18107} to #hal.device.promise<@__device_5>
%20413 = torch_c.from_builtin_tensor %20412 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20414 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18108 = arith.constant 1 : index
%dim_18109 = tensor.dim %20414, %c1_18108 : tensor<4x?x4096xf16>
%20415 = flow.tensor.transfer %20414 : tensor<4x?x4096xf16>{%dim_18109} to #hal.device.promise<@__device_5>
%20416 = torch_c.from_builtin_tensor %20415 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20417 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18110 = arith.constant 1 : index
%dim_18111 = tensor.dim %20417, %c1_18110 : tensor<4x?x4096xf16>
%20418 = flow.tensor.transfer %20417 : tensor<4x?x4096xf16>{%dim_18111} to #hal.device.promise<@__device_5>
%20419 = torch_c.from_builtin_tensor %20418 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20420 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18112 = arith.constant 1 : index
%dim_18113 = tensor.dim %20420, %c1_18112 : tensor<4x?x4096xf16>
%20421 = flow.tensor.transfer %20420 : tensor<4x?x4096xf16>{%dim_18113} to #hal.device.promise<@__device_5>
%20422 = torch_c.from_builtin_tensor %20421 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18114 = torch.constant.int 1
%20423 = torch.aten.add.Tensor %20404, %20407, %int1_18114 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18115 = torch.constant.int 1
%20424 = torch.aten.add.Tensor %20423, %20410, %int1_18115 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18116 = torch.constant.int 1
%20425 = torch.aten.add.Tensor %20424, %20413, %int1_18116 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18117 = torch.constant.int 1
%20426 = torch.aten.add.Tensor %20425, %20416, %int1_18117 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18118 = torch.constant.int 1
%20427 = torch.aten.add.Tensor %20426, %20249, %int1_18118 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18119 = torch.constant.int 1
%20428 = torch.aten.add.Tensor %20427, %20419, %int1_18119 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18120 = torch.constant.int 1
%20429 = torch.aten.add.Tensor %20428, %20422, %int1_18120 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20430 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18121 = arith.constant 1 : index
%dim_18122 = tensor.dim %20430, %c1_18121 : tensor<4x?x4096xf16>
%20431 = flow.tensor.transfer %20430 : tensor<4x?x4096xf16>{%dim_18122} to #hal.device.promise<@__device_6>
%20432 = torch_c.from_builtin_tensor %20431 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20433 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18123 = arith.constant 1 : index
%dim_18124 = tensor.dim %20433, %c1_18123 : tensor<4x?x4096xf16>
%20434 = flow.tensor.transfer %20433 : tensor<4x?x4096xf16>{%dim_18124} to #hal.device.promise<@__device_6>
%20435 = torch_c.from_builtin_tensor %20434 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20436 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18125 = arith.constant 1 : index
%dim_18126 = tensor.dim %20436, %c1_18125 : tensor<4x?x4096xf16>
%20437 = flow.tensor.transfer %20436 : tensor<4x?x4096xf16>{%dim_18126} to #hal.device.promise<@__device_6>
%20438 = torch_c.from_builtin_tensor %20437 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20439 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18127 = arith.constant 1 : index
%dim_18128 = tensor.dim %20439, %c1_18127 : tensor<4x?x4096xf16>
%20440 = flow.tensor.transfer %20439 : tensor<4x?x4096xf16>{%dim_18128} to #hal.device.promise<@__device_6>
%20441 = torch_c.from_builtin_tensor %20440 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20442 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18129 = arith.constant 1 : index
%dim_18130 = tensor.dim %20442, %c1_18129 : tensor<4x?x4096xf16>
%20443 = flow.tensor.transfer %20442 : tensor<4x?x4096xf16>{%dim_18130} to #hal.device.promise<@__device_6>
%20444 = torch_c.from_builtin_tensor %20443 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20445 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18131 = arith.constant 1 : index
%dim_18132 = tensor.dim %20445, %c1_18131 : tensor<4x?x4096xf16>
%20446 = flow.tensor.transfer %20445 : tensor<4x?x4096xf16>{%dim_18132} to #hal.device.promise<@__device_6>
%20447 = torch_c.from_builtin_tensor %20446 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20448 = torch_c.to_builtin_tensor %20261 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18133 = arith.constant 1 : index
%dim_18134 = tensor.dim %20448, %c1_18133 : tensor<4x?x4096xf16>
%20449 = flow.tensor.transfer %20448 : tensor<4x?x4096xf16>{%dim_18134} to #hal.device.promise<@__device_6>
%20450 = torch_c.from_builtin_tensor %20449 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18135 = torch.constant.int 1
%20451 = torch.aten.add.Tensor %20432, %20435, %int1_18135 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18136 = torch.constant.int 1
%20452 = torch.aten.add.Tensor %20451, %20438, %int1_18136 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18137 = torch.constant.int 1
%20453 = torch.aten.add.Tensor %20452, %20441, %int1_18137 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18138 = torch.constant.int 1
%20454 = torch.aten.add.Tensor %20453, %20444, %int1_18138 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18139 = torch.constant.int 1
%20455 = torch.aten.add.Tensor %20454, %20447, %int1_18139 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18140 = torch.constant.int 1
%20456 = torch.aten.add.Tensor %20455, %20255, %int1_18140 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18141 = torch.constant.int 1
%20457 = torch.aten.add.Tensor %20456, %20450, %int1_18141 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20458 = torch_c.to_builtin_tensor %20219 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18142 = arith.constant 1 : index
%dim_18143 = tensor.dim %20458, %c1_18142 : tensor<4x?x4096xf16>
%20459 = flow.tensor.transfer %20458 : tensor<4x?x4096xf16>{%dim_18143} to #hal.device.promise<@__device_7>
%20460 = torch_c.from_builtin_tensor %20459 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20461 = torch_c.to_builtin_tensor %20225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18144 = arith.constant 1 : index
%dim_18145 = tensor.dim %20461, %c1_18144 : tensor<4x?x4096xf16>
%20462 = flow.tensor.transfer %20461 : tensor<4x?x4096xf16>{%dim_18145} to #hal.device.promise<@__device_7>
%20463 = torch_c.from_builtin_tensor %20462 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20464 = torch_c.to_builtin_tensor %20231 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18146 = arith.constant 1 : index
%dim_18147 = tensor.dim %20464, %c1_18146 : tensor<4x?x4096xf16>
%20465 = flow.tensor.transfer %20464 : tensor<4x?x4096xf16>{%dim_18147} to #hal.device.promise<@__device_7>
%20466 = torch_c.from_builtin_tensor %20465 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20467 = torch_c.to_builtin_tensor %20237 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18148 = arith.constant 1 : index
%dim_18149 = tensor.dim %20467, %c1_18148 : tensor<4x?x4096xf16>
%20468 = flow.tensor.transfer %20467 : tensor<4x?x4096xf16>{%dim_18149} to #hal.device.promise<@__device_7>
%20469 = torch_c.from_builtin_tensor %20468 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20470 = torch_c.to_builtin_tensor %20243 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18150 = arith.constant 1 : index
%dim_18151 = tensor.dim %20470, %c1_18150 : tensor<4x?x4096xf16>
%20471 = flow.tensor.transfer %20470 : tensor<4x?x4096xf16>{%dim_18151} to #hal.device.promise<@__device_7>
%20472 = torch_c.from_builtin_tensor %20471 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20473 = torch_c.to_builtin_tensor %20249 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18152 = arith.constant 1 : index
%dim_18153 = tensor.dim %20473, %c1_18152 : tensor<4x?x4096xf16>
%20474 = flow.tensor.transfer %20473 : tensor<4x?x4096xf16>{%dim_18153} to #hal.device.promise<@__device_7>
%20475 = torch_c.from_builtin_tensor %20474 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20476 = torch_c.to_builtin_tensor %20255 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18154 = arith.constant 1 : index
%dim_18155 = tensor.dim %20476, %c1_18154 : tensor<4x?x4096xf16>
%20477 = flow.tensor.transfer %20476 : tensor<4x?x4096xf16>{%dim_18155} to #hal.device.promise<@__device_7>
%20478 = torch_c.from_builtin_tensor %20477 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18156 = torch.constant.int 1
%20479 = torch.aten.add.Tensor %20460, %20463, %int1_18156 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18157 = torch.constant.int 1
%20480 = torch.aten.add.Tensor %20479, %20466, %int1_18157 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18158 = torch.constant.int 1
%20481 = torch.aten.add.Tensor %20480, %20469, %int1_18158 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18159 = torch.constant.int 1
%20482 = torch.aten.add.Tensor %20481, %20472, %int1_18159 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18160 = torch.constant.int 1
%20483 = torch.aten.add.Tensor %20482, %20475, %int1_18160 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18161 = torch.constant.int 1
%20484 = torch.aten.add.Tensor %20483, %20478, %int1_18161 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18162 = torch.constant.int 1
%20485 = torch.aten.add.Tensor %20484, %20261, %int1_18162 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18163 = torch.constant.int 1
%20486 = torch.aten.add.Tensor %19145, %20289, %int1_18163 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18164 = torch.constant.int 1
%20487 = torch.aten.add.Tensor %19146, %20317, %int1_18164 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18165 = torch.constant.int 1
%20488 = torch.aten.add.Tensor %19147, %20345, %int1_18165 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18166 = torch.constant.int 1
%20489 = torch.aten.add.Tensor %19148, %20373, %int1_18166 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18167 = torch.constant.int 1
%20490 = torch.aten.add.Tensor %19149, %20401, %int1_18167 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18168 = torch.constant.int 1
%20491 = torch.aten.add.Tensor %19150, %20429, %int1_18168 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18169 = torch.constant.int 1
%20492 = torch.aten.add.Tensor %19151, %20457, %int1_18169 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18170 = torch.constant.int 1
%20493 = torch.aten.add.Tensor %19152, %20485, %int1_18170 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_18171 = torch.constant.int 6
%20494 = torch.prims.convert_element_type %20486, %int6_18171 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18172 = torch.constant.int 6
%20495 = torch.prims.convert_element_type %20487, %int6_18172 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18173 = torch.constant.int 6
%20496 = torch.prims.convert_element_type %20488, %int6_18173 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18174 = torch.constant.int 6
%20497 = torch.prims.convert_element_type %20489, %int6_18174 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18175 = torch.constant.int 6
%20498 = torch.prims.convert_element_type %20490, %int6_18175 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18176 = torch.constant.int 6
%20499 = torch.prims.convert_element_type %20491, %int6_18176 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18177 = torch.constant.int 6
%20500 = torch.prims.convert_element_type %20492, %int6_18177 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18178 = torch.constant.int 6
%20501 = torch.prims.convert_element_type %20493, %int6_18178 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18179 = torch.constant.int 2
%20502 = torch.aten.pow.Tensor_Scalar %20494, %int2_18179 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18180 = torch.constant.int 2
%20503 = torch.aten.pow.Tensor_Scalar %20495, %int2_18180 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18181 = torch.constant.int 2
%20504 = torch.aten.pow.Tensor_Scalar %20496, %int2_18181 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18182 = torch.constant.int 2
%20505 = torch.aten.pow.Tensor_Scalar %20497, %int2_18182 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18183 = torch.constant.int 2
%20506 = torch.aten.pow.Tensor_Scalar %20498, %int2_18183 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18184 = torch.constant.int 2
%20507 = torch.aten.pow.Tensor_Scalar %20499, %int2_18184 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18185 = torch.constant.int 2
%20508 = torch.aten.pow.Tensor_Scalar %20500, %int2_18185 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18186 = torch.constant.int 2
%20509 = torch.aten.pow.Tensor_Scalar %20501, %int2_18186 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_18187 = torch.constant.int -1
%20510 = torch.prim.ListConstruct %int-1_18187 : (!torch.int) -> !torch.list<int>
%true_18188 = torch.constant.bool true
%none_18189 = torch.constant.none
%20511 = torch.aten.mean.dim %20502, %20510, %true_18188, %none_18189 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18190 = torch.constant.int -1
%20512 = torch.prim.ListConstruct %int-1_18190 : (!torch.int) -> !torch.list<int>
%true_18191 = torch.constant.bool true
%none_18192 = torch.constant.none
%20513 = torch.aten.mean.dim %20503, %20512, %true_18191, %none_18192 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18193 = torch.constant.int -1
%20514 = torch.prim.ListConstruct %int-1_18193 : (!torch.int) -> !torch.list<int>
%true_18194 = torch.constant.bool true
%none_18195 = torch.constant.none
%20515 = torch.aten.mean.dim %20504, %20514, %true_18194, %none_18195 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18196 = torch.constant.int -1
%20516 = torch.prim.ListConstruct %int-1_18196 : (!torch.int) -> !torch.list<int>
%true_18197 = torch.constant.bool true
%none_18198 = torch.constant.none
%20517 = torch.aten.mean.dim %20505, %20516, %true_18197, %none_18198 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18199 = torch.constant.int -1
%20518 = torch.prim.ListConstruct %int-1_18199 : (!torch.int) -> !torch.list<int>
%true_18200 = torch.constant.bool true
%none_18201 = torch.constant.none
%20519 = torch.aten.mean.dim %20506, %20518, %true_18200, %none_18201 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18202 = torch.constant.int -1
%20520 = torch.prim.ListConstruct %int-1_18202 : (!torch.int) -> !torch.list<int>
%true_18203 = torch.constant.bool true
%none_18204 = torch.constant.none
%20521 = torch.aten.mean.dim %20507, %20520, %true_18203, %none_18204 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18205 = torch.constant.int -1
%20522 = torch.prim.ListConstruct %int-1_18205 : (!torch.int) -> !torch.list<int>
%true_18206 = torch.constant.bool true
%none_18207 = torch.constant.none
%20523 = torch.aten.mean.dim %20508, %20522, %true_18206, %none_18207 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18208 = torch.constant.int -1
%20524 = torch.prim.ListConstruct %int-1_18208 : (!torch.int) -> !torch.list<int>
%true_18209 = torch.constant.bool true
%none_18210 = torch.constant.none
%20525 = torch.aten.mean.dim %20509, %20524, %true_18209, %none_18210 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18211 = torch.constant.float 9.9999997473787516E-6
%int1_18212 = torch.constant.int 1
%20526 = torch.aten.add.Scalar %20511, %float9.999990e-06_18211, %int1_18212 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18213 = torch.constant.float 9.9999997473787516E-6
%int1_18214 = torch.constant.int 1
%20527 = torch.aten.add.Scalar %20513, %float9.999990e-06_18213, %int1_18214 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18215 = torch.constant.float 9.9999997473787516E-6
%int1_18216 = torch.constant.int 1
%20528 = torch.aten.add.Scalar %20515, %float9.999990e-06_18215, %int1_18216 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18217 = torch.constant.float 9.9999997473787516E-6
%int1_18218 = torch.constant.int 1
%20529 = torch.aten.add.Scalar %20517, %float9.999990e-06_18217, %int1_18218 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18219 = torch.constant.float 9.9999997473787516E-6
%int1_18220 = torch.constant.int 1
%20530 = torch.aten.add.Scalar %20519, %float9.999990e-06_18219, %int1_18220 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18221 = torch.constant.float 9.9999997473787516E-6
%int1_18222 = torch.constant.int 1
%20531 = torch.aten.add.Scalar %20521, %float9.999990e-06_18221, %int1_18222 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18223 = torch.constant.float 9.9999997473787516E-6
%int1_18224 = torch.constant.int 1
%20532 = torch.aten.add.Scalar %20523, %float9.999990e-06_18223, %int1_18224 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18225 = torch.constant.float 9.9999997473787516E-6
%int1_18226 = torch.constant.int 1
%20533 = torch.aten.add.Scalar %20525, %float9.999990e-06_18225, %int1_18226 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20534 = torch.aten.rsqrt %20526 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20535 = torch.aten.rsqrt %20527 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20536 = torch.aten.rsqrt %20528 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20537 = torch.aten.rsqrt %20529 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20538 = torch.aten.rsqrt %20530 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20539 = torch.aten.rsqrt %20531 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20540 = torch.aten.rsqrt %20532 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20541 = torch.aten.rsqrt %20533 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %20541, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%20542 = torch.aten.mul.Tensor %20494, %20534 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20543 = torch.aten.mul.Tensor %20495, %20535 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20544 = torch.aten.mul.Tensor %20496, %20536 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20545 = torch.aten.mul.Tensor %20497, %20537 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20546 = torch.aten.mul.Tensor %20498, %20538 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20547 = torch.aten.mul.Tensor %20499, %20539 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20548 = torch.aten.mul.Tensor %20500, %20540 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20549 = torch.aten.mul.Tensor %20501, %20541 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20550 = torch.aten.mul.Tensor %696, %20542 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20551 = torch.aten.mul.Tensor %697, %20543 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20552 = torch.aten.mul.Tensor %698, %20544 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20553 = torch.aten.mul.Tensor %699, %20545 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20554 = torch.aten.mul.Tensor %700, %20546 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20555 = torch.aten.mul.Tensor %701, %20547 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20556 = torch.aten.mul.Tensor %702, %20548 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%20557 = torch.aten.mul.Tensor %703, %20549 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %20557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_18227 = torch.constant.int 5
%20558 = torch.prims.convert_element_type %20550, %int5_18227 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18228 = torch.constant.int 5
%20559 = torch.prims.convert_element_type %20551, %int5_18228 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18229 = torch.constant.int 5
%20560 = torch.prims.convert_element_type %20552, %int5_18229 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18230 = torch.constant.int 5
%20561 = torch.prims.convert_element_type %20553, %int5_18230 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18231 = torch.constant.int 5
%20562 = torch.prims.convert_element_type %20554, %int5_18231 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18232 = torch.constant.int 5
%20563 = torch.prims.convert_element_type %20555, %int5_18232 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18233 = torch.constant.int 5
%20564 = torch.prims.convert_element_type %20556, %int5_18233 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18234 = torch.constant.int 5
%20565 = torch.prims.convert_element_type %20557, %int5_18234 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18235 = torch.constant.int 1
%int0_18236 = torch.constant.int 0
%20566 = torch.prim.ListConstruct %int1_18235, %int0_18236 : (!torch.int, !torch.int) -> !torch.list<int>
%20567 = torch.aten.permute %704, %20566 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18237 = torch.constant.int 1
%int0_18238 = torch.constant.int 0
%20568 = torch.prim.ListConstruct %int1_18237, %int0_18238 : (!torch.int, !torch.int) -> !torch.list<int>
%20569 = torch.aten.permute %705, %20568 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18239 = torch.constant.int 1
%int0_18240 = torch.constant.int 0
%20570 = torch.prim.ListConstruct %int1_18239, %int0_18240 : (!torch.int, !torch.int) -> !torch.list<int>
%20571 = torch.aten.permute %706, %20570 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18241 = torch.constant.int 1
%int0_18242 = torch.constant.int 0
%20572 = torch.prim.ListConstruct %int1_18241, %int0_18242 : (!torch.int, !torch.int) -> !torch.list<int>
%20573 = torch.aten.permute %707, %20572 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18243 = torch.constant.int 1
%int0_18244 = torch.constant.int 0
%20574 = torch.prim.ListConstruct %int1_18243, %int0_18244 : (!torch.int, !torch.int) -> !torch.list<int>
%20575 = torch.aten.permute %708, %20574 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18245 = torch.constant.int 1
%int0_18246 = torch.constant.int 0
%20576 = torch.prim.ListConstruct %int1_18245, %int0_18246 : (!torch.int, !torch.int) -> !torch.list<int>
%20577 = torch.aten.permute %709, %20576 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18247 = torch.constant.int 1
%int0_18248 = torch.constant.int 0
%20578 = torch.prim.ListConstruct %int1_18247, %int0_18248 : (!torch.int, !torch.int) -> !torch.list<int>
%20579 = torch.aten.permute %710, %20578 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18249 = torch.constant.int 1
%int0_18250 = torch.constant.int 0
%20580 = torch.prim.ListConstruct %int1_18249, %int0_18250 : (!torch.int, !torch.int) -> !torch.list<int>
%20581 = torch.aten.permute %711, %20580 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_18251 = torch.constant.int 4
%20582 = torch.aten.mul.int %int4_18251, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18252 = torch.constant.int 4096
%20583 = torch.prim.ListConstruct %20582, %int4096_18252 : (!torch.int, !torch.int) -> !torch.list<int>
%20584 = torch.aten.view %20558, %20583 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20584, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20585 = torch.aten.mm %20584, %20567 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20585, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18253 = torch.constant.int 4
%int1792_18254 = torch.constant.int 1792
%20586 = torch.prim.ListConstruct %int4_18253, %2482, %int1792_18254 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20587 = torch.aten.view %20585, %20586 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18255 = torch.constant.int 4
%20588 = torch.aten.mul.int %int4_18255, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18256 = torch.constant.int 4096
%20589 = torch.prim.ListConstruct %20588, %int4096_18256 : (!torch.int, !torch.int) -> !torch.list<int>
%20590 = torch.aten.view %20559, %20589 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20590, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20591 = torch.aten.mm %20590, %20569 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20591, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18257 = torch.constant.int 4
%int1792_18258 = torch.constant.int 1792
%20592 = torch.prim.ListConstruct %int4_18257, %2482, %int1792_18258 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20593 = torch.aten.view %20591, %20592 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18259 = torch.constant.int 4
%20594 = torch.aten.mul.int %int4_18259, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18260 = torch.constant.int 4096
%20595 = torch.prim.ListConstruct %20594, %int4096_18260 : (!torch.int, !torch.int) -> !torch.list<int>
%20596 = torch.aten.view %20560, %20595 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20596, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20597 = torch.aten.mm %20596, %20571 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20597, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18261 = torch.constant.int 4
%int1792_18262 = torch.constant.int 1792
%20598 = torch.prim.ListConstruct %int4_18261, %2482, %int1792_18262 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20599 = torch.aten.view %20597, %20598 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18263 = torch.constant.int 4
%20600 = torch.aten.mul.int %int4_18263, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18264 = torch.constant.int 4096
%20601 = torch.prim.ListConstruct %20600, %int4096_18264 : (!torch.int, !torch.int) -> !torch.list<int>
%20602 = torch.aten.view %20561, %20601 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20602, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20603 = torch.aten.mm %20602, %20573 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20603, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18265 = torch.constant.int 4
%int1792_18266 = torch.constant.int 1792
%20604 = torch.prim.ListConstruct %int4_18265, %2482, %int1792_18266 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20605 = torch.aten.view %20603, %20604 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18267 = torch.constant.int 4
%20606 = torch.aten.mul.int %int4_18267, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18268 = torch.constant.int 4096
%20607 = torch.prim.ListConstruct %20606, %int4096_18268 : (!torch.int, !torch.int) -> !torch.list<int>
%20608 = torch.aten.view %20562, %20607 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20608, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20609 = torch.aten.mm %20608, %20575 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20609, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18269 = torch.constant.int 4
%int1792_18270 = torch.constant.int 1792
%20610 = torch.prim.ListConstruct %int4_18269, %2482, %int1792_18270 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20611 = torch.aten.view %20609, %20610 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18271 = torch.constant.int 4
%20612 = torch.aten.mul.int %int4_18271, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18272 = torch.constant.int 4096
%20613 = torch.prim.ListConstruct %20612, %int4096_18272 : (!torch.int, !torch.int) -> !torch.list<int>
%20614 = torch.aten.view %20563, %20613 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20614, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20615 = torch.aten.mm %20614, %20577 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20615, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18273 = torch.constant.int 4
%int1792_18274 = torch.constant.int 1792
%20616 = torch.prim.ListConstruct %int4_18273, %2482, %int1792_18274 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20617 = torch.aten.view %20615, %20616 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18275 = torch.constant.int 4
%20618 = torch.aten.mul.int %int4_18275, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18276 = torch.constant.int 4096
%20619 = torch.prim.ListConstruct %20618, %int4096_18276 : (!torch.int, !torch.int) -> !torch.list<int>
%20620 = torch.aten.view %20564, %20619 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20620, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20621 = torch.aten.mm %20620, %20579 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20621, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18277 = torch.constant.int 4
%int1792_18278 = torch.constant.int 1792
%20622 = torch.prim.ListConstruct %int4_18277, %2482, %int1792_18278 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20623 = torch.aten.view %20621, %20622 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18279 = torch.constant.int 4
%20624 = torch.aten.mul.int %int4_18279, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18280 = torch.constant.int 4096
%20625 = torch.prim.ListConstruct %20624, %int4096_18280 : (!torch.int, !torch.int) -> !torch.list<int>
%20626 = torch.aten.view %20565, %20625 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20626, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20627 = torch.aten.mm %20626, %20581 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20627, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18281 = torch.constant.int 4
%int1792_18282 = torch.constant.int 1792
%20628 = torch.prim.ListConstruct %int4_18281, %2482, %int1792_18282 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20629 = torch.aten.view %20627, %20628 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20630 = torch.aten.silu %20587 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20631 = torch.aten.silu %20593 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20632 = torch.aten.silu %20599 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20633 = torch.aten.silu %20605 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20634 = torch.aten.silu %20611 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20635 = torch.aten.silu %20617 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20636 = torch.aten.silu %20623 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20637 = torch.aten.silu %20629 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_18283 = torch.constant.int 1
%int0_18284 = torch.constant.int 0
%20638 = torch.prim.ListConstruct %int1_18283, %int0_18284 : (!torch.int, !torch.int) -> !torch.list<int>
%20639 = torch.aten.permute %712, %20638 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18285 = torch.constant.int 1
%int0_18286 = torch.constant.int 0
%20640 = torch.prim.ListConstruct %int1_18285, %int0_18286 : (!torch.int, !torch.int) -> !torch.list<int>
%20641 = torch.aten.permute %713, %20640 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18287 = torch.constant.int 1
%int0_18288 = torch.constant.int 0
%20642 = torch.prim.ListConstruct %int1_18287, %int0_18288 : (!torch.int, !torch.int) -> !torch.list<int>
%20643 = torch.aten.permute %714, %20642 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18289 = torch.constant.int 1
%int0_18290 = torch.constant.int 0
%20644 = torch.prim.ListConstruct %int1_18289, %int0_18290 : (!torch.int, !torch.int) -> !torch.list<int>
%20645 = torch.aten.permute %715, %20644 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18291 = torch.constant.int 1
%int0_18292 = torch.constant.int 0
%20646 = torch.prim.ListConstruct %int1_18291, %int0_18292 : (!torch.int, !torch.int) -> !torch.list<int>
%20647 = torch.aten.permute %716, %20646 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18293 = torch.constant.int 1
%int0_18294 = torch.constant.int 0
%20648 = torch.prim.ListConstruct %int1_18293, %int0_18294 : (!torch.int, !torch.int) -> !torch.list<int>
%20649 = torch.aten.permute %717, %20648 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18295 = torch.constant.int 1
%int0_18296 = torch.constant.int 0
%20650 = torch.prim.ListConstruct %int1_18295, %int0_18296 : (!torch.int, !torch.int) -> !torch.list<int>
%20651 = torch.aten.permute %718, %20650 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_18297 = torch.constant.int 1
%int0_18298 = torch.constant.int 0
%20652 = torch.prim.ListConstruct %int1_18297, %int0_18298 : (!torch.int, !torch.int) -> !torch.list<int>
%20653 = torch.aten.permute %719, %20652 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_18299 = torch.constant.int 4
%20654 = torch.aten.mul.int %int4_18299, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18300 = torch.constant.int 4096
%20655 = torch.prim.ListConstruct %20654, %int4096_18300 : (!torch.int, !torch.int) -> !torch.list<int>
%20656 = torch.aten.view %20558, %20655 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20656, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20657 = torch.aten.mm %20656, %20639 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20657, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18301 = torch.constant.int 4
%int1792_18302 = torch.constant.int 1792
%20658 = torch.prim.ListConstruct %int4_18301, %2482, %int1792_18302 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20659 = torch.aten.view %20657, %20658 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18303 = torch.constant.int 4
%20660 = torch.aten.mul.int %int4_18303, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18304 = torch.constant.int 4096
%20661 = torch.prim.ListConstruct %20660, %int4096_18304 : (!torch.int, !torch.int) -> !torch.list<int>
%20662 = torch.aten.view %20559, %20661 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20662, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20663 = torch.aten.mm %20662, %20641 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20663, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18305 = torch.constant.int 4
%int1792_18306 = torch.constant.int 1792
%20664 = torch.prim.ListConstruct %int4_18305, %2482, %int1792_18306 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20665 = torch.aten.view %20663, %20664 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18307 = torch.constant.int 4
%20666 = torch.aten.mul.int %int4_18307, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18308 = torch.constant.int 4096
%20667 = torch.prim.ListConstruct %20666, %int4096_18308 : (!torch.int, !torch.int) -> !torch.list<int>
%20668 = torch.aten.view %20560, %20667 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20668, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20669 = torch.aten.mm %20668, %20643 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20669, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18309 = torch.constant.int 4
%int1792_18310 = torch.constant.int 1792
%20670 = torch.prim.ListConstruct %int4_18309, %2482, %int1792_18310 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20671 = torch.aten.view %20669, %20670 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18311 = torch.constant.int 4
%20672 = torch.aten.mul.int %int4_18311, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18312 = torch.constant.int 4096
%20673 = torch.prim.ListConstruct %20672, %int4096_18312 : (!torch.int, !torch.int) -> !torch.list<int>
%20674 = torch.aten.view %20561, %20673 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20674, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20675 = torch.aten.mm %20674, %20645 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20675, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18313 = torch.constant.int 4
%int1792_18314 = torch.constant.int 1792
%20676 = torch.prim.ListConstruct %int4_18313, %2482, %int1792_18314 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20677 = torch.aten.view %20675, %20676 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18315 = torch.constant.int 4
%20678 = torch.aten.mul.int %int4_18315, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18316 = torch.constant.int 4096
%20679 = torch.prim.ListConstruct %20678, %int4096_18316 : (!torch.int, !torch.int) -> !torch.list<int>
%20680 = torch.aten.view %20562, %20679 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20680, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20681 = torch.aten.mm %20680, %20647 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20681, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18317 = torch.constant.int 4
%int1792_18318 = torch.constant.int 1792
%20682 = torch.prim.ListConstruct %int4_18317, %2482, %int1792_18318 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20683 = torch.aten.view %20681, %20682 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18319 = torch.constant.int 4
%20684 = torch.aten.mul.int %int4_18319, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18320 = torch.constant.int 4096
%20685 = torch.prim.ListConstruct %20684, %int4096_18320 : (!torch.int, !torch.int) -> !torch.list<int>
%20686 = torch.aten.view %20563, %20685 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20686, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20687 = torch.aten.mm %20686, %20649 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20687, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18321 = torch.constant.int 4
%int1792_18322 = torch.constant.int 1792
%20688 = torch.prim.ListConstruct %int4_18321, %2482, %int1792_18322 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20689 = torch.aten.view %20687, %20688 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18323 = torch.constant.int 4
%20690 = torch.aten.mul.int %int4_18323, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18324 = torch.constant.int 4096
%20691 = torch.prim.ListConstruct %20690, %int4096_18324 : (!torch.int, !torch.int) -> !torch.list<int>
%20692 = torch.aten.view %20564, %20691 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20692, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20693 = torch.aten.mm %20692, %20651 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20693, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18325 = torch.constant.int 4
%int1792_18326 = torch.constant.int 1792
%20694 = torch.prim.ListConstruct %int4_18325, %2482, %int1792_18326 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20695 = torch.aten.view %20693, %20694 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_18327 = torch.constant.int 4
%20696 = torch.aten.mul.int %int4_18327, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18328 = torch.constant.int 4096
%20697 = torch.prim.ListConstruct %20696, %int4096_18328 : (!torch.int, !torch.int) -> !torch.list<int>
%20698 = torch.aten.view %20565, %20697 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20698, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%20699 = torch.aten.mm %20698, %20653 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20699, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_18329 = torch.constant.int 4
%int1792_18330 = torch.constant.int 1792
%20700 = torch.prim.ListConstruct %int4_18329, %2482, %int1792_18330 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20701 = torch.aten.view %20699, %20700 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20702 = torch.aten.mul.Tensor %20630, %20659 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20703 = torch.aten.mul.Tensor %20631, %20665 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20704 = torch.aten.mul.Tensor %20632, %20671 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20705 = torch.aten.mul.Tensor %20633, %20677 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20706 = torch.aten.mul.Tensor %20634, %20683 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20707 = torch.aten.mul.Tensor %20635, %20689 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20708 = torch.aten.mul.Tensor %20636, %20695 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%20709 = torch.aten.mul.Tensor %20637, %20701 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %20709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_18331 = torch.constant.int 1
%int0_18332 = torch.constant.int 0
%20710 = torch.prim.ListConstruct %int1_18331, %int0_18332 : (!torch.int, !torch.int) -> !torch.list<int>
%20711 = torch.aten.permute %720, %20710 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18333 = torch.constant.int 1
%int0_18334 = torch.constant.int 0
%20712 = torch.prim.ListConstruct %int1_18333, %int0_18334 : (!torch.int, !torch.int) -> !torch.list<int>
%20713 = torch.aten.permute %721, %20712 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18335 = torch.constant.int 1
%int0_18336 = torch.constant.int 0
%20714 = torch.prim.ListConstruct %int1_18335, %int0_18336 : (!torch.int, !torch.int) -> !torch.list<int>
%20715 = torch.aten.permute %722, %20714 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18337 = torch.constant.int 1
%int0_18338 = torch.constant.int 0
%20716 = torch.prim.ListConstruct %int1_18337, %int0_18338 : (!torch.int, !torch.int) -> !torch.list<int>
%20717 = torch.aten.permute %723, %20716 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18339 = torch.constant.int 1
%int0_18340 = torch.constant.int 0
%20718 = torch.prim.ListConstruct %int1_18339, %int0_18340 : (!torch.int, !torch.int) -> !torch.list<int>
%20719 = torch.aten.permute %724, %20718 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18341 = torch.constant.int 1
%int0_18342 = torch.constant.int 0
%20720 = torch.prim.ListConstruct %int1_18341, %int0_18342 : (!torch.int, !torch.int) -> !torch.list<int>
%20721 = torch.aten.permute %725, %20720 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18343 = torch.constant.int 1
%int0_18344 = torch.constant.int 0
%20722 = torch.prim.ListConstruct %int1_18343, %int0_18344 : (!torch.int, !torch.int) -> !torch.list<int>
%20723 = torch.aten.permute %726, %20722 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18345 = torch.constant.int 1
%int0_18346 = torch.constant.int 0
%20724 = torch.prim.ListConstruct %int1_18345, %int0_18346 : (!torch.int, !torch.int) -> !torch.list<int>
%20725 = torch.aten.permute %727, %20724 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_18347 = torch.constant.int 1
%20726 = torch.aten.size.int %20587, %int1_18347 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18348 = torch.constant.int 4
%20727 = torch.aten.mul.int %int4_18348, %20726 : !torch.int, !torch.int -> !torch.int
%int1792_18349 = torch.constant.int 1792
%20728 = torch.prim.ListConstruct %20727, %int1792_18349 : (!torch.int, !torch.int) -> !torch.list<int>
%20729 = torch.aten.view %20702, %20728 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20729, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20730 = torch.aten.mm %20729, %20711 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20730, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18350 = torch.constant.int 4
%int4096_18351 = torch.constant.int 4096
%20731 = torch.prim.ListConstruct %int4_18350, %20726, %int4096_18351 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20732 = torch.aten.view %20730, %20731 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18352 = torch.constant.int 1
%20733 = torch.aten.size.int %20593, %int1_18352 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18353 = torch.constant.int 4
%20734 = torch.aten.mul.int %int4_18353, %20733 : !torch.int, !torch.int -> !torch.int
%int1792_18354 = torch.constant.int 1792
%20735 = torch.prim.ListConstruct %20734, %int1792_18354 : (!torch.int, !torch.int) -> !torch.list<int>
%20736 = torch.aten.view %20703, %20735 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20736, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20737 = torch.aten.mm %20736, %20713 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20737, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18355 = torch.constant.int 4
%int4096_18356 = torch.constant.int 4096
%20738 = torch.prim.ListConstruct %int4_18355, %20733, %int4096_18356 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20739 = torch.aten.view %20737, %20738 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18357 = torch.constant.int 1
%20740 = torch.aten.size.int %20599, %int1_18357 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18358 = torch.constant.int 4
%20741 = torch.aten.mul.int %int4_18358, %20740 : !torch.int, !torch.int -> !torch.int
%int1792_18359 = torch.constant.int 1792
%20742 = torch.prim.ListConstruct %20741, %int1792_18359 : (!torch.int, !torch.int) -> !torch.list<int>
%20743 = torch.aten.view %20704, %20742 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20743, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20744 = torch.aten.mm %20743, %20715 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20744, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18360 = torch.constant.int 4
%int4096_18361 = torch.constant.int 4096
%20745 = torch.prim.ListConstruct %int4_18360, %20740, %int4096_18361 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20746 = torch.aten.view %20744, %20745 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18362 = torch.constant.int 1
%20747 = torch.aten.size.int %20605, %int1_18362 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18363 = torch.constant.int 4
%20748 = torch.aten.mul.int %int4_18363, %20747 : !torch.int, !torch.int -> !torch.int
%int1792_18364 = torch.constant.int 1792
%20749 = torch.prim.ListConstruct %20748, %int1792_18364 : (!torch.int, !torch.int) -> !torch.list<int>
%20750 = torch.aten.view %20705, %20749 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20750, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20751 = torch.aten.mm %20750, %20717 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20751, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18365 = torch.constant.int 4
%int4096_18366 = torch.constant.int 4096
%20752 = torch.prim.ListConstruct %int4_18365, %20747, %int4096_18366 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20753 = torch.aten.view %20751, %20752 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18367 = torch.constant.int 1
%20754 = torch.aten.size.int %20611, %int1_18367 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18368 = torch.constant.int 4
%20755 = torch.aten.mul.int %int4_18368, %20754 : !torch.int, !torch.int -> !torch.int
%int1792_18369 = torch.constant.int 1792
%20756 = torch.prim.ListConstruct %20755, %int1792_18369 : (!torch.int, !torch.int) -> !torch.list<int>
%20757 = torch.aten.view %20706, %20756 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20757, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20758 = torch.aten.mm %20757, %20719 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20758, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18370 = torch.constant.int 4
%int4096_18371 = torch.constant.int 4096
%20759 = torch.prim.ListConstruct %int4_18370, %20754, %int4096_18371 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20760 = torch.aten.view %20758, %20759 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18372 = torch.constant.int 1
%20761 = torch.aten.size.int %20617, %int1_18372 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18373 = torch.constant.int 4
%20762 = torch.aten.mul.int %int4_18373, %20761 : !torch.int, !torch.int -> !torch.int
%int1792_18374 = torch.constant.int 1792
%20763 = torch.prim.ListConstruct %20762, %int1792_18374 : (!torch.int, !torch.int) -> !torch.list<int>
%20764 = torch.aten.view %20707, %20763 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20764, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20765 = torch.aten.mm %20764, %20721 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20765, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18375 = torch.constant.int 4
%int4096_18376 = torch.constant.int 4096
%20766 = torch.prim.ListConstruct %int4_18375, %20761, %int4096_18376 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20767 = torch.aten.view %20765, %20766 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18377 = torch.constant.int 1
%20768 = torch.aten.size.int %20623, %int1_18377 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18378 = torch.constant.int 4
%20769 = torch.aten.mul.int %int4_18378, %20768 : !torch.int, !torch.int -> !torch.int
%int1792_18379 = torch.constant.int 1792
%20770 = torch.prim.ListConstruct %20769, %int1792_18379 : (!torch.int, !torch.int) -> !torch.list<int>
%20771 = torch.aten.view %20708, %20770 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20771, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20772 = torch.aten.mm %20771, %20723 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20772, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18380 = torch.constant.int 4
%int4096_18381 = torch.constant.int 4096
%20773 = torch.prim.ListConstruct %int4_18380, %20768, %int4096_18381 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20774 = torch.aten.view %20772, %20773 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18382 = torch.constant.int 1
%20775 = torch.aten.size.int %20629, %int1_18382 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_18383 = torch.constant.int 4
%20776 = torch.aten.mul.int %int4_18383, %20775 : !torch.int, !torch.int -> !torch.int
%int1792_18384 = torch.constant.int 1792
%20777 = torch.prim.ListConstruct %20776, %int1792_18384 : (!torch.int, !torch.int) -> !torch.list<int>
%20778 = torch.aten.view %20709, %20777 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %20778, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%20779 = torch.aten.mm %20778, %20725 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %20779, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_18385 = torch.constant.int 4
%int4096_18386 = torch.constant.int 4096
%20780 = torch.prim.ListConstruct %int4_18385, %20775, %int4096_18386 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%20781 = torch.aten.view %20779, %20780 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20782 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18387 = arith.constant 1 : index
%dim_18388 = tensor.dim %20782, %c1_18387 : tensor<4x?x4096xf16>
%20783 = flow.tensor.transfer %20782 : tensor<4x?x4096xf16>{%dim_18388} to #hal.device.promise<@__device_0>
%20784 = torch_c.from_builtin_tensor %20783 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20785 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18389 = arith.constant 1 : index
%dim_18390 = tensor.dim %20785, %c1_18389 : tensor<4x?x4096xf16>
%20786 = flow.tensor.transfer %20785 : tensor<4x?x4096xf16>{%dim_18390} to #hal.device.promise<@__device_0>
%20787 = torch_c.from_builtin_tensor %20786 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20788 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18391 = arith.constant 1 : index
%dim_18392 = tensor.dim %20788, %c1_18391 : tensor<4x?x4096xf16>
%20789 = flow.tensor.transfer %20788 : tensor<4x?x4096xf16>{%dim_18392} to #hal.device.promise<@__device_0>
%20790 = torch_c.from_builtin_tensor %20789 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20791 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18393 = arith.constant 1 : index
%dim_18394 = tensor.dim %20791, %c1_18393 : tensor<4x?x4096xf16>
%20792 = flow.tensor.transfer %20791 : tensor<4x?x4096xf16>{%dim_18394} to #hal.device.promise<@__device_0>
%20793 = torch_c.from_builtin_tensor %20792 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20794 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18395 = arith.constant 1 : index
%dim_18396 = tensor.dim %20794, %c1_18395 : tensor<4x?x4096xf16>
%20795 = flow.tensor.transfer %20794 : tensor<4x?x4096xf16>{%dim_18396} to #hal.device.promise<@__device_0>
%20796 = torch_c.from_builtin_tensor %20795 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20797 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18397 = arith.constant 1 : index
%dim_18398 = tensor.dim %20797, %c1_18397 : tensor<4x?x4096xf16>
%20798 = flow.tensor.transfer %20797 : tensor<4x?x4096xf16>{%dim_18398} to #hal.device.promise<@__device_0>
%20799 = torch_c.from_builtin_tensor %20798 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20800 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18399 = arith.constant 1 : index
%dim_18400 = tensor.dim %20800, %c1_18399 : tensor<4x?x4096xf16>
%20801 = flow.tensor.transfer %20800 : tensor<4x?x4096xf16>{%dim_18400} to #hal.device.promise<@__device_0>
%20802 = torch_c.from_builtin_tensor %20801 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18401 = torch.constant.int 1
%20803 = torch.aten.add.Tensor %20732, %20784, %int1_18401 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18402 = torch.constant.int 1
%20804 = torch.aten.add.Tensor %20803, %20787, %int1_18402 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18403 = torch.constant.int 1
%20805 = torch.aten.add.Tensor %20804, %20790, %int1_18403 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18404 = torch.constant.int 1
%20806 = torch.aten.add.Tensor %20805, %20793, %int1_18404 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18405 = torch.constant.int 1
%20807 = torch.aten.add.Tensor %20806, %20796, %int1_18405 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18406 = torch.constant.int 1
%20808 = torch.aten.add.Tensor %20807, %20799, %int1_18406 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18407 = torch.constant.int 1
%20809 = torch.aten.add.Tensor %20808, %20802, %int1_18407 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20810 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18408 = arith.constant 1 : index
%dim_18409 = tensor.dim %20810, %c1_18408 : tensor<4x?x4096xf16>
%20811 = flow.tensor.transfer %20810 : tensor<4x?x4096xf16>{%dim_18409} to #hal.device.promise<@__device_1>
%20812 = torch_c.from_builtin_tensor %20811 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20813 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18410 = arith.constant 1 : index
%dim_18411 = tensor.dim %20813, %c1_18410 : tensor<4x?x4096xf16>
%20814 = flow.tensor.transfer %20813 : tensor<4x?x4096xf16>{%dim_18411} to #hal.device.promise<@__device_1>
%20815 = torch_c.from_builtin_tensor %20814 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20816 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18412 = arith.constant 1 : index
%dim_18413 = tensor.dim %20816, %c1_18412 : tensor<4x?x4096xf16>
%20817 = flow.tensor.transfer %20816 : tensor<4x?x4096xf16>{%dim_18413} to #hal.device.promise<@__device_1>
%20818 = torch_c.from_builtin_tensor %20817 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20819 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18414 = arith.constant 1 : index
%dim_18415 = tensor.dim %20819, %c1_18414 : tensor<4x?x4096xf16>
%20820 = flow.tensor.transfer %20819 : tensor<4x?x4096xf16>{%dim_18415} to #hal.device.promise<@__device_1>
%20821 = torch_c.from_builtin_tensor %20820 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20822 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18416 = arith.constant 1 : index
%dim_18417 = tensor.dim %20822, %c1_18416 : tensor<4x?x4096xf16>
%20823 = flow.tensor.transfer %20822 : tensor<4x?x4096xf16>{%dim_18417} to #hal.device.promise<@__device_1>
%20824 = torch_c.from_builtin_tensor %20823 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20825 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18418 = arith.constant 1 : index
%dim_18419 = tensor.dim %20825, %c1_18418 : tensor<4x?x4096xf16>
%20826 = flow.tensor.transfer %20825 : tensor<4x?x4096xf16>{%dim_18419} to #hal.device.promise<@__device_1>
%20827 = torch_c.from_builtin_tensor %20826 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20828 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18420 = arith.constant 1 : index
%dim_18421 = tensor.dim %20828, %c1_18420 : tensor<4x?x4096xf16>
%20829 = flow.tensor.transfer %20828 : tensor<4x?x4096xf16>{%dim_18421} to #hal.device.promise<@__device_1>
%20830 = torch_c.from_builtin_tensor %20829 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18422 = torch.constant.int 1
%20831 = torch.aten.add.Tensor %20812, %20739, %int1_18422 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18423 = torch.constant.int 1
%20832 = torch.aten.add.Tensor %20831, %20815, %int1_18423 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18424 = torch.constant.int 1
%20833 = torch.aten.add.Tensor %20832, %20818, %int1_18424 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18425 = torch.constant.int 1
%20834 = torch.aten.add.Tensor %20833, %20821, %int1_18425 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18426 = torch.constant.int 1
%20835 = torch.aten.add.Tensor %20834, %20824, %int1_18426 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18427 = torch.constant.int 1
%20836 = torch.aten.add.Tensor %20835, %20827, %int1_18427 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18428 = torch.constant.int 1
%20837 = torch.aten.add.Tensor %20836, %20830, %int1_18428 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20838 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18429 = arith.constant 1 : index
%dim_18430 = tensor.dim %20838, %c1_18429 : tensor<4x?x4096xf16>
%20839 = flow.tensor.transfer %20838 : tensor<4x?x4096xf16>{%dim_18430} to #hal.device.promise<@__device_2>
%20840 = torch_c.from_builtin_tensor %20839 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20841 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18431 = arith.constant 1 : index
%dim_18432 = tensor.dim %20841, %c1_18431 : tensor<4x?x4096xf16>
%20842 = flow.tensor.transfer %20841 : tensor<4x?x4096xf16>{%dim_18432} to #hal.device.promise<@__device_2>
%20843 = torch_c.from_builtin_tensor %20842 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20844 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18433 = arith.constant 1 : index
%dim_18434 = tensor.dim %20844, %c1_18433 : tensor<4x?x4096xf16>
%20845 = flow.tensor.transfer %20844 : tensor<4x?x4096xf16>{%dim_18434} to #hal.device.promise<@__device_2>
%20846 = torch_c.from_builtin_tensor %20845 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20847 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18435 = arith.constant 1 : index
%dim_18436 = tensor.dim %20847, %c1_18435 : tensor<4x?x4096xf16>
%20848 = flow.tensor.transfer %20847 : tensor<4x?x4096xf16>{%dim_18436} to #hal.device.promise<@__device_2>
%20849 = torch_c.from_builtin_tensor %20848 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20850 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18437 = arith.constant 1 : index
%dim_18438 = tensor.dim %20850, %c1_18437 : tensor<4x?x4096xf16>
%20851 = flow.tensor.transfer %20850 : tensor<4x?x4096xf16>{%dim_18438} to #hal.device.promise<@__device_2>
%20852 = torch_c.from_builtin_tensor %20851 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20853 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18439 = arith.constant 1 : index
%dim_18440 = tensor.dim %20853, %c1_18439 : tensor<4x?x4096xf16>
%20854 = flow.tensor.transfer %20853 : tensor<4x?x4096xf16>{%dim_18440} to #hal.device.promise<@__device_2>
%20855 = torch_c.from_builtin_tensor %20854 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20856 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18441 = arith.constant 1 : index
%dim_18442 = tensor.dim %20856, %c1_18441 : tensor<4x?x4096xf16>
%20857 = flow.tensor.transfer %20856 : tensor<4x?x4096xf16>{%dim_18442} to #hal.device.promise<@__device_2>
%20858 = torch_c.from_builtin_tensor %20857 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18443 = torch.constant.int 1
%20859 = torch.aten.add.Tensor %20840, %20843, %int1_18443 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18444 = torch.constant.int 1
%20860 = torch.aten.add.Tensor %20859, %20746, %int1_18444 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18445 = torch.constant.int 1
%20861 = torch.aten.add.Tensor %20860, %20846, %int1_18445 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18446 = torch.constant.int 1
%20862 = torch.aten.add.Tensor %20861, %20849, %int1_18446 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18447 = torch.constant.int 1
%20863 = torch.aten.add.Tensor %20862, %20852, %int1_18447 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18448 = torch.constant.int 1
%20864 = torch.aten.add.Tensor %20863, %20855, %int1_18448 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18449 = torch.constant.int 1
%20865 = torch.aten.add.Tensor %20864, %20858, %int1_18449 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20866 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18450 = arith.constant 1 : index
%dim_18451 = tensor.dim %20866, %c1_18450 : tensor<4x?x4096xf16>
%20867 = flow.tensor.transfer %20866 : tensor<4x?x4096xf16>{%dim_18451} to #hal.device.promise<@__device_3>
%20868 = torch_c.from_builtin_tensor %20867 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20869 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18452 = arith.constant 1 : index
%dim_18453 = tensor.dim %20869, %c1_18452 : tensor<4x?x4096xf16>
%20870 = flow.tensor.transfer %20869 : tensor<4x?x4096xf16>{%dim_18453} to #hal.device.promise<@__device_3>
%20871 = torch_c.from_builtin_tensor %20870 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20872 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18454 = arith.constant 1 : index
%dim_18455 = tensor.dim %20872, %c1_18454 : tensor<4x?x4096xf16>
%20873 = flow.tensor.transfer %20872 : tensor<4x?x4096xf16>{%dim_18455} to #hal.device.promise<@__device_3>
%20874 = torch_c.from_builtin_tensor %20873 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20875 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18456 = arith.constant 1 : index
%dim_18457 = tensor.dim %20875, %c1_18456 : tensor<4x?x4096xf16>
%20876 = flow.tensor.transfer %20875 : tensor<4x?x4096xf16>{%dim_18457} to #hal.device.promise<@__device_3>
%20877 = torch_c.from_builtin_tensor %20876 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20878 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18458 = arith.constant 1 : index
%dim_18459 = tensor.dim %20878, %c1_18458 : tensor<4x?x4096xf16>
%20879 = flow.tensor.transfer %20878 : tensor<4x?x4096xf16>{%dim_18459} to #hal.device.promise<@__device_3>
%20880 = torch_c.from_builtin_tensor %20879 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20881 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18460 = arith.constant 1 : index
%dim_18461 = tensor.dim %20881, %c1_18460 : tensor<4x?x4096xf16>
%20882 = flow.tensor.transfer %20881 : tensor<4x?x4096xf16>{%dim_18461} to #hal.device.promise<@__device_3>
%20883 = torch_c.from_builtin_tensor %20882 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20884 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18462 = arith.constant 1 : index
%dim_18463 = tensor.dim %20884, %c1_18462 : tensor<4x?x4096xf16>
%20885 = flow.tensor.transfer %20884 : tensor<4x?x4096xf16>{%dim_18463} to #hal.device.promise<@__device_3>
%20886 = torch_c.from_builtin_tensor %20885 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18464 = torch.constant.int 1
%20887 = torch.aten.add.Tensor %20868, %20871, %int1_18464 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18465 = torch.constant.int 1
%20888 = torch.aten.add.Tensor %20887, %20874, %int1_18465 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18466 = torch.constant.int 1
%20889 = torch.aten.add.Tensor %20888, %20753, %int1_18466 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18467 = torch.constant.int 1
%20890 = torch.aten.add.Tensor %20889, %20877, %int1_18467 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18468 = torch.constant.int 1
%20891 = torch.aten.add.Tensor %20890, %20880, %int1_18468 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18469 = torch.constant.int 1
%20892 = torch.aten.add.Tensor %20891, %20883, %int1_18469 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18470 = torch.constant.int 1
%20893 = torch.aten.add.Tensor %20892, %20886, %int1_18470 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20894 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18471 = arith.constant 1 : index
%dim_18472 = tensor.dim %20894, %c1_18471 : tensor<4x?x4096xf16>
%20895 = flow.tensor.transfer %20894 : tensor<4x?x4096xf16>{%dim_18472} to #hal.device.promise<@__device_4>
%20896 = torch_c.from_builtin_tensor %20895 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20897 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18473 = arith.constant 1 : index
%dim_18474 = tensor.dim %20897, %c1_18473 : tensor<4x?x4096xf16>
%20898 = flow.tensor.transfer %20897 : tensor<4x?x4096xf16>{%dim_18474} to #hal.device.promise<@__device_4>
%20899 = torch_c.from_builtin_tensor %20898 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20900 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18475 = arith.constant 1 : index
%dim_18476 = tensor.dim %20900, %c1_18475 : tensor<4x?x4096xf16>
%20901 = flow.tensor.transfer %20900 : tensor<4x?x4096xf16>{%dim_18476} to #hal.device.promise<@__device_4>
%20902 = torch_c.from_builtin_tensor %20901 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20903 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18477 = arith.constant 1 : index
%dim_18478 = tensor.dim %20903, %c1_18477 : tensor<4x?x4096xf16>
%20904 = flow.tensor.transfer %20903 : tensor<4x?x4096xf16>{%dim_18478} to #hal.device.promise<@__device_4>
%20905 = torch_c.from_builtin_tensor %20904 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20906 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18479 = arith.constant 1 : index
%dim_18480 = tensor.dim %20906, %c1_18479 : tensor<4x?x4096xf16>
%20907 = flow.tensor.transfer %20906 : tensor<4x?x4096xf16>{%dim_18480} to #hal.device.promise<@__device_4>
%20908 = torch_c.from_builtin_tensor %20907 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20909 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18481 = arith.constant 1 : index
%dim_18482 = tensor.dim %20909, %c1_18481 : tensor<4x?x4096xf16>
%20910 = flow.tensor.transfer %20909 : tensor<4x?x4096xf16>{%dim_18482} to #hal.device.promise<@__device_4>
%20911 = torch_c.from_builtin_tensor %20910 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20912 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18483 = arith.constant 1 : index
%dim_18484 = tensor.dim %20912, %c1_18483 : tensor<4x?x4096xf16>
%20913 = flow.tensor.transfer %20912 : tensor<4x?x4096xf16>{%dim_18484} to #hal.device.promise<@__device_4>
%20914 = torch_c.from_builtin_tensor %20913 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18485 = torch.constant.int 1
%20915 = torch.aten.add.Tensor %20896, %20899, %int1_18485 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18486 = torch.constant.int 1
%20916 = torch.aten.add.Tensor %20915, %20902, %int1_18486 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18487 = torch.constant.int 1
%20917 = torch.aten.add.Tensor %20916, %20905, %int1_18487 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18488 = torch.constant.int 1
%20918 = torch.aten.add.Tensor %20917, %20760, %int1_18488 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18489 = torch.constant.int 1
%20919 = torch.aten.add.Tensor %20918, %20908, %int1_18489 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18490 = torch.constant.int 1
%20920 = torch.aten.add.Tensor %20919, %20911, %int1_18490 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18491 = torch.constant.int 1
%20921 = torch.aten.add.Tensor %20920, %20914, %int1_18491 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20922 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18492 = arith.constant 1 : index
%dim_18493 = tensor.dim %20922, %c1_18492 : tensor<4x?x4096xf16>
%20923 = flow.tensor.transfer %20922 : tensor<4x?x4096xf16>{%dim_18493} to #hal.device.promise<@__device_5>
%20924 = torch_c.from_builtin_tensor %20923 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20925 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18494 = arith.constant 1 : index
%dim_18495 = tensor.dim %20925, %c1_18494 : tensor<4x?x4096xf16>
%20926 = flow.tensor.transfer %20925 : tensor<4x?x4096xf16>{%dim_18495} to #hal.device.promise<@__device_5>
%20927 = torch_c.from_builtin_tensor %20926 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20928 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18496 = arith.constant 1 : index
%dim_18497 = tensor.dim %20928, %c1_18496 : tensor<4x?x4096xf16>
%20929 = flow.tensor.transfer %20928 : tensor<4x?x4096xf16>{%dim_18497} to #hal.device.promise<@__device_5>
%20930 = torch_c.from_builtin_tensor %20929 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20931 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18498 = arith.constant 1 : index
%dim_18499 = tensor.dim %20931, %c1_18498 : tensor<4x?x4096xf16>
%20932 = flow.tensor.transfer %20931 : tensor<4x?x4096xf16>{%dim_18499} to #hal.device.promise<@__device_5>
%20933 = torch_c.from_builtin_tensor %20932 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20934 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18500 = arith.constant 1 : index
%dim_18501 = tensor.dim %20934, %c1_18500 : tensor<4x?x4096xf16>
%20935 = flow.tensor.transfer %20934 : tensor<4x?x4096xf16>{%dim_18501} to #hal.device.promise<@__device_5>
%20936 = torch_c.from_builtin_tensor %20935 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20937 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18502 = arith.constant 1 : index
%dim_18503 = tensor.dim %20937, %c1_18502 : tensor<4x?x4096xf16>
%20938 = flow.tensor.transfer %20937 : tensor<4x?x4096xf16>{%dim_18503} to #hal.device.promise<@__device_5>
%20939 = torch_c.from_builtin_tensor %20938 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20940 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18504 = arith.constant 1 : index
%dim_18505 = tensor.dim %20940, %c1_18504 : tensor<4x?x4096xf16>
%20941 = flow.tensor.transfer %20940 : tensor<4x?x4096xf16>{%dim_18505} to #hal.device.promise<@__device_5>
%20942 = torch_c.from_builtin_tensor %20941 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18506 = torch.constant.int 1
%20943 = torch.aten.add.Tensor %20924, %20927, %int1_18506 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18507 = torch.constant.int 1
%20944 = torch.aten.add.Tensor %20943, %20930, %int1_18507 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18508 = torch.constant.int 1
%20945 = torch.aten.add.Tensor %20944, %20933, %int1_18508 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18509 = torch.constant.int 1
%20946 = torch.aten.add.Tensor %20945, %20936, %int1_18509 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18510 = torch.constant.int 1
%20947 = torch.aten.add.Tensor %20946, %20767, %int1_18510 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18511 = torch.constant.int 1
%20948 = torch.aten.add.Tensor %20947, %20939, %int1_18511 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18512 = torch.constant.int 1
%20949 = torch.aten.add.Tensor %20948, %20942, %int1_18512 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20950 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18513 = arith.constant 1 : index
%dim_18514 = tensor.dim %20950, %c1_18513 : tensor<4x?x4096xf16>
%20951 = flow.tensor.transfer %20950 : tensor<4x?x4096xf16>{%dim_18514} to #hal.device.promise<@__device_6>
%20952 = torch_c.from_builtin_tensor %20951 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20953 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18515 = arith.constant 1 : index
%dim_18516 = tensor.dim %20953, %c1_18515 : tensor<4x?x4096xf16>
%20954 = flow.tensor.transfer %20953 : tensor<4x?x4096xf16>{%dim_18516} to #hal.device.promise<@__device_6>
%20955 = torch_c.from_builtin_tensor %20954 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20956 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18517 = arith.constant 1 : index
%dim_18518 = tensor.dim %20956, %c1_18517 : tensor<4x?x4096xf16>
%20957 = flow.tensor.transfer %20956 : tensor<4x?x4096xf16>{%dim_18518} to #hal.device.promise<@__device_6>
%20958 = torch_c.from_builtin_tensor %20957 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20959 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18519 = arith.constant 1 : index
%dim_18520 = tensor.dim %20959, %c1_18519 : tensor<4x?x4096xf16>
%20960 = flow.tensor.transfer %20959 : tensor<4x?x4096xf16>{%dim_18520} to #hal.device.promise<@__device_6>
%20961 = torch_c.from_builtin_tensor %20960 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20962 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18521 = arith.constant 1 : index
%dim_18522 = tensor.dim %20962, %c1_18521 : tensor<4x?x4096xf16>
%20963 = flow.tensor.transfer %20962 : tensor<4x?x4096xf16>{%dim_18522} to #hal.device.promise<@__device_6>
%20964 = torch_c.from_builtin_tensor %20963 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20965 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18523 = arith.constant 1 : index
%dim_18524 = tensor.dim %20965, %c1_18523 : tensor<4x?x4096xf16>
%20966 = flow.tensor.transfer %20965 : tensor<4x?x4096xf16>{%dim_18524} to #hal.device.promise<@__device_6>
%20967 = torch_c.from_builtin_tensor %20966 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20968 = torch_c.to_builtin_tensor %20781 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18525 = arith.constant 1 : index
%dim_18526 = tensor.dim %20968, %c1_18525 : tensor<4x?x4096xf16>
%20969 = flow.tensor.transfer %20968 : tensor<4x?x4096xf16>{%dim_18526} to #hal.device.promise<@__device_6>
%20970 = torch_c.from_builtin_tensor %20969 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18527 = torch.constant.int 1
%20971 = torch.aten.add.Tensor %20952, %20955, %int1_18527 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18528 = torch.constant.int 1
%20972 = torch.aten.add.Tensor %20971, %20958, %int1_18528 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18529 = torch.constant.int 1
%20973 = torch.aten.add.Tensor %20972, %20961, %int1_18529 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18530 = torch.constant.int 1
%20974 = torch.aten.add.Tensor %20973, %20964, %int1_18530 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18531 = torch.constant.int 1
%20975 = torch.aten.add.Tensor %20974, %20967, %int1_18531 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18532 = torch.constant.int 1
%20976 = torch.aten.add.Tensor %20975, %20774, %int1_18532 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18533 = torch.constant.int 1
%20977 = torch.aten.add.Tensor %20976, %20970, %int1_18533 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20978 = torch_c.to_builtin_tensor %20732 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18534 = arith.constant 1 : index
%dim_18535 = tensor.dim %20978, %c1_18534 : tensor<4x?x4096xf16>
%20979 = flow.tensor.transfer %20978 : tensor<4x?x4096xf16>{%dim_18535} to #hal.device.promise<@__device_7>
%20980 = torch_c.from_builtin_tensor %20979 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20981 = torch_c.to_builtin_tensor %20739 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18536 = arith.constant 1 : index
%dim_18537 = tensor.dim %20981, %c1_18536 : tensor<4x?x4096xf16>
%20982 = flow.tensor.transfer %20981 : tensor<4x?x4096xf16>{%dim_18537} to #hal.device.promise<@__device_7>
%20983 = torch_c.from_builtin_tensor %20982 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20984 = torch_c.to_builtin_tensor %20746 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18538 = arith.constant 1 : index
%dim_18539 = tensor.dim %20984, %c1_18538 : tensor<4x?x4096xf16>
%20985 = flow.tensor.transfer %20984 : tensor<4x?x4096xf16>{%dim_18539} to #hal.device.promise<@__device_7>
%20986 = torch_c.from_builtin_tensor %20985 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20987 = torch_c.to_builtin_tensor %20753 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18540 = arith.constant 1 : index
%dim_18541 = tensor.dim %20987, %c1_18540 : tensor<4x?x4096xf16>
%20988 = flow.tensor.transfer %20987 : tensor<4x?x4096xf16>{%dim_18541} to #hal.device.promise<@__device_7>
%20989 = torch_c.from_builtin_tensor %20988 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20990 = torch_c.to_builtin_tensor %20760 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18542 = arith.constant 1 : index
%dim_18543 = tensor.dim %20990, %c1_18542 : tensor<4x?x4096xf16>
%20991 = flow.tensor.transfer %20990 : tensor<4x?x4096xf16>{%dim_18543} to #hal.device.promise<@__device_7>
%20992 = torch_c.from_builtin_tensor %20991 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20993 = torch_c.to_builtin_tensor %20767 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18544 = arith.constant 1 : index
%dim_18545 = tensor.dim %20993, %c1_18544 : tensor<4x?x4096xf16>
%20994 = flow.tensor.transfer %20993 : tensor<4x?x4096xf16>{%dim_18545} to #hal.device.promise<@__device_7>
%20995 = torch_c.from_builtin_tensor %20994 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%20996 = torch_c.to_builtin_tensor %20774 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_18546 = arith.constant 1 : index
%dim_18547 = tensor.dim %20996, %c1_18546 : tensor<4x?x4096xf16>
%20997 = flow.tensor.transfer %20996 : tensor<4x?x4096xf16>{%dim_18547} to #hal.device.promise<@__device_7>
%20998 = torch_c.from_builtin_tensor %20997 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18548 = torch.constant.int 1
%20999 = torch.aten.add.Tensor %20980, %20983, %int1_18548 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %20999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18549 = torch.constant.int 1
%21000 = torch.aten.add.Tensor %20999, %20986, %int1_18549 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18550 = torch.constant.int 1
%21001 = torch.aten.add.Tensor %21000, %20989, %int1_18550 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18551 = torch.constant.int 1
%21002 = torch.aten.add.Tensor %21001, %20992, %int1_18551 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18552 = torch.constant.int 1
%21003 = torch.aten.add.Tensor %21002, %20995, %int1_18552 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18553 = torch.constant.int 1
%21004 = torch.aten.add.Tensor %21003, %20998, %int1_18553 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18554 = torch.constant.int 1
%21005 = torch.aten.add.Tensor %21004, %20781, %int1_18554 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18555 = torch.constant.int 1
%21006 = torch.aten.add.Tensor %20486, %20809, %int1_18555 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18556 = torch.constant.int 1
%21007 = torch.aten.add.Tensor %20487, %20837, %int1_18556 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18557 = torch.constant.int 1
%21008 = torch.aten.add.Tensor %20488, %20865, %int1_18557 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18558 = torch.constant.int 1
%21009 = torch.aten.add.Tensor %20489, %20893, %int1_18558 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18559 = torch.constant.int 1
%21010 = torch.aten.add.Tensor %20490, %20921, %int1_18559 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18560 = torch.constant.int 1
%21011 = torch.aten.add.Tensor %20491, %20949, %int1_18560 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18561 = torch.constant.int 1
%21012 = torch.aten.add.Tensor %20492, %20977, %int1_18561 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18562 = torch.constant.int 1
%21013 = torch.aten.add.Tensor %20493, %21005, %int1_18562 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_18563 = torch.constant.int 6
%21014 = torch.prims.convert_element_type %21006, %int6_18563 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18564 = torch.constant.int 6
%21015 = torch.prims.convert_element_type %21007, %int6_18564 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18565 = torch.constant.int 6
%21016 = torch.prims.convert_element_type %21008, %int6_18565 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18566 = torch.constant.int 6
%21017 = torch.prims.convert_element_type %21009, %int6_18566 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18567 = torch.constant.int 6
%21018 = torch.prims.convert_element_type %21010, %int6_18567 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18568 = torch.constant.int 6
%21019 = torch.prims.convert_element_type %21011, %int6_18568 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18569 = torch.constant.int 6
%21020 = torch.prims.convert_element_type %21012, %int6_18569 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_18570 = torch.constant.int 6
%21021 = torch.prims.convert_element_type %21013, %int6_18570 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18571 = torch.constant.int 2
%21022 = torch.aten.pow.Tensor_Scalar %21014, %int2_18571 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18572 = torch.constant.int 2
%21023 = torch.aten.pow.Tensor_Scalar %21015, %int2_18572 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18573 = torch.constant.int 2
%21024 = torch.aten.pow.Tensor_Scalar %21016, %int2_18573 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18574 = torch.constant.int 2
%21025 = torch.aten.pow.Tensor_Scalar %21017, %int2_18574 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18575 = torch.constant.int 2
%21026 = torch.aten.pow.Tensor_Scalar %21018, %int2_18575 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18576 = torch.constant.int 2
%21027 = torch.aten.pow.Tensor_Scalar %21019, %int2_18576 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18577 = torch.constant.int 2
%21028 = torch.aten.pow.Tensor_Scalar %21020, %int2_18577 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_18578 = torch.constant.int 2
%21029 = torch.aten.pow.Tensor_Scalar %21021, %int2_18578 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_18579 = torch.constant.int -1
%21030 = torch.prim.ListConstruct %int-1_18579 : (!torch.int) -> !torch.list<int>
%true_18580 = torch.constant.bool true
%none_18581 = torch.constant.none
%21031 = torch.aten.mean.dim %21022, %21030, %true_18580, %none_18581 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18582 = torch.constant.int -1
%21032 = torch.prim.ListConstruct %int-1_18582 : (!torch.int) -> !torch.list<int>
%true_18583 = torch.constant.bool true
%none_18584 = torch.constant.none
%21033 = torch.aten.mean.dim %21023, %21032, %true_18583, %none_18584 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18585 = torch.constant.int -1
%21034 = torch.prim.ListConstruct %int-1_18585 : (!torch.int) -> !torch.list<int>
%true_18586 = torch.constant.bool true
%none_18587 = torch.constant.none
%21035 = torch.aten.mean.dim %21024, %21034, %true_18586, %none_18587 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18588 = torch.constant.int -1
%21036 = torch.prim.ListConstruct %int-1_18588 : (!torch.int) -> !torch.list<int>
%true_18589 = torch.constant.bool true
%none_18590 = torch.constant.none
%21037 = torch.aten.mean.dim %21025, %21036, %true_18589, %none_18590 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18591 = torch.constant.int -1
%21038 = torch.prim.ListConstruct %int-1_18591 : (!torch.int) -> !torch.list<int>
%true_18592 = torch.constant.bool true
%none_18593 = torch.constant.none
%21039 = torch.aten.mean.dim %21026, %21038, %true_18592, %none_18593 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18594 = torch.constant.int -1
%21040 = torch.prim.ListConstruct %int-1_18594 : (!torch.int) -> !torch.list<int>
%true_18595 = torch.constant.bool true
%none_18596 = torch.constant.none
%21041 = torch.aten.mean.dim %21027, %21040, %true_18595, %none_18596 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18597 = torch.constant.int -1
%21042 = torch.prim.ListConstruct %int-1_18597 : (!torch.int) -> !torch.list<int>
%true_18598 = torch.constant.bool true
%none_18599 = torch.constant.none
%21043 = torch.aten.mean.dim %21028, %21042, %true_18598, %none_18599 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_18600 = torch.constant.int -1
%21044 = torch.prim.ListConstruct %int-1_18600 : (!torch.int) -> !torch.list<int>
%true_18601 = torch.constant.bool true
%none_18602 = torch.constant.none
%21045 = torch.aten.mean.dim %21029, %21044, %true_18601, %none_18602 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18603 = torch.constant.float 9.9999997473787516E-6
%int1_18604 = torch.constant.int 1
%21046 = torch.aten.add.Scalar %21031, %float9.999990e-06_18603, %int1_18604 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18605 = torch.constant.float 9.9999997473787516E-6
%int1_18606 = torch.constant.int 1
%21047 = torch.aten.add.Scalar %21033, %float9.999990e-06_18605, %int1_18606 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18607 = torch.constant.float 9.9999997473787516E-6
%int1_18608 = torch.constant.int 1
%21048 = torch.aten.add.Scalar %21035, %float9.999990e-06_18607, %int1_18608 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18609 = torch.constant.float 9.9999997473787516E-6
%int1_18610 = torch.constant.int 1
%21049 = torch.aten.add.Scalar %21037, %float9.999990e-06_18609, %int1_18610 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18611 = torch.constant.float 9.9999997473787516E-6
%int1_18612 = torch.constant.int 1
%21050 = torch.aten.add.Scalar %21039, %float9.999990e-06_18611, %int1_18612 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18613 = torch.constant.float 9.9999997473787516E-6
%int1_18614 = torch.constant.int 1
%21051 = torch.aten.add.Scalar %21041, %float9.999990e-06_18613, %int1_18614 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18615 = torch.constant.float 9.9999997473787516E-6
%int1_18616 = torch.constant.int 1
%21052 = torch.aten.add.Scalar %21043, %float9.999990e-06_18615, %int1_18616 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_18617 = torch.constant.float 9.9999997473787516E-6
%int1_18618 = torch.constant.int 1
%21053 = torch.aten.add.Scalar %21045, %float9.999990e-06_18617, %int1_18618 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21054 = torch.aten.rsqrt %21046 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21055 = torch.aten.rsqrt %21047 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21056 = torch.aten.rsqrt %21048 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21057 = torch.aten.rsqrt %21049 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21058 = torch.aten.rsqrt %21050 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21059 = torch.aten.rsqrt %21051 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21060 = torch.aten.rsqrt %21052 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21061 = torch.aten.rsqrt %21053 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %21061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%21062 = torch.aten.mul.Tensor %21014, %21054 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21063 = torch.aten.mul.Tensor %21015, %21055 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21064 = torch.aten.mul.Tensor %21016, %21056 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21065 = torch.aten.mul.Tensor %21017, %21057 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21066 = torch.aten.mul.Tensor %21018, %21058 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21067 = torch.aten.mul.Tensor %21019, %21059 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21068 = torch.aten.mul.Tensor %21020, %21060 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21069 = torch.aten.mul.Tensor %21021, %21061 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21070 = torch.aten.mul.Tensor %728, %21062 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21071 = torch.aten.mul.Tensor %729, %21063 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21072 = torch.aten.mul.Tensor %730, %21064 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21073 = torch.aten.mul.Tensor %731, %21065 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21074 = torch.aten.mul.Tensor %732, %21066 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21075 = torch.aten.mul.Tensor %733, %21067 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21076 = torch.aten.mul.Tensor %734, %21068 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%21077 = torch.aten.mul.Tensor %735, %21069 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %21077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_18619 = torch.constant.int 5
%21078 = torch.prims.convert_element_type %21070, %int5_18619 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18620 = torch.constant.int 5
%21079 = torch.prims.convert_element_type %21071, %int5_18620 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18621 = torch.constant.int 5
%21080 = torch.prims.convert_element_type %21072, %int5_18621 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18622 = torch.constant.int 5
%21081 = torch.prims.convert_element_type %21073, %int5_18622 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18623 = torch.constant.int 5
%21082 = torch.prims.convert_element_type %21074, %int5_18623 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18624 = torch.constant.int 5
%21083 = torch.prims.convert_element_type %21075, %int5_18624 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18625 = torch.constant.int 5
%21084 = torch.prims.convert_element_type %21076, %int5_18625 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_18626 = torch.constant.int 5
%21085 = torch.prims.convert_element_type %21077, %int5_18626 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %21085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_18627 = torch.constant.int 1
%int0_18628 = torch.constant.int 0
%21086 = torch.prim.ListConstruct %int1_18627, %int0_18628 : (!torch.int, !torch.int) -> !torch.list<int>
%21087 = torch.aten.permute %736, %21086 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18629 = torch.constant.int 1
%int0_18630 = torch.constant.int 0
%21088 = torch.prim.ListConstruct %int1_18629, %int0_18630 : (!torch.int, !torch.int) -> !torch.list<int>
%21089 = torch.aten.permute %737, %21088 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18631 = torch.constant.int 1
%int0_18632 = torch.constant.int 0
%21090 = torch.prim.ListConstruct %int1_18631, %int0_18632 : (!torch.int, !torch.int) -> !torch.list<int>
%21091 = torch.aten.permute %738, %21090 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18633 = torch.constant.int 1
%int0_18634 = torch.constant.int 0
%21092 = torch.prim.ListConstruct %int1_18633, %int0_18634 : (!torch.int, !torch.int) -> !torch.list<int>
%21093 = torch.aten.permute %739, %21092 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18635 = torch.constant.int 1
%int0_18636 = torch.constant.int 0
%21094 = torch.prim.ListConstruct %int1_18635, %int0_18636 : (!torch.int, !torch.int) -> !torch.list<int>
%21095 = torch.aten.permute %740, %21094 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18637 = torch.constant.int 1
%int0_18638 = torch.constant.int 0
%21096 = torch.prim.ListConstruct %int1_18637, %int0_18638 : (!torch.int, !torch.int) -> !torch.list<int>
%21097 = torch.aten.permute %741, %21096 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18639 = torch.constant.int 1
%int0_18640 = torch.constant.int 0
%21098 = torch.prim.ListConstruct %int1_18639, %int0_18640 : (!torch.int, !torch.int) -> !torch.list<int>
%21099 = torch.aten.permute %742, %21098 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_18641 = torch.constant.int 1
%int0_18642 = torch.constant.int 0
%21100 = torch.prim.ListConstruct %int1_18641, %int0_18642 : (!torch.int, !torch.int) -> !torch.list<int>
%21101 = torch.aten.permute %743, %21100 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_18643 = torch.constant.int 4
%21102 = torch.aten.mul.int %int4_18643, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18644 = torch.constant.int 4096
%21103 = torch.prim.ListConstruct %21102, %int4096_18644 : (!torch.int, !torch.int) -> !torch.list<int>
%21104 = torch.aten.view %21078, %21103 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21104, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21105 = torch.aten.mm %21104, %21087 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21105, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18645 = torch.constant.int 4
%int512_18646 = torch.constant.int 512
%21106 = torch.prim.ListConstruct %int4_18645, %2482, %int512_18646 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21107 = torch.aten.view %21105, %21106 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18647 = torch.constant.int 4
%21108 = torch.aten.mul.int %int4_18647, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18648 = torch.constant.int 4096
%21109 = torch.prim.ListConstruct %21108, %int4096_18648 : (!torch.int, !torch.int) -> !torch.list<int>
%21110 = torch.aten.view %21079, %21109 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21110, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21111 = torch.aten.mm %21110, %21089 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21111, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18649 = torch.constant.int 4
%int512_18650 = torch.constant.int 512
%21112 = torch.prim.ListConstruct %int4_18649, %2482, %int512_18650 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21113 = torch.aten.view %21111, %21112 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18651 = torch.constant.int 4
%21114 = torch.aten.mul.int %int4_18651, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18652 = torch.constant.int 4096
%21115 = torch.prim.ListConstruct %21114, %int4096_18652 : (!torch.int, !torch.int) -> !torch.list<int>
%21116 = torch.aten.view %21080, %21115 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21116, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21117 = torch.aten.mm %21116, %21091 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21117, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18653 = torch.constant.int 4
%int512_18654 = torch.constant.int 512
%21118 = torch.prim.ListConstruct %int4_18653, %2482, %int512_18654 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21119 = torch.aten.view %21117, %21118 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18655 = torch.constant.int 4
%21120 = torch.aten.mul.int %int4_18655, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18656 = torch.constant.int 4096
%21121 = torch.prim.ListConstruct %21120, %int4096_18656 : (!torch.int, !torch.int) -> !torch.list<int>
%21122 = torch.aten.view %21081, %21121 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21122, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21123 = torch.aten.mm %21122, %21093 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21123, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18657 = torch.constant.int 4
%int512_18658 = torch.constant.int 512
%21124 = torch.prim.ListConstruct %int4_18657, %2482, %int512_18658 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21125 = torch.aten.view %21123, %21124 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18659 = torch.constant.int 4
%21126 = torch.aten.mul.int %int4_18659, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18660 = torch.constant.int 4096
%21127 = torch.prim.ListConstruct %21126, %int4096_18660 : (!torch.int, !torch.int) -> !torch.list<int>
%21128 = torch.aten.view %21082, %21127 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21128, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21129 = torch.aten.mm %21128, %21095 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21129, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18661 = torch.constant.int 4
%int512_18662 = torch.constant.int 512
%21130 = torch.prim.ListConstruct %int4_18661, %2482, %int512_18662 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21131 = torch.aten.view %21129, %21130 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18663 = torch.constant.int 4
%21132 = torch.aten.mul.int %int4_18663, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18664 = torch.constant.int 4096
%21133 = torch.prim.ListConstruct %21132, %int4096_18664 : (!torch.int, !torch.int) -> !torch.list<int>
%21134 = torch.aten.view %21083, %21133 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21134, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21135 = torch.aten.mm %21134, %21097 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21135, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18665 = torch.constant.int 4
%int512_18666 = torch.constant.int 512
%21136 = torch.prim.ListConstruct %int4_18665, %2482, %int512_18666 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21137 = torch.aten.view %21135, %21136 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18667 = torch.constant.int 4
%21138 = torch.aten.mul.int %int4_18667, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18668 = torch.constant.int 4096
%21139 = torch.prim.ListConstruct %21138, %int4096_18668 : (!torch.int, !torch.int) -> !torch.list<int>
%21140 = torch.aten.view %21084, %21139 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21140, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21141 = torch.aten.mm %21140, %21099 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21141, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18669 = torch.constant.int 4
%int512_18670 = torch.constant.int 512
%21142 = torch.prim.ListConstruct %int4_18669, %2482, %int512_18670 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21143 = torch.aten.view %21141, %21142 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_18671 = torch.constant.int 4
%21144 = torch.aten.mul.int %int4_18671, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18672 = torch.constant.int 4096
%21145 = torch.prim.ListConstruct %21144, %int4096_18672 : (!torch.int, !torch.int) -> !torch.list<int>
%21146 = torch.aten.view %21085, %21145 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21146, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21147 = torch.aten.mm %21146, %21101 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %21147, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_18673 = torch.constant.int 4
%int512_18674 = torch.constant.int 512
%21148 = torch.prim.ListConstruct %int4_18673, %2482, %int512_18674 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21149 = torch.aten.view %21147, %21148 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %21149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_18675 = torch.constant.int 1
%int0_18676 = torch.constant.int 0
%21150 = torch.prim.ListConstruct %int1_18675, %int0_18676 : (!torch.int, !torch.int) -> !torch.list<int>
%21151 = torch.aten.permute %744, %21150 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18677 = torch.constant.int 1
%int0_18678 = torch.constant.int 0
%21152 = torch.prim.ListConstruct %int1_18677, %int0_18678 : (!torch.int, !torch.int) -> !torch.list<int>
%21153 = torch.aten.permute %745, %21152 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18679 = torch.constant.int 1
%int0_18680 = torch.constant.int 0
%21154 = torch.prim.ListConstruct %int1_18679, %int0_18680 : (!torch.int, !torch.int) -> !torch.list<int>
%21155 = torch.aten.permute %746, %21154 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18681 = torch.constant.int 1
%int0_18682 = torch.constant.int 0
%21156 = torch.prim.ListConstruct %int1_18681, %int0_18682 : (!torch.int, !torch.int) -> !torch.list<int>
%21157 = torch.aten.permute %747, %21156 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18683 = torch.constant.int 1
%int0_18684 = torch.constant.int 0
%21158 = torch.prim.ListConstruct %int1_18683, %int0_18684 : (!torch.int, !torch.int) -> !torch.list<int>
%21159 = torch.aten.permute %748, %21158 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18685 = torch.constant.int 1
%int0_18686 = torch.constant.int 0
%21160 = torch.prim.ListConstruct %int1_18685, %int0_18686 : (!torch.int, !torch.int) -> !torch.list<int>
%21161 = torch.aten.permute %749, %21160 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18687 = torch.constant.int 1
%int0_18688 = torch.constant.int 0
%21162 = torch.prim.ListConstruct %int1_18687, %int0_18688 : (!torch.int, !torch.int) -> !torch.list<int>
%21163 = torch.aten.permute %750, %21162 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18689 = torch.constant.int 1
%int0_18690 = torch.constant.int 0
%21164 = torch.prim.ListConstruct %int1_18689, %int0_18690 : (!torch.int, !torch.int) -> !torch.list<int>
%21165 = torch.aten.permute %751, %21164 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_18691 = torch.constant.int 4
%21166 = torch.aten.mul.int %int4_18691, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18692 = torch.constant.int 4096
%21167 = torch.prim.ListConstruct %21166, %int4096_18692 : (!torch.int, !torch.int) -> !torch.list<int>
%21168 = torch.aten.view %21078, %21167 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21168, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21169 = torch.aten.mm %21168, %21151 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21169, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18693 = torch.constant.int 4
%int128_18694 = torch.constant.int 128
%21170 = torch.prim.ListConstruct %int4_18693, %2482, %int128_18694 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21171 = torch.aten.view %21169, %21170 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18695 = torch.constant.int 4
%21172 = torch.aten.mul.int %int4_18695, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18696 = torch.constant.int 4096
%21173 = torch.prim.ListConstruct %21172, %int4096_18696 : (!torch.int, !torch.int) -> !torch.list<int>
%21174 = torch.aten.view %21079, %21173 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21174, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21175 = torch.aten.mm %21174, %21153 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21175, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18697 = torch.constant.int 4
%int128_18698 = torch.constant.int 128
%21176 = torch.prim.ListConstruct %int4_18697, %2482, %int128_18698 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21177 = torch.aten.view %21175, %21176 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18699 = torch.constant.int 4
%21178 = torch.aten.mul.int %int4_18699, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18700 = torch.constant.int 4096
%21179 = torch.prim.ListConstruct %21178, %int4096_18700 : (!torch.int, !torch.int) -> !torch.list<int>
%21180 = torch.aten.view %21080, %21179 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21180, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21181 = torch.aten.mm %21180, %21155 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21181, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18701 = torch.constant.int 4
%int128_18702 = torch.constant.int 128
%21182 = torch.prim.ListConstruct %int4_18701, %2482, %int128_18702 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21183 = torch.aten.view %21181, %21182 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18703 = torch.constant.int 4
%21184 = torch.aten.mul.int %int4_18703, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18704 = torch.constant.int 4096
%21185 = torch.prim.ListConstruct %21184, %int4096_18704 : (!torch.int, !torch.int) -> !torch.list<int>
%21186 = torch.aten.view %21081, %21185 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21186, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21187 = torch.aten.mm %21186, %21157 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21187, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18705 = torch.constant.int 4
%int128_18706 = torch.constant.int 128
%21188 = torch.prim.ListConstruct %int4_18705, %2482, %int128_18706 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21189 = torch.aten.view %21187, %21188 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18707 = torch.constant.int 4
%21190 = torch.aten.mul.int %int4_18707, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18708 = torch.constant.int 4096
%21191 = torch.prim.ListConstruct %21190, %int4096_18708 : (!torch.int, !torch.int) -> !torch.list<int>
%21192 = torch.aten.view %21082, %21191 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21192, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21193 = torch.aten.mm %21192, %21159 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21193, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18709 = torch.constant.int 4
%int128_18710 = torch.constant.int 128
%21194 = torch.prim.ListConstruct %int4_18709, %2482, %int128_18710 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21195 = torch.aten.view %21193, %21194 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18711 = torch.constant.int 4
%21196 = torch.aten.mul.int %int4_18711, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18712 = torch.constant.int 4096
%21197 = torch.prim.ListConstruct %21196, %int4096_18712 : (!torch.int, !torch.int) -> !torch.list<int>
%21198 = torch.aten.view %21083, %21197 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21198, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21199 = torch.aten.mm %21198, %21161 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21199, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18713 = torch.constant.int 4
%int128_18714 = torch.constant.int 128
%21200 = torch.prim.ListConstruct %int4_18713, %2482, %int128_18714 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21201 = torch.aten.view %21199, %21200 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18715 = torch.constant.int 4
%21202 = torch.aten.mul.int %int4_18715, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18716 = torch.constant.int 4096
%21203 = torch.prim.ListConstruct %21202, %int4096_18716 : (!torch.int, !torch.int) -> !torch.list<int>
%21204 = torch.aten.view %21084, %21203 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21204, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21205 = torch.aten.mm %21204, %21163 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21205, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18717 = torch.constant.int 4
%int128_18718 = torch.constant.int 128
%21206 = torch.prim.ListConstruct %int4_18717, %2482, %int128_18718 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21207 = torch.aten.view %21205, %21206 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18719 = torch.constant.int 4
%21208 = torch.aten.mul.int %int4_18719, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18720 = torch.constant.int 4096
%21209 = torch.prim.ListConstruct %21208, %int4096_18720 : (!torch.int, !torch.int) -> !torch.list<int>
%21210 = torch.aten.view %21085, %21209 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21210, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21211 = torch.aten.mm %21210, %21165 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21211, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18721 = torch.constant.int 4
%int128_18722 = torch.constant.int 128
%21212 = torch.prim.ListConstruct %int4_18721, %2482, %int128_18722 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21213 = torch.aten.view %21211, %21212 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_18723 = torch.constant.int 1
%int0_18724 = torch.constant.int 0
%21214 = torch.prim.ListConstruct %int1_18723, %int0_18724 : (!torch.int, !torch.int) -> !torch.list<int>
%21215 = torch.aten.permute %752, %21214 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18725 = torch.constant.int 1
%int0_18726 = torch.constant.int 0
%21216 = torch.prim.ListConstruct %int1_18725, %int0_18726 : (!torch.int, !torch.int) -> !torch.list<int>
%21217 = torch.aten.permute %753, %21216 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18727 = torch.constant.int 1
%int0_18728 = torch.constant.int 0
%21218 = torch.prim.ListConstruct %int1_18727, %int0_18728 : (!torch.int, !torch.int) -> !torch.list<int>
%21219 = torch.aten.permute %754, %21218 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18729 = torch.constant.int 1
%int0_18730 = torch.constant.int 0
%21220 = torch.prim.ListConstruct %int1_18729, %int0_18730 : (!torch.int, !torch.int) -> !torch.list<int>
%21221 = torch.aten.permute %755, %21220 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18731 = torch.constant.int 1
%int0_18732 = torch.constant.int 0
%21222 = torch.prim.ListConstruct %int1_18731, %int0_18732 : (!torch.int, !torch.int) -> !torch.list<int>
%21223 = torch.aten.permute %756, %21222 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18733 = torch.constant.int 1
%int0_18734 = torch.constant.int 0
%21224 = torch.prim.ListConstruct %int1_18733, %int0_18734 : (!torch.int, !torch.int) -> !torch.list<int>
%21225 = torch.aten.permute %757, %21224 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18735 = torch.constant.int 1
%int0_18736 = torch.constant.int 0
%21226 = torch.prim.ListConstruct %int1_18735, %int0_18736 : (!torch.int, !torch.int) -> !torch.list<int>
%21227 = torch.aten.permute %758, %21226 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_18737 = torch.constant.int 1
%int0_18738 = torch.constant.int 0
%21228 = torch.prim.ListConstruct %int1_18737, %int0_18738 : (!torch.int, !torch.int) -> !torch.list<int>
%21229 = torch.aten.permute %759, %21228 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_18739 = torch.constant.int 4
%21230 = torch.aten.mul.int %int4_18739, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18740 = torch.constant.int 4096
%21231 = torch.prim.ListConstruct %21230, %int4096_18740 : (!torch.int, !torch.int) -> !torch.list<int>
%21232 = torch.aten.view %21078, %21231 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21232, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21233 = torch.aten.mm %21232, %21215 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21233, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18741 = torch.constant.int 4
%int128_18742 = torch.constant.int 128
%21234 = torch.prim.ListConstruct %int4_18741, %2482, %int128_18742 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21235 = torch.aten.view %21233, %21234 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18743 = torch.constant.int 4
%21236 = torch.aten.mul.int %int4_18743, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18744 = torch.constant.int 4096
%21237 = torch.prim.ListConstruct %21236, %int4096_18744 : (!torch.int, !torch.int) -> !torch.list<int>
%21238 = torch.aten.view %21079, %21237 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21238, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21239 = torch.aten.mm %21238, %21217 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21239, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18745 = torch.constant.int 4
%int128_18746 = torch.constant.int 128
%21240 = torch.prim.ListConstruct %int4_18745, %2482, %int128_18746 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21241 = torch.aten.view %21239, %21240 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18747 = torch.constant.int 4
%21242 = torch.aten.mul.int %int4_18747, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18748 = torch.constant.int 4096
%21243 = torch.prim.ListConstruct %21242, %int4096_18748 : (!torch.int, !torch.int) -> !torch.list<int>
%21244 = torch.aten.view %21080, %21243 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21244, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21245 = torch.aten.mm %21244, %21219 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21245, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18749 = torch.constant.int 4
%int128_18750 = torch.constant.int 128
%21246 = torch.prim.ListConstruct %int4_18749, %2482, %int128_18750 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21247 = torch.aten.view %21245, %21246 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18751 = torch.constant.int 4
%21248 = torch.aten.mul.int %int4_18751, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18752 = torch.constant.int 4096
%21249 = torch.prim.ListConstruct %21248, %int4096_18752 : (!torch.int, !torch.int) -> !torch.list<int>
%21250 = torch.aten.view %21081, %21249 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21250, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21251 = torch.aten.mm %21250, %21221 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21251, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18753 = torch.constant.int 4
%int128_18754 = torch.constant.int 128
%21252 = torch.prim.ListConstruct %int4_18753, %2482, %int128_18754 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21253 = torch.aten.view %21251, %21252 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18755 = torch.constant.int 4
%21254 = torch.aten.mul.int %int4_18755, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18756 = torch.constant.int 4096
%21255 = torch.prim.ListConstruct %21254, %int4096_18756 : (!torch.int, !torch.int) -> !torch.list<int>
%21256 = torch.aten.view %21082, %21255 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21256, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21257 = torch.aten.mm %21256, %21223 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21257, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18757 = torch.constant.int 4
%int128_18758 = torch.constant.int 128
%21258 = torch.prim.ListConstruct %int4_18757, %2482, %int128_18758 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21259 = torch.aten.view %21257, %21258 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18759 = torch.constant.int 4
%21260 = torch.aten.mul.int %int4_18759, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18760 = torch.constant.int 4096
%21261 = torch.prim.ListConstruct %21260, %int4096_18760 : (!torch.int, !torch.int) -> !torch.list<int>
%21262 = torch.aten.view %21083, %21261 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21262, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21263 = torch.aten.mm %21262, %21225 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21263, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18761 = torch.constant.int 4
%int128_18762 = torch.constant.int 128
%21264 = torch.prim.ListConstruct %int4_18761, %2482, %int128_18762 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21265 = torch.aten.view %21263, %21264 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18763 = torch.constant.int 4
%21266 = torch.aten.mul.int %int4_18763, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18764 = torch.constant.int 4096
%21267 = torch.prim.ListConstruct %21266, %int4096_18764 : (!torch.int, !torch.int) -> !torch.list<int>
%21268 = torch.aten.view %21084, %21267 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21268, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21269 = torch.aten.mm %21268, %21227 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21269, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18765 = torch.constant.int 4
%int128_18766 = torch.constant.int 128
%21270 = torch.prim.ListConstruct %int4_18765, %2482, %int128_18766 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21271 = torch.aten.view %21269, %21270 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18767 = torch.constant.int 4
%21272 = torch.aten.mul.int %int4_18767, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_18768 = torch.constant.int 4096
%21273 = torch.prim.ListConstruct %21272, %int4096_18768 : (!torch.int, !torch.int) -> !torch.list<int>
%21274 = torch.aten.view %21085, %21273 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %21274, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%21275 = torch.aten.mm %21274, %21229 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %21275, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_18769 = torch.constant.int 4
%int128_18770 = torch.constant.int 128
%21276 = torch.prim.ListConstruct %int4_18769, %2482, %int128_18770 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21277 = torch.aten.view %21275, %21276 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %21277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_18771 = torch.constant.int 4
%int4_18772 = torch.constant.int 4
%int128_18773 = torch.constant.int 128
%21278 = torch.prim.ListConstruct %int4_18771, %2482, %int4_18772, %int128_18773 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21279 = torch.aten.view %21107, %21278 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18774 = torch.constant.int 4
%int4_18775 = torch.constant.int 4
%int128_18776 = torch.constant.int 128
%21280 = torch.prim.ListConstruct %int4_18774, %2482, %int4_18775, %int128_18776 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21281 = torch.aten.view %21113, %21280 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18777 = torch.constant.int 4
%int4_18778 = torch.constant.int 4
%int128_18779 = torch.constant.int 128
%21282 = torch.prim.ListConstruct %int4_18777, %2482, %int4_18778, %int128_18779 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21283 = torch.aten.view %21119, %21282 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18780 = torch.constant.int 4
%int4_18781 = torch.constant.int 4
%int128_18782 = torch.constant.int 128
%21284 = torch.prim.ListConstruct %int4_18780, %2482, %int4_18781, %int128_18782 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21285 = torch.aten.view %21125, %21284 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18783 = torch.constant.int 4
%int4_18784 = torch.constant.int 4
%int128_18785 = torch.constant.int 128
%21286 = torch.prim.ListConstruct %int4_18783, %2482, %int4_18784, %int128_18785 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21287 = torch.aten.view %21131, %21286 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18786 = torch.constant.int 4
%int4_18787 = torch.constant.int 4
%int128_18788 = torch.constant.int 128
%21288 = torch.prim.ListConstruct %int4_18786, %2482, %int4_18787, %int128_18788 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21289 = torch.aten.view %21137, %21288 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18789 = torch.constant.int 4
%int4_18790 = torch.constant.int 4
%int128_18791 = torch.constant.int 128
%21290 = torch.prim.ListConstruct %int4_18789, %2482, %int4_18790, %int128_18791 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21291 = torch.aten.view %21143, %21290 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18792 = torch.constant.int 4
%int4_18793 = torch.constant.int 4
%int128_18794 = torch.constant.int 128
%21292 = torch.prim.ListConstruct %int4_18792, %2482, %int4_18793, %int128_18794 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21293 = torch.aten.view %21149, %21292 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_18795 = torch.constant.int 4
%int1_18796 = torch.constant.int 1
%int128_18797 = torch.constant.int 128
%21294 = torch.prim.ListConstruct %int4_18795, %2482, %int1_18796, %int128_18797 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21295 = torch.aten.view %21171, %21294 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18798 = torch.constant.int 4
%int1_18799 = torch.constant.int 1
%int128_18800 = torch.constant.int 128
%21296 = torch.prim.ListConstruct %int4_18798, %2482, %int1_18799, %int128_18800 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21297 = torch.aten.view %21177, %21296 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18801 = torch.constant.int 4
%int1_18802 = torch.constant.int 1
%int128_18803 = torch.constant.int 128
%21298 = torch.prim.ListConstruct %int4_18801, %2482, %int1_18802, %int128_18803 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21299 = torch.aten.view %21183, %21298 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18804 = torch.constant.int 4
%int1_18805 = torch.constant.int 1
%int128_18806 = torch.constant.int 128
%21300 = torch.prim.ListConstruct %int4_18804, %2482, %int1_18805, %int128_18806 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21301 = torch.aten.view %21189, %21300 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18807 = torch.constant.int 4
%int1_18808 = torch.constant.int 1
%int128_18809 = torch.constant.int 128
%21302 = torch.prim.ListConstruct %int4_18807, %2482, %int1_18808, %int128_18809 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21303 = torch.aten.view %21195, %21302 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18810 = torch.constant.int 4
%int1_18811 = torch.constant.int 1
%int128_18812 = torch.constant.int 128
%21304 = torch.prim.ListConstruct %int4_18810, %2482, %int1_18811, %int128_18812 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21305 = torch.aten.view %21201, %21304 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18813 = torch.constant.int 4
%int1_18814 = torch.constant.int 1
%int128_18815 = torch.constant.int 128
%21306 = torch.prim.ListConstruct %int4_18813, %2482, %int1_18814, %int128_18815 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21307 = torch.aten.view %21207, %21306 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18816 = torch.constant.int 4
%int1_18817 = torch.constant.int 1
%int128_18818 = torch.constant.int 128
%21308 = torch.prim.ListConstruct %int4_18816, %2482, %int1_18817, %int128_18818 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21309 = torch.aten.view %21213, %21308 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18819 = torch.constant.int 4
%int1_18820 = torch.constant.int 1
%int128_18821 = torch.constant.int 128
%21310 = torch.prim.ListConstruct %int4_18819, %2482, %int1_18820, %int128_18821 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21311 = torch.aten.view %21235, %21310 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18822 = torch.constant.int 4
%int1_18823 = torch.constant.int 1
%int128_18824 = torch.constant.int 128
%21312 = torch.prim.ListConstruct %int4_18822, %2482, %int1_18823, %int128_18824 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21313 = torch.aten.view %21241, %21312 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18825 = torch.constant.int 4
%int1_18826 = torch.constant.int 1
%int128_18827 = torch.constant.int 128
%21314 = torch.prim.ListConstruct %int4_18825, %2482, %int1_18826, %int128_18827 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21315 = torch.aten.view %21247, %21314 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18828 = torch.constant.int 4
%int1_18829 = torch.constant.int 1
%int128_18830 = torch.constant.int 128
%21316 = torch.prim.ListConstruct %int4_18828, %2482, %int1_18829, %int128_18830 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21317 = torch.aten.view %21253, %21316 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18831 = torch.constant.int 4
%int1_18832 = torch.constant.int 1
%int128_18833 = torch.constant.int 128
%21318 = torch.prim.ListConstruct %int4_18831, %2482, %int1_18832, %int128_18833 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21319 = torch.aten.view %21259, %21318 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18834 = torch.constant.int 4
%int1_18835 = torch.constant.int 1
%int128_18836 = torch.constant.int 128
%21320 = torch.prim.ListConstruct %int4_18834, %2482, %int1_18835, %int128_18836 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21321 = torch.aten.view %21265, %21320 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18837 = torch.constant.int 4
%int1_18838 = torch.constant.int 1
%int128_18839 = torch.constant.int 128
%21322 = torch.prim.ListConstruct %int4_18837, %2482, %int1_18838, %int128_18839 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21323 = torch.aten.view %21271, %21322 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_18840 = torch.constant.int 4
%int1_18841 = torch.constant.int 1
%int128_18842 = torch.constant.int 128
%21324 = torch.prim.ListConstruct %int4_18840, %2482, %int1_18841, %int128_18842 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21325 = torch.aten.view %21277, %21324 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_18843 = torch.constant.int 131072
%none_18844 = torch.constant.none
%none_18845 = torch.constant.none
%cpu_18846 = torch.constant.device "cpu"
%false_18847 = torch.constant.bool false
%21326 = torch.aten.arange %int131072_18843, %none_18844, %none_18845, %cpu_18846, %false_18847 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_18848 = torch.constant.int 0
%int128_18849 = torch.constant.int 128
%int2_18850 = torch.constant.int 2
%none_18851 = torch.constant.none
%none_18852 = torch.constant.none
%cpu_18853 = torch.constant.device "cpu"
%false_18854 = torch.constant.bool false
%21327 = torch.aten.arange.start_step %int0_18848, %int128_18849, %int2_18850, %none_18851, %none_18852, %cpu_18853, %false_18854 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_18855 = torch.constant.int 0
%int0_18856 = torch.constant.int 0
%int64_18857 = torch.constant.int 64
%int1_18858 = torch.constant.int 1
%21328 = torch.aten.slice.Tensor %21327, %int0_18855, %int0_18856, %int64_18857, %int1_18858 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_18859 = torch.constant.int 6
%21329 = torch.prims.convert_element_type %21328, %int6_18859 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_18860 = torch.constant.int 128
%21330 = torch.aten.div.Scalar %21329, %int128_18860 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_18861 = torch.constant.float 5.000000e+05
%21331 = torch.aten.pow.Scalar %float5.000000e05_18861, %21330 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%21332 = torch.aten.reciprocal %21331 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_18862 = torch.constant.float 1.000000e+00
%21333 = torch.aten.mul.Scalar %21332, %float1.000000e00_18862 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_18863 = torch.constant.int 131072
%int1_18864 = torch.constant.int 1
%21334 = torch.prim.ListConstruct %int131072_18863, %int1_18864 : (!torch.int, !torch.int) -> !torch.list<int>
%21335 = torch.aten.view %21326, %21334 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%21336 = torch.aten.mul.Tensor %21335, %21333 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%21337 = torch.aten.cos %21336 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%21338 = torch.aten.sin %21336 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%21339 = torch.aten.complex %21337, %21338 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%21340 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21341 = flow.tensor.transfer %21340 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%21342 = torch_c.from_builtin_tensor %21341 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21343 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21344 = flow.tensor.transfer %21343 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%21345 = torch_c.from_builtin_tensor %21344 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21346 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21347 = flow.tensor.transfer %21346 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%21348 = torch_c.from_builtin_tensor %21347 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21349 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21350 = flow.tensor.transfer %21349 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%21351 = torch_c.from_builtin_tensor %21350 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21352 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21353 = flow.tensor.transfer %21352 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%21354 = torch_c.from_builtin_tensor %21353 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21355 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21356 = flow.tensor.transfer %21355 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%21357 = torch_c.from_builtin_tensor %21356 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21358 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21359 = flow.tensor.transfer %21358 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%21360 = torch_c.from_builtin_tensor %21359 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21361 = torch_c.to_builtin_tensor %21339 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21362 = flow.tensor.transfer %21361 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%21363 = torch_c.from_builtin_tensor %21362 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_18865 = torch.constant.int 1
%21364 = torch.aten.size.int %21107, %int1_18865 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18866 = torch.constant.int 0
%21365 = torch.aten.add.int %int0_18866, %21364 : !torch.int, !torch.int -> !torch.int
%int0_18867 = torch.constant.int 0
%int0_18868 = torch.constant.int 0
%int1_18869 = torch.constant.int 1
%21366 = torch.aten.slice.Tensor %21342, %int0_18867, %int0_18868, %21365, %int1_18869 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21366, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18870 = torch.constant.int 1
%int0_18871 = torch.constant.int 0
%int9223372036854775807_18872 = torch.constant.int 9223372036854775807
%int1_18873 = torch.constant.int 1
%21367 = torch.aten.slice.Tensor %21366, %int1_18870, %int0_18871, %int9223372036854775807_18872, %int1_18873 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21367, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18874 = torch.constant.int 0
%21368 = torch.aten.unsqueeze %21367, %int0_18874 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21368, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18875 = torch.constant.int 2
%21369 = torch.aten.unsqueeze %21368, %int2_18875 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21369, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18876 = torch.constant.int 3
%int0_18877 = torch.constant.int 0
%int9223372036854775807_18878 = torch.constant.int 9223372036854775807
%int1_18879 = torch.constant.int 1
%21370 = torch.aten.slice.Tensor %21369, %int3_18876, %int0_18877, %int9223372036854775807_18878, %int1_18879 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21370, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21371 = torch_c.to_builtin_tensor %21279 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_18880 = arith.constant 1 : index
%dim_18881 = tensor.dim %21371, %c1_18880 : tensor<4x?x4x128xf16>
%21372 = flow.tensor.bitcast %21371 : tensor<4x?x4x128xf16>{%dim_18881} -> tensor<4x?x4x64xcomplex<f16>>{%dim_18881}
%21373 = torch_c.from_builtin_tensor %21372 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21374 = torch.aten.mul.Tensor %21373, %21370 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21375 = torch_c.to_builtin_tensor %21374 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_18882 = arith.constant 1 : index
%dim_18883 = tensor.dim %21375, %c1_18882 : tensor<4x?x4x64xcomplex<f32>>
%21376 = flow.tensor.bitcast %21375 : tensor<4x?x4x64xcomplex<f32>>{%dim_18883} -> tensor<4x?x4x128xf32>{%dim_18883}
%21377 = torch_c.from_builtin_tensor %21376 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_18884 = torch.constant.int 5
%21378 = torch.prims.convert_element_type %21377, %int5_18884 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_18885 = torch.constant.int 1
%21379 = torch.aten.size.int %21113, %int1_18885 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18886 = torch.constant.int 0
%21380 = torch.aten.add.int %int0_18886, %21379 : !torch.int, !torch.int -> !torch.int
%int0_18887 = torch.constant.int 0
%int0_18888 = torch.constant.int 0
%int1_18889 = torch.constant.int 1
%21381 = torch.aten.slice.Tensor %21345, %int0_18887, %int0_18888, %21380, %int1_18889 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21381, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18890 = torch.constant.int 1
%int0_18891 = torch.constant.int 0
%int9223372036854775807_18892 = torch.constant.int 9223372036854775807
%int1_18893 = torch.constant.int 1
%21382 = torch.aten.slice.Tensor %21381, %int1_18890, %int0_18891, %int9223372036854775807_18892, %int1_18893 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21382, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18894 = torch.constant.int 0
%21383 = torch.aten.unsqueeze %21382, %int0_18894 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21383, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18895 = torch.constant.int 2
%21384 = torch.aten.unsqueeze %21383, %int2_18895 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21384, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18896 = torch.constant.int 3
%int0_18897 = torch.constant.int 0
%int9223372036854775807_18898 = torch.constant.int 9223372036854775807
%int1_18899 = torch.constant.int 1
%21385 = torch.aten.slice.Tensor %21384, %int3_18896, %int0_18897, %int9223372036854775807_18898, %int1_18899 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21385, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21386 = torch_c.to_builtin_tensor %21281 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_18900 = arith.constant 1 : index
%dim_18901 = tensor.dim %21386, %c1_18900 : tensor<4x?x4x128xf16>
%21387 = flow.tensor.bitcast %21386 : tensor<4x?x4x128xf16>{%dim_18901} -> tensor<4x?x4x64xcomplex<f16>>{%dim_18901}
%21388 = torch_c.from_builtin_tensor %21387 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21389 = torch.aten.mul.Tensor %21388, %21385 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21390 = torch_c.to_builtin_tensor %21389 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_18902 = arith.constant 1 : index
%dim_18903 = tensor.dim %21390, %c1_18902 : tensor<4x?x4x64xcomplex<f32>>
%21391 = flow.tensor.bitcast %21390 : tensor<4x?x4x64xcomplex<f32>>{%dim_18903} -> tensor<4x?x4x128xf32>{%dim_18903}
%21392 = torch_c.from_builtin_tensor %21391 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_18904 = torch.constant.int 5
%21393 = torch.prims.convert_element_type %21392, %int5_18904 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_18905 = torch.constant.int 1
%21394 = torch.aten.size.int %21119, %int1_18905 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18906 = torch.constant.int 0
%21395 = torch.aten.add.int %int0_18906, %21394 : !torch.int, !torch.int -> !torch.int
%int0_18907 = torch.constant.int 0
%int0_18908 = torch.constant.int 0
%int1_18909 = torch.constant.int 1
%21396 = torch.aten.slice.Tensor %21348, %int0_18907, %int0_18908, %21395, %int1_18909 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21396, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18910 = torch.constant.int 1
%int0_18911 = torch.constant.int 0
%int9223372036854775807_18912 = torch.constant.int 9223372036854775807
%int1_18913 = torch.constant.int 1
%21397 = torch.aten.slice.Tensor %21396, %int1_18910, %int0_18911, %int9223372036854775807_18912, %int1_18913 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21397, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18914 = torch.constant.int 0
%21398 = torch.aten.unsqueeze %21397, %int0_18914 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21398, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18915 = torch.constant.int 2
%21399 = torch.aten.unsqueeze %21398, %int2_18915 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21399, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18916 = torch.constant.int 3
%int0_18917 = torch.constant.int 0
%int9223372036854775807_18918 = torch.constant.int 9223372036854775807
%int1_18919 = torch.constant.int 1
%21400 = torch.aten.slice.Tensor %21399, %int3_18916, %int0_18917, %int9223372036854775807_18918, %int1_18919 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21400, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21401 = torch_c.to_builtin_tensor %21283 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_18920 = arith.constant 1 : index
%dim_18921 = tensor.dim %21401, %c1_18920 : tensor<4x?x4x128xf16>
%21402 = flow.tensor.bitcast %21401 : tensor<4x?x4x128xf16>{%dim_18921} -> tensor<4x?x4x64xcomplex<f16>>{%dim_18921}
%21403 = torch_c.from_builtin_tensor %21402 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21404 = torch.aten.mul.Tensor %21403, %21400 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21405 = torch_c.to_builtin_tensor %21404 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_18922 = arith.constant 1 : index
%dim_18923 = tensor.dim %21405, %c1_18922 : tensor<4x?x4x64xcomplex<f32>>
%21406 = flow.tensor.bitcast %21405 : tensor<4x?x4x64xcomplex<f32>>{%dim_18923} -> tensor<4x?x4x128xf32>{%dim_18923}
%21407 = torch_c.from_builtin_tensor %21406 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_18924 = torch.constant.int 5
%21408 = torch.prims.convert_element_type %21407, %int5_18924 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_18925 = torch.constant.int 1
%21409 = torch.aten.size.int %21125, %int1_18925 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18926 = torch.constant.int 0
%21410 = torch.aten.add.int %int0_18926, %21409 : !torch.int, !torch.int -> !torch.int
%int0_18927 = torch.constant.int 0
%int0_18928 = torch.constant.int 0
%int1_18929 = torch.constant.int 1
%21411 = torch.aten.slice.Tensor %21351, %int0_18927, %int0_18928, %21410, %int1_18929 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21411, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18930 = torch.constant.int 1
%int0_18931 = torch.constant.int 0
%int9223372036854775807_18932 = torch.constant.int 9223372036854775807
%int1_18933 = torch.constant.int 1
%21412 = torch.aten.slice.Tensor %21411, %int1_18930, %int0_18931, %int9223372036854775807_18932, %int1_18933 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21412, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18934 = torch.constant.int 0
%21413 = torch.aten.unsqueeze %21412, %int0_18934 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21413, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18935 = torch.constant.int 2
%21414 = torch.aten.unsqueeze %21413, %int2_18935 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21414, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18936 = torch.constant.int 3
%int0_18937 = torch.constant.int 0
%int9223372036854775807_18938 = torch.constant.int 9223372036854775807
%int1_18939 = torch.constant.int 1
%21415 = torch.aten.slice.Tensor %21414, %int3_18936, %int0_18937, %int9223372036854775807_18938, %int1_18939 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21415, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21416 = torch_c.to_builtin_tensor %21285 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_18940 = arith.constant 1 : index
%dim_18941 = tensor.dim %21416, %c1_18940 : tensor<4x?x4x128xf16>
%21417 = flow.tensor.bitcast %21416 : tensor<4x?x4x128xf16>{%dim_18941} -> tensor<4x?x4x64xcomplex<f16>>{%dim_18941}
%21418 = torch_c.from_builtin_tensor %21417 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21419 = torch.aten.mul.Tensor %21418, %21415 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21420 = torch_c.to_builtin_tensor %21419 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_18942 = arith.constant 1 : index
%dim_18943 = tensor.dim %21420, %c1_18942 : tensor<4x?x4x64xcomplex<f32>>
%21421 = flow.tensor.bitcast %21420 : tensor<4x?x4x64xcomplex<f32>>{%dim_18943} -> tensor<4x?x4x128xf32>{%dim_18943}
%21422 = torch_c.from_builtin_tensor %21421 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_18944 = torch.constant.int 5
%21423 = torch.prims.convert_element_type %21422, %int5_18944 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_18945 = torch.constant.int 1
%21424 = torch.aten.size.int %21131, %int1_18945 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18946 = torch.constant.int 0
%21425 = torch.aten.add.int %int0_18946, %21424 : !torch.int, !torch.int -> !torch.int
%int0_18947 = torch.constant.int 0
%int0_18948 = torch.constant.int 0
%int1_18949 = torch.constant.int 1
%21426 = torch.aten.slice.Tensor %21354, %int0_18947, %int0_18948, %21425, %int1_18949 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21426, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18950 = torch.constant.int 1
%int0_18951 = torch.constant.int 0
%int9223372036854775807_18952 = torch.constant.int 9223372036854775807
%int1_18953 = torch.constant.int 1
%21427 = torch.aten.slice.Tensor %21426, %int1_18950, %int0_18951, %int9223372036854775807_18952, %int1_18953 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21427, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18954 = torch.constant.int 0
%21428 = torch.aten.unsqueeze %21427, %int0_18954 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21428, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18955 = torch.constant.int 2
%21429 = torch.aten.unsqueeze %21428, %int2_18955 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21429, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18956 = torch.constant.int 3
%int0_18957 = torch.constant.int 0
%int9223372036854775807_18958 = torch.constant.int 9223372036854775807
%int1_18959 = torch.constant.int 1
%21430 = torch.aten.slice.Tensor %21429, %int3_18956, %int0_18957, %int9223372036854775807_18958, %int1_18959 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21430, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21431 = torch_c.to_builtin_tensor %21287 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_18960 = arith.constant 1 : index
%dim_18961 = tensor.dim %21431, %c1_18960 : tensor<4x?x4x128xf16>
%21432 = flow.tensor.bitcast %21431 : tensor<4x?x4x128xf16>{%dim_18961} -> tensor<4x?x4x64xcomplex<f16>>{%dim_18961}
%21433 = torch_c.from_builtin_tensor %21432 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21434 = torch.aten.mul.Tensor %21433, %21430 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21435 = torch_c.to_builtin_tensor %21434 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_18962 = arith.constant 1 : index
%dim_18963 = tensor.dim %21435, %c1_18962 : tensor<4x?x4x64xcomplex<f32>>
%21436 = flow.tensor.bitcast %21435 : tensor<4x?x4x64xcomplex<f32>>{%dim_18963} -> tensor<4x?x4x128xf32>{%dim_18963}
%21437 = torch_c.from_builtin_tensor %21436 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_18964 = torch.constant.int 5
%21438 = torch.prims.convert_element_type %21437, %int5_18964 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_18965 = torch.constant.int 1
%21439 = torch.aten.size.int %21137, %int1_18965 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18966 = torch.constant.int 0
%21440 = torch.aten.add.int %int0_18966, %21439 : !torch.int, !torch.int -> !torch.int
%int0_18967 = torch.constant.int 0
%int0_18968 = torch.constant.int 0
%int1_18969 = torch.constant.int 1
%21441 = torch.aten.slice.Tensor %21357, %int0_18967, %int0_18968, %21440, %int1_18969 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21441, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18970 = torch.constant.int 1
%int0_18971 = torch.constant.int 0
%int9223372036854775807_18972 = torch.constant.int 9223372036854775807
%int1_18973 = torch.constant.int 1
%21442 = torch.aten.slice.Tensor %21441, %int1_18970, %int0_18971, %int9223372036854775807_18972, %int1_18973 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21442, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18974 = torch.constant.int 0
%21443 = torch.aten.unsqueeze %21442, %int0_18974 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21443, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18975 = torch.constant.int 2
%21444 = torch.aten.unsqueeze %21443, %int2_18975 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21444, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18976 = torch.constant.int 3
%int0_18977 = torch.constant.int 0
%int9223372036854775807_18978 = torch.constant.int 9223372036854775807
%int1_18979 = torch.constant.int 1
%21445 = torch.aten.slice.Tensor %21444, %int3_18976, %int0_18977, %int9223372036854775807_18978, %int1_18979 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21445, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21446 = torch_c.to_builtin_tensor %21289 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_18980 = arith.constant 1 : index
%dim_18981 = tensor.dim %21446, %c1_18980 : tensor<4x?x4x128xf16>
%21447 = flow.tensor.bitcast %21446 : tensor<4x?x4x128xf16>{%dim_18981} -> tensor<4x?x4x64xcomplex<f16>>{%dim_18981}
%21448 = torch_c.from_builtin_tensor %21447 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21449 = torch.aten.mul.Tensor %21448, %21445 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21450 = torch_c.to_builtin_tensor %21449 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_18982 = arith.constant 1 : index
%dim_18983 = tensor.dim %21450, %c1_18982 : tensor<4x?x4x64xcomplex<f32>>
%21451 = flow.tensor.bitcast %21450 : tensor<4x?x4x64xcomplex<f32>>{%dim_18983} -> tensor<4x?x4x128xf32>{%dim_18983}
%21452 = torch_c.from_builtin_tensor %21451 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_18984 = torch.constant.int 5
%21453 = torch.prims.convert_element_type %21452, %int5_18984 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_18985 = torch.constant.int 1
%21454 = torch.aten.size.int %21143, %int1_18985 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_18986 = torch.constant.int 0
%21455 = torch.aten.add.int %int0_18986, %21454 : !torch.int, !torch.int -> !torch.int
%int0_18987 = torch.constant.int 0
%int0_18988 = torch.constant.int 0
%int1_18989 = torch.constant.int 1
%21456 = torch.aten.slice.Tensor %21360, %int0_18987, %int0_18988, %21455, %int1_18989 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21456, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_18990 = torch.constant.int 1
%int0_18991 = torch.constant.int 0
%int9223372036854775807_18992 = torch.constant.int 9223372036854775807
%int1_18993 = torch.constant.int 1
%21457 = torch.aten.slice.Tensor %21456, %int1_18990, %int0_18991, %int9223372036854775807_18992, %int1_18993 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21457, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_18994 = torch.constant.int 0
%21458 = torch.aten.unsqueeze %21457, %int0_18994 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21458, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_18995 = torch.constant.int 2
%21459 = torch.aten.unsqueeze %21458, %int2_18995 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21459, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_18996 = torch.constant.int 3
%int0_18997 = torch.constant.int 0
%int9223372036854775807_18998 = torch.constant.int 9223372036854775807
%int1_18999 = torch.constant.int 1
%21460 = torch.aten.slice.Tensor %21459, %int3_18996, %int0_18997, %int9223372036854775807_18998, %int1_18999 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21460, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21461 = torch_c.to_builtin_tensor %21291 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_19000 = arith.constant 1 : index
%dim_19001 = tensor.dim %21461, %c1_19000 : tensor<4x?x4x128xf16>
%21462 = flow.tensor.bitcast %21461 : tensor<4x?x4x128xf16>{%dim_19001} -> tensor<4x?x4x64xcomplex<f16>>{%dim_19001}
%21463 = torch_c.from_builtin_tensor %21462 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21464 = torch.aten.mul.Tensor %21463, %21460 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21465 = torch_c.to_builtin_tensor %21464 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_19002 = arith.constant 1 : index
%dim_19003 = tensor.dim %21465, %c1_19002 : tensor<4x?x4x64xcomplex<f32>>
%21466 = flow.tensor.bitcast %21465 : tensor<4x?x4x64xcomplex<f32>>{%dim_19003} -> tensor<4x?x4x128xf32>{%dim_19003}
%21467 = torch_c.from_builtin_tensor %21466 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_19004 = torch.constant.int 5
%21468 = torch.prims.convert_element_type %21467, %int5_19004 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19005 = torch.constant.int 1
%21469 = torch.aten.size.int %21149, %int1_19005 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_19006 = torch.constant.int 0
%21470 = torch.aten.add.int %int0_19006, %21469 : !torch.int, !torch.int -> !torch.int
%int0_19007 = torch.constant.int 0
%int0_19008 = torch.constant.int 0
%int1_19009 = torch.constant.int 1
%21471 = torch.aten.slice.Tensor %21363, %int0_19007, %int0_19008, %21470, %int1_19009 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21471, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19010 = torch.constant.int 1
%int0_19011 = torch.constant.int 0
%int9223372036854775807_19012 = torch.constant.int 9223372036854775807
%int1_19013 = torch.constant.int 1
%21472 = torch.aten.slice.Tensor %21471, %int1_19010, %int0_19011, %int9223372036854775807_19012, %int1_19013 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21472, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19014 = torch.constant.int 0
%21473 = torch.aten.unsqueeze %21472, %int0_19014 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21473, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19015 = torch.constant.int 2
%21474 = torch.aten.unsqueeze %21473, %int2_19015 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21474, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19016 = torch.constant.int 3
%int0_19017 = torch.constant.int 0
%int9223372036854775807_19018 = torch.constant.int 9223372036854775807
%int1_19019 = torch.constant.int 1
%21475 = torch.aten.slice.Tensor %21474, %int3_19016, %int0_19017, %int9223372036854775807_19018, %int1_19019 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21475, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21476 = torch_c.to_builtin_tensor %21293 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_19020 = arith.constant 1 : index
%dim_19021 = tensor.dim %21476, %c1_19020 : tensor<4x?x4x128xf16>
%21477 = flow.tensor.bitcast %21476 : tensor<4x?x4x128xf16>{%dim_19021} -> tensor<4x?x4x64xcomplex<f16>>{%dim_19021}
%21478 = torch_c.from_builtin_tensor %21477 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %21478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%21479 = torch.aten.mul.Tensor %21478, %21475 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %21479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%21480 = torch_c.to_builtin_tensor %21479 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_19022 = arith.constant 1 : index
%dim_19023 = tensor.dim %21480, %c1_19022 : tensor<4x?x4x64xcomplex<f32>>
%21481 = flow.tensor.bitcast %21480 : tensor<4x?x4x64xcomplex<f32>>{%dim_19023} -> tensor<4x?x4x128xf32>{%dim_19023}
%21482 = torch_c.from_builtin_tensor %21481 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %21482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_19024 = torch.constant.int 5
%21483 = torch.prims.convert_element_type %21482, %int5_19024 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_19025 = torch.constant.int 131072
%none_19026 = torch.constant.none
%none_19027 = torch.constant.none
%cpu_19028 = torch.constant.device "cpu"
%false_19029 = torch.constant.bool false
%21484 = torch.aten.arange %int131072_19025, %none_19026, %none_19027, %cpu_19028, %false_19029 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_19030 = torch.constant.int 0
%int128_19031 = torch.constant.int 128
%int2_19032 = torch.constant.int 2
%none_19033 = torch.constant.none
%none_19034 = torch.constant.none
%cpu_19035 = torch.constant.device "cpu"
%false_19036 = torch.constant.bool false
%21485 = torch.aten.arange.start_step %int0_19030, %int128_19031, %int2_19032, %none_19033, %none_19034, %cpu_19035, %false_19036 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_19037 = torch.constant.int 0
%int0_19038 = torch.constant.int 0
%int64_19039 = torch.constant.int 64
%int1_19040 = torch.constant.int 1
%21486 = torch.aten.slice.Tensor %21485, %int0_19037, %int0_19038, %int64_19039, %int1_19040 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_19041 = torch.constant.int 6
%21487 = torch.prims.convert_element_type %21486, %int6_19041 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_19042 = torch.constant.int 128
%21488 = torch.aten.div.Scalar %21487, %int128_19042 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_19043 = torch.constant.float 5.000000e+05
%21489 = torch.aten.pow.Scalar %float5.000000e05_19043, %21488 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%21490 = torch.aten.reciprocal %21489 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_19044 = torch.constant.float 1.000000e+00
%21491 = torch.aten.mul.Scalar %21490, %float1.000000e00_19044 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_19045 = torch.constant.int 131072
%int1_19046 = torch.constant.int 1
%21492 = torch.prim.ListConstruct %int131072_19045, %int1_19046 : (!torch.int, !torch.int) -> !torch.list<int>
%21493 = torch.aten.view %21484, %21492 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%21494 = torch.aten.mul.Tensor %21493, %21491 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%21495 = torch.aten.cos %21494 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%21496 = torch.aten.sin %21494 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%21497 = torch.aten.complex %21495, %21496 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%21498 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21499 = flow.tensor.transfer %21498 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%21500 = torch_c.from_builtin_tensor %21499 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21501 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21502 = flow.tensor.transfer %21501 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%21503 = torch_c.from_builtin_tensor %21502 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21504 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21505 = flow.tensor.transfer %21504 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%21506 = torch_c.from_builtin_tensor %21505 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21507 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21508 = flow.tensor.transfer %21507 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%21509 = torch_c.from_builtin_tensor %21508 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21510 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21511 = flow.tensor.transfer %21510 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%21512 = torch_c.from_builtin_tensor %21511 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21513 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21514 = flow.tensor.transfer %21513 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%21515 = torch_c.from_builtin_tensor %21514 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21516 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21517 = flow.tensor.transfer %21516 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%21518 = torch_c.from_builtin_tensor %21517 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%21519 = torch_c.to_builtin_tensor %21497 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%21520 = flow.tensor.transfer %21519 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%21521 = torch_c.from_builtin_tensor %21520 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_19047 = torch.constant.int 1
%21522 = torch.aten.size.int %21171, %int1_19047 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19048 = torch.constant.int 0
%21523 = torch.aten.add.int %int0_19048, %21522 : !torch.int, !torch.int -> !torch.int
%int0_19049 = torch.constant.int 0
%int0_19050 = torch.constant.int 0
%int1_19051 = torch.constant.int 1
%21524 = torch.aten.slice.Tensor %21500, %int0_19049, %int0_19050, %21523, %int1_19051 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21524, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19052 = torch.constant.int 1
%int0_19053 = torch.constant.int 0
%int9223372036854775807_19054 = torch.constant.int 9223372036854775807
%int1_19055 = torch.constant.int 1
%21525 = torch.aten.slice.Tensor %21524, %int1_19052, %int0_19053, %int9223372036854775807_19054, %int1_19055 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21525, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19056 = torch.constant.int 0
%21526 = torch.aten.unsqueeze %21525, %int0_19056 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21526, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19057 = torch.constant.int 2
%21527 = torch.aten.unsqueeze %21526, %int2_19057 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21527, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19058 = torch.constant.int 3
%int0_19059 = torch.constant.int 0
%int9223372036854775807_19060 = torch.constant.int 9223372036854775807
%int1_19061 = torch.constant.int 1
%21528 = torch.aten.slice.Tensor %21527, %int3_19058, %int0_19059, %int9223372036854775807_19060, %int1_19061 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21528, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21529 = torch_c.to_builtin_tensor %21295 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19062 = arith.constant 1 : index
%dim_19063 = tensor.dim %21529, %c1_19062 : tensor<4x?x1x128xf16>
%21530 = flow.tensor.bitcast %21529 : tensor<4x?x1x128xf16>{%dim_19063} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19063}
%21531 = torch_c.from_builtin_tensor %21530 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21532 = torch.aten.mul.Tensor %21531, %21528 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21533 = torch_c.to_builtin_tensor %21532 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19064 = arith.constant 1 : index
%dim_19065 = tensor.dim %21533, %c1_19064 : tensor<4x?x1x64xcomplex<f32>>
%21534 = flow.tensor.bitcast %21533 : tensor<4x?x1x64xcomplex<f32>>{%dim_19065} -> tensor<4x?x1x128xf32>{%dim_19065}
%21535 = torch_c.from_builtin_tensor %21534 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19066 = torch.constant.int 5
%21536 = torch.prims.convert_element_type %21535, %int5_19066 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19067 = torch.constant.int 1
%21537 = torch.aten.size.int %21177, %int1_19067 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19068 = torch.constant.int 0
%21538 = torch.aten.add.int %int0_19068, %21537 : !torch.int, !torch.int -> !torch.int
%int0_19069 = torch.constant.int 0
%int0_19070 = torch.constant.int 0
%int1_19071 = torch.constant.int 1
%21539 = torch.aten.slice.Tensor %21503, %int0_19069, %int0_19070, %21538, %int1_19071 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21539, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19072 = torch.constant.int 1
%int0_19073 = torch.constant.int 0
%int9223372036854775807_19074 = torch.constant.int 9223372036854775807
%int1_19075 = torch.constant.int 1
%21540 = torch.aten.slice.Tensor %21539, %int1_19072, %int0_19073, %int9223372036854775807_19074, %int1_19075 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21540, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19076 = torch.constant.int 0
%21541 = torch.aten.unsqueeze %21540, %int0_19076 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21541, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19077 = torch.constant.int 2
%21542 = torch.aten.unsqueeze %21541, %int2_19077 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21542, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19078 = torch.constant.int 3
%int0_19079 = torch.constant.int 0
%int9223372036854775807_19080 = torch.constant.int 9223372036854775807
%int1_19081 = torch.constant.int 1
%21543 = torch.aten.slice.Tensor %21542, %int3_19078, %int0_19079, %int9223372036854775807_19080, %int1_19081 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21543, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21544 = torch_c.to_builtin_tensor %21297 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19082 = arith.constant 1 : index
%dim_19083 = tensor.dim %21544, %c1_19082 : tensor<4x?x1x128xf16>
%21545 = flow.tensor.bitcast %21544 : tensor<4x?x1x128xf16>{%dim_19083} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19083}
%21546 = torch_c.from_builtin_tensor %21545 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21547 = torch.aten.mul.Tensor %21546, %21543 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21548 = torch_c.to_builtin_tensor %21547 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19084 = arith.constant 1 : index
%dim_19085 = tensor.dim %21548, %c1_19084 : tensor<4x?x1x64xcomplex<f32>>
%21549 = flow.tensor.bitcast %21548 : tensor<4x?x1x64xcomplex<f32>>{%dim_19085} -> tensor<4x?x1x128xf32>{%dim_19085}
%21550 = torch_c.from_builtin_tensor %21549 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19086 = torch.constant.int 5
%21551 = torch.prims.convert_element_type %21550, %int5_19086 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19087 = torch.constant.int 1
%21552 = torch.aten.size.int %21183, %int1_19087 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19088 = torch.constant.int 0
%21553 = torch.aten.add.int %int0_19088, %21552 : !torch.int, !torch.int -> !torch.int
%int0_19089 = torch.constant.int 0
%int0_19090 = torch.constant.int 0
%int1_19091 = torch.constant.int 1
%21554 = torch.aten.slice.Tensor %21506, %int0_19089, %int0_19090, %21553, %int1_19091 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21554, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19092 = torch.constant.int 1
%int0_19093 = torch.constant.int 0
%int9223372036854775807_19094 = torch.constant.int 9223372036854775807
%int1_19095 = torch.constant.int 1
%21555 = torch.aten.slice.Tensor %21554, %int1_19092, %int0_19093, %int9223372036854775807_19094, %int1_19095 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21555, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19096 = torch.constant.int 0
%21556 = torch.aten.unsqueeze %21555, %int0_19096 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21556, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19097 = torch.constant.int 2
%21557 = torch.aten.unsqueeze %21556, %int2_19097 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21557, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19098 = torch.constant.int 3
%int0_19099 = torch.constant.int 0
%int9223372036854775807_19100 = torch.constant.int 9223372036854775807
%int1_19101 = torch.constant.int 1
%21558 = torch.aten.slice.Tensor %21557, %int3_19098, %int0_19099, %int9223372036854775807_19100, %int1_19101 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21558, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21559 = torch_c.to_builtin_tensor %21299 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19102 = arith.constant 1 : index
%dim_19103 = tensor.dim %21559, %c1_19102 : tensor<4x?x1x128xf16>
%21560 = flow.tensor.bitcast %21559 : tensor<4x?x1x128xf16>{%dim_19103} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19103}
%21561 = torch_c.from_builtin_tensor %21560 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21562 = torch.aten.mul.Tensor %21561, %21558 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21563 = torch_c.to_builtin_tensor %21562 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19104 = arith.constant 1 : index
%dim_19105 = tensor.dim %21563, %c1_19104 : tensor<4x?x1x64xcomplex<f32>>
%21564 = flow.tensor.bitcast %21563 : tensor<4x?x1x64xcomplex<f32>>{%dim_19105} -> tensor<4x?x1x128xf32>{%dim_19105}
%21565 = torch_c.from_builtin_tensor %21564 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19106 = torch.constant.int 5
%21566 = torch.prims.convert_element_type %21565, %int5_19106 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19107 = torch.constant.int 1
%21567 = torch.aten.size.int %21189, %int1_19107 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19108 = torch.constant.int 0
%21568 = torch.aten.add.int %int0_19108, %21567 : !torch.int, !torch.int -> !torch.int
%int0_19109 = torch.constant.int 0
%int0_19110 = torch.constant.int 0
%int1_19111 = torch.constant.int 1
%21569 = torch.aten.slice.Tensor %21509, %int0_19109, %int0_19110, %21568, %int1_19111 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21569, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19112 = torch.constant.int 1
%int0_19113 = torch.constant.int 0
%int9223372036854775807_19114 = torch.constant.int 9223372036854775807
%int1_19115 = torch.constant.int 1
%21570 = torch.aten.slice.Tensor %21569, %int1_19112, %int0_19113, %int9223372036854775807_19114, %int1_19115 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21570, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19116 = torch.constant.int 0
%21571 = torch.aten.unsqueeze %21570, %int0_19116 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21571, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19117 = torch.constant.int 2
%21572 = torch.aten.unsqueeze %21571, %int2_19117 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21572, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19118 = torch.constant.int 3
%int0_19119 = torch.constant.int 0
%int9223372036854775807_19120 = torch.constant.int 9223372036854775807
%int1_19121 = torch.constant.int 1
%21573 = torch.aten.slice.Tensor %21572, %int3_19118, %int0_19119, %int9223372036854775807_19120, %int1_19121 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21573, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21574 = torch_c.to_builtin_tensor %21301 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19122 = arith.constant 1 : index
%dim_19123 = tensor.dim %21574, %c1_19122 : tensor<4x?x1x128xf16>
%21575 = flow.tensor.bitcast %21574 : tensor<4x?x1x128xf16>{%dim_19123} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19123}
%21576 = torch_c.from_builtin_tensor %21575 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21577 = torch.aten.mul.Tensor %21576, %21573 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21578 = torch_c.to_builtin_tensor %21577 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19124 = arith.constant 1 : index
%dim_19125 = tensor.dim %21578, %c1_19124 : tensor<4x?x1x64xcomplex<f32>>
%21579 = flow.tensor.bitcast %21578 : tensor<4x?x1x64xcomplex<f32>>{%dim_19125} -> tensor<4x?x1x128xf32>{%dim_19125}
%21580 = torch_c.from_builtin_tensor %21579 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19126 = torch.constant.int 5
%21581 = torch.prims.convert_element_type %21580, %int5_19126 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19127 = torch.constant.int 1
%21582 = torch.aten.size.int %21195, %int1_19127 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19128 = torch.constant.int 0
%21583 = torch.aten.add.int %int0_19128, %21582 : !torch.int, !torch.int -> !torch.int
%int0_19129 = torch.constant.int 0
%int0_19130 = torch.constant.int 0
%int1_19131 = torch.constant.int 1
%21584 = torch.aten.slice.Tensor %21512, %int0_19129, %int0_19130, %21583, %int1_19131 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21584, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19132 = torch.constant.int 1
%int0_19133 = torch.constant.int 0
%int9223372036854775807_19134 = torch.constant.int 9223372036854775807
%int1_19135 = torch.constant.int 1
%21585 = torch.aten.slice.Tensor %21584, %int1_19132, %int0_19133, %int9223372036854775807_19134, %int1_19135 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21585, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19136 = torch.constant.int 0
%21586 = torch.aten.unsqueeze %21585, %int0_19136 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21586, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19137 = torch.constant.int 2
%21587 = torch.aten.unsqueeze %21586, %int2_19137 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21587, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19138 = torch.constant.int 3
%int0_19139 = torch.constant.int 0
%int9223372036854775807_19140 = torch.constant.int 9223372036854775807
%int1_19141 = torch.constant.int 1
%21588 = torch.aten.slice.Tensor %21587, %int3_19138, %int0_19139, %int9223372036854775807_19140, %int1_19141 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21588, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21589 = torch_c.to_builtin_tensor %21303 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19142 = arith.constant 1 : index
%dim_19143 = tensor.dim %21589, %c1_19142 : tensor<4x?x1x128xf16>
%21590 = flow.tensor.bitcast %21589 : tensor<4x?x1x128xf16>{%dim_19143} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19143}
%21591 = torch_c.from_builtin_tensor %21590 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21592 = torch.aten.mul.Tensor %21591, %21588 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21593 = torch_c.to_builtin_tensor %21592 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19144 = arith.constant 1 : index
%dim_19145 = tensor.dim %21593, %c1_19144 : tensor<4x?x1x64xcomplex<f32>>
%21594 = flow.tensor.bitcast %21593 : tensor<4x?x1x64xcomplex<f32>>{%dim_19145} -> tensor<4x?x1x128xf32>{%dim_19145}
%21595 = torch_c.from_builtin_tensor %21594 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19146 = torch.constant.int 5
%21596 = torch.prims.convert_element_type %21595, %int5_19146 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19147 = torch.constant.int 1
%21597 = torch.aten.size.int %21201, %int1_19147 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19148 = torch.constant.int 0
%21598 = torch.aten.add.int %int0_19148, %21597 : !torch.int, !torch.int -> !torch.int
%int0_19149 = torch.constant.int 0
%int0_19150 = torch.constant.int 0
%int1_19151 = torch.constant.int 1
%21599 = torch.aten.slice.Tensor %21515, %int0_19149, %int0_19150, %21598, %int1_19151 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21599, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19152 = torch.constant.int 1
%int0_19153 = torch.constant.int 0
%int9223372036854775807_19154 = torch.constant.int 9223372036854775807
%int1_19155 = torch.constant.int 1
%21600 = torch.aten.slice.Tensor %21599, %int1_19152, %int0_19153, %int9223372036854775807_19154, %int1_19155 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21600, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19156 = torch.constant.int 0
%21601 = torch.aten.unsqueeze %21600, %int0_19156 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21601, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19157 = torch.constant.int 2
%21602 = torch.aten.unsqueeze %21601, %int2_19157 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21602, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19158 = torch.constant.int 3
%int0_19159 = torch.constant.int 0
%int9223372036854775807_19160 = torch.constant.int 9223372036854775807
%int1_19161 = torch.constant.int 1
%21603 = torch.aten.slice.Tensor %21602, %int3_19158, %int0_19159, %int9223372036854775807_19160, %int1_19161 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21603, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21604 = torch_c.to_builtin_tensor %21305 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19162 = arith.constant 1 : index
%dim_19163 = tensor.dim %21604, %c1_19162 : tensor<4x?x1x128xf16>
%21605 = flow.tensor.bitcast %21604 : tensor<4x?x1x128xf16>{%dim_19163} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19163}
%21606 = torch_c.from_builtin_tensor %21605 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21607 = torch.aten.mul.Tensor %21606, %21603 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21608 = torch_c.to_builtin_tensor %21607 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19164 = arith.constant 1 : index
%dim_19165 = tensor.dim %21608, %c1_19164 : tensor<4x?x1x64xcomplex<f32>>
%21609 = flow.tensor.bitcast %21608 : tensor<4x?x1x64xcomplex<f32>>{%dim_19165} -> tensor<4x?x1x128xf32>{%dim_19165}
%21610 = torch_c.from_builtin_tensor %21609 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19166 = torch.constant.int 5
%21611 = torch.prims.convert_element_type %21610, %int5_19166 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19167 = torch.constant.int 1
%21612 = torch.aten.size.int %21207, %int1_19167 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19168 = torch.constant.int 0
%21613 = torch.aten.add.int %int0_19168, %21612 : !torch.int, !torch.int -> !torch.int
%int0_19169 = torch.constant.int 0
%int0_19170 = torch.constant.int 0
%int1_19171 = torch.constant.int 1
%21614 = torch.aten.slice.Tensor %21518, %int0_19169, %int0_19170, %21613, %int1_19171 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21614, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19172 = torch.constant.int 1
%int0_19173 = torch.constant.int 0
%int9223372036854775807_19174 = torch.constant.int 9223372036854775807
%int1_19175 = torch.constant.int 1
%21615 = torch.aten.slice.Tensor %21614, %int1_19172, %int0_19173, %int9223372036854775807_19174, %int1_19175 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21615, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19176 = torch.constant.int 0
%21616 = torch.aten.unsqueeze %21615, %int0_19176 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21616, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19177 = torch.constant.int 2
%21617 = torch.aten.unsqueeze %21616, %int2_19177 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21617, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19178 = torch.constant.int 3
%int0_19179 = torch.constant.int 0
%int9223372036854775807_19180 = torch.constant.int 9223372036854775807
%int1_19181 = torch.constant.int 1
%21618 = torch.aten.slice.Tensor %21617, %int3_19178, %int0_19179, %int9223372036854775807_19180, %int1_19181 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21618, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21619 = torch_c.to_builtin_tensor %21307 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19182 = arith.constant 1 : index
%dim_19183 = tensor.dim %21619, %c1_19182 : tensor<4x?x1x128xf16>
%21620 = flow.tensor.bitcast %21619 : tensor<4x?x1x128xf16>{%dim_19183} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19183}
%21621 = torch_c.from_builtin_tensor %21620 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21622 = torch.aten.mul.Tensor %21621, %21618 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21623 = torch_c.to_builtin_tensor %21622 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19184 = arith.constant 1 : index
%dim_19185 = tensor.dim %21623, %c1_19184 : tensor<4x?x1x64xcomplex<f32>>
%21624 = flow.tensor.bitcast %21623 : tensor<4x?x1x64xcomplex<f32>>{%dim_19185} -> tensor<4x?x1x128xf32>{%dim_19185}
%21625 = torch_c.from_builtin_tensor %21624 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19186 = torch.constant.int 5
%21626 = torch.prims.convert_element_type %21625, %int5_19186 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_19187 = torch.constant.int 1
%21627 = torch.aten.size.int %21213, %int1_19187 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_19188 = torch.constant.int 0
%21628 = torch.aten.add.int %int0_19188, %21627 : !torch.int, !torch.int -> !torch.int
%int0_19189 = torch.constant.int 0
%int0_19190 = torch.constant.int 0
%int1_19191 = torch.constant.int 1
%21629 = torch.aten.slice.Tensor %21521, %int0_19189, %int0_19190, %21628, %int1_19191 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21629, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_19192 = torch.constant.int 1
%int0_19193 = torch.constant.int 0
%int9223372036854775807_19194 = torch.constant.int 9223372036854775807
%int1_19195 = torch.constant.int 1
%21630 = torch.aten.slice.Tensor %21629, %int1_19192, %int0_19193, %int9223372036854775807_19194, %int1_19195 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %21630, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_19196 = torch.constant.int 0
%21631 = torch.aten.unsqueeze %21630, %int0_19196 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %21631, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_19197 = torch.constant.int 2
%21632 = torch.aten.unsqueeze %21631, %int2_19197 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21632, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_19198 = torch.constant.int 3
%int0_19199 = torch.constant.int 0
%int9223372036854775807_19200 = torch.constant.int 9223372036854775807
%int1_19201 = torch.constant.int 1
%21633 = torch.aten.slice.Tensor %21632, %int3_19198, %int0_19199, %int9223372036854775807_19200, %int1_19201 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21633, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%21634 = torch_c.to_builtin_tensor %21309 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_19202 = arith.constant 1 : index
%dim_19203 = tensor.dim %21634, %c1_19202 : tensor<4x?x1x128xf16>
%21635 = flow.tensor.bitcast %21634 : tensor<4x?x1x128xf16>{%dim_19203} -> tensor<4x?x1x64xcomplex<f16>>{%dim_19203}
%21636 = torch_c.from_builtin_tensor %21635 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %21636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%21637 = torch.aten.mul.Tensor %21636, %21633 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %21637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%21638 = torch_c.to_builtin_tensor %21637 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_19204 = arith.constant 1 : index
%dim_19205 = tensor.dim %21638, %c1_19204 : tensor<4x?x1x64xcomplex<f32>>
%21639 = flow.tensor.bitcast %21638 : tensor<4x?x1x64xcomplex<f32>>{%dim_19205} -> tensor<4x?x1x128xf32>{%dim_19205}
%21640 = torch_c.from_builtin_tensor %21639 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %21640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_19206 = torch.constant.int 5
%21641 = torch.prims.convert_element_type %21640, %int5_19206 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %21641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_19207 = torch.constant.int 64
%21642 = torch.aten.mul.Scalar %2364, %int64_19207 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21642, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19208 = torch.constant.int 64
%21643 = torch.aten.mul.Scalar %2367, %int64_19208 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21643, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19209 = torch.constant.int 64
%21644 = torch.aten.mul.Scalar %2370, %int64_19209 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21644, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19210 = torch.constant.int 64
%21645 = torch.aten.mul.Scalar %2373, %int64_19210 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21645, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19211 = torch.constant.int 64
%21646 = torch.aten.mul.Scalar %2376, %int64_19211 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21646, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19212 = torch.constant.int 64
%21647 = torch.aten.mul.Scalar %2379, %int64_19212 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21647, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19213 = torch.constant.int 64
%21648 = torch.aten.mul.Scalar %2382, %int64_19213 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21648, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_19214 = torch.constant.int 64
%21649 = torch.aten.mul.Scalar %2385, %int64_19214 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21649, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20 = torch.constant.int 20
%int1_19215 = torch.constant.int 1
%21650 = torch.aten.add.Scalar %21642, %int20, %int1_19215 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21650, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19216 = torch.constant.int 20
%int1_19217 = torch.constant.int 1
%21651 = torch.aten.add.Scalar %21643, %int20_19216, %int1_19217 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21651, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19218 = torch.constant.int 20
%int1_19219 = torch.constant.int 1
%21652 = torch.aten.add.Scalar %21644, %int20_19218, %int1_19219 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21652, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19220 = torch.constant.int 20
%int1_19221 = torch.constant.int 1
%21653 = torch.aten.add.Scalar %21645, %int20_19220, %int1_19221 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21653, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19222 = torch.constant.int 20
%int1_19223 = torch.constant.int 1
%21654 = torch.aten.add.Scalar %21646, %int20_19222, %int1_19223 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21654, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19224 = torch.constant.int 20
%int1_19225 = torch.constant.int 1
%21655 = torch.aten.add.Scalar %21647, %int20_19224, %int1_19225 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21655, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19226 = torch.constant.int 20
%int1_19227 = torch.constant.int 1
%21656 = torch.aten.add.Scalar %21648, %int20_19226, %int1_19227 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21656, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int20_19228 = torch.constant.int 20
%int1_19229 = torch.constant.int 1
%21657 = torch.aten.add.Scalar %21649, %int20_19228, %int1_19229 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21657, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_19230 = torch.constant.int 4
%int16_19231 = torch.constant.int 16
%int1_19232 = torch.constant.int 1
%int128_19233 = torch.constant.int 128
%21658 = torch.prim.ListConstruct %int4_19230, %3095, %int16_19231, %int1_19232, %int128_19233 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21659 = torch.aten.view %21536, %21658 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21659, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19234 = torch.constant.int 4
%int16_19235 = torch.constant.int 16
%int1_19236 = torch.constant.int 1
%int128_19237 = torch.constant.int 128
%21660 = torch.prim.ListConstruct %int4_19234, %3095, %int16_19235, %int1_19236, %int128_19237 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21661 = torch.aten.view %21551, %21660 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21661, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19238 = torch.constant.int 4
%int16_19239 = torch.constant.int 16
%int1_19240 = torch.constant.int 1
%int128_19241 = torch.constant.int 128
%21662 = torch.prim.ListConstruct %int4_19238, %3095, %int16_19239, %int1_19240, %int128_19241 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21663 = torch.aten.view %21566, %21662 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21663, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19242 = torch.constant.int 4
%int16_19243 = torch.constant.int 16
%int1_19244 = torch.constant.int 1
%int128_19245 = torch.constant.int 128
%21664 = torch.prim.ListConstruct %int4_19242, %3095, %int16_19243, %int1_19244, %int128_19245 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21665 = torch.aten.view %21581, %21664 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21665, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19246 = torch.constant.int 4
%int16_19247 = torch.constant.int 16
%int1_19248 = torch.constant.int 1
%int128_19249 = torch.constant.int 128
%21666 = torch.prim.ListConstruct %int4_19246, %3095, %int16_19247, %int1_19248, %int128_19249 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21667 = torch.aten.view %21596, %21666 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21667, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19250 = torch.constant.int 4
%int16_19251 = torch.constant.int 16
%int1_19252 = torch.constant.int 1
%int128_19253 = torch.constant.int 128
%21668 = torch.prim.ListConstruct %int4_19250, %3095, %int16_19251, %int1_19252, %int128_19253 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21669 = torch.aten.view %21611, %21668 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21669, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19254 = torch.constant.int 4
%int16_19255 = torch.constant.int 16
%int1_19256 = torch.constant.int 1
%int128_19257 = torch.constant.int 128
%21670 = torch.prim.ListConstruct %int4_19254, %3095, %int16_19255, %int1_19256, %int128_19257 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21671 = torch.aten.view %21626, %21670 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21671, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19258 = torch.constant.int 4
%int16_19259 = torch.constant.int 16
%int1_19260 = torch.constant.int 1
%int128_19261 = torch.constant.int 128
%21672 = torch.prim.ListConstruct %int4_19258, %3095, %int16_19259, %int1_19260, %int128_19261 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21673 = torch.aten.view %21641, %21672 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21673, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19262 = torch.constant.int 4
%21674 = torch.aten.mul.int %int4_19262, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19263 = torch.constant.int 16
%int1_19264 = torch.constant.int 1
%int128_19265 = torch.constant.int 128
%21675 = torch.prim.ListConstruct %21674, %int16_19263, %int1_19264, %int128_19265 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21676 = torch.aten.view %21659, %21675 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21676, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19266 = torch.constant.int 4
%21677 = torch.aten.mul.int %int4_19266, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19267 = torch.constant.int 16
%int1_19268 = torch.constant.int 1
%int128_19269 = torch.constant.int 128
%21678 = torch.prim.ListConstruct %21677, %int16_19267, %int1_19268, %int128_19269 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21679 = torch.aten.view %21661, %21678 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21679, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19270 = torch.constant.int 4
%21680 = torch.aten.mul.int %int4_19270, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19271 = torch.constant.int 16
%int1_19272 = torch.constant.int 1
%int128_19273 = torch.constant.int 128
%21681 = torch.prim.ListConstruct %21680, %int16_19271, %int1_19272, %int128_19273 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21682 = torch.aten.view %21663, %21681 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21682, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19274 = torch.constant.int 4
%21683 = torch.aten.mul.int %int4_19274, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19275 = torch.constant.int 16
%int1_19276 = torch.constant.int 1
%int128_19277 = torch.constant.int 128
%21684 = torch.prim.ListConstruct %21683, %int16_19275, %int1_19276, %int128_19277 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21685 = torch.aten.view %21665, %21684 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21685, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19278 = torch.constant.int 4
%21686 = torch.aten.mul.int %int4_19278, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19279 = torch.constant.int 16
%int1_19280 = torch.constant.int 1
%int128_19281 = torch.constant.int 128
%21687 = torch.prim.ListConstruct %21686, %int16_19279, %int1_19280, %int128_19281 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21688 = torch.aten.view %21667, %21687 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21688, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19282 = torch.constant.int 4
%21689 = torch.aten.mul.int %int4_19282, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19283 = torch.constant.int 16
%int1_19284 = torch.constant.int 1
%int128_19285 = torch.constant.int 128
%21690 = torch.prim.ListConstruct %21689, %int16_19283, %int1_19284, %int128_19285 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21691 = torch.aten.view %21669, %21690 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21691, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19286 = torch.constant.int 4
%21692 = torch.aten.mul.int %int4_19286, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19287 = torch.constant.int 16
%int1_19288 = torch.constant.int 1
%int128_19289 = torch.constant.int 128
%21693 = torch.prim.ListConstruct %21692, %int16_19287, %int1_19288, %int128_19289 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21694 = torch.aten.view %21671, %21693 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21694, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19290 = torch.constant.int 4
%21695 = torch.aten.mul.int %int4_19290, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19291 = torch.constant.int 16
%int1_19292 = torch.constant.int 1
%int128_19293 = torch.constant.int 128
%21696 = torch.prim.ListConstruct %21695, %int16_19291, %int1_19292, %int128_19293 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21697 = torch.aten.view %21673, %21696 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21697, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19294 = torch.constant.int 4
%21698 = torch.aten.mul.int %int4_19294, %3095 : !torch.int, !torch.int -> !torch.int
%21699 = torch.prim.ListConstruct %21698 : (!torch.int) -> !torch.list<int>
%21700 = torch.aten.view %21650, %21699 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21700, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19295 = torch.constant.int 4
%21701 = torch.aten.mul.int %int4_19295, %3095 : !torch.int, !torch.int -> !torch.int
%21702 = torch.prim.ListConstruct %21701 : (!torch.int) -> !torch.list<int>
%21703 = torch.aten.view %21651, %21702 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21703, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19296 = torch.constant.int 4
%21704 = torch.aten.mul.int %int4_19296, %3095 : !torch.int, !torch.int -> !torch.int
%21705 = torch.prim.ListConstruct %21704 : (!torch.int) -> !torch.list<int>
%21706 = torch.aten.view %21652, %21705 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21706, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19297 = torch.constant.int 4
%21707 = torch.aten.mul.int %int4_19297, %3095 : !torch.int, !torch.int -> !torch.int
%21708 = torch.prim.ListConstruct %21707 : (!torch.int) -> !torch.list<int>
%21709 = torch.aten.view %21653, %21708 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21709, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19298 = torch.constant.int 4
%21710 = torch.aten.mul.int %int4_19298, %3095 : !torch.int, !torch.int -> !torch.int
%21711 = torch.prim.ListConstruct %21710 : (!torch.int) -> !torch.list<int>
%21712 = torch.aten.view %21654, %21711 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21712, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19299 = torch.constant.int 4
%21713 = torch.aten.mul.int %int4_19299, %3095 : !torch.int, !torch.int -> !torch.int
%21714 = torch.prim.ListConstruct %21713 : (!torch.int) -> !torch.list<int>
%21715 = torch.aten.view %21655, %21714 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21715, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19300 = torch.constant.int 4
%21716 = torch.aten.mul.int %int4_19300, %3095 : !torch.int, !torch.int -> !torch.int
%21717 = torch.prim.ListConstruct %21716 : (!torch.int) -> !torch.list<int>
%21718 = torch.aten.view %21656, %21717 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21718, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19301 = torch.constant.int 4
%21719 = torch.aten.mul.int %int4_19301, %3095 : !torch.int, !torch.int -> !torch.int
%21720 = torch.prim.ListConstruct %21719 : (!torch.int) -> !torch.list<int>
%21721 = torch.aten.view %21657, %21720 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21721, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19302 = torch.constant.int 4
%int16_19303 = torch.constant.int 16
%int1_19304 = torch.constant.int 1
%int128_19305 = torch.constant.int 128
%21722 = torch.prim.ListConstruct %int4_19302, %3095, %int16_19303, %int1_19304, %int128_19305 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21723 = torch.aten.view %21311, %21722 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21723, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19306 = torch.constant.int 4
%int16_19307 = torch.constant.int 16
%int1_19308 = torch.constant.int 1
%int128_19309 = torch.constant.int 128
%21724 = torch.prim.ListConstruct %int4_19306, %3095, %int16_19307, %int1_19308, %int128_19309 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21725 = torch.aten.view %21313, %21724 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21725, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19310 = torch.constant.int 4
%int16_19311 = torch.constant.int 16
%int1_19312 = torch.constant.int 1
%int128_19313 = torch.constant.int 128
%21726 = torch.prim.ListConstruct %int4_19310, %3095, %int16_19311, %int1_19312, %int128_19313 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21727 = torch.aten.view %21315, %21726 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21727, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19314 = torch.constant.int 4
%int16_19315 = torch.constant.int 16
%int1_19316 = torch.constant.int 1
%int128_19317 = torch.constant.int 128
%21728 = torch.prim.ListConstruct %int4_19314, %3095, %int16_19315, %int1_19316, %int128_19317 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21729 = torch.aten.view %21317, %21728 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21729, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19318 = torch.constant.int 4
%int16_19319 = torch.constant.int 16
%int1_19320 = torch.constant.int 1
%int128_19321 = torch.constant.int 128
%21730 = torch.prim.ListConstruct %int4_19318, %3095, %int16_19319, %int1_19320, %int128_19321 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21731 = torch.aten.view %21319, %21730 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21731, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19322 = torch.constant.int 4
%int16_19323 = torch.constant.int 16
%int1_19324 = torch.constant.int 1
%int128_19325 = torch.constant.int 128
%21732 = torch.prim.ListConstruct %int4_19322, %3095, %int16_19323, %int1_19324, %int128_19325 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21733 = torch.aten.view %21321, %21732 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21733, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19326 = torch.constant.int 4
%int16_19327 = torch.constant.int 16
%int1_19328 = torch.constant.int 1
%int128_19329 = torch.constant.int 128
%21734 = torch.prim.ListConstruct %int4_19326, %3095, %int16_19327, %int1_19328, %int128_19329 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21735 = torch.aten.view %21323, %21734 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21735, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19330 = torch.constant.int 4
%int16_19331 = torch.constant.int 16
%int1_19332 = torch.constant.int 1
%int128_19333 = torch.constant.int 128
%21736 = torch.prim.ListConstruct %int4_19330, %3095, %int16_19331, %int1_19332, %int128_19333 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21737 = torch.aten.view %21325, %21736 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %21737, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_19334 = torch.constant.int 4
%21738 = torch.aten.mul.int %int4_19334, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19335 = torch.constant.int 16
%int1_19336 = torch.constant.int 1
%int128_19337 = torch.constant.int 128
%21739 = torch.prim.ListConstruct %21738, %int16_19335, %int1_19336, %int128_19337 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21740 = torch.aten.view %21723, %21739 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21740, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19338 = torch.constant.int 4
%21741 = torch.aten.mul.int %int4_19338, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19339 = torch.constant.int 16
%int1_19340 = torch.constant.int 1
%int128_19341 = torch.constant.int 128
%21742 = torch.prim.ListConstruct %21741, %int16_19339, %int1_19340, %int128_19341 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21743 = torch.aten.view %21725, %21742 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21743, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19342 = torch.constant.int 4
%21744 = torch.aten.mul.int %int4_19342, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19343 = torch.constant.int 16
%int1_19344 = torch.constant.int 1
%int128_19345 = torch.constant.int 128
%21745 = torch.prim.ListConstruct %21744, %int16_19343, %int1_19344, %int128_19345 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21746 = torch.aten.view %21727, %21745 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21746, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19346 = torch.constant.int 4
%21747 = torch.aten.mul.int %int4_19346, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19347 = torch.constant.int 16
%int1_19348 = torch.constant.int 1
%int128_19349 = torch.constant.int 128
%21748 = torch.prim.ListConstruct %21747, %int16_19347, %int1_19348, %int128_19349 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21749 = torch.aten.view %21729, %21748 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21749, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19350 = torch.constant.int 4
%21750 = torch.aten.mul.int %int4_19350, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19351 = torch.constant.int 16
%int1_19352 = torch.constant.int 1
%int128_19353 = torch.constant.int 128
%21751 = torch.prim.ListConstruct %21750, %int16_19351, %int1_19352, %int128_19353 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21752 = torch.aten.view %21731, %21751 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21752, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19354 = torch.constant.int 4
%21753 = torch.aten.mul.int %int4_19354, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19355 = torch.constant.int 16
%int1_19356 = torch.constant.int 1
%int128_19357 = torch.constant.int 128
%21754 = torch.prim.ListConstruct %21753, %int16_19355, %int1_19356, %int128_19357 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21755 = torch.aten.view %21733, %21754 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21755, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19358 = torch.constant.int 4
%21756 = torch.aten.mul.int %int4_19358, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19359 = torch.constant.int 16
%int1_19360 = torch.constant.int 1
%int128_19361 = torch.constant.int 128
%21757 = torch.prim.ListConstruct %21756, %int16_19359, %int1_19360, %int128_19361 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21758 = torch.aten.view %21735, %21757 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21758, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_19362 = torch.constant.int 4
%21759 = torch.aten.mul.int %int4_19362, %3095 : !torch.int, !torch.int -> !torch.int
%int16_19363 = torch.constant.int 16
%int1_19364 = torch.constant.int 1
%int128_19365 = torch.constant.int 128
%21760 = torch.prim.ListConstruct %21759, %int16_19363, %int1_19364, %int128_19365 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21761 = torch.aten.view %21737, %21760 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21761, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_19366 = torch.constant.int 1
%int1_19367 = torch.constant.int 1
%21762 = torch.aten.add.Scalar %21650, %int1_19366, %int1_19367 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21762, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19368 = torch.constant.int 1
%int1_19369 = torch.constant.int 1
%21763 = torch.aten.add.Scalar %21651, %int1_19368, %int1_19369 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21763, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19370 = torch.constant.int 1
%int1_19371 = torch.constant.int 1
%21764 = torch.aten.add.Scalar %21652, %int1_19370, %int1_19371 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21764, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19372 = torch.constant.int 1
%int1_19373 = torch.constant.int 1
%21765 = torch.aten.add.Scalar %21653, %int1_19372, %int1_19373 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21765, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19374 = torch.constant.int 1
%int1_19375 = torch.constant.int 1
%21766 = torch.aten.add.Scalar %21654, %int1_19374, %int1_19375 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21766, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19376 = torch.constant.int 1
%int1_19377 = torch.constant.int 1
%21767 = torch.aten.add.Scalar %21655, %int1_19376, %int1_19377 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21767, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19378 = torch.constant.int 1
%int1_19379 = torch.constant.int 1
%21768 = torch.aten.add.Scalar %21656, %int1_19378, %int1_19379 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21768, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_19380 = torch.constant.int 1
%int1_19381 = torch.constant.int 1
%21769 = torch.aten.add.Scalar %21657, %int1_19380, %int1_19381 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %21769, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_19382 = torch.constant.int 4
%21770 = torch.aten.mul.int %int4_19382, %3095 : !torch.int, !torch.int -> !torch.int
%21771 = torch.prim.ListConstruct %21770 : (!torch.int) -> !torch.list<int>
%21772 = torch.aten.view %21762, %21771 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21772, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19383 = torch.constant.int 4
%21773 = torch.aten.mul.int %int4_19383, %3095 : !torch.int, !torch.int -> !torch.int
%21774 = torch.prim.ListConstruct %21773 : (!torch.int) -> !torch.list<int>
%21775 = torch.aten.view %21763, %21774 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21775, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19384 = torch.constant.int 4
%21776 = torch.aten.mul.int %int4_19384, %3095 : !torch.int, !torch.int -> !torch.int
%21777 = torch.prim.ListConstruct %21776 : (!torch.int) -> !torch.list<int>
%21778 = torch.aten.view %21764, %21777 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21778, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19385 = torch.constant.int 4
%21779 = torch.aten.mul.int %int4_19385, %3095 : !torch.int, !torch.int -> !torch.int
%21780 = torch.prim.ListConstruct %21779 : (!torch.int) -> !torch.list<int>
%21781 = torch.aten.view %21765, %21780 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21781, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19386 = torch.constant.int 4
%21782 = torch.aten.mul.int %int4_19386, %3095 : !torch.int, !torch.int -> !torch.int
%21783 = torch.prim.ListConstruct %21782 : (!torch.int) -> !torch.list<int>
%21784 = torch.aten.view %21766, %21783 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21784, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19387 = torch.constant.int 4
%21785 = torch.aten.mul.int %int4_19387, %3095 : !torch.int, !torch.int -> !torch.int
%21786 = torch.prim.ListConstruct %21785 : (!torch.int) -> !torch.list<int>
%21787 = torch.aten.view %21767, %21786 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21787, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19388 = torch.constant.int 4
%21788 = torch.aten.mul.int %int4_19388, %3095 : !torch.int, !torch.int -> !torch.int
%21789 = torch.prim.ListConstruct %21788 : (!torch.int) -> !torch.list<int>
%21790 = torch.aten.view %21768, %21789 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21790, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_19389 = torch.constant.int 4
%21791 = torch.aten.mul.int %int4_19389, %3095 : !torch.int, !torch.int -> !torch.int
%21792 = torch.prim.ListConstruct %21791 : (!torch.int) -> !torch.list<int>
%21793 = torch.aten.view %21769, %21792 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21793, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%21794 = torch.prim.ListConstruct %21700, %21772 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19390 = torch.constant.int 0
%21795 = torch.aten.cat %21794, %int0_19390 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21795, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21796 = torch.prim.ListConstruct %21703, %21775 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19391 = torch.constant.int 0
%21797 = torch.aten.cat %21796, %int0_19391 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21797, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21798 = torch.prim.ListConstruct %21706, %21778 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19392 = torch.constant.int 0
%21799 = torch.aten.cat %21798, %int0_19392 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21799, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21800 = torch.prim.ListConstruct %21709, %21781 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19393 = torch.constant.int 0
%21801 = torch.aten.cat %21800, %int0_19393 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21801, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21802 = torch.prim.ListConstruct %21712, %21784 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19394 = torch.constant.int 0
%21803 = torch.aten.cat %21802, %int0_19394 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21803, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21804 = torch.prim.ListConstruct %21715, %21787 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19395 = torch.constant.int 0
%21805 = torch.aten.cat %21804, %int0_19395 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21805, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21806 = torch.prim.ListConstruct %21718, %21790 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19396 = torch.constant.int 0
%21807 = torch.aten.cat %21806, %int0_19396 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21807, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21808 = torch.prim.ListConstruct %21721, %21793 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_19397 = torch.constant.int 0
%21809 = torch.aten.cat %21808, %int0_19397 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %21809, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%21810 = torch.prim.ListConstruct %21676, %21740 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19398 = torch.constant.int 0
%21811 = torch.aten.cat %21810, %int0_19398 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21811, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21812 = torch.prim.ListConstruct %21679, %21743 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19399 = torch.constant.int 0
%21813 = torch.aten.cat %21812, %int0_19399 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21813, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21814 = torch.prim.ListConstruct %21682, %21746 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19400 = torch.constant.int 0
%21815 = torch.aten.cat %21814, %int0_19400 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21815, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21816 = torch.prim.ListConstruct %21685, %21749 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19401 = torch.constant.int 0
%21817 = torch.aten.cat %21816, %int0_19401 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21817, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21818 = torch.prim.ListConstruct %21688, %21752 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19402 = torch.constant.int 0
%21819 = torch.aten.cat %21818, %int0_19402 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21819, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21820 = torch.prim.ListConstruct %21691, %21755 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19403 = torch.constant.int 0
%21821 = torch.aten.cat %21820, %int0_19403 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21821, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21822 = torch.prim.ListConstruct %21694, %21758 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19404 = torch.constant.int 0
%21823 = torch.aten.cat %21822, %int0_19404 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21823, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21824 = torch.prim.ListConstruct %21697, %21761 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_19405 = torch.constant.int 0
%21825 = torch.aten.cat %21824, %int0_19405 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21825, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19406 = torch.constant.int 32
%int2_19407 = torch.constant.int 2
%int16_19408 = torch.constant.int 16
%int1_19409 = torch.constant.int 1
%int128_19410 = torch.constant.int 128
%21826 = torch.prim.ListConstruct %3023, %int32_19406, %int2_19407, %int16_19408, %int1_19409, %int128_19410 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21827 = torch.aten.view %19976, %21826 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21827, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19411 = torch.constant.int 32
%21828 = torch.aten.mul.int %3023, %int32_19411 : !torch.int, !torch.int -> !torch.int
%int2_19412 = torch.constant.int 2
%21829 = torch.aten.mul.int %21828, %int2_19412 : !torch.int, !torch.int -> !torch.int
%int16_19413 = torch.constant.int 16
%int1_19414 = torch.constant.int 1
%int128_19415 = torch.constant.int 128
%21830 = torch.prim.ListConstruct %21829, %int16_19413, %int1_19414, %int128_19415 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21831 = torch.aten.view %21827, %21830 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21831, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21832 = torch.prim.ListConstruct %21795 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19416 = torch.constant.bool false
%21833 = torch.aten.index_put %21831, %21832, %21811, %false_19416 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21833, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19417 = torch.constant.int 32
%int2_19418 = torch.constant.int 2
%int16_19419 = torch.constant.int 16
%int1_19420 = torch.constant.int 1
%int128_19421 = torch.constant.int 128
%21834 = torch.prim.ListConstruct %3023, %int32_19417, %int2_19418, %int16_19419, %int1_19420, %int128_19421 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21835 = torch.aten.view %21833, %21834 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21835, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19422 = torch.constant.int 131072
%21836 = torch.prim.ListConstruct %3023, %int131072_19422 : (!torch.int, !torch.int) -> !torch.list<int>
%21837 = torch.aten.view %21835, %21836 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21837, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19423 = torch.constant.int 32
%int2_19424 = torch.constant.int 2
%int16_19425 = torch.constant.int 16
%int1_19426 = torch.constant.int 1
%int128_19427 = torch.constant.int 128
%21838 = torch.prim.ListConstruct %3026, %int32_19423, %int2_19424, %int16_19425, %int1_19426, %int128_19427 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21839 = torch.aten.view %19988, %21838 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21839, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19428 = torch.constant.int 32
%21840 = torch.aten.mul.int %3026, %int32_19428 : !torch.int, !torch.int -> !torch.int
%int2_19429 = torch.constant.int 2
%21841 = torch.aten.mul.int %21840, %int2_19429 : !torch.int, !torch.int -> !torch.int
%int16_19430 = torch.constant.int 16
%int1_19431 = torch.constant.int 1
%int128_19432 = torch.constant.int 128
%21842 = torch.prim.ListConstruct %21841, %int16_19430, %int1_19431, %int128_19432 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21843 = torch.aten.view %21839, %21842 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21843, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21844 = torch.prim.ListConstruct %21797 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19433 = torch.constant.bool false
%21845 = torch.aten.index_put %21843, %21844, %21813, %false_19433 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21845, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19434 = torch.constant.int 32
%int2_19435 = torch.constant.int 2
%int16_19436 = torch.constant.int 16
%int1_19437 = torch.constant.int 1
%int128_19438 = torch.constant.int 128
%21846 = torch.prim.ListConstruct %3026, %int32_19434, %int2_19435, %int16_19436, %int1_19437, %int128_19438 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21847 = torch.aten.view %21845, %21846 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21847, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19439 = torch.constant.int 131072
%21848 = torch.prim.ListConstruct %3026, %int131072_19439 : (!torch.int, !torch.int) -> !torch.list<int>
%21849 = torch.aten.view %21847, %21848 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21849, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19440 = torch.constant.int 32
%int2_19441 = torch.constant.int 2
%int16_19442 = torch.constant.int 16
%int1_19443 = torch.constant.int 1
%int128_19444 = torch.constant.int 128
%21850 = torch.prim.ListConstruct %3029, %int32_19440, %int2_19441, %int16_19442, %int1_19443, %int128_19444 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21851 = torch.aten.view %20000, %21850 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21851, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19445 = torch.constant.int 32
%21852 = torch.aten.mul.int %3029, %int32_19445 : !torch.int, !torch.int -> !torch.int
%int2_19446 = torch.constant.int 2
%21853 = torch.aten.mul.int %21852, %int2_19446 : !torch.int, !torch.int -> !torch.int
%int16_19447 = torch.constant.int 16
%int1_19448 = torch.constant.int 1
%int128_19449 = torch.constant.int 128
%21854 = torch.prim.ListConstruct %21853, %int16_19447, %int1_19448, %int128_19449 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21855 = torch.aten.view %21851, %21854 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21855, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21856 = torch.prim.ListConstruct %21799 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19450 = torch.constant.bool false
%21857 = torch.aten.index_put %21855, %21856, %21815, %false_19450 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21857, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19451 = torch.constant.int 32
%int2_19452 = torch.constant.int 2
%int16_19453 = torch.constant.int 16
%int1_19454 = torch.constant.int 1
%int128_19455 = torch.constant.int 128
%21858 = torch.prim.ListConstruct %3029, %int32_19451, %int2_19452, %int16_19453, %int1_19454, %int128_19455 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21859 = torch.aten.view %21857, %21858 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21859, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19456 = torch.constant.int 131072
%21860 = torch.prim.ListConstruct %3029, %int131072_19456 : (!torch.int, !torch.int) -> !torch.list<int>
%21861 = torch.aten.view %21859, %21860 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21861, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19457 = torch.constant.int 32
%int2_19458 = torch.constant.int 2
%int16_19459 = torch.constant.int 16
%int1_19460 = torch.constant.int 1
%int128_19461 = torch.constant.int 128
%21862 = torch.prim.ListConstruct %3032, %int32_19457, %int2_19458, %int16_19459, %int1_19460, %int128_19461 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21863 = torch.aten.view %20012, %21862 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21863, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19462 = torch.constant.int 32
%21864 = torch.aten.mul.int %3032, %int32_19462 : !torch.int, !torch.int -> !torch.int
%int2_19463 = torch.constant.int 2
%21865 = torch.aten.mul.int %21864, %int2_19463 : !torch.int, !torch.int -> !torch.int
%int16_19464 = torch.constant.int 16
%int1_19465 = torch.constant.int 1
%int128_19466 = torch.constant.int 128
%21866 = torch.prim.ListConstruct %21865, %int16_19464, %int1_19465, %int128_19466 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21867 = torch.aten.view %21863, %21866 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21867, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21868 = torch.prim.ListConstruct %21801 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19467 = torch.constant.bool false
%21869 = torch.aten.index_put %21867, %21868, %21817, %false_19467 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21869, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19468 = torch.constant.int 32
%int2_19469 = torch.constant.int 2
%int16_19470 = torch.constant.int 16
%int1_19471 = torch.constant.int 1
%int128_19472 = torch.constant.int 128
%21870 = torch.prim.ListConstruct %3032, %int32_19468, %int2_19469, %int16_19470, %int1_19471, %int128_19472 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21871 = torch.aten.view %21869, %21870 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21871, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19473 = torch.constant.int 131072
%21872 = torch.prim.ListConstruct %3032, %int131072_19473 : (!torch.int, !torch.int) -> !torch.list<int>
%21873 = torch.aten.view %21871, %21872 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21873, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19474 = torch.constant.int 32
%int2_19475 = torch.constant.int 2
%int16_19476 = torch.constant.int 16
%int1_19477 = torch.constant.int 1
%int128_19478 = torch.constant.int 128
%21874 = torch.prim.ListConstruct %3035, %int32_19474, %int2_19475, %int16_19476, %int1_19477, %int128_19478 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21875 = torch.aten.view %20024, %21874 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21875, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19479 = torch.constant.int 32
%21876 = torch.aten.mul.int %3035, %int32_19479 : !torch.int, !torch.int -> !torch.int
%int2_19480 = torch.constant.int 2
%21877 = torch.aten.mul.int %21876, %int2_19480 : !torch.int, !torch.int -> !torch.int
%int16_19481 = torch.constant.int 16
%int1_19482 = torch.constant.int 1
%int128_19483 = torch.constant.int 128
%21878 = torch.prim.ListConstruct %21877, %int16_19481, %int1_19482, %int128_19483 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21879 = torch.aten.view %21875, %21878 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21879, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21880 = torch.prim.ListConstruct %21803 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19484 = torch.constant.bool false
%21881 = torch.aten.index_put %21879, %21880, %21819, %false_19484 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21881, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19485 = torch.constant.int 32
%int2_19486 = torch.constant.int 2
%int16_19487 = torch.constant.int 16
%int1_19488 = torch.constant.int 1
%int128_19489 = torch.constant.int 128
%21882 = torch.prim.ListConstruct %3035, %int32_19485, %int2_19486, %int16_19487, %int1_19488, %int128_19489 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21883 = torch.aten.view %21881, %21882 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21883, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19490 = torch.constant.int 131072
%21884 = torch.prim.ListConstruct %3035, %int131072_19490 : (!torch.int, !torch.int) -> !torch.list<int>
%21885 = torch.aten.view %21883, %21884 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21885, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19491 = torch.constant.int 32
%int2_19492 = torch.constant.int 2
%int16_19493 = torch.constant.int 16
%int1_19494 = torch.constant.int 1
%int128_19495 = torch.constant.int 128
%21886 = torch.prim.ListConstruct %3038, %int32_19491, %int2_19492, %int16_19493, %int1_19494, %int128_19495 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21887 = torch.aten.view %20036, %21886 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21887, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19496 = torch.constant.int 32
%21888 = torch.aten.mul.int %3038, %int32_19496 : !torch.int, !torch.int -> !torch.int
%int2_19497 = torch.constant.int 2
%21889 = torch.aten.mul.int %21888, %int2_19497 : !torch.int, !torch.int -> !torch.int
%int16_19498 = torch.constant.int 16
%int1_19499 = torch.constant.int 1
%int128_19500 = torch.constant.int 128
%21890 = torch.prim.ListConstruct %21889, %int16_19498, %int1_19499, %int128_19500 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21891 = torch.aten.view %21887, %21890 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21891, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21892 = torch.prim.ListConstruct %21805 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19501 = torch.constant.bool false
%21893 = torch.aten.index_put %21891, %21892, %21821, %false_19501 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21893, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19502 = torch.constant.int 32
%int2_19503 = torch.constant.int 2
%int16_19504 = torch.constant.int 16
%int1_19505 = torch.constant.int 1
%int128_19506 = torch.constant.int 128
%21894 = torch.prim.ListConstruct %3038, %int32_19502, %int2_19503, %int16_19504, %int1_19505, %int128_19506 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21895 = torch.aten.view %21893, %21894 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21895, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19507 = torch.constant.int 131072
%21896 = torch.prim.ListConstruct %3038, %int131072_19507 : (!torch.int, !torch.int) -> !torch.list<int>
%21897 = torch.aten.view %21895, %21896 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21897, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19508 = torch.constant.int 32
%int2_19509 = torch.constant.int 2
%int16_19510 = torch.constant.int 16
%int1_19511 = torch.constant.int 1
%int128_19512 = torch.constant.int 128
%21898 = torch.prim.ListConstruct %3041, %int32_19508, %int2_19509, %int16_19510, %int1_19511, %int128_19512 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21899 = torch.aten.view %20048, %21898 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21899, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19513 = torch.constant.int 32
%21900 = torch.aten.mul.int %3041, %int32_19513 : !torch.int, !torch.int -> !torch.int
%int2_19514 = torch.constant.int 2
%21901 = torch.aten.mul.int %21900, %int2_19514 : !torch.int, !torch.int -> !torch.int
%int16_19515 = torch.constant.int 16
%int1_19516 = torch.constant.int 1
%int128_19517 = torch.constant.int 128
%21902 = torch.prim.ListConstruct %21901, %int16_19515, %int1_19516, %int128_19517 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21903 = torch.aten.view %21899, %21902 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21903, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21904 = torch.prim.ListConstruct %21807 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19518 = torch.constant.bool false
%21905 = torch.aten.index_put %21903, %21904, %21823, %false_19518 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21905, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19519 = torch.constant.int 32
%int2_19520 = torch.constant.int 2
%int16_19521 = torch.constant.int 16
%int1_19522 = torch.constant.int 1
%int128_19523 = torch.constant.int 128
%21906 = torch.prim.ListConstruct %3041, %int32_19519, %int2_19520, %int16_19521, %int1_19522, %int128_19523 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21907 = torch.aten.view %21905, %21906 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21907, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19524 = torch.constant.int 131072
%21908 = torch.prim.ListConstruct %3041, %int131072_19524 : (!torch.int, !torch.int) -> !torch.list<int>
%21909 = torch.aten.view %21907, %21908 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21909, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_19525 = torch.constant.int 32
%int2_19526 = torch.constant.int 2
%int16_19527 = torch.constant.int 16
%int1_19528 = torch.constant.int 1
%int128_19529 = torch.constant.int 128
%21910 = torch.prim.ListConstruct %3044, %int32_19525, %int2_19526, %int16_19527, %int1_19528, %int128_19529 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21911 = torch.aten.view %20060, %21910 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21911, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_19530 = torch.constant.int 32
%21912 = torch.aten.mul.int %3044, %int32_19530 : !torch.int, !torch.int -> !torch.int
%int2_19531 = torch.constant.int 2
%21913 = torch.aten.mul.int %21912, %int2_19531 : !torch.int, !torch.int -> !torch.int
%int16_19532 = torch.constant.int 16
%int1_19533 = torch.constant.int 1
%int128_19534 = torch.constant.int 128
%21914 = torch.prim.ListConstruct %21913, %int16_19532, %int1_19533, %int128_19534 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21915 = torch.aten.view %21911, %21914 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21915, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%21916 = torch.prim.ListConstruct %21809 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_19535 = torch.constant.bool false
%21917 = torch.aten.index_put %21915, %21916, %21825, %false_19535 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %21917, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_19536 = torch.constant.int 32
%int2_19537 = torch.constant.int 2
%int16_19538 = torch.constant.int 16
%int1_19539 = torch.constant.int 1
%int128_19540 = torch.constant.int 128
%21918 = torch.prim.ListConstruct %3044, %int32_19536, %int2_19537, %int16_19538, %int1_19539, %int128_19540 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21919 = torch.aten.view %21917, %21918 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %21919, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_19541 = torch.constant.int 131072
%21920 = torch.prim.ListConstruct %3044, %int131072_19541 : (!torch.int, !torch.int) -> !torch.list<int>
%21921 = torch.aten.view %21919, %21920 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %21921, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_19542 = torch.constant.int -2
%21922 = torch.aten.unsqueeze %21536, %int-2_19542 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19543 = torch.constant.int -2
%21923 = torch.aten.unsqueeze %21551, %int-2_19543 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19544 = torch.constant.int -2
%21924 = torch.aten.unsqueeze %21566, %int-2_19544 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19545 = torch.constant.int -2
%21925 = torch.aten.unsqueeze %21581, %int-2_19545 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19546 = torch.constant.int -2
%21926 = torch.aten.unsqueeze %21596, %int-2_19546 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19547 = torch.constant.int -2
%21927 = torch.aten.unsqueeze %21611, %int-2_19547 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19548 = torch.constant.int -2
%21928 = torch.aten.unsqueeze %21626, %int-2_19548 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19549 = torch.constant.int -2
%21929 = torch.aten.unsqueeze %21641, %int-2_19549 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_19550 = torch.constant.int 4
%int1_19551 = torch.constant.int 1
%int4_19552 = torch.constant.int 4
%int128_19553 = torch.constant.int 128
%21930 = torch.prim.ListConstruct %int4_19550, %21522, %int1_19551, %int4_19552, %int128_19553 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19554 = torch.constant.bool false
%21931 = torch.aten.expand %21922, %21930, %false_19554 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19555 = torch.constant.int 4
%int1_19556 = torch.constant.int 1
%int4_19557 = torch.constant.int 4
%int128_19558 = torch.constant.int 128
%21932 = torch.prim.ListConstruct %int4_19555, %21522, %int1_19556, %int4_19557, %int128_19558 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19559 = torch.constant.bool false
%21933 = torch.aten.expand %21923, %21932, %false_19559 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19560 = torch.constant.int 4
%int1_19561 = torch.constant.int 1
%int4_19562 = torch.constant.int 4
%int128_19563 = torch.constant.int 128
%21934 = torch.prim.ListConstruct %int4_19560, %21522, %int1_19561, %int4_19562, %int128_19563 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19564 = torch.constant.bool false
%21935 = torch.aten.expand %21924, %21934, %false_19564 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19565 = torch.constant.int 4
%int1_19566 = torch.constant.int 1
%int4_19567 = torch.constant.int 4
%int128_19568 = torch.constant.int 128
%21936 = torch.prim.ListConstruct %int4_19565, %21522, %int1_19566, %int4_19567, %int128_19568 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19569 = torch.constant.bool false
%21937 = torch.aten.expand %21925, %21936, %false_19569 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19570 = torch.constant.int 4
%int1_19571 = torch.constant.int 1
%int4_19572 = torch.constant.int 4
%int128_19573 = torch.constant.int 128
%21938 = torch.prim.ListConstruct %int4_19570, %21522, %int1_19571, %int4_19572, %int128_19573 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19574 = torch.constant.bool false
%21939 = torch.aten.expand %21926, %21938, %false_19574 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19575 = torch.constant.int 4
%int1_19576 = torch.constant.int 1
%int4_19577 = torch.constant.int 4
%int128_19578 = torch.constant.int 128
%21940 = torch.prim.ListConstruct %int4_19575, %21522, %int1_19576, %int4_19577, %int128_19578 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19579 = torch.constant.bool false
%21941 = torch.aten.expand %21927, %21940, %false_19579 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19580 = torch.constant.int 4
%int1_19581 = torch.constant.int 1
%int4_19582 = torch.constant.int 4
%int128_19583 = torch.constant.int 128
%21942 = torch.prim.ListConstruct %int4_19580, %21522, %int1_19581, %int4_19582, %int128_19583 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19584 = torch.constant.bool false
%21943 = torch.aten.expand %21928, %21942, %false_19584 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19585 = torch.constant.int 4
%int1_19586 = torch.constant.int 1
%int4_19587 = torch.constant.int 4
%int128_19588 = torch.constant.int 128
%21944 = torch.prim.ListConstruct %int4_19585, %21522, %int1_19586, %int4_19587, %int128_19588 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19589 = torch.constant.bool false
%21945 = torch.aten.expand %21929, %21944, %false_19589 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19590 = torch.constant.int 4
%int4_19591 = torch.constant.int 4
%int128_19592 = torch.constant.int 128
%21946 = torch.prim.ListConstruct %int4_19590, %21522, %int4_19591, %int128_19592 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21947 = torch.aten.view %21931, %21946 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19593 = torch.constant.int 4
%int4_19594 = torch.constant.int 4
%int128_19595 = torch.constant.int 128
%21948 = torch.prim.ListConstruct %int4_19593, %21522, %int4_19594, %int128_19595 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21949 = torch.aten.view %21933, %21948 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19596 = torch.constant.int 4
%int4_19597 = torch.constant.int 4
%int128_19598 = torch.constant.int 128
%21950 = torch.prim.ListConstruct %int4_19596, %21522, %int4_19597, %int128_19598 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21951 = torch.aten.view %21935, %21950 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19599 = torch.constant.int 4
%int4_19600 = torch.constant.int 4
%int128_19601 = torch.constant.int 128
%21952 = torch.prim.ListConstruct %int4_19599, %21522, %int4_19600, %int128_19601 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21953 = torch.aten.view %21937, %21952 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19602 = torch.constant.int 4
%int4_19603 = torch.constant.int 4
%int128_19604 = torch.constant.int 128
%21954 = torch.prim.ListConstruct %int4_19602, %21522, %int4_19603, %int128_19604 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21955 = torch.aten.view %21939, %21954 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19605 = torch.constant.int 4
%int4_19606 = torch.constant.int 4
%int128_19607 = torch.constant.int 128
%21956 = torch.prim.ListConstruct %int4_19605, %21522, %int4_19606, %int128_19607 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21957 = torch.aten.view %21941, %21956 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19608 = torch.constant.int 4
%int4_19609 = torch.constant.int 4
%int128_19610 = torch.constant.int 128
%21958 = torch.prim.ListConstruct %int4_19608, %21522, %int4_19609, %int128_19610 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21959 = torch.aten.view %21943, %21958 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19611 = torch.constant.int 4
%int4_19612 = torch.constant.int 4
%int128_19613 = torch.constant.int 128
%21960 = torch.prim.ListConstruct %int4_19611, %21522, %int4_19612, %int128_19613 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21961 = torch.aten.view %21945, %21960 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_19614 = torch.constant.int -2
%21962 = torch.aten.unsqueeze %21311, %int-2_19614 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19615 = torch.constant.int -2
%21963 = torch.aten.unsqueeze %21313, %int-2_19615 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19616 = torch.constant.int -2
%21964 = torch.aten.unsqueeze %21315, %int-2_19616 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19617 = torch.constant.int -2
%21965 = torch.aten.unsqueeze %21317, %int-2_19617 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19618 = torch.constant.int -2
%21966 = torch.aten.unsqueeze %21319, %int-2_19618 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19619 = torch.constant.int -2
%21967 = torch.aten.unsqueeze %21321, %int-2_19619 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19620 = torch.constant.int -2
%21968 = torch.aten.unsqueeze %21323, %int-2_19620 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_19621 = torch.constant.int -2
%21969 = torch.aten.unsqueeze %21325, %int-2_19621 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %21969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_19622 = torch.constant.int 1
%21970 = torch.aten.size.int %21235, %int1_19622 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_19623 = torch.constant.int 4
%int1_19624 = torch.constant.int 1
%int4_19625 = torch.constant.int 4
%int128_19626 = torch.constant.int 128
%21971 = torch.prim.ListConstruct %int4_19623, %21970, %int1_19624, %int4_19625, %int128_19626 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19627 = torch.constant.bool false
%21972 = torch.aten.expand %21962, %21971, %false_19627 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19628 = torch.constant.int 4
%int1_19629 = torch.constant.int 1
%int4_19630 = torch.constant.int 4
%int128_19631 = torch.constant.int 128
%21973 = torch.prim.ListConstruct %int4_19628, %21970, %int1_19629, %int4_19630, %int128_19631 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19632 = torch.constant.bool false
%21974 = torch.aten.expand %21963, %21973, %false_19632 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19633 = torch.constant.int 4
%int1_19634 = torch.constant.int 1
%int4_19635 = torch.constant.int 4
%int128_19636 = torch.constant.int 128
%21975 = torch.prim.ListConstruct %int4_19633, %21970, %int1_19634, %int4_19635, %int128_19636 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19637 = torch.constant.bool false
%21976 = torch.aten.expand %21964, %21975, %false_19637 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19638 = torch.constant.int 4
%int1_19639 = torch.constant.int 1
%int4_19640 = torch.constant.int 4
%int128_19641 = torch.constant.int 128
%21977 = torch.prim.ListConstruct %int4_19638, %21970, %int1_19639, %int4_19640, %int128_19641 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19642 = torch.constant.bool false
%21978 = torch.aten.expand %21965, %21977, %false_19642 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19643 = torch.constant.int 4
%int1_19644 = torch.constant.int 1
%int4_19645 = torch.constant.int 4
%int128_19646 = torch.constant.int 128
%21979 = torch.prim.ListConstruct %int4_19643, %21970, %int1_19644, %int4_19645, %int128_19646 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19647 = torch.constant.bool false
%21980 = torch.aten.expand %21966, %21979, %false_19647 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19648 = torch.constant.int 4
%int1_19649 = torch.constant.int 1
%int4_19650 = torch.constant.int 4
%int128_19651 = torch.constant.int 128
%21981 = torch.prim.ListConstruct %int4_19648, %21970, %int1_19649, %int4_19650, %int128_19651 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19652 = torch.constant.bool false
%21982 = torch.aten.expand %21967, %21981, %false_19652 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19653 = torch.constant.int 4
%int1_19654 = torch.constant.int 1
%int4_19655 = torch.constant.int 4
%int128_19656 = torch.constant.int 128
%21983 = torch.prim.ListConstruct %int4_19653, %21970, %int1_19654, %int4_19655, %int128_19656 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19657 = torch.constant.bool false
%21984 = torch.aten.expand %21968, %21983, %false_19657 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19658 = torch.constant.int 4
%int1_19659 = torch.constant.int 1
%int4_19660 = torch.constant.int 4
%int128_19661 = torch.constant.int 128
%21985 = torch.prim.ListConstruct %int4_19658, %21970, %int1_19659, %int4_19660, %int128_19661 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_19662 = torch.constant.bool false
%21986 = torch.aten.expand %21969, %21985, %false_19662 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %21986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_19663 = torch.constant.int 4
%int4_19664 = torch.constant.int 4
%int128_19665 = torch.constant.int 128
%21987 = torch.prim.ListConstruct %int4_19663, %21970, %int4_19664, %int128_19665 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21988 = torch.aten.view %21972, %21987 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19666 = torch.constant.int 4
%int4_19667 = torch.constant.int 4
%int128_19668 = torch.constant.int 128
%21989 = torch.prim.ListConstruct %int4_19666, %21970, %int4_19667, %int128_19668 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21990 = torch.aten.view %21974, %21989 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19669 = torch.constant.int 4
%int4_19670 = torch.constant.int 4
%int128_19671 = torch.constant.int 128
%21991 = torch.prim.ListConstruct %int4_19669, %21970, %int4_19670, %int128_19671 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21992 = torch.aten.view %21976, %21991 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19672 = torch.constant.int 4
%int4_19673 = torch.constant.int 4
%int128_19674 = torch.constant.int 128
%21993 = torch.prim.ListConstruct %int4_19672, %21970, %int4_19673, %int128_19674 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21994 = torch.aten.view %21978, %21993 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19675 = torch.constant.int 4
%int4_19676 = torch.constant.int 4
%int128_19677 = torch.constant.int 128
%21995 = torch.prim.ListConstruct %int4_19675, %21970, %int4_19676, %int128_19677 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21996 = torch.aten.view %21980, %21995 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19678 = torch.constant.int 4
%int4_19679 = torch.constant.int 4
%int128_19680 = torch.constant.int 128
%21997 = torch.prim.ListConstruct %int4_19678, %21970, %int4_19679, %int128_19680 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%21998 = torch.aten.view %21982, %21997 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %21998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19681 = torch.constant.int 4
%int4_19682 = torch.constant.int 4
%int128_19683 = torch.constant.int 128
%21999 = torch.prim.ListConstruct %int4_19681, %21970, %int4_19682, %int128_19683 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22000 = torch.aten.view %21984, %21999 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19684 = torch.constant.int 4
%int4_19685 = torch.constant.int 4
%int128_19686 = torch.constant.int 128
%22001 = torch.prim.ListConstruct %int4_19684, %21970, %int4_19685, %int128_19686 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22002 = torch.aten.view %21986, %22001 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19687 = torch.constant.int 1
%int2_19688 = torch.constant.int 2
%22003 = torch.aten.transpose.int %21378, %int1_19687, %int2_19688 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22003, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19689 = torch.constant.int 1
%int2_19690 = torch.constant.int 2
%22004 = torch.aten.transpose.int %21393, %int1_19689, %int2_19690 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22004, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19691 = torch.constant.int 1
%int2_19692 = torch.constant.int 2
%22005 = torch.aten.transpose.int %21408, %int1_19691, %int2_19692 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22005, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19693 = torch.constant.int 1
%int2_19694 = torch.constant.int 2
%22006 = torch.aten.transpose.int %21423, %int1_19693, %int2_19694 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22006, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19695 = torch.constant.int 1
%int2_19696 = torch.constant.int 2
%22007 = torch.aten.transpose.int %21438, %int1_19695, %int2_19696 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22007, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19697 = torch.constant.int 1
%int2_19698 = torch.constant.int 2
%22008 = torch.aten.transpose.int %21453, %int1_19697, %int2_19698 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22008, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19699 = torch.constant.int 1
%int2_19700 = torch.constant.int 2
%22009 = torch.aten.transpose.int %21468, %int1_19699, %int2_19700 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22009, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19701 = torch.constant.int 1
%int2_19702 = torch.constant.int 2
%22010 = torch.aten.transpose.int %21483, %int1_19701, %int2_19702 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22010, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19703 = torch.constant.int 1
%int2_19704 = torch.constant.int 2
%22011 = torch.aten.transpose.int %21947, %int1_19703, %int2_19704 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22011, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19705 = torch.constant.int 1
%int2_19706 = torch.constant.int 2
%22012 = torch.aten.transpose.int %21949, %int1_19705, %int2_19706 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22012, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19707 = torch.constant.int 1
%int2_19708 = torch.constant.int 2
%22013 = torch.aten.transpose.int %21951, %int1_19707, %int2_19708 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22013, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19709 = torch.constant.int 1
%int2_19710 = torch.constant.int 2
%22014 = torch.aten.transpose.int %21953, %int1_19709, %int2_19710 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22014, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19711 = torch.constant.int 1
%int2_19712 = torch.constant.int 2
%22015 = torch.aten.transpose.int %21955, %int1_19711, %int2_19712 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22015, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19713 = torch.constant.int 1
%int2_19714 = torch.constant.int 2
%22016 = torch.aten.transpose.int %21957, %int1_19713, %int2_19714 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22016, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19715 = torch.constant.int 1
%int2_19716 = torch.constant.int 2
%22017 = torch.aten.transpose.int %21959, %int1_19715, %int2_19716 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22017, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19717 = torch.constant.int 1
%int2_19718 = torch.constant.int 2
%22018 = torch.aten.transpose.int %21961, %int1_19717, %int2_19718 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22018, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19719 = torch.constant.int 1
%int2_19720 = torch.constant.int 2
%22019 = torch.aten.transpose.int %21988, %int1_19719, %int2_19720 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22019, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19721 = torch.constant.int 1
%int2_19722 = torch.constant.int 2
%22020 = torch.aten.transpose.int %21990, %int1_19721, %int2_19722 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22020, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19723 = torch.constant.int 1
%int2_19724 = torch.constant.int 2
%22021 = torch.aten.transpose.int %21992, %int1_19723, %int2_19724 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22021, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19725 = torch.constant.int 1
%int2_19726 = torch.constant.int 2
%22022 = torch.aten.transpose.int %21994, %int1_19725, %int2_19726 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22022, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19727 = torch.constant.int 1
%int2_19728 = torch.constant.int 2
%22023 = torch.aten.transpose.int %21996, %int1_19727, %int2_19728 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22023, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19729 = torch.constant.int 1
%int2_19730 = torch.constant.int 2
%22024 = torch.aten.transpose.int %21998, %int1_19729, %int2_19730 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22024, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19731 = torch.constant.int 1
%int2_19732 = torch.constant.int 2
%22025 = torch.aten.transpose.int %22000, %int1_19731, %int2_19732 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22025, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19733 = torch.constant.int 1
%int2_19734 = torch.constant.int 2
%22026 = torch.aten.transpose.int %22002, %int1_19733, %int2_19734 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %22026, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19735 = torch.constant.float 0.000000e+00
%true_19736 = torch.constant.bool true
%none_19737 = torch.constant.none
%none_19738 = torch.constant.none
%22027:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22003, %22011, %22019, %float0.000000e00_19735, %true_19736, %none_19737, %none_19738) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22027#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19739 = torch.constant.float 0.000000e+00
%true_19740 = torch.constant.bool true
%none_19741 = torch.constant.none
%none_19742 = torch.constant.none
%22028:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22004, %22012, %22020, %float0.000000e00_19739, %true_19740, %none_19741, %none_19742) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22028#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19743 = torch.constant.float 0.000000e+00
%true_19744 = torch.constant.bool true
%none_19745 = torch.constant.none
%none_19746 = torch.constant.none
%22029:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22005, %22013, %22021, %float0.000000e00_19743, %true_19744, %none_19745, %none_19746) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22029#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19747 = torch.constant.float 0.000000e+00
%true_19748 = torch.constant.bool true
%none_19749 = torch.constant.none
%none_19750 = torch.constant.none
%22030:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22006, %22014, %22022, %float0.000000e00_19747, %true_19748, %none_19749, %none_19750) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22030#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19751 = torch.constant.float 0.000000e+00
%true_19752 = torch.constant.bool true
%none_19753 = torch.constant.none
%none_19754 = torch.constant.none
%22031:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22007, %22015, %22023, %float0.000000e00_19751, %true_19752, %none_19753, %none_19754) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22031#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19755 = torch.constant.float 0.000000e+00
%true_19756 = torch.constant.bool true
%none_19757 = torch.constant.none
%none_19758 = torch.constant.none
%22032:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22008, %22016, %22024, %float0.000000e00_19755, %true_19756, %none_19757, %none_19758) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22032#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19759 = torch.constant.float 0.000000e+00
%true_19760 = torch.constant.bool true
%none_19761 = torch.constant.none
%none_19762 = torch.constant.none
%22033:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22009, %22017, %22025, %float0.000000e00_19759, %true_19760, %none_19761, %none_19762) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22033#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_19763 = torch.constant.float 0.000000e+00
%true_19764 = torch.constant.bool true
%none_19765 = torch.constant.none
%none_19766 = torch.constant.none
%22034:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%22010, %22018, %22026, %float0.000000e00_19763, %true_19764, %none_19765, %none_19766) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %22034#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_19767 = torch.constant.int 1
%int2_19768 = torch.constant.int 2
%22035 = torch.aten.transpose.int %22027#0, %int1_19767, %int2_19768 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19769 = torch.constant.int 1
%int2_19770 = torch.constant.int 2
%22036 = torch.aten.transpose.int %22028#0, %int1_19769, %int2_19770 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19771 = torch.constant.int 1
%int2_19772 = torch.constant.int 2
%22037 = torch.aten.transpose.int %22029#0, %int1_19771, %int2_19772 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19773 = torch.constant.int 1
%int2_19774 = torch.constant.int 2
%22038 = torch.aten.transpose.int %22030#0, %int1_19773, %int2_19774 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19775 = torch.constant.int 1
%int2_19776 = torch.constant.int 2
%22039 = torch.aten.transpose.int %22031#0, %int1_19775, %int2_19776 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19777 = torch.constant.int 1
%int2_19778 = torch.constant.int 2
%22040 = torch.aten.transpose.int %22032#0, %int1_19777, %int2_19778 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19779 = torch.constant.int 1
%int2_19780 = torch.constant.int 2
%22041 = torch.aten.transpose.int %22033#0, %int1_19779, %int2_19780 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_19781 = torch.constant.int 1
%int2_19782 = torch.constant.int 2
%22042 = torch.aten.transpose.int %22034#0, %int1_19781, %int2_19782 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %22042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_19783 = torch.constant.int 4
%int512_19784 = torch.constant.int 512
%22043 = torch.prim.ListConstruct %int4_19783, %21364, %int512_19784 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22044 = torch.aten.view %22035, %22043 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19785 = torch.constant.int 4
%int512_19786 = torch.constant.int 512
%22045 = torch.prim.ListConstruct %int4_19785, %21379, %int512_19786 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22046 = torch.aten.view %22036, %22045 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19787 = torch.constant.int 4
%int512_19788 = torch.constant.int 512
%22047 = torch.prim.ListConstruct %int4_19787, %21394, %int512_19788 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22048 = torch.aten.view %22037, %22047 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19789 = torch.constant.int 4
%int512_19790 = torch.constant.int 512
%22049 = torch.prim.ListConstruct %int4_19789, %21409, %int512_19790 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22050 = torch.aten.view %22038, %22049 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19791 = torch.constant.int 4
%int512_19792 = torch.constant.int 512
%22051 = torch.prim.ListConstruct %int4_19791, %21424, %int512_19792 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22052 = torch.aten.view %22039, %22051 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19793 = torch.constant.int 4
%int512_19794 = torch.constant.int 512
%22053 = torch.prim.ListConstruct %int4_19793, %21439, %int512_19794 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22054 = torch.aten.view %22040, %22053 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19795 = torch.constant.int 4
%int512_19796 = torch.constant.int 512
%22055 = torch.prim.ListConstruct %int4_19795, %21454, %int512_19796 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22056 = torch.aten.view %22041, %22055 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_19797 = torch.constant.int 4
%int512_19798 = torch.constant.int 512
%22057 = torch.prim.ListConstruct %int4_19797, %21469, %int512_19798 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22058 = torch.aten.view %22042, %22057 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_19799 = torch.constant.int 1
%int0_19800 = torch.constant.int 0
%22059 = torch.prim.ListConstruct %int1_19799, %int0_19800 : (!torch.int, !torch.int) -> !torch.list<int>
%22060 = torch.aten.permute %760, %22059 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19801 = torch.constant.int 1
%int0_19802 = torch.constant.int 0
%22061 = torch.prim.ListConstruct %int1_19801, %int0_19802 : (!torch.int, !torch.int) -> !torch.list<int>
%22062 = torch.aten.permute %761, %22061 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19803 = torch.constant.int 1
%int0_19804 = torch.constant.int 0
%22063 = torch.prim.ListConstruct %int1_19803, %int0_19804 : (!torch.int, !torch.int) -> !torch.list<int>
%22064 = torch.aten.permute %762, %22063 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19805 = torch.constant.int 1
%int0_19806 = torch.constant.int 0
%22065 = torch.prim.ListConstruct %int1_19805, %int0_19806 : (!torch.int, !torch.int) -> !torch.list<int>
%22066 = torch.aten.permute %763, %22065 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19807 = torch.constant.int 1
%int0_19808 = torch.constant.int 0
%22067 = torch.prim.ListConstruct %int1_19807, %int0_19808 : (!torch.int, !torch.int) -> !torch.list<int>
%22068 = torch.aten.permute %764, %22067 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19809 = torch.constant.int 1
%int0_19810 = torch.constant.int 0
%22069 = torch.prim.ListConstruct %int1_19809, %int0_19810 : (!torch.int, !torch.int) -> !torch.list<int>
%22070 = torch.aten.permute %765, %22069 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19811 = torch.constant.int 1
%int0_19812 = torch.constant.int 0
%22071 = torch.prim.ListConstruct %int1_19811, %int0_19812 : (!torch.int, !torch.int) -> !torch.list<int>
%22072 = torch.aten.permute %766, %22071 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_19813 = torch.constant.int 1
%int0_19814 = torch.constant.int 0
%22073 = torch.prim.ListConstruct %int1_19813, %int0_19814 : (!torch.int, !torch.int) -> !torch.list<int>
%22074 = torch.aten.permute %767, %22073 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_19815 = torch.constant.int 4
%22075 = torch.aten.mul.int %int4_19815, %21364 : !torch.int, !torch.int -> !torch.int
%int512_19816 = torch.constant.int 512
%22076 = torch.prim.ListConstruct %22075, %int512_19816 : (!torch.int, !torch.int) -> !torch.list<int>
%22077 = torch.aten.view %22044, %22076 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22077, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22078 = torch.aten.mm %22077, %22060 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22078, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19817 = torch.constant.int 4
%int4096_19818 = torch.constant.int 4096
%22079 = torch.prim.ListConstruct %int4_19817, %21364, %int4096_19818 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22080 = torch.aten.view %22078, %22079 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19819 = torch.constant.int 4
%22081 = torch.aten.mul.int %int4_19819, %21379 : !torch.int, !torch.int -> !torch.int
%int512_19820 = torch.constant.int 512
%22082 = torch.prim.ListConstruct %22081, %int512_19820 : (!torch.int, !torch.int) -> !torch.list<int>
%22083 = torch.aten.view %22046, %22082 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22083, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22084 = torch.aten.mm %22083, %22062 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22084, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19821 = torch.constant.int 4
%int4096_19822 = torch.constant.int 4096
%22085 = torch.prim.ListConstruct %int4_19821, %21379, %int4096_19822 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22086 = torch.aten.view %22084, %22085 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19823 = torch.constant.int 4
%22087 = torch.aten.mul.int %int4_19823, %21394 : !torch.int, !torch.int -> !torch.int
%int512_19824 = torch.constant.int 512
%22088 = torch.prim.ListConstruct %22087, %int512_19824 : (!torch.int, !torch.int) -> !torch.list<int>
%22089 = torch.aten.view %22048, %22088 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22089, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22090 = torch.aten.mm %22089, %22064 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22090, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19825 = torch.constant.int 4
%int4096_19826 = torch.constant.int 4096
%22091 = torch.prim.ListConstruct %int4_19825, %21394, %int4096_19826 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22092 = torch.aten.view %22090, %22091 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19827 = torch.constant.int 4
%22093 = torch.aten.mul.int %int4_19827, %21409 : !torch.int, !torch.int -> !torch.int
%int512_19828 = torch.constant.int 512
%22094 = torch.prim.ListConstruct %22093, %int512_19828 : (!torch.int, !torch.int) -> !torch.list<int>
%22095 = torch.aten.view %22050, %22094 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22095, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22096 = torch.aten.mm %22095, %22066 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22096, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19829 = torch.constant.int 4
%int4096_19830 = torch.constant.int 4096
%22097 = torch.prim.ListConstruct %int4_19829, %21409, %int4096_19830 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22098 = torch.aten.view %22096, %22097 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19831 = torch.constant.int 4
%22099 = torch.aten.mul.int %int4_19831, %21424 : !torch.int, !torch.int -> !torch.int
%int512_19832 = torch.constant.int 512
%22100 = torch.prim.ListConstruct %22099, %int512_19832 : (!torch.int, !torch.int) -> !torch.list<int>
%22101 = torch.aten.view %22052, %22100 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22101, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22102 = torch.aten.mm %22101, %22068 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22102, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19833 = torch.constant.int 4
%int4096_19834 = torch.constant.int 4096
%22103 = torch.prim.ListConstruct %int4_19833, %21424, %int4096_19834 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22104 = torch.aten.view %22102, %22103 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19835 = torch.constant.int 4
%22105 = torch.aten.mul.int %int4_19835, %21439 : !torch.int, !torch.int -> !torch.int
%int512_19836 = torch.constant.int 512
%22106 = torch.prim.ListConstruct %22105, %int512_19836 : (!torch.int, !torch.int) -> !torch.list<int>
%22107 = torch.aten.view %22054, %22106 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22107, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22108 = torch.aten.mm %22107, %22070 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22108, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19837 = torch.constant.int 4
%int4096_19838 = torch.constant.int 4096
%22109 = torch.prim.ListConstruct %int4_19837, %21439, %int4096_19838 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22110 = torch.aten.view %22108, %22109 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19839 = torch.constant.int 4
%22111 = torch.aten.mul.int %int4_19839, %21454 : !torch.int, !torch.int -> !torch.int
%int512_19840 = torch.constant.int 512
%22112 = torch.prim.ListConstruct %22111, %int512_19840 : (!torch.int, !torch.int) -> !torch.list<int>
%22113 = torch.aten.view %22056, %22112 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22113, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22114 = torch.aten.mm %22113, %22072 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22114, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19841 = torch.constant.int 4
%int4096_19842 = torch.constant.int 4096
%22115 = torch.prim.ListConstruct %int4_19841, %21454, %int4096_19842 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22116 = torch.aten.view %22114, %22115 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_19843 = torch.constant.int 4
%22117 = torch.aten.mul.int %int4_19843, %21469 : !torch.int, !torch.int -> !torch.int
%int512_19844 = torch.constant.int 512
%22118 = torch.prim.ListConstruct %22117, %int512_19844 : (!torch.int, !torch.int) -> !torch.list<int>
%22119 = torch.aten.view %22058, %22118 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22119, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%22120 = torch.aten.mm %22119, %22074 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22120, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_19845 = torch.constant.int 4
%int4096_19846 = torch.constant.int 4096
%22121 = torch.prim.ListConstruct %int4_19845, %21469, %int4096_19846 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22122 = torch.aten.view %22120, %22121 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22123 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19847 = arith.constant 1 : index
%dim_19848 = tensor.dim %22123, %c1_19847 : tensor<4x?x4096xf16>
%22124 = flow.tensor.transfer %22123 : tensor<4x?x4096xf16>{%dim_19848} to #hal.device.promise<@__device_0>
%22125 = torch_c.from_builtin_tensor %22124 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22126 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19849 = arith.constant 1 : index
%dim_19850 = tensor.dim %22126, %c1_19849 : tensor<4x?x4096xf16>
%22127 = flow.tensor.transfer %22126 : tensor<4x?x4096xf16>{%dim_19850} to #hal.device.promise<@__device_0>
%22128 = torch_c.from_builtin_tensor %22127 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22129 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19851 = arith.constant 1 : index
%dim_19852 = tensor.dim %22129, %c1_19851 : tensor<4x?x4096xf16>
%22130 = flow.tensor.transfer %22129 : tensor<4x?x4096xf16>{%dim_19852} to #hal.device.promise<@__device_0>
%22131 = torch_c.from_builtin_tensor %22130 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22132 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19853 = arith.constant 1 : index
%dim_19854 = tensor.dim %22132, %c1_19853 : tensor<4x?x4096xf16>
%22133 = flow.tensor.transfer %22132 : tensor<4x?x4096xf16>{%dim_19854} to #hal.device.promise<@__device_0>
%22134 = torch_c.from_builtin_tensor %22133 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22135 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19855 = arith.constant 1 : index
%dim_19856 = tensor.dim %22135, %c1_19855 : tensor<4x?x4096xf16>
%22136 = flow.tensor.transfer %22135 : tensor<4x?x4096xf16>{%dim_19856} to #hal.device.promise<@__device_0>
%22137 = torch_c.from_builtin_tensor %22136 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22138 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19857 = arith.constant 1 : index
%dim_19858 = tensor.dim %22138, %c1_19857 : tensor<4x?x4096xf16>
%22139 = flow.tensor.transfer %22138 : tensor<4x?x4096xf16>{%dim_19858} to #hal.device.promise<@__device_0>
%22140 = torch_c.from_builtin_tensor %22139 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22141 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19859 = arith.constant 1 : index
%dim_19860 = tensor.dim %22141, %c1_19859 : tensor<4x?x4096xf16>
%22142 = flow.tensor.transfer %22141 : tensor<4x?x4096xf16>{%dim_19860} to #hal.device.promise<@__device_0>
%22143 = torch_c.from_builtin_tensor %22142 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19861 = torch.constant.int 1
%22144 = torch.aten.add.Tensor %22080, %22125, %int1_19861 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19862 = torch.constant.int 1
%22145 = torch.aten.add.Tensor %22144, %22128, %int1_19862 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19863 = torch.constant.int 1
%22146 = torch.aten.add.Tensor %22145, %22131, %int1_19863 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19864 = torch.constant.int 1
%22147 = torch.aten.add.Tensor %22146, %22134, %int1_19864 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19865 = torch.constant.int 1
%22148 = torch.aten.add.Tensor %22147, %22137, %int1_19865 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19866 = torch.constant.int 1
%22149 = torch.aten.add.Tensor %22148, %22140, %int1_19866 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19867 = torch.constant.int 1
%22150 = torch.aten.add.Tensor %22149, %22143, %int1_19867 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22151 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19868 = arith.constant 1 : index
%dim_19869 = tensor.dim %22151, %c1_19868 : tensor<4x?x4096xf16>
%22152 = flow.tensor.transfer %22151 : tensor<4x?x4096xf16>{%dim_19869} to #hal.device.promise<@__device_1>
%22153 = torch_c.from_builtin_tensor %22152 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22154 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19870 = arith.constant 1 : index
%dim_19871 = tensor.dim %22154, %c1_19870 : tensor<4x?x4096xf16>
%22155 = flow.tensor.transfer %22154 : tensor<4x?x4096xf16>{%dim_19871} to #hal.device.promise<@__device_1>
%22156 = torch_c.from_builtin_tensor %22155 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22157 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19872 = arith.constant 1 : index
%dim_19873 = tensor.dim %22157, %c1_19872 : tensor<4x?x4096xf16>
%22158 = flow.tensor.transfer %22157 : tensor<4x?x4096xf16>{%dim_19873} to #hal.device.promise<@__device_1>
%22159 = torch_c.from_builtin_tensor %22158 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22160 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19874 = arith.constant 1 : index
%dim_19875 = tensor.dim %22160, %c1_19874 : tensor<4x?x4096xf16>
%22161 = flow.tensor.transfer %22160 : tensor<4x?x4096xf16>{%dim_19875} to #hal.device.promise<@__device_1>
%22162 = torch_c.from_builtin_tensor %22161 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22163 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19876 = arith.constant 1 : index
%dim_19877 = tensor.dim %22163, %c1_19876 : tensor<4x?x4096xf16>
%22164 = flow.tensor.transfer %22163 : tensor<4x?x4096xf16>{%dim_19877} to #hal.device.promise<@__device_1>
%22165 = torch_c.from_builtin_tensor %22164 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22166 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19878 = arith.constant 1 : index
%dim_19879 = tensor.dim %22166, %c1_19878 : tensor<4x?x4096xf16>
%22167 = flow.tensor.transfer %22166 : tensor<4x?x4096xf16>{%dim_19879} to #hal.device.promise<@__device_1>
%22168 = torch_c.from_builtin_tensor %22167 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22169 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19880 = arith.constant 1 : index
%dim_19881 = tensor.dim %22169, %c1_19880 : tensor<4x?x4096xf16>
%22170 = flow.tensor.transfer %22169 : tensor<4x?x4096xf16>{%dim_19881} to #hal.device.promise<@__device_1>
%22171 = torch_c.from_builtin_tensor %22170 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19882 = torch.constant.int 1
%22172 = torch.aten.add.Tensor %22153, %22086, %int1_19882 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19883 = torch.constant.int 1
%22173 = torch.aten.add.Tensor %22172, %22156, %int1_19883 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19884 = torch.constant.int 1
%22174 = torch.aten.add.Tensor %22173, %22159, %int1_19884 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19885 = torch.constant.int 1
%22175 = torch.aten.add.Tensor %22174, %22162, %int1_19885 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19886 = torch.constant.int 1
%22176 = torch.aten.add.Tensor %22175, %22165, %int1_19886 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19887 = torch.constant.int 1
%22177 = torch.aten.add.Tensor %22176, %22168, %int1_19887 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19888 = torch.constant.int 1
%22178 = torch.aten.add.Tensor %22177, %22171, %int1_19888 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22179 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19889 = arith.constant 1 : index
%dim_19890 = tensor.dim %22179, %c1_19889 : tensor<4x?x4096xf16>
%22180 = flow.tensor.transfer %22179 : tensor<4x?x4096xf16>{%dim_19890} to #hal.device.promise<@__device_2>
%22181 = torch_c.from_builtin_tensor %22180 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22182 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19891 = arith.constant 1 : index
%dim_19892 = tensor.dim %22182, %c1_19891 : tensor<4x?x4096xf16>
%22183 = flow.tensor.transfer %22182 : tensor<4x?x4096xf16>{%dim_19892} to #hal.device.promise<@__device_2>
%22184 = torch_c.from_builtin_tensor %22183 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22185 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19893 = arith.constant 1 : index
%dim_19894 = tensor.dim %22185, %c1_19893 : tensor<4x?x4096xf16>
%22186 = flow.tensor.transfer %22185 : tensor<4x?x4096xf16>{%dim_19894} to #hal.device.promise<@__device_2>
%22187 = torch_c.from_builtin_tensor %22186 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22188 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19895 = arith.constant 1 : index
%dim_19896 = tensor.dim %22188, %c1_19895 : tensor<4x?x4096xf16>
%22189 = flow.tensor.transfer %22188 : tensor<4x?x4096xf16>{%dim_19896} to #hal.device.promise<@__device_2>
%22190 = torch_c.from_builtin_tensor %22189 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22191 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19897 = arith.constant 1 : index
%dim_19898 = tensor.dim %22191, %c1_19897 : tensor<4x?x4096xf16>
%22192 = flow.tensor.transfer %22191 : tensor<4x?x4096xf16>{%dim_19898} to #hal.device.promise<@__device_2>
%22193 = torch_c.from_builtin_tensor %22192 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22194 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19899 = arith.constant 1 : index
%dim_19900 = tensor.dim %22194, %c1_19899 : tensor<4x?x4096xf16>
%22195 = flow.tensor.transfer %22194 : tensor<4x?x4096xf16>{%dim_19900} to #hal.device.promise<@__device_2>
%22196 = torch_c.from_builtin_tensor %22195 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22197 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19901 = arith.constant 1 : index
%dim_19902 = tensor.dim %22197, %c1_19901 : tensor<4x?x4096xf16>
%22198 = flow.tensor.transfer %22197 : tensor<4x?x4096xf16>{%dim_19902} to #hal.device.promise<@__device_2>
%22199 = torch_c.from_builtin_tensor %22198 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19903 = torch.constant.int 1
%22200 = torch.aten.add.Tensor %22181, %22184, %int1_19903 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19904 = torch.constant.int 1
%22201 = torch.aten.add.Tensor %22200, %22092, %int1_19904 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19905 = torch.constant.int 1
%22202 = torch.aten.add.Tensor %22201, %22187, %int1_19905 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19906 = torch.constant.int 1
%22203 = torch.aten.add.Tensor %22202, %22190, %int1_19906 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19907 = torch.constant.int 1
%22204 = torch.aten.add.Tensor %22203, %22193, %int1_19907 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19908 = torch.constant.int 1
%22205 = torch.aten.add.Tensor %22204, %22196, %int1_19908 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19909 = torch.constant.int 1
%22206 = torch.aten.add.Tensor %22205, %22199, %int1_19909 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22207 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19910 = arith.constant 1 : index
%dim_19911 = tensor.dim %22207, %c1_19910 : tensor<4x?x4096xf16>
%22208 = flow.tensor.transfer %22207 : tensor<4x?x4096xf16>{%dim_19911} to #hal.device.promise<@__device_3>
%22209 = torch_c.from_builtin_tensor %22208 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22210 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19912 = arith.constant 1 : index
%dim_19913 = tensor.dim %22210, %c1_19912 : tensor<4x?x4096xf16>
%22211 = flow.tensor.transfer %22210 : tensor<4x?x4096xf16>{%dim_19913} to #hal.device.promise<@__device_3>
%22212 = torch_c.from_builtin_tensor %22211 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22213 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19914 = arith.constant 1 : index
%dim_19915 = tensor.dim %22213, %c1_19914 : tensor<4x?x4096xf16>
%22214 = flow.tensor.transfer %22213 : tensor<4x?x4096xf16>{%dim_19915} to #hal.device.promise<@__device_3>
%22215 = torch_c.from_builtin_tensor %22214 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22216 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19916 = arith.constant 1 : index
%dim_19917 = tensor.dim %22216, %c1_19916 : tensor<4x?x4096xf16>
%22217 = flow.tensor.transfer %22216 : tensor<4x?x4096xf16>{%dim_19917} to #hal.device.promise<@__device_3>
%22218 = torch_c.from_builtin_tensor %22217 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22219 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19918 = arith.constant 1 : index
%dim_19919 = tensor.dim %22219, %c1_19918 : tensor<4x?x4096xf16>
%22220 = flow.tensor.transfer %22219 : tensor<4x?x4096xf16>{%dim_19919} to #hal.device.promise<@__device_3>
%22221 = torch_c.from_builtin_tensor %22220 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22222 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19920 = arith.constant 1 : index
%dim_19921 = tensor.dim %22222, %c1_19920 : tensor<4x?x4096xf16>
%22223 = flow.tensor.transfer %22222 : tensor<4x?x4096xf16>{%dim_19921} to #hal.device.promise<@__device_3>
%22224 = torch_c.from_builtin_tensor %22223 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22225 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19922 = arith.constant 1 : index
%dim_19923 = tensor.dim %22225, %c1_19922 : tensor<4x?x4096xf16>
%22226 = flow.tensor.transfer %22225 : tensor<4x?x4096xf16>{%dim_19923} to #hal.device.promise<@__device_3>
%22227 = torch_c.from_builtin_tensor %22226 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19924 = torch.constant.int 1
%22228 = torch.aten.add.Tensor %22209, %22212, %int1_19924 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19925 = torch.constant.int 1
%22229 = torch.aten.add.Tensor %22228, %22215, %int1_19925 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19926 = torch.constant.int 1
%22230 = torch.aten.add.Tensor %22229, %22098, %int1_19926 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19927 = torch.constant.int 1
%22231 = torch.aten.add.Tensor %22230, %22218, %int1_19927 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19928 = torch.constant.int 1
%22232 = torch.aten.add.Tensor %22231, %22221, %int1_19928 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19929 = torch.constant.int 1
%22233 = torch.aten.add.Tensor %22232, %22224, %int1_19929 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19930 = torch.constant.int 1
%22234 = torch.aten.add.Tensor %22233, %22227, %int1_19930 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22235 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19931 = arith.constant 1 : index
%dim_19932 = tensor.dim %22235, %c1_19931 : tensor<4x?x4096xf16>
%22236 = flow.tensor.transfer %22235 : tensor<4x?x4096xf16>{%dim_19932} to #hal.device.promise<@__device_4>
%22237 = torch_c.from_builtin_tensor %22236 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22238 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19933 = arith.constant 1 : index
%dim_19934 = tensor.dim %22238, %c1_19933 : tensor<4x?x4096xf16>
%22239 = flow.tensor.transfer %22238 : tensor<4x?x4096xf16>{%dim_19934} to #hal.device.promise<@__device_4>
%22240 = torch_c.from_builtin_tensor %22239 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22241 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19935 = arith.constant 1 : index
%dim_19936 = tensor.dim %22241, %c1_19935 : tensor<4x?x4096xf16>
%22242 = flow.tensor.transfer %22241 : tensor<4x?x4096xf16>{%dim_19936} to #hal.device.promise<@__device_4>
%22243 = torch_c.from_builtin_tensor %22242 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22244 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19937 = arith.constant 1 : index
%dim_19938 = tensor.dim %22244, %c1_19937 : tensor<4x?x4096xf16>
%22245 = flow.tensor.transfer %22244 : tensor<4x?x4096xf16>{%dim_19938} to #hal.device.promise<@__device_4>
%22246 = torch_c.from_builtin_tensor %22245 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22247 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19939 = arith.constant 1 : index
%dim_19940 = tensor.dim %22247, %c1_19939 : tensor<4x?x4096xf16>
%22248 = flow.tensor.transfer %22247 : tensor<4x?x4096xf16>{%dim_19940} to #hal.device.promise<@__device_4>
%22249 = torch_c.from_builtin_tensor %22248 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22250 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19941 = arith.constant 1 : index
%dim_19942 = tensor.dim %22250, %c1_19941 : tensor<4x?x4096xf16>
%22251 = flow.tensor.transfer %22250 : tensor<4x?x4096xf16>{%dim_19942} to #hal.device.promise<@__device_4>
%22252 = torch_c.from_builtin_tensor %22251 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22253 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19943 = arith.constant 1 : index
%dim_19944 = tensor.dim %22253, %c1_19943 : tensor<4x?x4096xf16>
%22254 = flow.tensor.transfer %22253 : tensor<4x?x4096xf16>{%dim_19944} to #hal.device.promise<@__device_4>
%22255 = torch_c.from_builtin_tensor %22254 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19945 = torch.constant.int 1
%22256 = torch.aten.add.Tensor %22237, %22240, %int1_19945 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19946 = torch.constant.int 1
%22257 = torch.aten.add.Tensor %22256, %22243, %int1_19946 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19947 = torch.constant.int 1
%22258 = torch.aten.add.Tensor %22257, %22246, %int1_19947 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19948 = torch.constant.int 1
%22259 = torch.aten.add.Tensor %22258, %22104, %int1_19948 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19949 = torch.constant.int 1
%22260 = torch.aten.add.Tensor %22259, %22249, %int1_19949 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19950 = torch.constant.int 1
%22261 = torch.aten.add.Tensor %22260, %22252, %int1_19950 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19951 = torch.constant.int 1
%22262 = torch.aten.add.Tensor %22261, %22255, %int1_19951 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22263 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19952 = arith.constant 1 : index
%dim_19953 = tensor.dim %22263, %c1_19952 : tensor<4x?x4096xf16>
%22264 = flow.tensor.transfer %22263 : tensor<4x?x4096xf16>{%dim_19953} to #hal.device.promise<@__device_5>
%22265 = torch_c.from_builtin_tensor %22264 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22266 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19954 = arith.constant 1 : index
%dim_19955 = tensor.dim %22266, %c1_19954 : tensor<4x?x4096xf16>
%22267 = flow.tensor.transfer %22266 : tensor<4x?x4096xf16>{%dim_19955} to #hal.device.promise<@__device_5>
%22268 = torch_c.from_builtin_tensor %22267 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22269 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19956 = arith.constant 1 : index
%dim_19957 = tensor.dim %22269, %c1_19956 : tensor<4x?x4096xf16>
%22270 = flow.tensor.transfer %22269 : tensor<4x?x4096xf16>{%dim_19957} to #hal.device.promise<@__device_5>
%22271 = torch_c.from_builtin_tensor %22270 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22272 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19958 = arith.constant 1 : index
%dim_19959 = tensor.dim %22272, %c1_19958 : tensor<4x?x4096xf16>
%22273 = flow.tensor.transfer %22272 : tensor<4x?x4096xf16>{%dim_19959} to #hal.device.promise<@__device_5>
%22274 = torch_c.from_builtin_tensor %22273 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22275 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19960 = arith.constant 1 : index
%dim_19961 = tensor.dim %22275, %c1_19960 : tensor<4x?x4096xf16>
%22276 = flow.tensor.transfer %22275 : tensor<4x?x4096xf16>{%dim_19961} to #hal.device.promise<@__device_5>
%22277 = torch_c.from_builtin_tensor %22276 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22278 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19962 = arith.constant 1 : index
%dim_19963 = tensor.dim %22278, %c1_19962 : tensor<4x?x4096xf16>
%22279 = flow.tensor.transfer %22278 : tensor<4x?x4096xf16>{%dim_19963} to #hal.device.promise<@__device_5>
%22280 = torch_c.from_builtin_tensor %22279 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22281 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19964 = arith.constant 1 : index
%dim_19965 = tensor.dim %22281, %c1_19964 : tensor<4x?x4096xf16>
%22282 = flow.tensor.transfer %22281 : tensor<4x?x4096xf16>{%dim_19965} to #hal.device.promise<@__device_5>
%22283 = torch_c.from_builtin_tensor %22282 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19966 = torch.constant.int 1
%22284 = torch.aten.add.Tensor %22265, %22268, %int1_19966 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19967 = torch.constant.int 1
%22285 = torch.aten.add.Tensor %22284, %22271, %int1_19967 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19968 = torch.constant.int 1
%22286 = torch.aten.add.Tensor %22285, %22274, %int1_19968 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19969 = torch.constant.int 1
%22287 = torch.aten.add.Tensor %22286, %22277, %int1_19969 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19970 = torch.constant.int 1
%22288 = torch.aten.add.Tensor %22287, %22110, %int1_19970 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19971 = torch.constant.int 1
%22289 = torch.aten.add.Tensor %22288, %22280, %int1_19971 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19972 = torch.constant.int 1
%22290 = torch.aten.add.Tensor %22289, %22283, %int1_19972 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22291 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19973 = arith.constant 1 : index
%dim_19974 = tensor.dim %22291, %c1_19973 : tensor<4x?x4096xf16>
%22292 = flow.tensor.transfer %22291 : tensor<4x?x4096xf16>{%dim_19974} to #hal.device.promise<@__device_6>
%22293 = torch_c.from_builtin_tensor %22292 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22294 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19975 = arith.constant 1 : index
%dim_19976 = tensor.dim %22294, %c1_19975 : tensor<4x?x4096xf16>
%22295 = flow.tensor.transfer %22294 : tensor<4x?x4096xf16>{%dim_19976} to #hal.device.promise<@__device_6>
%22296 = torch_c.from_builtin_tensor %22295 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22297 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19977 = arith.constant 1 : index
%dim_19978 = tensor.dim %22297, %c1_19977 : tensor<4x?x4096xf16>
%22298 = flow.tensor.transfer %22297 : tensor<4x?x4096xf16>{%dim_19978} to #hal.device.promise<@__device_6>
%22299 = torch_c.from_builtin_tensor %22298 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22300 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19979 = arith.constant 1 : index
%dim_19980 = tensor.dim %22300, %c1_19979 : tensor<4x?x4096xf16>
%22301 = flow.tensor.transfer %22300 : tensor<4x?x4096xf16>{%dim_19980} to #hal.device.promise<@__device_6>
%22302 = torch_c.from_builtin_tensor %22301 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22303 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19981 = arith.constant 1 : index
%dim_19982 = tensor.dim %22303, %c1_19981 : tensor<4x?x4096xf16>
%22304 = flow.tensor.transfer %22303 : tensor<4x?x4096xf16>{%dim_19982} to #hal.device.promise<@__device_6>
%22305 = torch_c.from_builtin_tensor %22304 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22306 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19983 = arith.constant 1 : index
%dim_19984 = tensor.dim %22306, %c1_19983 : tensor<4x?x4096xf16>
%22307 = flow.tensor.transfer %22306 : tensor<4x?x4096xf16>{%dim_19984} to #hal.device.promise<@__device_6>
%22308 = torch_c.from_builtin_tensor %22307 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22309 = torch_c.to_builtin_tensor %22122 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19985 = arith.constant 1 : index
%dim_19986 = tensor.dim %22309, %c1_19985 : tensor<4x?x4096xf16>
%22310 = flow.tensor.transfer %22309 : tensor<4x?x4096xf16>{%dim_19986} to #hal.device.promise<@__device_6>
%22311 = torch_c.from_builtin_tensor %22310 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19987 = torch.constant.int 1
%22312 = torch.aten.add.Tensor %22293, %22296, %int1_19987 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19988 = torch.constant.int 1
%22313 = torch.aten.add.Tensor %22312, %22299, %int1_19988 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19989 = torch.constant.int 1
%22314 = torch.aten.add.Tensor %22313, %22302, %int1_19989 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19990 = torch.constant.int 1
%22315 = torch.aten.add.Tensor %22314, %22305, %int1_19990 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19991 = torch.constant.int 1
%22316 = torch.aten.add.Tensor %22315, %22308, %int1_19991 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19992 = torch.constant.int 1
%22317 = torch.aten.add.Tensor %22316, %22116, %int1_19992 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_19993 = torch.constant.int 1
%22318 = torch.aten.add.Tensor %22317, %22311, %int1_19993 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22319 = torch_c.to_builtin_tensor %22080 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19994 = arith.constant 1 : index
%dim_19995 = tensor.dim %22319, %c1_19994 : tensor<4x?x4096xf16>
%22320 = flow.tensor.transfer %22319 : tensor<4x?x4096xf16>{%dim_19995} to #hal.device.promise<@__device_7>
%22321 = torch_c.from_builtin_tensor %22320 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22322 = torch_c.to_builtin_tensor %22086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19996 = arith.constant 1 : index
%dim_19997 = tensor.dim %22322, %c1_19996 : tensor<4x?x4096xf16>
%22323 = flow.tensor.transfer %22322 : tensor<4x?x4096xf16>{%dim_19997} to #hal.device.promise<@__device_7>
%22324 = torch_c.from_builtin_tensor %22323 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22325 = torch_c.to_builtin_tensor %22092 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_19998 = arith.constant 1 : index
%dim_19999 = tensor.dim %22325, %c1_19998 : tensor<4x?x4096xf16>
%22326 = flow.tensor.transfer %22325 : tensor<4x?x4096xf16>{%dim_19999} to #hal.device.promise<@__device_7>
%22327 = torch_c.from_builtin_tensor %22326 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22328 = torch_c.to_builtin_tensor %22098 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20000 = arith.constant 1 : index
%dim_20001 = tensor.dim %22328, %c1_20000 : tensor<4x?x4096xf16>
%22329 = flow.tensor.transfer %22328 : tensor<4x?x4096xf16>{%dim_20001} to #hal.device.promise<@__device_7>
%22330 = torch_c.from_builtin_tensor %22329 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22331 = torch_c.to_builtin_tensor %22104 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20002 = arith.constant 1 : index
%dim_20003 = tensor.dim %22331, %c1_20002 : tensor<4x?x4096xf16>
%22332 = flow.tensor.transfer %22331 : tensor<4x?x4096xf16>{%dim_20003} to #hal.device.promise<@__device_7>
%22333 = torch_c.from_builtin_tensor %22332 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22334 = torch_c.to_builtin_tensor %22110 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20004 = arith.constant 1 : index
%dim_20005 = tensor.dim %22334, %c1_20004 : tensor<4x?x4096xf16>
%22335 = flow.tensor.transfer %22334 : tensor<4x?x4096xf16>{%dim_20005} to #hal.device.promise<@__device_7>
%22336 = torch_c.from_builtin_tensor %22335 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22337 = torch_c.to_builtin_tensor %22116 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20006 = arith.constant 1 : index
%dim_20007 = tensor.dim %22337, %c1_20006 : tensor<4x?x4096xf16>
%22338 = flow.tensor.transfer %22337 : tensor<4x?x4096xf16>{%dim_20007} to #hal.device.promise<@__device_7>
%22339 = torch_c.from_builtin_tensor %22338 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20008 = torch.constant.int 1
%22340 = torch.aten.add.Tensor %22321, %22324, %int1_20008 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20009 = torch.constant.int 1
%22341 = torch.aten.add.Tensor %22340, %22327, %int1_20009 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20010 = torch.constant.int 1
%22342 = torch.aten.add.Tensor %22341, %22330, %int1_20010 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20011 = torch.constant.int 1
%22343 = torch.aten.add.Tensor %22342, %22333, %int1_20011 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20012 = torch.constant.int 1
%22344 = torch.aten.add.Tensor %22343, %22336, %int1_20012 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20013 = torch.constant.int 1
%22345 = torch.aten.add.Tensor %22344, %22339, %int1_20013 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20014 = torch.constant.int 1
%22346 = torch.aten.add.Tensor %22345, %22122, %int1_20014 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20015 = torch.constant.int 1
%22347 = torch.aten.add.Tensor %21006, %22150, %int1_20015 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20016 = torch.constant.int 1
%22348 = torch.aten.add.Tensor %21007, %22178, %int1_20016 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20017 = torch.constant.int 1
%22349 = torch.aten.add.Tensor %21008, %22206, %int1_20017 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20018 = torch.constant.int 1
%22350 = torch.aten.add.Tensor %21009, %22234, %int1_20018 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20019 = torch.constant.int 1
%22351 = torch.aten.add.Tensor %21010, %22262, %int1_20019 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20020 = torch.constant.int 1
%22352 = torch.aten.add.Tensor %21011, %22290, %int1_20020 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20021 = torch.constant.int 1
%22353 = torch.aten.add.Tensor %21012, %22318, %int1_20021 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20022 = torch.constant.int 1
%22354 = torch.aten.add.Tensor %21013, %22346, %int1_20022 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_20023 = torch.constant.int 6
%22355 = torch.prims.convert_element_type %22347, %int6_20023 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20024 = torch.constant.int 6
%22356 = torch.prims.convert_element_type %22348, %int6_20024 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20025 = torch.constant.int 6
%22357 = torch.prims.convert_element_type %22349, %int6_20025 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20026 = torch.constant.int 6
%22358 = torch.prims.convert_element_type %22350, %int6_20026 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20027 = torch.constant.int 6
%22359 = torch.prims.convert_element_type %22351, %int6_20027 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20028 = torch.constant.int 6
%22360 = torch.prims.convert_element_type %22352, %int6_20028 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20029 = torch.constant.int 6
%22361 = torch.prims.convert_element_type %22353, %int6_20029 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20030 = torch.constant.int 6
%22362 = torch.prims.convert_element_type %22354, %int6_20030 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20031 = torch.constant.int 2
%22363 = torch.aten.pow.Tensor_Scalar %22355, %int2_20031 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20032 = torch.constant.int 2
%22364 = torch.aten.pow.Tensor_Scalar %22356, %int2_20032 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20033 = torch.constant.int 2
%22365 = torch.aten.pow.Tensor_Scalar %22357, %int2_20033 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20034 = torch.constant.int 2
%22366 = torch.aten.pow.Tensor_Scalar %22358, %int2_20034 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20035 = torch.constant.int 2
%22367 = torch.aten.pow.Tensor_Scalar %22359, %int2_20035 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20036 = torch.constant.int 2
%22368 = torch.aten.pow.Tensor_Scalar %22360, %int2_20036 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20037 = torch.constant.int 2
%22369 = torch.aten.pow.Tensor_Scalar %22361, %int2_20037 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20038 = torch.constant.int 2
%22370 = torch.aten.pow.Tensor_Scalar %22362, %int2_20038 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_20039 = torch.constant.int -1
%22371 = torch.prim.ListConstruct %int-1_20039 : (!torch.int) -> !torch.list<int>
%true_20040 = torch.constant.bool true
%none_20041 = torch.constant.none
%22372 = torch.aten.mean.dim %22363, %22371, %true_20040, %none_20041 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20042 = torch.constant.int -1
%22373 = torch.prim.ListConstruct %int-1_20042 : (!torch.int) -> !torch.list<int>
%true_20043 = torch.constant.bool true
%none_20044 = torch.constant.none
%22374 = torch.aten.mean.dim %22364, %22373, %true_20043, %none_20044 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20045 = torch.constant.int -1
%22375 = torch.prim.ListConstruct %int-1_20045 : (!torch.int) -> !torch.list<int>
%true_20046 = torch.constant.bool true
%none_20047 = torch.constant.none
%22376 = torch.aten.mean.dim %22365, %22375, %true_20046, %none_20047 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20048 = torch.constant.int -1
%22377 = torch.prim.ListConstruct %int-1_20048 : (!torch.int) -> !torch.list<int>
%true_20049 = torch.constant.bool true
%none_20050 = torch.constant.none
%22378 = torch.aten.mean.dim %22366, %22377, %true_20049, %none_20050 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20051 = torch.constant.int -1
%22379 = torch.prim.ListConstruct %int-1_20051 : (!torch.int) -> !torch.list<int>
%true_20052 = torch.constant.bool true
%none_20053 = torch.constant.none
%22380 = torch.aten.mean.dim %22367, %22379, %true_20052, %none_20053 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20054 = torch.constant.int -1
%22381 = torch.prim.ListConstruct %int-1_20054 : (!torch.int) -> !torch.list<int>
%true_20055 = torch.constant.bool true
%none_20056 = torch.constant.none
%22382 = torch.aten.mean.dim %22368, %22381, %true_20055, %none_20056 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20057 = torch.constant.int -1
%22383 = torch.prim.ListConstruct %int-1_20057 : (!torch.int) -> !torch.list<int>
%true_20058 = torch.constant.bool true
%none_20059 = torch.constant.none
%22384 = torch.aten.mean.dim %22369, %22383, %true_20058, %none_20059 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20060 = torch.constant.int -1
%22385 = torch.prim.ListConstruct %int-1_20060 : (!torch.int) -> !torch.list<int>
%true_20061 = torch.constant.bool true
%none_20062 = torch.constant.none
%22386 = torch.aten.mean.dim %22370, %22385, %true_20061, %none_20062 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20063 = torch.constant.float 9.9999997473787516E-6
%int1_20064 = torch.constant.int 1
%22387 = torch.aten.add.Scalar %22372, %float9.999990e-06_20063, %int1_20064 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20065 = torch.constant.float 9.9999997473787516E-6
%int1_20066 = torch.constant.int 1
%22388 = torch.aten.add.Scalar %22374, %float9.999990e-06_20065, %int1_20066 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20067 = torch.constant.float 9.9999997473787516E-6
%int1_20068 = torch.constant.int 1
%22389 = torch.aten.add.Scalar %22376, %float9.999990e-06_20067, %int1_20068 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20069 = torch.constant.float 9.9999997473787516E-6
%int1_20070 = torch.constant.int 1
%22390 = torch.aten.add.Scalar %22378, %float9.999990e-06_20069, %int1_20070 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20071 = torch.constant.float 9.9999997473787516E-6
%int1_20072 = torch.constant.int 1
%22391 = torch.aten.add.Scalar %22380, %float9.999990e-06_20071, %int1_20072 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20073 = torch.constant.float 9.9999997473787516E-6
%int1_20074 = torch.constant.int 1
%22392 = torch.aten.add.Scalar %22382, %float9.999990e-06_20073, %int1_20074 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20075 = torch.constant.float 9.9999997473787516E-6
%int1_20076 = torch.constant.int 1
%22393 = torch.aten.add.Scalar %22384, %float9.999990e-06_20075, %int1_20076 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20077 = torch.constant.float 9.9999997473787516E-6
%int1_20078 = torch.constant.int 1
%22394 = torch.aten.add.Scalar %22386, %float9.999990e-06_20077, %int1_20078 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22395 = torch.aten.rsqrt %22387 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22396 = torch.aten.rsqrt %22388 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22397 = torch.aten.rsqrt %22389 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22398 = torch.aten.rsqrt %22390 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22399 = torch.aten.rsqrt %22391 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22400 = torch.aten.rsqrt %22392 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22401 = torch.aten.rsqrt %22393 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22402 = torch.aten.rsqrt %22394 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22402, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22403 = torch.aten.mul.Tensor %22355, %22395 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22404 = torch.aten.mul.Tensor %22356, %22396 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22405 = torch.aten.mul.Tensor %22357, %22397 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22406 = torch.aten.mul.Tensor %22358, %22398 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22407 = torch.aten.mul.Tensor %22359, %22399 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22408 = torch.aten.mul.Tensor %22360, %22400 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22409 = torch.aten.mul.Tensor %22361, %22401 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22410 = torch.aten.mul.Tensor %22362, %22402 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22411 = torch.aten.mul.Tensor %768, %22403 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22412 = torch.aten.mul.Tensor %769, %22404 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22413 = torch.aten.mul.Tensor %770, %22405 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22414 = torch.aten.mul.Tensor %771, %22406 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22415 = torch.aten.mul.Tensor %772, %22407 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22416 = torch.aten.mul.Tensor %773, %22408 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22417 = torch.aten.mul.Tensor %774, %22409 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22418 = torch.aten.mul.Tensor %775, %22410 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_20079 = torch.constant.int 5
%22419 = torch.prims.convert_element_type %22411, %int5_20079 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20080 = torch.constant.int 5
%22420 = torch.prims.convert_element_type %22412, %int5_20080 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20081 = torch.constant.int 5
%22421 = torch.prims.convert_element_type %22413, %int5_20081 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20082 = torch.constant.int 5
%22422 = torch.prims.convert_element_type %22414, %int5_20082 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20083 = torch.constant.int 5
%22423 = torch.prims.convert_element_type %22415, %int5_20083 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20084 = torch.constant.int 5
%22424 = torch.prims.convert_element_type %22416, %int5_20084 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20085 = torch.constant.int 5
%22425 = torch.prims.convert_element_type %22417, %int5_20085 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20086 = torch.constant.int 5
%22426 = torch.prims.convert_element_type %22418, %int5_20086 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20087 = torch.constant.int 1
%int0_20088 = torch.constant.int 0
%22427 = torch.prim.ListConstruct %int1_20087, %int0_20088 : (!torch.int, !torch.int) -> !torch.list<int>
%22428 = torch.aten.permute %776, %22427 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20089 = torch.constant.int 1
%int0_20090 = torch.constant.int 0
%22429 = torch.prim.ListConstruct %int1_20089, %int0_20090 : (!torch.int, !torch.int) -> !torch.list<int>
%22430 = torch.aten.permute %777, %22429 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20091 = torch.constant.int 1
%int0_20092 = torch.constant.int 0
%22431 = torch.prim.ListConstruct %int1_20091, %int0_20092 : (!torch.int, !torch.int) -> !torch.list<int>
%22432 = torch.aten.permute %778, %22431 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20093 = torch.constant.int 1
%int0_20094 = torch.constant.int 0
%22433 = torch.prim.ListConstruct %int1_20093, %int0_20094 : (!torch.int, !torch.int) -> !torch.list<int>
%22434 = torch.aten.permute %779, %22433 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20095 = torch.constant.int 1
%int0_20096 = torch.constant.int 0
%22435 = torch.prim.ListConstruct %int1_20095, %int0_20096 : (!torch.int, !torch.int) -> !torch.list<int>
%22436 = torch.aten.permute %780, %22435 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20097 = torch.constant.int 1
%int0_20098 = torch.constant.int 0
%22437 = torch.prim.ListConstruct %int1_20097, %int0_20098 : (!torch.int, !torch.int) -> !torch.list<int>
%22438 = torch.aten.permute %781, %22437 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20099 = torch.constant.int 1
%int0_20100 = torch.constant.int 0
%22439 = torch.prim.ListConstruct %int1_20099, %int0_20100 : (!torch.int, !torch.int) -> !torch.list<int>
%22440 = torch.aten.permute %782, %22439 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20101 = torch.constant.int 1
%int0_20102 = torch.constant.int 0
%22441 = torch.prim.ListConstruct %int1_20101, %int0_20102 : (!torch.int, !torch.int) -> !torch.list<int>
%22442 = torch.aten.permute %783, %22441 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_20103 = torch.constant.int 4
%22443 = torch.aten.mul.int %int4_20103, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20104 = torch.constant.int 4096
%22444 = torch.prim.ListConstruct %22443, %int4096_20104 : (!torch.int, !torch.int) -> !torch.list<int>
%22445 = torch.aten.view %22419, %22444 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22445, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22446 = torch.aten.mm %22445, %22428 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22446, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20105 = torch.constant.int 4
%int1792_20106 = torch.constant.int 1792
%22447 = torch.prim.ListConstruct %int4_20105, %2482, %int1792_20106 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22448 = torch.aten.view %22446, %22447 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20107 = torch.constant.int 4
%22449 = torch.aten.mul.int %int4_20107, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20108 = torch.constant.int 4096
%22450 = torch.prim.ListConstruct %22449, %int4096_20108 : (!torch.int, !torch.int) -> !torch.list<int>
%22451 = torch.aten.view %22420, %22450 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22451, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22452 = torch.aten.mm %22451, %22430 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22452, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20109 = torch.constant.int 4
%int1792_20110 = torch.constant.int 1792
%22453 = torch.prim.ListConstruct %int4_20109, %2482, %int1792_20110 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22454 = torch.aten.view %22452, %22453 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20111 = torch.constant.int 4
%22455 = torch.aten.mul.int %int4_20111, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20112 = torch.constant.int 4096
%22456 = torch.prim.ListConstruct %22455, %int4096_20112 : (!torch.int, !torch.int) -> !torch.list<int>
%22457 = torch.aten.view %22421, %22456 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22457, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22458 = torch.aten.mm %22457, %22432 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22458, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20113 = torch.constant.int 4
%int1792_20114 = torch.constant.int 1792
%22459 = torch.prim.ListConstruct %int4_20113, %2482, %int1792_20114 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22460 = torch.aten.view %22458, %22459 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20115 = torch.constant.int 4
%22461 = torch.aten.mul.int %int4_20115, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20116 = torch.constant.int 4096
%22462 = torch.prim.ListConstruct %22461, %int4096_20116 : (!torch.int, !torch.int) -> !torch.list<int>
%22463 = torch.aten.view %22422, %22462 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22463, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22464 = torch.aten.mm %22463, %22434 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22464, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20117 = torch.constant.int 4
%int1792_20118 = torch.constant.int 1792
%22465 = torch.prim.ListConstruct %int4_20117, %2482, %int1792_20118 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22466 = torch.aten.view %22464, %22465 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20119 = torch.constant.int 4
%22467 = torch.aten.mul.int %int4_20119, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20120 = torch.constant.int 4096
%22468 = torch.prim.ListConstruct %22467, %int4096_20120 : (!torch.int, !torch.int) -> !torch.list<int>
%22469 = torch.aten.view %22423, %22468 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22469, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22470 = torch.aten.mm %22469, %22436 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22470, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20121 = torch.constant.int 4
%int1792_20122 = torch.constant.int 1792
%22471 = torch.prim.ListConstruct %int4_20121, %2482, %int1792_20122 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22472 = torch.aten.view %22470, %22471 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20123 = torch.constant.int 4
%22473 = torch.aten.mul.int %int4_20123, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20124 = torch.constant.int 4096
%22474 = torch.prim.ListConstruct %22473, %int4096_20124 : (!torch.int, !torch.int) -> !torch.list<int>
%22475 = torch.aten.view %22424, %22474 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22475, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22476 = torch.aten.mm %22475, %22438 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22476, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20125 = torch.constant.int 4
%int1792_20126 = torch.constant.int 1792
%22477 = torch.prim.ListConstruct %int4_20125, %2482, %int1792_20126 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22478 = torch.aten.view %22476, %22477 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20127 = torch.constant.int 4
%22479 = torch.aten.mul.int %int4_20127, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20128 = torch.constant.int 4096
%22480 = torch.prim.ListConstruct %22479, %int4096_20128 : (!torch.int, !torch.int) -> !torch.list<int>
%22481 = torch.aten.view %22425, %22480 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22481, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22482 = torch.aten.mm %22481, %22440 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22482, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20129 = torch.constant.int 4
%int1792_20130 = torch.constant.int 1792
%22483 = torch.prim.ListConstruct %int4_20129, %2482, %int1792_20130 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22484 = torch.aten.view %22482, %22483 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20131 = torch.constant.int 4
%22485 = torch.aten.mul.int %int4_20131, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20132 = torch.constant.int 4096
%22486 = torch.prim.ListConstruct %22485, %int4096_20132 : (!torch.int, !torch.int) -> !torch.list<int>
%22487 = torch.aten.view %22426, %22486 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22487, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22488 = torch.aten.mm %22487, %22442 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22488, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20133 = torch.constant.int 4
%int1792_20134 = torch.constant.int 1792
%22489 = torch.prim.ListConstruct %int4_20133, %2482, %int1792_20134 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22490 = torch.aten.view %22488, %22489 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22491 = torch.aten.silu %22448 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22492 = torch.aten.silu %22454 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22493 = torch.aten.silu %22460 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22494 = torch.aten.silu %22466 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22495 = torch.aten.silu %22472 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22496 = torch.aten.silu %22478 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22497 = torch.aten.silu %22484 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22498 = torch.aten.silu %22490 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_20135 = torch.constant.int 1
%int0_20136 = torch.constant.int 0
%22499 = torch.prim.ListConstruct %int1_20135, %int0_20136 : (!torch.int, !torch.int) -> !torch.list<int>
%22500 = torch.aten.permute %784, %22499 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20137 = torch.constant.int 1
%int0_20138 = torch.constant.int 0
%22501 = torch.prim.ListConstruct %int1_20137, %int0_20138 : (!torch.int, !torch.int) -> !torch.list<int>
%22502 = torch.aten.permute %785, %22501 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20139 = torch.constant.int 1
%int0_20140 = torch.constant.int 0
%22503 = torch.prim.ListConstruct %int1_20139, %int0_20140 : (!torch.int, !torch.int) -> !torch.list<int>
%22504 = torch.aten.permute %786, %22503 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20141 = torch.constant.int 1
%int0_20142 = torch.constant.int 0
%22505 = torch.prim.ListConstruct %int1_20141, %int0_20142 : (!torch.int, !torch.int) -> !torch.list<int>
%22506 = torch.aten.permute %787, %22505 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20143 = torch.constant.int 1
%int0_20144 = torch.constant.int 0
%22507 = torch.prim.ListConstruct %int1_20143, %int0_20144 : (!torch.int, !torch.int) -> !torch.list<int>
%22508 = torch.aten.permute %788, %22507 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20145 = torch.constant.int 1
%int0_20146 = torch.constant.int 0
%22509 = torch.prim.ListConstruct %int1_20145, %int0_20146 : (!torch.int, !torch.int) -> !torch.list<int>
%22510 = torch.aten.permute %789, %22509 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20147 = torch.constant.int 1
%int0_20148 = torch.constant.int 0
%22511 = torch.prim.ListConstruct %int1_20147, %int0_20148 : (!torch.int, !torch.int) -> !torch.list<int>
%22512 = torch.aten.permute %790, %22511 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_20149 = torch.constant.int 1
%int0_20150 = torch.constant.int 0
%22513 = torch.prim.ListConstruct %int1_20149, %int0_20150 : (!torch.int, !torch.int) -> !torch.list<int>
%22514 = torch.aten.permute %791, %22513 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_20151 = torch.constant.int 4
%22515 = torch.aten.mul.int %int4_20151, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20152 = torch.constant.int 4096
%22516 = torch.prim.ListConstruct %22515, %int4096_20152 : (!torch.int, !torch.int) -> !torch.list<int>
%22517 = torch.aten.view %22419, %22516 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22517, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22518 = torch.aten.mm %22517, %22500 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22518, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20153 = torch.constant.int 4
%int1792_20154 = torch.constant.int 1792
%22519 = torch.prim.ListConstruct %int4_20153, %2482, %int1792_20154 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22520 = torch.aten.view %22518, %22519 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20155 = torch.constant.int 4
%22521 = torch.aten.mul.int %int4_20155, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20156 = torch.constant.int 4096
%22522 = torch.prim.ListConstruct %22521, %int4096_20156 : (!torch.int, !torch.int) -> !torch.list<int>
%22523 = torch.aten.view %22420, %22522 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22523, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22524 = torch.aten.mm %22523, %22502 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22524, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20157 = torch.constant.int 4
%int1792_20158 = torch.constant.int 1792
%22525 = torch.prim.ListConstruct %int4_20157, %2482, %int1792_20158 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22526 = torch.aten.view %22524, %22525 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20159 = torch.constant.int 4
%22527 = torch.aten.mul.int %int4_20159, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20160 = torch.constant.int 4096
%22528 = torch.prim.ListConstruct %22527, %int4096_20160 : (!torch.int, !torch.int) -> !torch.list<int>
%22529 = torch.aten.view %22421, %22528 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22529, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22530 = torch.aten.mm %22529, %22504 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22530, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20161 = torch.constant.int 4
%int1792_20162 = torch.constant.int 1792
%22531 = torch.prim.ListConstruct %int4_20161, %2482, %int1792_20162 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22532 = torch.aten.view %22530, %22531 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20163 = torch.constant.int 4
%22533 = torch.aten.mul.int %int4_20163, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20164 = torch.constant.int 4096
%22534 = torch.prim.ListConstruct %22533, %int4096_20164 : (!torch.int, !torch.int) -> !torch.list<int>
%22535 = torch.aten.view %22422, %22534 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22535, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22536 = torch.aten.mm %22535, %22506 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22536, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20165 = torch.constant.int 4
%int1792_20166 = torch.constant.int 1792
%22537 = torch.prim.ListConstruct %int4_20165, %2482, %int1792_20166 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22538 = torch.aten.view %22536, %22537 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20167 = torch.constant.int 4
%22539 = torch.aten.mul.int %int4_20167, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20168 = torch.constant.int 4096
%22540 = torch.prim.ListConstruct %22539, %int4096_20168 : (!torch.int, !torch.int) -> !torch.list<int>
%22541 = torch.aten.view %22423, %22540 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22541, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22542 = torch.aten.mm %22541, %22508 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22542, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20169 = torch.constant.int 4
%int1792_20170 = torch.constant.int 1792
%22543 = torch.prim.ListConstruct %int4_20169, %2482, %int1792_20170 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22544 = torch.aten.view %22542, %22543 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20171 = torch.constant.int 4
%22545 = torch.aten.mul.int %int4_20171, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20172 = torch.constant.int 4096
%22546 = torch.prim.ListConstruct %22545, %int4096_20172 : (!torch.int, !torch.int) -> !torch.list<int>
%22547 = torch.aten.view %22424, %22546 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22547, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22548 = torch.aten.mm %22547, %22510 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22548, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20173 = torch.constant.int 4
%int1792_20174 = torch.constant.int 1792
%22549 = torch.prim.ListConstruct %int4_20173, %2482, %int1792_20174 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22550 = torch.aten.view %22548, %22549 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20175 = torch.constant.int 4
%22551 = torch.aten.mul.int %int4_20175, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20176 = torch.constant.int 4096
%22552 = torch.prim.ListConstruct %22551, %int4096_20176 : (!torch.int, !torch.int) -> !torch.list<int>
%22553 = torch.aten.view %22425, %22552 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22553, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22554 = torch.aten.mm %22553, %22512 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22554, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20177 = torch.constant.int 4
%int1792_20178 = torch.constant.int 1792
%22555 = torch.prim.ListConstruct %int4_20177, %2482, %int1792_20178 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22556 = torch.aten.view %22554, %22555 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_20179 = torch.constant.int 4
%22557 = torch.aten.mul.int %int4_20179, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20180 = torch.constant.int 4096
%22558 = torch.prim.ListConstruct %22557, %int4096_20180 : (!torch.int, !torch.int) -> !torch.list<int>
%22559 = torch.aten.view %22426, %22558 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22559, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22560 = torch.aten.mm %22559, %22514 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22560, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_20181 = torch.constant.int 4
%int1792_20182 = torch.constant.int 1792
%22561 = torch.prim.ListConstruct %int4_20181, %2482, %int1792_20182 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22562 = torch.aten.view %22560, %22561 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22563 = torch.aten.mul.Tensor %22491, %22520 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22564 = torch.aten.mul.Tensor %22492, %22526 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22565 = torch.aten.mul.Tensor %22493, %22532 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22566 = torch.aten.mul.Tensor %22494, %22538 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22567 = torch.aten.mul.Tensor %22495, %22544 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22568 = torch.aten.mul.Tensor %22496, %22550 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22569 = torch.aten.mul.Tensor %22497, %22556 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%22570 = torch.aten.mul.Tensor %22498, %22562 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %22570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_20183 = torch.constant.int 1
%int0_20184 = torch.constant.int 0
%22571 = torch.prim.ListConstruct %int1_20183, %int0_20184 : (!torch.int, !torch.int) -> !torch.list<int>
%22572 = torch.aten.permute %792, %22571 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20185 = torch.constant.int 1
%int0_20186 = torch.constant.int 0
%22573 = torch.prim.ListConstruct %int1_20185, %int0_20186 : (!torch.int, !torch.int) -> !torch.list<int>
%22574 = torch.aten.permute %793, %22573 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20187 = torch.constant.int 1
%int0_20188 = torch.constant.int 0
%22575 = torch.prim.ListConstruct %int1_20187, %int0_20188 : (!torch.int, !torch.int) -> !torch.list<int>
%22576 = torch.aten.permute %794, %22575 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20189 = torch.constant.int 1
%int0_20190 = torch.constant.int 0
%22577 = torch.prim.ListConstruct %int1_20189, %int0_20190 : (!torch.int, !torch.int) -> !torch.list<int>
%22578 = torch.aten.permute %795, %22577 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20191 = torch.constant.int 1
%int0_20192 = torch.constant.int 0
%22579 = torch.prim.ListConstruct %int1_20191, %int0_20192 : (!torch.int, !torch.int) -> !torch.list<int>
%22580 = torch.aten.permute %796, %22579 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20193 = torch.constant.int 1
%int0_20194 = torch.constant.int 0
%22581 = torch.prim.ListConstruct %int1_20193, %int0_20194 : (!torch.int, !torch.int) -> !torch.list<int>
%22582 = torch.aten.permute %797, %22581 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20195 = torch.constant.int 1
%int0_20196 = torch.constant.int 0
%22583 = torch.prim.ListConstruct %int1_20195, %int0_20196 : (!torch.int, !torch.int) -> !torch.list<int>
%22584 = torch.aten.permute %798, %22583 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20197 = torch.constant.int 1
%int0_20198 = torch.constant.int 0
%22585 = torch.prim.ListConstruct %int1_20197, %int0_20198 : (!torch.int, !torch.int) -> !torch.list<int>
%22586 = torch.aten.permute %799, %22585 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_20199 = torch.constant.int 1
%22587 = torch.aten.size.int %22448, %int1_20199 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20200 = torch.constant.int 4
%22588 = torch.aten.mul.int %int4_20200, %22587 : !torch.int, !torch.int -> !torch.int
%int1792_20201 = torch.constant.int 1792
%22589 = torch.prim.ListConstruct %22588, %int1792_20201 : (!torch.int, !torch.int) -> !torch.list<int>
%22590 = torch.aten.view %22563, %22589 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22590, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22591 = torch.aten.mm %22590, %22572 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22591, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20202 = torch.constant.int 4
%int4096_20203 = torch.constant.int 4096
%22592 = torch.prim.ListConstruct %int4_20202, %22587, %int4096_20203 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22593 = torch.aten.view %22591, %22592 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20204 = torch.constant.int 1
%22594 = torch.aten.size.int %22454, %int1_20204 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20205 = torch.constant.int 4
%22595 = torch.aten.mul.int %int4_20205, %22594 : !torch.int, !torch.int -> !torch.int
%int1792_20206 = torch.constant.int 1792
%22596 = torch.prim.ListConstruct %22595, %int1792_20206 : (!torch.int, !torch.int) -> !torch.list<int>
%22597 = torch.aten.view %22564, %22596 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22597, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22598 = torch.aten.mm %22597, %22574 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22598, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20207 = torch.constant.int 4
%int4096_20208 = torch.constant.int 4096
%22599 = torch.prim.ListConstruct %int4_20207, %22594, %int4096_20208 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22600 = torch.aten.view %22598, %22599 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20209 = torch.constant.int 1
%22601 = torch.aten.size.int %22460, %int1_20209 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20210 = torch.constant.int 4
%22602 = torch.aten.mul.int %int4_20210, %22601 : !torch.int, !torch.int -> !torch.int
%int1792_20211 = torch.constant.int 1792
%22603 = torch.prim.ListConstruct %22602, %int1792_20211 : (!torch.int, !torch.int) -> !torch.list<int>
%22604 = torch.aten.view %22565, %22603 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22604, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22605 = torch.aten.mm %22604, %22576 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22605, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20212 = torch.constant.int 4
%int4096_20213 = torch.constant.int 4096
%22606 = torch.prim.ListConstruct %int4_20212, %22601, %int4096_20213 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22607 = torch.aten.view %22605, %22606 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20214 = torch.constant.int 1
%22608 = torch.aten.size.int %22466, %int1_20214 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20215 = torch.constant.int 4
%22609 = torch.aten.mul.int %int4_20215, %22608 : !torch.int, !torch.int -> !torch.int
%int1792_20216 = torch.constant.int 1792
%22610 = torch.prim.ListConstruct %22609, %int1792_20216 : (!torch.int, !torch.int) -> !torch.list<int>
%22611 = torch.aten.view %22566, %22610 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22611, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22612 = torch.aten.mm %22611, %22578 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22612, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20217 = torch.constant.int 4
%int4096_20218 = torch.constant.int 4096
%22613 = torch.prim.ListConstruct %int4_20217, %22608, %int4096_20218 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22614 = torch.aten.view %22612, %22613 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20219 = torch.constant.int 1
%22615 = torch.aten.size.int %22472, %int1_20219 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20220 = torch.constant.int 4
%22616 = torch.aten.mul.int %int4_20220, %22615 : !torch.int, !torch.int -> !torch.int
%int1792_20221 = torch.constant.int 1792
%22617 = torch.prim.ListConstruct %22616, %int1792_20221 : (!torch.int, !torch.int) -> !torch.list<int>
%22618 = torch.aten.view %22567, %22617 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22618, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22619 = torch.aten.mm %22618, %22580 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22619, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20222 = torch.constant.int 4
%int4096_20223 = torch.constant.int 4096
%22620 = torch.prim.ListConstruct %int4_20222, %22615, %int4096_20223 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22621 = torch.aten.view %22619, %22620 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20224 = torch.constant.int 1
%22622 = torch.aten.size.int %22478, %int1_20224 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20225 = torch.constant.int 4
%22623 = torch.aten.mul.int %int4_20225, %22622 : !torch.int, !torch.int -> !torch.int
%int1792_20226 = torch.constant.int 1792
%22624 = torch.prim.ListConstruct %22623, %int1792_20226 : (!torch.int, !torch.int) -> !torch.list<int>
%22625 = torch.aten.view %22568, %22624 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22625, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22626 = torch.aten.mm %22625, %22582 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22626, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20227 = torch.constant.int 4
%int4096_20228 = torch.constant.int 4096
%22627 = torch.prim.ListConstruct %int4_20227, %22622, %int4096_20228 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22628 = torch.aten.view %22626, %22627 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20229 = torch.constant.int 1
%22629 = torch.aten.size.int %22484, %int1_20229 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20230 = torch.constant.int 4
%22630 = torch.aten.mul.int %int4_20230, %22629 : !torch.int, !torch.int -> !torch.int
%int1792_20231 = torch.constant.int 1792
%22631 = torch.prim.ListConstruct %22630, %int1792_20231 : (!torch.int, !torch.int) -> !torch.list<int>
%22632 = torch.aten.view %22569, %22631 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22632, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22633 = torch.aten.mm %22632, %22584 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22633, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20232 = torch.constant.int 4
%int4096_20233 = torch.constant.int 4096
%22634 = torch.prim.ListConstruct %int4_20232, %22629, %int4096_20233 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22635 = torch.aten.view %22633, %22634 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20234 = torch.constant.int 1
%22636 = torch.aten.size.int %22490, %int1_20234 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_20235 = torch.constant.int 4
%22637 = torch.aten.mul.int %int4_20235, %22636 : !torch.int, !torch.int -> !torch.int
%int1792_20236 = torch.constant.int 1792
%22638 = torch.prim.ListConstruct %22637, %int1792_20236 : (!torch.int, !torch.int) -> !torch.list<int>
%22639 = torch.aten.view %22570, %22638 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %22639, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%22640 = torch.aten.mm %22639, %22586 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22640, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_20237 = torch.constant.int 4
%int4096_20238 = torch.constant.int 4096
%22641 = torch.prim.ListConstruct %int4_20237, %22636, %int4096_20238 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22642 = torch.aten.view %22640, %22641 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22643 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20239 = arith.constant 1 : index
%dim_20240 = tensor.dim %22643, %c1_20239 : tensor<4x?x4096xf16>
%22644 = flow.tensor.transfer %22643 : tensor<4x?x4096xf16>{%dim_20240} to #hal.device.promise<@__device_0>
%22645 = torch_c.from_builtin_tensor %22644 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22646 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20241 = arith.constant 1 : index
%dim_20242 = tensor.dim %22646, %c1_20241 : tensor<4x?x4096xf16>
%22647 = flow.tensor.transfer %22646 : tensor<4x?x4096xf16>{%dim_20242} to #hal.device.promise<@__device_0>
%22648 = torch_c.from_builtin_tensor %22647 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22649 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20243 = arith.constant 1 : index
%dim_20244 = tensor.dim %22649, %c1_20243 : tensor<4x?x4096xf16>
%22650 = flow.tensor.transfer %22649 : tensor<4x?x4096xf16>{%dim_20244} to #hal.device.promise<@__device_0>
%22651 = torch_c.from_builtin_tensor %22650 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22652 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20245 = arith.constant 1 : index
%dim_20246 = tensor.dim %22652, %c1_20245 : tensor<4x?x4096xf16>
%22653 = flow.tensor.transfer %22652 : tensor<4x?x4096xf16>{%dim_20246} to #hal.device.promise<@__device_0>
%22654 = torch_c.from_builtin_tensor %22653 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22655 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20247 = arith.constant 1 : index
%dim_20248 = tensor.dim %22655, %c1_20247 : tensor<4x?x4096xf16>
%22656 = flow.tensor.transfer %22655 : tensor<4x?x4096xf16>{%dim_20248} to #hal.device.promise<@__device_0>
%22657 = torch_c.from_builtin_tensor %22656 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22658 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20249 = arith.constant 1 : index
%dim_20250 = tensor.dim %22658, %c1_20249 : tensor<4x?x4096xf16>
%22659 = flow.tensor.transfer %22658 : tensor<4x?x4096xf16>{%dim_20250} to #hal.device.promise<@__device_0>
%22660 = torch_c.from_builtin_tensor %22659 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22661 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20251 = arith.constant 1 : index
%dim_20252 = tensor.dim %22661, %c1_20251 : tensor<4x?x4096xf16>
%22662 = flow.tensor.transfer %22661 : tensor<4x?x4096xf16>{%dim_20252} to #hal.device.promise<@__device_0>
%22663 = torch_c.from_builtin_tensor %22662 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20253 = torch.constant.int 1
%22664 = torch.aten.add.Tensor %22593, %22645, %int1_20253 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20254 = torch.constant.int 1
%22665 = torch.aten.add.Tensor %22664, %22648, %int1_20254 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20255 = torch.constant.int 1
%22666 = torch.aten.add.Tensor %22665, %22651, %int1_20255 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20256 = torch.constant.int 1
%22667 = torch.aten.add.Tensor %22666, %22654, %int1_20256 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20257 = torch.constant.int 1
%22668 = torch.aten.add.Tensor %22667, %22657, %int1_20257 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20258 = torch.constant.int 1
%22669 = torch.aten.add.Tensor %22668, %22660, %int1_20258 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20259 = torch.constant.int 1
%22670 = torch.aten.add.Tensor %22669, %22663, %int1_20259 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22671 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20260 = arith.constant 1 : index
%dim_20261 = tensor.dim %22671, %c1_20260 : tensor<4x?x4096xf16>
%22672 = flow.tensor.transfer %22671 : tensor<4x?x4096xf16>{%dim_20261} to #hal.device.promise<@__device_1>
%22673 = torch_c.from_builtin_tensor %22672 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22674 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20262 = arith.constant 1 : index
%dim_20263 = tensor.dim %22674, %c1_20262 : tensor<4x?x4096xf16>
%22675 = flow.tensor.transfer %22674 : tensor<4x?x4096xf16>{%dim_20263} to #hal.device.promise<@__device_1>
%22676 = torch_c.from_builtin_tensor %22675 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22677 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20264 = arith.constant 1 : index
%dim_20265 = tensor.dim %22677, %c1_20264 : tensor<4x?x4096xf16>
%22678 = flow.tensor.transfer %22677 : tensor<4x?x4096xf16>{%dim_20265} to #hal.device.promise<@__device_1>
%22679 = torch_c.from_builtin_tensor %22678 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22680 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20266 = arith.constant 1 : index
%dim_20267 = tensor.dim %22680, %c1_20266 : tensor<4x?x4096xf16>
%22681 = flow.tensor.transfer %22680 : tensor<4x?x4096xf16>{%dim_20267} to #hal.device.promise<@__device_1>
%22682 = torch_c.from_builtin_tensor %22681 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22683 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20268 = arith.constant 1 : index
%dim_20269 = tensor.dim %22683, %c1_20268 : tensor<4x?x4096xf16>
%22684 = flow.tensor.transfer %22683 : tensor<4x?x4096xf16>{%dim_20269} to #hal.device.promise<@__device_1>
%22685 = torch_c.from_builtin_tensor %22684 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22686 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20270 = arith.constant 1 : index
%dim_20271 = tensor.dim %22686, %c1_20270 : tensor<4x?x4096xf16>
%22687 = flow.tensor.transfer %22686 : tensor<4x?x4096xf16>{%dim_20271} to #hal.device.promise<@__device_1>
%22688 = torch_c.from_builtin_tensor %22687 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22689 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20272 = arith.constant 1 : index
%dim_20273 = tensor.dim %22689, %c1_20272 : tensor<4x?x4096xf16>
%22690 = flow.tensor.transfer %22689 : tensor<4x?x4096xf16>{%dim_20273} to #hal.device.promise<@__device_1>
%22691 = torch_c.from_builtin_tensor %22690 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20274 = torch.constant.int 1
%22692 = torch.aten.add.Tensor %22673, %22600, %int1_20274 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20275 = torch.constant.int 1
%22693 = torch.aten.add.Tensor %22692, %22676, %int1_20275 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20276 = torch.constant.int 1
%22694 = torch.aten.add.Tensor %22693, %22679, %int1_20276 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20277 = torch.constant.int 1
%22695 = torch.aten.add.Tensor %22694, %22682, %int1_20277 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20278 = torch.constant.int 1
%22696 = torch.aten.add.Tensor %22695, %22685, %int1_20278 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20279 = torch.constant.int 1
%22697 = torch.aten.add.Tensor %22696, %22688, %int1_20279 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20280 = torch.constant.int 1
%22698 = torch.aten.add.Tensor %22697, %22691, %int1_20280 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22699 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20281 = arith.constant 1 : index
%dim_20282 = tensor.dim %22699, %c1_20281 : tensor<4x?x4096xf16>
%22700 = flow.tensor.transfer %22699 : tensor<4x?x4096xf16>{%dim_20282} to #hal.device.promise<@__device_2>
%22701 = torch_c.from_builtin_tensor %22700 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22702 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20283 = arith.constant 1 : index
%dim_20284 = tensor.dim %22702, %c1_20283 : tensor<4x?x4096xf16>
%22703 = flow.tensor.transfer %22702 : tensor<4x?x4096xf16>{%dim_20284} to #hal.device.promise<@__device_2>
%22704 = torch_c.from_builtin_tensor %22703 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22705 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20285 = arith.constant 1 : index
%dim_20286 = tensor.dim %22705, %c1_20285 : tensor<4x?x4096xf16>
%22706 = flow.tensor.transfer %22705 : tensor<4x?x4096xf16>{%dim_20286} to #hal.device.promise<@__device_2>
%22707 = torch_c.from_builtin_tensor %22706 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22708 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20287 = arith.constant 1 : index
%dim_20288 = tensor.dim %22708, %c1_20287 : tensor<4x?x4096xf16>
%22709 = flow.tensor.transfer %22708 : tensor<4x?x4096xf16>{%dim_20288} to #hal.device.promise<@__device_2>
%22710 = torch_c.from_builtin_tensor %22709 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22711 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20289 = arith.constant 1 : index
%dim_20290 = tensor.dim %22711, %c1_20289 : tensor<4x?x4096xf16>
%22712 = flow.tensor.transfer %22711 : tensor<4x?x4096xf16>{%dim_20290} to #hal.device.promise<@__device_2>
%22713 = torch_c.from_builtin_tensor %22712 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22714 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20291 = arith.constant 1 : index
%dim_20292 = tensor.dim %22714, %c1_20291 : tensor<4x?x4096xf16>
%22715 = flow.tensor.transfer %22714 : tensor<4x?x4096xf16>{%dim_20292} to #hal.device.promise<@__device_2>
%22716 = torch_c.from_builtin_tensor %22715 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22717 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20293 = arith.constant 1 : index
%dim_20294 = tensor.dim %22717, %c1_20293 : tensor<4x?x4096xf16>
%22718 = flow.tensor.transfer %22717 : tensor<4x?x4096xf16>{%dim_20294} to #hal.device.promise<@__device_2>
%22719 = torch_c.from_builtin_tensor %22718 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20295 = torch.constant.int 1
%22720 = torch.aten.add.Tensor %22701, %22704, %int1_20295 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20296 = torch.constant.int 1
%22721 = torch.aten.add.Tensor %22720, %22607, %int1_20296 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20297 = torch.constant.int 1
%22722 = torch.aten.add.Tensor %22721, %22707, %int1_20297 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20298 = torch.constant.int 1
%22723 = torch.aten.add.Tensor %22722, %22710, %int1_20298 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20299 = torch.constant.int 1
%22724 = torch.aten.add.Tensor %22723, %22713, %int1_20299 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20300 = torch.constant.int 1
%22725 = torch.aten.add.Tensor %22724, %22716, %int1_20300 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20301 = torch.constant.int 1
%22726 = torch.aten.add.Tensor %22725, %22719, %int1_20301 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22727 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20302 = arith.constant 1 : index
%dim_20303 = tensor.dim %22727, %c1_20302 : tensor<4x?x4096xf16>
%22728 = flow.tensor.transfer %22727 : tensor<4x?x4096xf16>{%dim_20303} to #hal.device.promise<@__device_3>
%22729 = torch_c.from_builtin_tensor %22728 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22730 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20304 = arith.constant 1 : index
%dim_20305 = tensor.dim %22730, %c1_20304 : tensor<4x?x4096xf16>
%22731 = flow.tensor.transfer %22730 : tensor<4x?x4096xf16>{%dim_20305} to #hal.device.promise<@__device_3>
%22732 = torch_c.from_builtin_tensor %22731 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22733 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20306 = arith.constant 1 : index
%dim_20307 = tensor.dim %22733, %c1_20306 : tensor<4x?x4096xf16>
%22734 = flow.tensor.transfer %22733 : tensor<4x?x4096xf16>{%dim_20307} to #hal.device.promise<@__device_3>
%22735 = torch_c.from_builtin_tensor %22734 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22736 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20308 = arith.constant 1 : index
%dim_20309 = tensor.dim %22736, %c1_20308 : tensor<4x?x4096xf16>
%22737 = flow.tensor.transfer %22736 : tensor<4x?x4096xf16>{%dim_20309} to #hal.device.promise<@__device_3>
%22738 = torch_c.from_builtin_tensor %22737 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22739 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20310 = arith.constant 1 : index
%dim_20311 = tensor.dim %22739, %c1_20310 : tensor<4x?x4096xf16>
%22740 = flow.tensor.transfer %22739 : tensor<4x?x4096xf16>{%dim_20311} to #hal.device.promise<@__device_3>
%22741 = torch_c.from_builtin_tensor %22740 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22742 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20312 = arith.constant 1 : index
%dim_20313 = tensor.dim %22742, %c1_20312 : tensor<4x?x4096xf16>
%22743 = flow.tensor.transfer %22742 : tensor<4x?x4096xf16>{%dim_20313} to #hal.device.promise<@__device_3>
%22744 = torch_c.from_builtin_tensor %22743 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22745 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20314 = arith.constant 1 : index
%dim_20315 = tensor.dim %22745, %c1_20314 : tensor<4x?x4096xf16>
%22746 = flow.tensor.transfer %22745 : tensor<4x?x4096xf16>{%dim_20315} to #hal.device.promise<@__device_3>
%22747 = torch_c.from_builtin_tensor %22746 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20316 = torch.constant.int 1
%22748 = torch.aten.add.Tensor %22729, %22732, %int1_20316 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20317 = torch.constant.int 1
%22749 = torch.aten.add.Tensor %22748, %22735, %int1_20317 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20318 = torch.constant.int 1
%22750 = torch.aten.add.Tensor %22749, %22614, %int1_20318 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20319 = torch.constant.int 1
%22751 = torch.aten.add.Tensor %22750, %22738, %int1_20319 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20320 = torch.constant.int 1
%22752 = torch.aten.add.Tensor %22751, %22741, %int1_20320 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20321 = torch.constant.int 1
%22753 = torch.aten.add.Tensor %22752, %22744, %int1_20321 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20322 = torch.constant.int 1
%22754 = torch.aten.add.Tensor %22753, %22747, %int1_20322 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22755 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20323 = arith.constant 1 : index
%dim_20324 = tensor.dim %22755, %c1_20323 : tensor<4x?x4096xf16>
%22756 = flow.tensor.transfer %22755 : tensor<4x?x4096xf16>{%dim_20324} to #hal.device.promise<@__device_4>
%22757 = torch_c.from_builtin_tensor %22756 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22758 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20325 = arith.constant 1 : index
%dim_20326 = tensor.dim %22758, %c1_20325 : tensor<4x?x4096xf16>
%22759 = flow.tensor.transfer %22758 : tensor<4x?x4096xf16>{%dim_20326} to #hal.device.promise<@__device_4>
%22760 = torch_c.from_builtin_tensor %22759 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22761 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20327 = arith.constant 1 : index
%dim_20328 = tensor.dim %22761, %c1_20327 : tensor<4x?x4096xf16>
%22762 = flow.tensor.transfer %22761 : tensor<4x?x4096xf16>{%dim_20328} to #hal.device.promise<@__device_4>
%22763 = torch_c.from_builtin_tensor %22762 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22764 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20329 = arith.constant 1 : index
%dim_20330 = tensor.dim %22764, %c1_20329 : tensor<4x?x4096xf16>
%22765 = flow.tensor.transfer %22764 : tensor<4x?x4096xf16>{%dim_20330} to #hal.device.promise<@__device_4>
%22766 = torch_c.from_builtin_tensor %22765 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22767 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20331 = arith.constant 1 : index
%dim_20332 = tensor.dim %22767, %c1_20331 : tensor<4x?x4096xf16>
%22768 = flow.tensor.transfer %22767 : tensor<4x?x4096xf16>{%dim_20332} to #hal.device.promise<@__device_4>
%22769 = torch_c.from_builtin_tensor %22768 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22770 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20333 = arith.constant 1 : index
%dim_20334 = tensor.dim %22770, %c1_20333 : tensor<4x?x4096xf16>
%22771 = flow.tensor.transfer %22770 : tensor<4x?x4096xf16>{%dim_20334} to #hal.device.promise<@__device_4>
%22772 = torch_c.from_builtin_tensor %22771 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22773 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20335 = arith.constant 1 : index
%dim_20336 = tensor.dim %22773, %c1_20335 : tensor<4x?x4096xf16>
%22774 = flow.tensor.transfer %22773 : tensor<4x?x4096xf16>{%dim_20336} to #hal.device.promise<@__device_4>
%22775 = torch_c.from_builtin_tensor %22774 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20337 = torch.constant.int 1
%22776 = torch.aten.add.Tensor %22757, %22760, %int1_20337 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20338 = torch.constant.int 1
%22777 = torch.aten.add.Tensor %22776, %22763, %int1_20338 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20339 = torch.constant.int 1
%22778 = torch.aten.add.Tensor %22777, %22766, %int1_20339 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20340 = torch.constant.int 1
%22779 = torch.aten.add.Tensor %22778, %22621, %int1_20340 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20341 = torch.constant.int 1
%22780 = torch.aten.add.Tensor %22779, %22769, %int1_20341 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20342 = torch.constant.int 1
%22781 = torch.aten.add.Tensor %22780, %22772, %int1_20342 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20343 = torch.constant.int 1
%22782 = torch.aten.add.Tensor %22781, %22775, %int1_20343 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22783 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20344 = arith.constant 1 : index
%dim_20345 = tensor.dim %22783, %c1_20344 : tensor<4x?x4096xf16>
%22784 = flow.tensor.transfer %22783 : tensor<4x?x4096xf16>{%dim_20345} to #hal.device.promise<@__device_5>
%22785 = torch_c.from_builtin_tensor %22784 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22786 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20346 = arith.constant 1 : index
%dim_20347 = tensor.dim %22786, %c1_20346 : tensor<4x?x4096xf16>
%22787 = flow.tensor.transfer %22786 : tensor<4x?x4096xf16>{%dim_20347} to #hal.device.promise<@__device_5>
%22788 = torch_c.from_builtin_tensor %22787 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22789 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20348 = arith.constant 1 : index
%dim_20349 = tensor.dim %22789, %c1_20348 : tensor<4x?x4096xf16>
%22790 = flow.tensor.transfer %22789 : tensor<4x?x4096xf16>{%dim_20349} to #hal.device.promise<@__device_5>
%22791 = torch_c.from_builtin_tensor %22790 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22792 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20350 = arith.constant 1 : index
%dim_20351 = tensor.dim %22792, %c1_20350 : tensor<4x?x4096xf16>
%22793 = flow.tensor.transfer %22792 : tensor<4x?x4096xf16>{%dim_20351} to #hal.device.promise<@__device_5>
%22794 = torch_c.from_builtin_tensor %22793 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22795 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20352 = arith.constant 1 : index
%dim_20353 = tensor.dim %22795, %c1_20352 : tensor<4x?x4096xf16>
%22796 = flow.tensor.transfer %22795 : tensor<4x?x4096xf16>{%dim_20353} to #hal.device.promise<@__device_5>
%22797 = torch_c.from_builtin_tensor %22796 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22798 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20354 = arith.constant 1 : index
%dim_20355 = tensor.dim %22798, %c1_20354 : tensor<4x?x4096xf16>
%22799 = flow.tensor.transfer %22798 : tensor<4x?x4096xf16>{%dim_20355} to #hal.device.promise<@__device_5>
%22800 = torch_c.from_builtin_tensor %22799 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22801 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20356 = arith.constant 1 : index
%dim_20357 = tensor.dim %22801, %c1_20356 : tensor<4x?x4096xf16>
%22802 = flow.tensor.transfer %22801 : tensor<4x?x4096xf16>{%dim_20357} to #hal.device.promise<@__device_5>
%22803 = torch_c.from_builtin_tensor %22802 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20358 = torch.constant.int 1
%22804 = torch.aten.add.Tensor %22785, %22788, %int1_20358 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20359 = torch.constant.int 1
%22805 = torch.aten.add.Tensor %22804, %22791, %int1_20359 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20360 = torch.constant.int 1
%22806 = torch.aten.add.Tensor %22805, %22794, %int1_20360 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20361 = torch.constant.int 1
%22807 = torch.aten.add.Tensor %22806, %22797, %int1_20361 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20362 = torch.constant.int 1
%22808 = torch.aten.add.Tensor %22807, %22628, %int1_20362 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20363 = torch.constant.int 1
%22809 = torch.aten.add.Tensor %22808, %22800, %int1_20363 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20364 = torch.constant.int 1
%22810 = torch.aten.add.Tensor %22809, %22803, %int1_20364 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22811 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20365 = arith.constant 1 : index
%dim_20366 = tensor.dim %22811, %c1_20365 : tensor<4x?x4096xf16>
%22812 = flow.tensor.transfer %22811 : tensor<4x?x4096xf16>{%dim_20366} to #hal.device.promise<@__device_6>
%22813 = torch_c.from_builtin_tensor %22812 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22814 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20367 = arith.constant 1 : index
%dim_20368 = tensor.dim %22814, %c1_20367 : tensor<4x?x4096xf16>
%22815 = flow.tensor.transfer %22814 : tensor<4x?x4096xf16>{%dim_20368} to #hal.device.promise<@__device_6>
%22816 = torch_c.from_builtin_tensor %22815 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22817 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20369 = arith.constant 1 : index
%dim_20370 = tensor.dim %22817, %c1_20369 : tensor<4x?x4096xf16>
%22818 = flow.tensor.transfer %22817 : tensor<4x?x4096xf16>{%dim_20370} to #hal.device.promise<@__device_6>
%22819 = torch_c.from_builtin_tensor %22818 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22820 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20371 = arith.constant 1 : index
%dim_20372 = tensor.dim %22820, %c1_20371 : tensor<4x?x4096xf16>
%22821 = flow.tensor.transfer %22820 : tensor<4x?x4096xf16>{%dim_20372} to #hal.device.promise<@__device_6>
%22822 = torch_c.from_builtin_tensor %22821 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22823 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20373 = arith.constant 1 : index
%dim_20374 = tensor.dim %22823, %c1_20373 : tensor<4x?x4096xf16>
%22824 = flow.tensor.transfer %22823 : tensor<4x?x4096xf16>{%dim_20374} to #hal.device.promise<@__device_6>
%22825 = torch_c.from_builtin_tensor %22824 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22826 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20375 = arith.constant 1 : index
%dim_20376 = tensor.dim %22826, %c1_20375 : tensor<4x?x4096xf16>
%22827 = flow.tensor.transfer %22826 : tensor<4x?x4096xf16>{%dim_20376} to #hal.device.promise<@__device_6>
%22828 = torch_c.from_builtin_tensor %22827 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22829 = torch_c.to_builtin_tensor %22642 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20377 = arith.constant 1 : index
%dim_20378 = tensor.dim %22829, %c1_20377 : tensor<4x?x4096xf16>
%22830 = flow.tensor.transfer %22829 : tensor<4x?x4096xf16>{%dim_20378} to #hal.device.promise<@__device_6>
%22831 = torch_c.from_builtin_tensor %22830 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20379 = torch.constant.int 1
%22832 = torch.aten.add.Tensor %22813, %22816, %int1_20379 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20380 = torch.constant.int 1
%22833 = torch.aten.add.Tensor %22832, %22819, %int1_20380 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20381 = torch.constant.int 1
%22834 = torch.aten.add.Tensor %22833, %22822, %int1_20381 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20382 = torch.constant.int 1
%22835 = torch.aten.add.Tensor %22834, %22825, %int1_20382 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20383 = torch.constant.int 1
%22836 = torch.aten.add.Tensor %22835, %22828, %int1_20383 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20384 = torch.constant.int 1
%22837 = torch.aten.add.Tensor %22836, %22635, %int1_20384 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20385 = torch.constant.int 1
%22838 = torch.aten.add.Tensor %22837, %22831, %int1_20385 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22839 = torch_c.to_builtin_tensor %22593 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20386 = arith.constant 1 : index
%dim_20387 = tensor.dim %22839, %c1_20386 : tensor<4x?x4096xf16>
%22840 = flow.tensor.transfer %22839 : tensor<4x?x4096xf16>{%dim_20387} to #hal.device.promise<@__device_7>
%22841 = torch_c.from_builtin_tensor %22840 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22842 = torch_c.to_builtin_tensor %22600 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20388 = arith.constant 1 : index
%dim_20389 = tensor.dim %22842, %c1_20388 : tensor<4x?x4096xf16>
%22843 = flow.tensor.transfer %22842 : tensor<4x?x4096xf16>{%dim_20389} to #hal.device.promise<@__device_7>
%22844 = torch_c.from_builtin_tensor %22843 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22845 = torch_c.to_builtin_tensor %22607 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20390 = arith.constant 1 : index
%dim_20391 = tensor.dim %22845, %c1_20390 : tensor<4x?x4096xf16>
%22846 = flow.tensor.transfer %22845 : tensor<4x?x4096xf16>{%dim_20391} to #hal.device.promise<@__device_7>
%22847 = torch_c.from_builtin_tensor %22846 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22848 = torch_c.to_builtin_tensor %22614 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20392 = arith.constant 1 : index
%dim_20393 = tensor.dim %22848, %c1_20392 : tensor<4x?x4096xf16>
%22849 = flow.tensor.transfer %22848 : tensor<4x?x4096xf16>{%dim_20393} to #hal.device.promise<@__device_7>
%22850 = torch_c.from_builtin_tensor %22849 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22851 = torch_c.to_builtin_tensor %22621 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20394 = arith.constant 1 : index
%dim_20395 = tensor.dim %22851, %c1_20394 : tensor<4x?x4096xf16>
%22852 = flow.tensor.transfer %22851 : tensor<4x?x4096xf16>{%dim_20395} to #hal.device.promise<@__device_7>
%22853 = torch_c.from_builtin_tensor %22852 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22854 = torch_c.to_builtin_tensor %22628 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20396 = arith.constant 1 : index
%dim_20397 = tensor.dim %22854, %c1_20396 : tensor<4x?x4096xf16>
%22855 = flow.tensor.transfer %22854 : tensor<4x?x4096xf16>{%dim_20397} to #hal.device.promise<@__device_7>
%22856 = torch_c.from_builtin_tensor %22855 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%22857 = torch_c.to_builtin_tensor %22635 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_20398 = arith.constant 1 : index
%dim_20399 = tensor.dim %22857, %c1_20398 : tensor<4x?x4096xf16>
%22858 = flow.tensor.transfer %22857 : tensor<4x?x4096xf16>{%dim_20399} to #hal.device.promise<@__device_7>
%22859 = torch_c.from_builtin_tensor %22858 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20400 = torch.constant.int 1
%22860 = torch.aten.add.Tensor %22841, %22844, %int1_20400 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20401 = torch.constant.int 1
%22861 = torch.aten.add.Tensor %22860, %22847, %int1_20401 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20402 = torch.constant.int 1
%22862 = torch.aten.add.Tensor %22861, %22850, %int1_20402 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20403 = torch.constant.int 1
%22863 = torch.aten.add.Tensor %22862, %22853, %int1_20403 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20404 = torch.constant.int 1
%22864 = torch.aten.add.Tensor %22863, %22856, %int1_20404 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20405 = torch.constant.int 1
%22865 = torch.aten.add.Tensor %22864, %22859, %int1_20405 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20406 = torch.constant.int 1
%22866 = torch.aten.add.Tensor %22865, %22642, %int1_20406 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20407 = torch.constant.int 1
%22867 = torch.aten.add.Tensor %22347, %22670, %int1_20407 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20408 = torch.constant.int 1
%22868 = torch.aten.add.Tensor %22348, %22698, %int1_20408 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20409 = torch.constant.int 1
%22869 = torch.aten.add.Tensor %22349, %22726, %int1_20409 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20410 = torch.constant.int 1
%22870 = torch.aten.add.Tensor %22350, %22754, %int1_20410 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20411 = torch.constant.int 1
%22871 = torch.aten.add.Tensor %22351, %22782, %int1_20411 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20412 = torch.constant.int 1
%22872 = torch.aten.add.Tensor %22352, %22810, %int1_20412 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20413 = torch.constant.int 1
%22873 = torch.aten.add.Tensor %22353, %22838, %int1_20413 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20414 = torch.constant.int 1
%22874 = torch.aten.add.Tensor %22354, %22866, %int1_20414 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_20415 = torch.constant.int 6
%22875 = torch.prims.convert_element_type %22867, %int6_20415 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20416 = torch.constant.int 6
%22876 = torch.prims.convert_element_type %22868, %int6_20416 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20417 = torch.constant.int 6
%22877 = torch.prims.convert_element_type %22869, %int6_20417 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20418 = torch.constant.int 6
%22878 = torch.prims.convert_element_type %22870, %int6_20418 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20419 = torch.constant.int 6
%22879 = torch.prims.convert_element_type %22871, %int6_20419 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20420 = torch.constant.int 6
%22880 = torch.prims.convert_element_type %22872, %int6_20420 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20421 = torch.constant.int 6
%22881 = torch.prims.convert_element_type %22873, %int6_20421 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_20422 = torch.constant.int 6
%22882 = torch.prims.convert_element_type %22874, %int6_20422 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20423 = torch.constant.int 2
%22883 = torch.aten.pow.Tensor_Scalar %22875, %int2_20423 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20424 = torch.constant.int 2
%22884 = torch.aten.pow.Tensor_Scalar %22876, %int2_20424 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20425 = torch.constant.int 2
%22885 = torch.aten.pow.Tensor_Scalar %22877, %int2_20425 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20426 = torch.constant.int 2
%22886 = torch.aten.pow.Tensor_Scalar %22878, %int2_20426 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20427 = torch.constant.int 2
%22887 = torch.aten.pow.Tensor_Scalar %22879, %int2_20427 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20428 = torch.constant.int 2
%22888 = torch.aten.pow.Tensor_Scalar %22880, %int2_20428 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20429 = torch.constant.int 2
%22889 = torch.aten.pow.Tensor_Scalar %22881, %int2_20429 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_20430 = torch.constant.int 2
%22890 = torch.aten.pow.Tensor_Scalar %22882, %int2_20430 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_20431 = torch.constant.int -1
%22891 = torch.prim.ListConstruct %int-1_20431 : (!torch.int) -> !torch.list<int>
%true_20432 = torch.constant.bool true
%none_20433 = torch.constant.none
%22892 = torch.aten.mean.dim %22883, %22891, %true_20432, %none_20433 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20434 = torch.constant.int -1
%22893 = torch.prim.ListConstruct %int-1_20434 : (!torch.int) -> !torch.list<int>
%true_20435 = torch.constant.bool true
%none_20436 = torch.constant.none
%22894 = torch.aten.mean.dim %22884, %22893, %true_20435, %none_20436 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20437 = torch.constant.int -1
%22895 = torch.prim.ListConstruct %int-1_20437 : (!torch.int) -> !torch.list<int>
%true_20438 = torch.constant.bool true
%none_20439 = torch.constant.none
%22896 = torch.aten.mean.dim %22885, %22895, %true_20438, %none_20439 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20440 = torch.constant.int -1
%22897 = torch.prim.ListConstruct %int-1_20440 : (!torch.int) -> !torch.list<int>
%true_20441 = torch.constant.bool true
%none_20442 = torch.constant.none
%22898 = torch.aten.mean.dim %22886, %22897, %true_20441, %none_20442 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20443 = torch.constant.int -1
%22899 = torch.prim.ListConstruct %int-1_20443 : (!torch.int) -> !torch.list<int>
%true_20444 = torch.constant.bool true
%none_20445 = torch.constant.none
%22900 = torch.aten.mean.dim %22887, %22899, %true_20444, %none_20445 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20446 = torch.constant.int -1
%22901 = torch.prim.ListConstruct %int-1_20446 : (!torch.int) -> !torch.list<int>
%true_20447 = torch.constant.bool true
%none_20448 = torch.constant.none
%22902 = torch.aten.mean.dim %22888, %22901, %true_20447, %none_20448 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20449 = torch.constant.int -1
%22903 = torch.prim.ListConstruct %int-1_20449 : (!torch.int) -> !torch.list<int>
%true_20450 = torch.constant.bool true
%none_20451 = torch.constant.none
%22904 = torch.aten.mean.dim %22889, %22903, %true_20450, %none_20451 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_20452 = torch.constant.int -1
%22905 = torch.prim.ListConstruct %int-1_20452 : (!torch.int) -> !torch.list<int>
%true_20453 = torch.constant.bool true
%none_20454 = torch.constant.none
%22906 = torch.aten.mean.dim %22890, %22905, %true_20453, %none_20454 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20455 = torch.constant.float 9.9999997473787516E-6
%int1_20456 = torch.constant.int 1
%22907 = torch.aten.add.Scalar %22892, %float9.999990e-06_20455, %int1_20456 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20457 = torch.constant.float 9.9999997473787516E-6
%int1_20458 = torch.constant.int 1
%22908 = torch.aten.add.Scalar %22894, %float9.999990e-06_20457, %int1_20458 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20459 = torch.constant.float 9.9999997473787516E-6
%int1_20460 = torch.constant.int 1
%22909 = torch.aten.add.Scalar %22896, %float9.999990e-06_20459, %int1_20460 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20461 = torch.constant.float 9.9999997473787516E-6
%int1_20462 = torch.constant.int 1
%22910 = torch.aten.add.Scalar %22898, %float9.999990e-06_20461, %int1_20462 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20463 = torch.constant.float 9.9999997473787516E-6
%int1_20464 = torch.constant.int 1
%22911 = torch.aten.add.Scalar %22900, %float9.999990e-06_20463, %int1_20464 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20465 = torch.constant.float 9.9999997473787516E-6
%int1_20466 = torch.constant.int 1
%22912 = torch.aten.add.Scalar %22902, %float9.999990e-06_20465, %int1_20466 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20467 = torch.constant.float 9.9999997473787516E-6
%int1_20468 = torch.constant.int 1
%22913 = torch.aten.add.Scalar %22904, %float9.999990e-06_20467, %int1_20468 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_20469 = torch.constant.float 9.9999997473787516E-6
%int1_20470 = torch.constant.int 1
%22914 = torch.aten.add.Scalar %22906, %float9.999990e-06_20469, %int1_20470 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22915 = torch.aten.rsqrt %22907 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22916 = torch.aten.rsqrt %22908 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22917 = torch.aten.rsqrt %22909 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22918 = torch.aten.rsqrt %22910 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22919 = torch.aten.rsqrt %22911 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22920 = torch.aten.rsqrt %22912 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22921 = torch.aten.rsqrt %22913 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22922 = torch.aten.rsqrt %22914 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %22922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%22923 = torch.aten.mul.Tensor %22875, %22915 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22924 = torch.aten.mul.Tensor %22876, %22916 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22925 = torch.aten.mul.Tensor %22877, %22917 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22926 = torch.aten.mul.Tensor %22878, %22918 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22927 = torch.aten.mul.Tensor %22879, %22919 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22928 = torch.aten.mul.Tensor %22880, %22920 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22929 = torch.aten.mul.Tensor %22881, %22921 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22930 = torch.aten.mul.Tensor %22882, %22922 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22931 = torch.aten.mul.Tensor %800, %22923 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22932 = torch.aten.mul.Tensor %801, %22924 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22933 = torch.aten.mul.Tensor %802, %22925 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22934 = torch.aten.mul.Tensor %803, %22926 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22935 = torch.aten.mul.Tensor %804, %22927 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22936 = torch.aten.mul.Tensor %805, %22928 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22937 = torch.aten.mul.Tensor %806, %22929 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%22938 = torch.aten.mul.Tensor %807, %22930 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %22938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_20471 = torch.constant.int 5
%22939 = torch.prims.convert_element_type %22931, %int5_20471 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20472 = torch.constant.int 5
%22940 = torch.prims.convert_element_type %22932, %int5_20472 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20473 = torch.constant.int 5
%22941 = torch.prims.convert_element_type %22933, %int5_20473 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20474 = torch.constant.int 5
%22942 = torch.prims.convert_element_type %22934, %int5_20474 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20475 = torch.constant.int 5
%22943 = torch.prims.convert_element_type %22935, %int5_20475 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20476 = torch.constant.int 5
%22944 = torch.prims.convert_element_type %22936, %int5_20476 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20477 = torch.constant.int 5
%22945 = torch.prims.convert_element_type %22937, %int5_20477 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_20478 = torch.constant.int 5
%22946 = torch.prims.convert_element_type %22938, %int5_20478 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %22946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_20479 = torch.constant.int 1
%int0_20480 = torch.constant.int 0
%22947 = torch.prim.ListConstruct %int1_20479, %int0_20480 : (!torch.int, !torch.int) -> !torch.list<int>
%22948 = torch.aten.permute %808, %22947 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20481 = torch.constant.int 1
%int0_20482 = torch.constant.int 0
%22949 = torch.prim.ListConstruct %int1_20481, %int0_20482 : (!torch.int, !torch.int) -> !torch.list<int>
%22950 = torch.aten.permute %809, %22949 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20483 = torch.constant.int 1
%int0_20484 = torch.constant.int 0
%22951 = torch.prim.ListConstruct %int1_20483, %int0_20484 : (!torch.int, !torch.int) -> !torch.list<int>
%22952 = torch.aten.permute %810, %22951 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20485 = torch.constant.int 1
%int0_20486 = torch.constant.int 0
%22953 = torch.prim.ListConstruct %int1_20485, %int0_20486 : (!torch.int, !torch.int) -> !torch.list<int>
%22954 = torch.aten.permute %811, %22953 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20487 = torch.constant.int 1
%int0_20488 = torch.constant.int 0
%22955 = torch.prim.ListConstruct %int1_20487, %int0_20488 : (!torch.int, !torch.int) -> !torch.list<int>
%22956 = torch.aten.permute %812, %22955 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20489 = torch.constant.int 1
%int0_20490 = torch.constant.int 0
%22957 = torch.prim.ListConstruct %int1_20489, %int0_20490 : (!torch.int, !torch.int) -> !torch.list<int>
%22958 = torch.aten.permute %813, %22957 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20491 = torch.constant.int 1
%int0_20492 = torch.constant.int 0
%22959 = torch.prim.ListConstruct %int1_20491, %int0_20492 : (!torch.int, !torch.int) -> !torch.list<int>
%22960 = torch.aten.permute %814, %22959 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_20493 = torch.constant.int 1
%int0_20494 = torch.constant.int 0
%22961 = torch.prim.ListConstruct %int1_20493, %int0_20494 : (!torch.int, !torch.int) -> !torch.list<int>
%22962 = torch.aten.permute %815, %22961 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_20495 = torch.constant.int 4
%22963 = torch.aten.mul.int %int4_20495, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20496 = torch.constant.int 4096
%22964 = torch.prim.ListConstruct %22963, %int4096_20496 : (!torch.int, !torch.int) -> !torch.list<int>
%22965 = torch.aten.view %22939, %22964 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22965, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22966 = torch.aten.mm %22965, %22948 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22966, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20497 = torch.constant.int 4
%int512_20498 = torch.constant.int 512
%22967 = torch.prim.ListConstruct %int4_20497, %2482, %int512_20498 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22968 = torch.aten.view %22966, %22967 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20499 = torch.constant.int 4
%22969 = torch.aten.mul.int %int4_20499, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20500 = torch.constant.int 4096
%22970 = torch.prim.ListConstruct %22969, %int4096_20500 : (!torch.int, !torch.int) -> !torch.list<int>
%22971 = torch.aten.view %22940, %22970 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22971, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22972 = torch.aten.mm %22971, %22950 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22972, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20501 = torch.constant.int 4
%int512_20502 = torch.constant.int 512
%22973 = torch.prim.ListConstruct %int4_20501, %2482, %int512_20502 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22974 = torch.aten.view %22972, %22973 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20503 = torch.constant.int 4
%22975 = torch.aten.mul.int %int4_20503, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20504 = torch.constant.int 4096
%22976 = torch.prim.ListConstruct %22975, %int4096_20504 : (!torch.int, !torch.int) -> !torch.list<int>
%22977 = torch.aten.view %22941, %22976 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22977, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22978 = torch.aten.mm %22977, %22952 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22978, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20505 = torch.constant.int 4
%int512_20506 = torch.constant.int 512
%22979 = torch.prim.ListConstruct %int4_20505, %2482, %int512_20506 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22980 = torch.aten.view %22978, %22979 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20507 = torch.constant.int 4
%22981 = torch.aten.mul.int %int4_20507, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20508 = torch.constant.int 4096
%22982 = torch.prim.ListConstruct %22981, %int4096_20508 : (!torch.int, !torch.int) -> !torch.list<int>
%22983 = torch.aten.view %22942, %22982 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22983, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22984 = torch.aten.mm %22983, %22954 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22984, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20509 = torch.constant.int 4
%int512_20510 = torch.constant.int 512
%22985 = torch.prim.ListConstruct %int4_20509, %2482, %int512_20510 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22986 = torch.aten.view %22984, %22985 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20511 = torch.constant.int 4
%22987 = torch.aten.mul.int %int4_20511, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20512 = torch.constant.int 4096
%22988 = torch.prim.ListConstruct %22987, %int4096_20512 : (!torch.int, !torch.int) -> !torch.list<int>
%22989 = torch.aten.view %22943, %22988 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22989, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22990 = torch.aten.mm %22989, %22956 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22990, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20513 = torch.constant.int 4
%int512_20514 = torch.constant.int 512
%22991 = torch.prim.ListConstruct %int4_20513, %2482, %int512_20514 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22992 = torch.aten.view %22990, %22991 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20515 = torch.constant.int 4
%22993 = torch.aten.mul.int %int4_20515, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20516 = torch.constant.int 4096
%22994 = torch.prim.ListConstruct %22993, %int4096_20516 : (!torch.int, !torch.int) -> !torch.list<int>
%22995 = torch.aten.view %22944, %22994 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %22995, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%22996 = torch.aten.mm %22995, %22958 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %22996, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20517 = torch.constant.int 4
%int512_20518 = torch.constant.int 512
%22997 = torch.prim.ListConstruct %int4_20517, %2482, %int512_20518 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%22998 = torch.aten.view %22996, %22997 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %22998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20519 = torch.constant.int 4
%22999 = torch.aten.mul.int %int4_20519, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20520 = torch.constant.int 4096
%23000 = torch.prim.ListConstruct %22999, %int4096_20520 : (!torch.int, !torch.int) -> !torch.list<int>
%23001 = torch.aten.view %22945, %23000 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23001, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23002 = torch.aten.mm %23001, %22960 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23002, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20521 = torch.constant.int 4
%int512_20522 = torch.constant.int 512
%23003 = torch.prim.ListConstruct %int4_20521, %2482, %int512_20522 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23004 = torch.aten.view %23002, %23003 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_20523 = torch.constant.int 4
%23005 = torch.aten.mul.int %int4_20523, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20524 = torch.constant.int 4096
%23006 = torch.prim.ListConstruct %23005, %int4096_20524 : (!torch.int, !torch.int) -> !torch.list<int>
%23007 = torch.aten.view %22946, %23006 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23007, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23008 = torch.aten.mm %23007, %22962 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23008, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_20525 = torch.constant.int 4
%int512_20526 = torch.constant.int 512
%23009 = torch.prim.ListConstruct %int4_20525, %2482, %int512_20526 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23010 = torch.aten.view %23008, %23009 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_20527 = torch.constant.int 1
%int0_20528 = torch.constant.int 0
%23011 = torch.prim.ListConstruct %int1_20527, %int0_20528 : (!torch.int, !torch.int) -> !torch.list<int>
%23012 = torch.aten.permute %816, %23011 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20529 = torch.constant.int 1
%int0_20530 = torch.constant.int 0
%23013 = torch.prim.ListConstruct %int1_20529, %int0_20530 : (!torch.int, !torch.int) -> !torch.list<int>
%23014 = torch.aten.permute %817, %23013 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20531 = torch.constant.int 1
%int0_20532 = torch.constant.int 0
%23015 = torch.prim.ListConstruct %int1_20531, %int0_20532 : (!torch.int, !torch.int) -> !torch.list<int>
%23016 = torch.aten.permute %818, %23015 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20533 = torch.constant.int 1
%int0_20534 = torch.constant.int 0
%23017 = torch.prim.ListConstruct %int1_20533, %int0_20534 : (!torch.int, !torch.int) -> !torch.list<int>
%23018 = torch.aten.permute %819, %23017 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20535 = torch.constant.int 1
%int0_20536 = torch.constant.int 0
%23019 = torch.prim.ListConstruct %int1_20535, %int0_20536 : (!torch.int, !torch.int) -> !torch.list<int>
%23020 = torch.aten.permute %820, %23019 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20537 = torch.constant.int 1
%int0_20538 = torch.constant.int 0
%23021 = torch.prim.ListConstruct %int1_20537, %int0_20538 : (!torch.int, !torch.int) -> !torch.list<int>
%23022 = torch.aten.permute %821, %23021 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20539 = torch.constant.int 1
%int0_20540 = torch.constant.int 0
%23023 = torch.prim.ListConstruct %int1_20539, %int0_20540 : (!torch.int, !torch.int) -> !torch.list<int>
%23024 = torch.aten.permute %822, %23023 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20541 = torch.constant.int 1
%int0_20542 = torch.constant.int 0
%23025 = torch.prim.ListConstruct %int1_20541, %int0_20542 : (!torch.int, !torch.int) -> !torch.list<int>
%23026 = torch.aten.permute %823, %23025 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_20543 = torch.constant.int 4
%23027 = torch.aten.mul.int %int4_20543, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20544 = torch.constant.int 4096
%23028 = torch.prim.ListConstruct %23027, %int4096_20544 : (!torch.int, !torch.int) -> !torch.list<int>
%23029 = torch.aten.view %22939, %23028 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23029, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23030 = torch.aten.mm %23029, %23012 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23030, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20545 = torch.constant.int 4
%int128_20546 = torch.constant.int 128
%23031 = torch.prim.ListConstruct %int4_20545, %2482, %int128_20546 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23032 = torch.aten.view %23030, %23031 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20547 = torch.constant.int 4
%23033 = torch.aten.mul.int %int4_20547, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20548 = torch.constant.int 4096
%23034 = torch.prim.ListConstruct %23033, %int4096_20548 : (!torch.int, !torch.int) -> !torch.list<int>
%23035 = torch.aten.view %22940, %23034 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23035, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23036 = torch.aten.mm %23035, %23014 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23036, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20549 = torch.constant.int 4
%int128_20550 = torch.constant.int 128
%23037 = torch.prim.ListConstruct %int4_20549, %2482, %int128_20550 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23038 = torch.aten.view %23036, %23037 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20551 = torch.constant.int 4
%23039 = torch.aten.mul.int %int4_20551, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20552 = torch.constant.int 4096
%23040 = torch.prim.ListConstruct %23039, %int4096_20552 : (!torch.int, !torch.int) -> !torch.list<int>
%23041 = torch.aten.view %22941, %23040 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23041, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23042 = torch.aten.mm %23041, %23016 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23042, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20553 = torch.constant.int 4
%int128_20554 = torch.constant.int 128
%23043 = torch.prim.ListConstruct %int4_20553, %2482, %int128_20554 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23044 = torch.aten.view %23042, %23043 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20555 = torch.constant.int 4
%23045 = torch.aten.mul.int %int4_20555, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20556 = torch.constant.int 4096
%23046 = torch.prim.ListConstruct %23045, %int4096_20556 : (!torch.int, !torch.int) -> !torch.list<int>
%23047 = torch.aten.view %22942, %23046 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23047, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23048 = torch.aten.mm %23047, %23018 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23048, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20557 = torch.constant.int 4
%int128_20558 = torch.constant.int 128
%23049 = torch.prim.ListConstruct %int4_20557, %2482, %int128_20558 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23050 = torch.aten.view %23048, %23049 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20559 = torch.constant.int 4
%23051 = torch.aten.mul.int %int4_20559, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20560 = torch.constant.int 4096
%23052 = torch.prim.ListConstruct %23051, %int4096_20560 : (!torch.int, !torch.int) -> !torch.list<int>
%23053 = torch.aten.view %22943, %23052 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23053, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23054 = torch.aten.mm %23053, %23020 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23054, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20561 = torch.constant.int 4
%int128_20562 = torch.constant.int 128
%23055 = torch.prim.ListConstruct %int4_20561, %2482, %int128_20562 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23056 = torch.aten.view %23054, %23055 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20563 = torch.constant.int 4
%23057 = torch.aten.mul.int %int4_20563, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20564 = torch.constant.int 4096
%23058 = torch.prim.ListConstruct %23057, %int4096_20564 : (!torch.int, !torch.int) -> !torch.list<int>
%23059 = torch.aten.view %22944, %23058 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23059, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23060 = torch.aten.mm %23059, %23022 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23060, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20565 = torch.constant.int 4
%int128_20566 = torch.constant.int 128
%23061 = torch.prim.ListConstruct %int4_20565, %2482, %int128_20566 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23062 = torch.aten.view %23060, %23061 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20567 = torch.constant.int 4
%23063 = torch.aten.mul.int %int4_20567, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20568 = torch.constant.int 4096
%23064 = torch.prim.ListConstruct %23063, %int4096_20568 : (!torch.int, !torch.int) -> !torch.list<int>
%23065 = torch.aten.view %22945, %23064 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23065, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23066 = torch.aten.mm %23065, %23024 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23066, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20569 = torch.constant.int 4
%int128_20570 = torch.constant.int 128
%23067 = torch.prim.ListConstruct %int4_20569, %2482, %int128_20570 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23068 = torch.aten.view %23066, %23067 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20571 = torch.constant.int 4
%23069 = torch.aten.mul.int %int4_20571, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20572 = torch.constant.int 4096
%23070 = torch.prim.ListConstruct %23069, %int4096_20572 : (!torch.int, !torch.int) -> !torch.list<int>
%23071 = torch.aten.view %22946, %23070 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23071, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23072 = torch.aten.mm %23071, %23026 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23072, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20573 = torch.constant.int 4
%int128_20574 = torch.constant.int 128
%23073 = torch.prim.ListConstruct %int4_20573, %2482, %int128_20574 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23074 = torch.aten.view %23072, %23073 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_20575 = torch.constant.int 1
%int0_20576 = torch.constant.int 0
%23075 = torch.prim.ListConstruct %int1_20575, %int0_20576 : (!torch.int, !torch.int) -> !torch.list<int>
%23076 = torch.aten.permute %824, %23075 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20577 = torch.constant.int 1
%int0_20578 = torch.constant.int 0
%23077 = torch.prim.ListConstruct %int1_20577, %int0_20578 : (!torch.int, !torch.int) -> !torch.list<int>
%23078 = torch.aten.permute %825, %23077 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20579 = torch.constant.int 1
%int0_20580 = torch.constant.int 0
%23079 = torch.prim.ListConstruct %int1_20579, %int0_20580 : (!torch.int, !torch.int) -> !torch.list<int>
%23080 = torch.aten.permute %826, %23079 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20581 = torch.constant.int 1
%int0_20582 = torch.constant.int 0
%23081 = torch.prim.ListConstruct %int1_20581, %int0_20582 : (!torch.int, !torch.int) -> !torch.list<int>
%23082 = torch.aten.permute %827, %23081 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20583 = torch.constant.int 1
%int0_20584 = torch.constant.int 0
%23083 = torch.prim.ListConstruct %int1_20583, %int0_20584 : (!torch.int, !torch.int) -> !torch.list<int>
%23084 = torch.aten.permute %828, %23083 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20585 = torch.constant.int 1
%int0_20586 = torch.constant.int 0
%23085 = torch.prim.ListConstruct %int1_20585, %int0_20586 : (!torch.int, !torch.int) -> !torch.list<int>
%23086 = torch.aten.permute %829, %23085 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20587 = torch.constant.int 1
%int0_20588 = torch.constant.int 0
%23087 = torch.prim.ListConstruct %int1_20587, %int0_20588 : (!torch.int, !torch.int) -> !torch.list<int>
%23088 = torch.aten.permute %830, %23087 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_20589 = torch.constant.int 1
%int0_20590 = torch.constant.int 0
%23089 = torch.prim.ListConstruct %int1_20589, %int0_20590 : (!torch.int, !torch.int) -> !torch.list<int>
%23090 = torch.aten.permute %831, %23089 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_20591 = torch.constant.int 4
%23091 = torch.aten.mul.int %int4_20591, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20592 = torch.constant.int 4096
%23092 = torch.prim.ListConstruct %23091, %int4096_20592 : (!torch.int, !torch.int) -> !torch.list<int>
%23093 = torch.aten.view %22939, %23092 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23093, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23094 = torch.aten.mm %23093, %23076 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23094, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20593 = torch.constant.int 4
%int128_20594 = torch.constant.int 128
%23095 = torch.prim.ListConstruct %int4_20593, %2482, %int128_20594 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23096 = torch.aten.view %23094, %23095 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20595 = torch.constant.int 4
%23097 = torch.aten.mul.int %int4_20595, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20596 = torch.constant.int 4096
%23098 = torch.prim.ListConstruct %23097, %int4096_20596 : (!torch.int, !torch.int) -> !torch.list<int>
%23099 = torch.aten.view %22940, %23098 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23099, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23100 = torch.aten.mm %23099, %23078 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23100, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20597 = torch.constant.int 4
%int128_20598 = torch.constant.int 128
%23101 = torch.prim.ListConstruct %int4_20597, %2482, %int128_20598 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23102 = torch.aten.view %23100, %23101 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20599 = torch.constant.int 4
%23103 = torch.aten.mul.int %int4_20599, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20600 = torch.constant.int 4096
%23104 = torch.prim.ListConstruct %23103, %int4096_20600 : (!torch.int, !torch.int) -> !torch.list<int>
%23105 = torch.aten.view %22941, %23104 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23105, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23106 = torch.aten.mm %23105, %23080 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23106, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20601 = torch.constant.int 4
%int128_20602 = torch.constant.int 128
%23107 = torch.prim.ListConstruct %int4_20601, %2482, %int128_20602 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23108 = torch.aten.view %23106, %23107 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20603 = torch.constant.int 4
%23109 = torch.aten.mul.int %int4_20603, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20604 = torch.constant.int 4096
%23110 = torch.prim.ListConstruct %23109, %int4096_20604 : (!torch.int, !torch.int) -> !torch.list<int>
%23111 = torch.aten.view %22942, %23110 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23111, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23112 = torch.aten.mm %23111, %23082 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23112, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20605 = torch.constant.int 4
%int128_20606 = torch.constant.int 128
%23113 = torch.prim.ListConstruct %int4_20605, %2482, %int128_20606 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23114 = torch.aten.view %23112, %23113 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20607 = torch.constant.int 4
%23115 = torch.aten.mul.int %int4_20607, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20608 = torch.constant.int 4096
%23116 = torch.prim.ListConstruct %23115, %int4096_20608 : (!torch.int, !torch.int) -> !torch.list<int>
%23117 = torch.aten.view %22943, %23116 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23117, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23118 = torch.aten.mm %23117, %23084 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23118, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20609 = torch.constant.int 4
%int128_20610 = torch.constant.int 128
%23119 = torch.prim.ListConstruct %int4_20609, %2482, %int128_20610 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23120 = torch.aten.view %23118, %23119 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20611 = torch.constant.int 4
%23121 = torch.aten.mul.int %int4_20611, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20612 = torch.constant.int 4096
%23122 = torch.prim.ListConstruct %23121, %int4096_20612 : (!torch.int, !torch.int) -> !torch.list<int>
%23123 = torch.aten.view %22944, %23122 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23123, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23124 = torch.aten.mm %23123, %23086 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23124, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20613 = torch.constant.int 4
%int128_20614 = torch.constant.int 128
%23125 = torch.prim.ListConstruct %int4_20613, %2482, %int128_20614 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23126 = torch.aten.view %23124, %23125 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20615 = torch.constant.int 4
%23127 = torch.aten.mul.int %int4_20615, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20616 = torch.constant.int 4096
%23128 = torch.prim.ListConstruct %23127, %int4096_20616 : (!torch.int, !torch.int) -> !torch.list<int>
%23129 = torch.aten.view %22945, %23128 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23129, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23130 = torch.aten.mm %23129, %23088 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23130, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20617 = torch.constant.int 4
%int128_20618 = torch.constant.int 128
%23131 = torch.prim.ListConstruct %int4_20617, %2482, %int128_20618 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23132 = torch.aten.view %23130, %23131 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20619 = torch.constant.int 4
%23133 = torch.aten.mul.int %int4_20619, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_20620 = torch.constant.int 4096
%23134 = torch.prim.ListConstruct %23133, %int4096_20620 : (!torch.int, !torch.int) -> !torch.list<int>
%23135 = torch.aten.view %22946, %23134 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23135, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%23136 = torch.aten.mm %23135, %23090 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %23136, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_20621 = torch.constant.int 4
%int128_20622 = torch.constant.int 128
%23137 = torch.prim.ListConstruct %int4_20621, %2482, %int128_20622 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23138 = torch.aten.view %23136, %23137 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %23138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_20623 = torch.constant.int 4
%int4_20624 = torch.constant.int 4
%int128_20625 = torch.constant.int 128
%23139 = torch.prim.ListConstruct %int4_20623, %2482, %int4_20624, %int128_20625 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23140 = torch.aten.view %22968, %23139 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20626 = torch.constant.int 4
%int4_20627 = torch.constant.int 4
%int128_20628 = torch.constant.int 128
%23141 = torch.prim.ListConstruct %int4_20626, %2482, %int4_20627, %int128_20628 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23142 = torch.aten.view %22974, %23141 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20629 = torch.constant.int 4
%int4_20630 = torch.constant.int 4
%int128_20631 = torch.constant.int 128
%23143 = torch.prim.ListConstruct %int4_20629, %2482, %int4_20630, %int128_20631 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23144 = torch.aten.view %22980, %23143 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20632 = torch.constant.int 4
%int4_20633 = torch.constant.int 4
%int128_20634 = torch.constant.int 128
%23145 = torch.prim.ListConstruct %int4_20632, %2482, %int4_20633, %int128_20634 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23146 = torch.aten.view %22986, %23145 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20635 = torch.constant.int 4
%int4_20636 = torch.constant.int 4
%int128_20637 = torch.constant.int 128
%23147 = torch.prim.ListConstruct %int4_20635, %2482, %int4_20636, %int128_20637 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23148 = torch.aten.view %22992, %23147 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20638 = torch.constant.int 4
%int4_20639 = torch.constant.int 4
%int128_20640 = torch.constant.int 128
%23149 = torch.prim.ListConstruct %int4_20638, %2482, %int4_20639, %int128_20640 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23150 = torch.aten.view %22998, %23149 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20641 = torch.constant.int 4
%int4_20642 = torch.constant.int 4
%int128_20643 = torch.constant.int 128
%23151 = torch.prim.ListConstruct %int4_20641, %2482, %int4_20642, %int128_20643 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23152 = torch.aten.view %23004, %23151 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20644 = torch.constant.int 4
%int4_20645 = torch.constant.int 4
%int128_20646 = torch.constant.int 128
%23153 = torch.prim.ListConstruct %int4_20644, %2482, %int4_20645, %int128_20646 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23154 = torch.aten.view %23010, %23153 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_20647 = torch.constant.int 4
%int1_20648 = torch.constant.int 1
%int128_20649 = torch.constant.int 128
%23155 = torch.prim.ListConstruct %int4_20647, %2482, %int1_20648, %int128_20649 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23156 = torch.aten.view %23032, %23155 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20650 = torch.constant.int 4
%int1_20651 = torch.constant.int 1
%int128_20652 = torch.constant.int 128
%23157 = torch.prim.ListConstruct %int4_20650, %2482, %int1_20651, %int128_20652 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23158 = torch.aten.view %23038, %23157 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20653 = torch.constant.int 4
%int1_20654 = torch.constant.int 1
%int128_20655 = torch.constant.int 128
%23159 = torch.prim.ListConstruct %int4_20653, %2482, %int1_20654, %int128_20655 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23160 = torch.aten.view %23044, %23159 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20656 = torch.constant.int 4
%int1_20657 = torch.constant.int 1
%int128_20658 = torch.constant.int 128
%23161 = torch.prim.ListConstruct %int4_20656, %2482, %int1_20657, %int128_20658 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23162 = torch.aten.view %23050, %23161 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20659 = torch.constant.int 4
%int1_20660 = torch.constant.int 1
%int128_20661 = torch.constant.int 128
%23163 = torch.prim.ListConstruct %int4_20659, %2482, %int1_20660, %int128_20661 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23164 = torch.aten.view %23056, %23163 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20662 = torch.constant.int 4
%int1_20663 = torch.constant.int 1
%int128_20664 = torch.constant.int 128
%23165 = torch.prim.ListConstruct %int4_20662, %2482, %int1_20663, %int128_20664 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23166 = torch.aten.view %23062, %23165 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20665 = torch.constant.int 4
%int1_20666 = torch.constant.int 1
%int128_20667 = torch.constant.int 128
%23167 = torch.prim.ListConstruct %int4_20665, %2482, %int1_20666, %int128_20667 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23168 = torch.aten.view %23068, %23167 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20668 = torch.constant.int 4
%int1_20669 = torch.constant.int 1
%int128_20670 = torch.constant.int 128
%23169 = torch.prim.ListConstruct %int4_20668, %2482, %int1_20669, %int128_20670 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23170 = torch.aten.view %23074, %23169 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20671 = torch.constant.int 4
%int1_20672 = torch.constant.int 1
%int128_20673 = torch.constant.int 128
%23171 = torch.prim.ListConstruct %int4_20671, %2482, %int1_20672, %int128_20673 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23172 = torch.aten.view %23096, %23171 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20674 = torch.constant.int 4
%int1_20675 = torch.constant.int 1
%int128_20676 = torch.constant.int 128
%23173 = torch.prim.ListConstruct %int4_20674, %2482, %int1_20675, %int128_20676 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23174 = torch.aten.view %23102, %23173 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20677 = torch.constant.int 4
%int1_20678 = torch.constant.int 1
%int128_20679 = torch.constant.int 128
%23175 = torch.prim.ListConstruct %int4_20677, %2482, %int1_20678, %int128_20679 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23176 = torch.aten.view %23108, %23175 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20680 = torch.constant.int 4
%int1_20681 = torch.constant.int 1
%int128_20682 = torch.constant.int 128
%23177 = torch.prim.ListConstruct %int4_20680, %2482, %int1_20681, %int128_20682 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23178 = torch.aten.view %23114, %23177 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20683 = torch.constant.int 4
%int1_20684 = torch.constant.int 1
%int128_20685 = torch.constant.int 128
%23179 = torch.prim.ListConstruct %int4_20683, %2482, %int1_20684, %int128_20685 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23180 = torch.aten.view %23120, %23179 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20686 = torch.constant.int 4
%int1_20687 = torch.constant.int 1
%int128_20688 = torch.constant.int 128
%23181 = torch.prim.ListConstruct %int4_20686, %2482, %int1_20687, %int128_20688 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23182 = torch.aten.view %23126, %23181 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20689 = torch.constant.int 4
%int1_20690 = torch.constant.int 1
%int128_20691 = torch.constant.int 128
%23183 = torch.prim.ListConstruct %int4_20689, %2482, %int1_20690, %int128_20691 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23184 = torch.aten.view %23132, %23183 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_20692 = torch.constant.int 4
%int1_20693 = torch.constant.int 1
%int128_20694 = torch.constant.int 128
%23185 = torch.prim.ListConstruct %int4_20692, %2482, %int1_20693, %int128_20694 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23186 = torch.aten.view %23138, %23185 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_20695 = torch.constant.int 131072
%none_20696 = torch.constant.none
%none_20697 = torch.constant.none
%cpu_20698 = torch.constant.device "cpu"
%false_20699 = torch.constant.bool false
%23187 = torch.aten.arange %int131072_20695, %none_20696, %none_20697, %cpu_20698, %false_20699 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_20700 = torch.constant.int 0
%int128_20701 = torch.constant.int 128
%int2_20702 = torch.constant.int 2
%none_20703 = torch.constant.none
%none_20704 = torch.constant.none
%cpu_20705 = torch.constant.device "cpu"
%false_20706 = torch.constant.bool false
%23188 = torch.aten.arange.start_step %int0_20700, %int128_20701, %int2_20702, %none_20703, %none_20704, %cpu_20705, %false_20706 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_20707 = torch.constant.int 0
%int0_20708 = torch.constant.int 0
%int64_20709 = torch.constant.int 64
%int1_20710 = torch.constant.int 1
%23189 = torch.aten.slice.Tensor %23188, %int0_20707, %int0_20708, %int64_20709, %int1_20710 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_20711 = torch.constant.int 6
%23190 = torch.prims.convert_element_type %23189, %int6_20711 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_20712 = torch.constant.int 128
%23191 = torch.aten.div.Scalar %23190, %int128_20712 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_20713 = torch.constant.float 5.000000e+05
%23192 = torch.aten.pow.Scalar %float5.000000e05_20713, %23191 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%23193 = torch.aten.reciprocal %23192 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_20714 = torch.constant.float 1.000000e+00
%23194 = torch.aten.mul.Scalar %23193, %float1.000000e00_20714 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_20715 = torch.constant.int 131072
%int1_20716 = torch.constant.int 1
%23195 = torch.prim.ListConstruct %int131072_20715, %int1_20716 : (!torch.int, !torch.int) -> !torch.list<int>
%23196 = torch.aten.view %23187, %23195 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%23197 = torch.aten.mul.Tensor %23196, %23194 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%23198 = torch.aten.cos %23197 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%23199 = torch.aten.sin %23197 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%23200 = torch.aten.complex %23198, %23199 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%23201 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23202 = flow.tensor.transfer %23201 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%23203 = torch_c.from_builtin_tensor %23202 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23204 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23205 = flow.tensor.transfer %23204 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%23206 = torch_c.from_builtin_tensor %23205 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23207 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23208 = flow.tensor.transfer %23207 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%23209 = torch_c.from_builtin_tensor %23208 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23210 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23211 = flow.tensor.transfer %23210 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%23212 = torch_c.from_builtin_tensor %23211 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23213 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23214 = flow.tensor.transfer %23213 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%23215 = torch_c.from_builtin_tensor %23214 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23216 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23217 = flow.tensor.transfer %23216 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%23218 = torch_c.from_builtin_tensor %23217 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23219 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23220 = flow.tensor.transfer %23219 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%23221 = torch_c.from_builtin_tensor %23220 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23222 = torch_c.to_builtin_tensor %23200 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23223 = flow.tensor.transfer %23222 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%23224 = torch_c.from_builtin_tensor %23223 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_20717 = torch.constant.int 1
%23225 = torch.aten.size.int %22968, %int1_20717 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20718 = torch.constant.int 0
%23226 = torch.aten.add.int %int0_20718, %23225 : !torch.int, !torch.int -> !torch.int
%int0_20719 = torch.constant.int 0
%int0_20720 = torch.constant.int 0
%int1_20721 = torch.constant.int 1
%23227 = torch.aten.slice.Tensor %23203, %int0_20719, %int0_20720, %23226, %int1_20721 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23227, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20722 = torch.constant.int 1
%int0_20723 = torch.constant.int 0
%int9223372036854775807_20724 = torch.constant.int 9223372036854775807
%int1_20725 = torch.constant.int 1
%23228 = torch.aten.slice.Tensor %23227, %int1_20722, %int0_20723, %int9223372036854775807_20724, %int1_20725 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23228, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20726 = torch.constant.int 0
%23229 = torch.aten.unsqueeze %23228, %int0_20726 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23229, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20727 = torch.constant.int 2
%23230 = torch.aten.unsqueeze %23229, %int2_20727 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23230, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20728 = torch.constant.int 3
%int0_20729 = torch.constant.int 0
%int9223372036854775807_20730 = torch.constant.int 9223372036854775807
%int1_20731 = torch.constant.int 1
%23231 = torch.aten.slice.Tensor %23230, %int3_20728, %int0_20729, %int9223372036854775807_20730, %int1_20731 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23231, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23232 = torch_c.to_builtin_tensor %23140 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20732 = arith.constant 1 : index
%dim_20733 = tensor.dim %23232, %c1_20732 : tensor<4x?x4x128xf16>
%23233 = flow.tensor.bitcast %23232 : tensor<4x?x4x128xf16>{%dim_20733} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20733}
%23234 = torch_c.from_builtin_tensor %23233 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23235 = torch.aten.mul.Tensor %23234, %23231 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23236 = torch_c.to_builtin_tensor %23235 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20734 = arith.constant 1 : index
%dim_20735 = tensor.dim %23236, %c1_20734 : tensor<4x?x4x64xcomplex<f32>>
%23237 = flow.tensor.bitcast %23236 : tensor<4x?x4x64xcomplex<f32>>{%dim_20735} -> tensor<4x?x4x128xf32>{%dim_20735}
%23238 = torch_c.from_builtin_tensor %23237 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20736 = torch.constant.int 5
%23239 = torch.prims.convert_element_type %23238, %int5_20736 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20737 = torch.constant.int 1
%23240 = torch.aten.size.int %22974, %int1_20737 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20738 = torch.constant.int 0
%23241 = torch.aten.add.int %int0_20738, %23240 : !torch.int, !torch.int -> !torch.int
%int0_20739 = torch.constant.int 0
%int0_20740 = torch.constant.int 0
%int1_20741 = torch.constant.int 1
%23242 = torch.aten.slice.Tensor %23206, %int0_20739, %int0_20740, %23241, %int1_20741 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23242, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20742 = torch.constant.int 1
%int0_20743 = torch.constant.int 0
%int9223372036854775807_20744 = torch.constant.int 9223372036854775807
%int1_20745 = torch.constant.int 1
%23243 = torch.aten.slice.Tensor %23242, %int1_20742, %int0_20743, %int9223372036854775807_20744, %int1_20745 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23243, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20746 = torch.constant.int 0
%23244 = torch.aten.unsqueeze %23243, %int0_20746 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23244, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20747 = torch.constant.int 2
%23245 = torch.aten.unsqueeze %23244, %int2_20747 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23245, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20748 = torch.constant.int 3
%int0_20749 = torch.constant.int 0
%int9223372036854775807_20750 = torch.constant.int 9223372036854775807
%int1_20751 = torch.constant.int 1
%23246 = torch.aten.slice.Tensor %23245, %int3_20748, %int0_20749, %int9223372036854775807_20750, %int1_20751 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23246, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23247 = torch_c.to_builtin_tensor %23142 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20752 = arith.constant 1 : index
%dim_20753 = tensor.dim %23247, %c1_20752 : tensor<4x?x4x128xf16>
%23248 = flow.tensor.bitcast %23247 : tensor<4x?x4x128xf16>{%dim_20753} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20753}
%23249 = torch_c.from_builtin_tensor %23248 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23250 = torch.aten.mul.Tensor %23249, %23246 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23251 = torch_c.to_builtin_tensor %23250 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20754 = arith.constant 1 : index
%dim_20755 = tensor.dim %23251, %c1_20754 : tensor<4x?x4x64xcomplex<f32>>
%23252 = flow.tensor.bitcast %23251 : tensor<4x?x4x64xcomplex<f32>>{%dim_20755} -> tensor<4x?x4x128xf32>{%dim_20755}
%23253 = torch_c.from_builtin_tensor %23252 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20756 = torch.constant.int 5
%23254 = torch.prims.convert_element_type %23253, %int5_20756 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20757 = torch.constant.int 1
%23255 = torch.aten.size.int %22980, %int1_20757 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20758 = torch.constant.int 0
%23256 = torch.aten.add.int %int0_20758, %23255 : !torch.int, !torch.int -> !torch.int
%int0_20759 = torch.constant.int 0
%int0_20760 = torch.constant.int 0
%int1_20761 = torch.constant.int 1
%23257 = torch.aten.slice.Tensor %23209, %int0_20759, %int0_20760, %23256, %int1_20761 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23257, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20762 = torch.constant.int 1
%int0_20763 = torch.constant.int 0
%int9223372036854775807_20764 = torch.constant.int 9223372036854775807
%int1_20765 = torch.constant.int 1
%23258 = torch.aten.slice.Tensor %23257, %int1_20762, %int0_20763, %int9223372036854775807_20764, %int1_20765 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23258, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20766 = torch.constant.int 0
%23259 = torch.aten.unsqueeze %23258, %int0_20766 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23259, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20767 = torch.constant.int 2
%23260 = torch.aten.unsqueeze %23259, %int2_20767 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23260, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20768 = torch.constant.int 3
%int0_20769 = torch.constant.int 0
%int9223372036854775807_20770 = torch.constant.int 9223372036854775807
%int1_20771 = torch.constant.int 1
%23261 = torch.aten.slice.Tensor %23260, %int3_20768, %int0_20769, %int9223372036854775807_20770, %int1_20771 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23261, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23262 = torch_c.to_builtin_tensor %23144 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20772 = arith.constant 1 : index
%dim_20773 = tensor.dim %23262, %c1_20772 : tensor<4x?x4x128xf16>
%23263 = flow.tensor.bitcast %23262 : tensor<4x?x4x128xf16>{%dim_20773} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20773}
%23264 = torch_c.from_builtin_tensor %23263 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23265 = torch.aten.mul.Tensor %23264, %23261 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23266 = torch_c.to_builtin_tensor %23265 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20774 = arith.constant 1 : index
%dim_20775 = tensor.dim %23266, %c1_20774 : tensor<4x?x4x64xcomplex<f32>>
%23267 = flow.tensor.bitcast %23266 : tensor<4x?x4x64xcomplex<f32>>{%dim_20775} -> tensor<4x?x4x128xf32>{%dim_20775}
%23268 = torch_c.from_builtin_tensor %23267 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20776 = torch.constant.int 5
%23269 = torch.prims.convert_element_type %23268, %int5_20776 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20777 = torch.constant.int 1
%23270 = torch.aten.size.int %22986, %int1_20777 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20778 = torch.constant.int 0
%23271 = torch.aten.add.int %int0_20778, %23270 : !torch.int, !torch.int -> !torch.int
%int0_20779 = torch.constant.int 0
%int0_20780 = torch.constant.int 0
%int1_20781 = torch.constant.int 1
%23272 = torch.aten.slice.Tensor %23212, %int0_20779, %int0_20780, %23271, %int1_20781 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23272, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20782 = torch.constant.int 1
%int0_20783 = torch.constant.int 0
%int9223372036854775807_20784 = torch.constant.int 9223372036854775807
%int1_20785 = torch.constant.int 1
%23273 = torch.aten.slice.Tensor %23272, %int1_20782, %int0_20783, %int9223372036854775807_20784, %int1_20785 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23273, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20786 = torch.constant.int 0
%23274 = torch.aten.unsqueeze %23273, %int0_20786 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23274, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20787 = torch.constant.int 2
%23275 = torch.aten.unsqueeze %23274, %int2_20787 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23275, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20788 = torch.constant.int 3
%int0_20789 = torch.constant.int 0
%int9223372036854775807_20790 = torch.constant.int 9223372036854775807
%int1_20791 = torch.constant.int 1
%23276 = torch.aten.slice.Tensor %23275, %int3_20788, %int0_20789, %int9223372036854775807_20790, %int1_20791 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23276, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23277 = torch_c.to_builtin_tensor %23146 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20792 = arith.constant 1 : index
%dim_20793 = tensor.dim %23277, %c1_20792 : tensor<4x?x4x128xf16>
%23278 = flow.tensor.bitcast %23277 : tensor<4x?x4x128xf16>{%dim_20793} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20793}
%23279 = torch_c.from_builtin_tensor %23278 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23280 = torch.aten.mul.Tensor %23279, %23276 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23281 = torch_c.to_builtin_tensor %23280 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20794 = arith.constant 1 : index
%dim_20795 = tensor.dim %23281, %c1_20794 : tensor<4x?x4x64xcomplex<f32>>
%23282 = flow.tensor.bitcast %23281 : tensor<4x?x4x64xcomplex<f32>>{%dim_20795} -> tensor<4x?x4x128xf32>{%dim_20795}
%23283 = torch_c.from_builtin_tensor %23282 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20796 = torch.constant.int 5
%23284 = torch.prims.convert_element_type %23283, %int5_20796 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20797 = torch.constant.int 1
%23285 = torch.aten.size.int %22992, %int1_20797 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20798 = torch.constant.int 0
%23286 = torch.aten.add.int %int0_20798, %23285 : !torch.int, !torch.int -> !torch.int
%int0_20799 = torch.constant.int 0
%int0_20800 = torch.constant.int 0
%int1_20801 = torch.constant.int 1
%23287 = torch.aten.slice.Tensor %23215, %int0_20799, %int0_20800, %23286, %int1_20801 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23287, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20802 = torch.constant.int 1
%int0_20803 = torch.constant.int 0
%int9223372036854775807_20804 = torch.constant.int 9223372036854775807
%int1_20805 = torch.constant.int 1
%23288 = torch.aten.slice.Tensor %23287, %int1_20802, %int0_20803, %int9223372036854775807_20804, %int1_20805 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23288, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20806 = torch.constant.int 0
%23289 = torch.aten.unsqueeze %23288, %int0_20806 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23289, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20807 = torch.constant.int 2
%23290 = torch.aten.unsqueeze %23289, %int2_20807 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23290, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20808 = torch.constant.int 3
%int0_20809 = torch.constant.int 0
%int9223372036854775807_20810 = torch.constant.int 9223372036854775807
%int1_20811 = torch.constant.int 1
%23291 = torch.aten.slice.Tensor %23290, %int3_20808, %int0_20809, %int9223372036854775807_20810, %int1_20811 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23291, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23292 = torch_c.to_builtin_tensor %23148 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20812 = arith.constant 1 : index
%dim_20813 = tensor.dim %23292, %c1_20812 : tensor<4x?x4x128xf16>
%23293 = flow.tensor.bitcast %23292 : tensor<4x?x4x128xf16>{%dim_20813} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20813}
%23294 = torch_c.from_builtin_tensor %23293 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23295 = torch.aten.mul.Tensor %23294, %23291 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23296 = torch_c.to_builtin_tensor %23295 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20814 = arith.constant 1 : index
%dim_20815 = tensor.dim %23296, %c1_20814 : tensor<4x?x4x64xcomplex<f32>>
%23297 = flow.tensor.bitcast %23296 : tensor<4x?x4x64xcomplex<f32>>{%dim_20815} -> tensor<4x?x4x128xf32>{%dim_20815}
%23298 = torch_c.from_builtin_tensor %23297 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20816 = torch.constant.int 5
%23299 = torch.prims.convert_element_type %23298, %int5_20816 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20817 = torch.constant.int 1
%23300 = torch.aten.size.int %22998, %int1_20817 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20818 = torch.constant.int 0
%23301 = torch.aten.add.int %int0_20818, %23300 : !torch.int, !torch.int -> !torch.int
%int0_20819 = torch.constant.int 0
%int0_20820 = torch.constant.int 0
%int1_20821 = torch.constant.int 1
%23302 = torch.aten.slice.Tensor %23218, %int0_20819, %int0_20820, %23301, %int1_20821 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23302, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20822 = torch.constant.int 1
%int0_20823 = torch.constant.int 0
%int9223372036854775807_20824 = torch.constant.int 9223372036854775807
%int1_20825 = torch.constant.int 1
%23303 = torch.aten.slice.Tensor %23302, %int1_20822, %int0_20823, %int9223372036854775807_20824, %int1_20825 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23303, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20826 = torch.constant.int 0
%23304 = torch.aten.unsqueeze %23303, %int0_20826 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23304, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20827 = torch.constant.int 2
%23305 = torch.aten.unsqueeze %23304, %int2_20827 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23305, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20828 = torch.constant.int 3
%int0_20829 = torch.constant.int 0
%int9223372036854775807_20830 = torch.constant.int 9223372036854775807
%int1_20831 = torch.constant.int 1
%23306 = torch.aten.slice.Tensor %23305, %int3_20828, %int0_20829, %int9223372036854775807_20830, %int1_20831 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23306, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23307 = torch_c.to_builtin_tensor %23150 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20832 = arith.constant 1 : index
%dim_20833 = tensor.dim %23307, %c1_20832 : tensor<4x?x4x128xf16>
%23308 = flow.tensor.bitcast %23307 : tensor<4x?x4x128xf16>{%dim_20833} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20833}
%23309 = torch_c.from_builtin_tensor %23308 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23310 = torch.aten.mul.Tensor %23309, %23306 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23311 = torch_c.to_builtin_tensor %23310 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20834 = arith.constant 1 : index
%dim_20835 = tensor.dim %23311, %c1_20834 : tensor<4x?x4x64xcomplex<f32>>
%23312 = flow.tensor.bitcast %23311 : tensor<4x?x4x64xcomplex<f32>>{%dim_20835} -> tensor<4x?x4x128xf32>{%dim_20835}
%23313 = torch_c.from_builtin_tensor %23312 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20836 = torch.constant.int 5
%23314 = torch.prims.convert_element_type %23313, %int5_20836 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20837 = torch.constant.int 1
%23315 = torch.aten.size.int %23004, %int1_20837 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20838 = torch.constant.int 0
%23316 = torch.aten.add.int %int0_20838, %23315 : !torch.int, !torch.int -> !torch.int
%int0_20839 = torch.constant.int 0
%int0_20840 = torch.constant.int 0
%int1_20841 = torch.constant.int 1
%23317 = torch.aten.slice.Tensor %23221, %int0_20839, %int0_20840, %23316, %int1_20841 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23317, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20842 = torch.constant.int 1
%int0_20843 = torch.constant.int 0
%int9223372036854775807_20844 = torch.constant.int 9223372036854775807
%int1_20845 = torch.constant.int 1
%23318 = torch.aten.slice.Tensor %23317, %int1_20842, %int0_20843, %int9223372036854775807_20844, %int1_20845 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23318, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20846 = torch.constant.int 0
%23319 = torch.aten.unsqueeze %23318, %int0_20846 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23319, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20847 = torch.constant.int 2
%23320 = torch.aten.unsqueeze %23319, %int2_20847 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23320, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20848 = torch.constant.int 3
%int0_20849 = torch.constant.int 0
%int9223372036854775807_20850 = torch.constant.int 9223372036854775807
%int1_20851 = torch.constant.int 1
%23321 = torch.aten.slice.Tensor %23320, %int3_20848, %int0_20849, %int9223372036854775807_20850, %int1_20851 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23321, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23322 = torch_c.to_builtin_tensor %23152 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20852 = arith.constant 1 : index
%dim_20853 = tensor.dim %23322, %c1_20852 : tensor<4x?x4x128xf16>
%23323 = flow.tensor.bitcast %23322 : tensor<4x?x4x128xf16>{%dim_20853} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20853}
%23324 = torch_c.from_builtin_tensor %23323 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23325 = torch.aten.mul.Tensor %23324, %23321 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23326 = torch_c.to_builtin_tensor %23325 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20854 = arith.constant 1 : index
%dim_20855 = tensor.dim %23326, %c1_20854 : tensor<4x?x4x64xcomplex<f32>>
%23327 = flow.tensor.bitcast %23326 : tensor<4x?x4x64xcomplex<f32>>{%dim_20855} -> tensor<4x?x4x128xf32>{%dim_20855}
%23328 = torch_c.from_builtin_tensor %23327 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20856 = torch.constant.int 5
%23329 = torch.prims.convert_element_type %23328, %int5_20856 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_20857 = torch.constant.int 1
%23330 = torch.aten.size.int %23010, %int1_20857 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_20858 = torch.constant.int 0
%23331 = torch.aten.add.int %int0_20858, %23330 : !torch.int, !torch.int -> !torch.int
%int0_20859 = torch.constant.int 0
%int0_20860 = torch.constant.int 0
%int1_20861 = torch.constant.int 1
%23332 = torch.aten.slice.Tensor %23224, %int0_20859, %int0_20860, %23331, %int1_20861 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23332, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20862 = torch.constant.int 1
%int0_20863 = torch.constant.int 0
%int9223372036854775807_20864 = torch.constant.int 9223372036854775807
%int1_20865 = torch.constant.int 1
%23333 = torch.aten.slice.Tensor %23332, %int1_20862, %int0_20863, %int9223372036854775807_20864, %int1_20865 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23333, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20866 = torch.constant.int 0
%23334 = torch.aten.unsqueeze %23333, %int0_20866 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23334, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20867 = torch.constant.int 2
%23335 = torch.aten.unsqueeze %23334, %int2_20867 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23335, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20868 = torch.constant.int 3
%int0_20869 = torch.constant.int 0
%int9223372036854775807_20870 = torch.constant.int 9223372036854775807
%int1_20871 = torch.constant.int 1
%23336 = torch.aten.slice.Tensor %23335, %int3_20868, %int0_20869, %int9223372036854775807_20870, %int1_20871 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23336, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23337 = torch_c.to_builtin_tensor %23154 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_20872 = arith.constant 1 : index
%dim_20873 = tensor.dim %23337, %c1_20872 : tensor<4x?x4x128xf16>
%23338 = flow.tensor.bitcast %23337 : tensor<4x?x4x128xf16>{%dim_20873} -> tensor<4x?x4x64xcomplex<f16>>{%dim_20873}
%23339 = torch_c.from_builtin_tensor %23338 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %23339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%23340 = torch.aten.mul.Tensor %23339, %23336 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %23340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%23341 = torch_c.to_builtin_tensor %23340 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_20874 = arith.constant 1 : index
%dim_20875 = tensor.dim %23341, %c1_20874 : tensor<4x?x4x64xcomplex<f32>>
%23342 = flow.tensor.bitcast %23341 : tensor<4x?x4x64xcomplex<f32>>{%dim_20875} -> tensor<4x?x4x128xf32>{%dim_20875}
%23343 = torch_c.from_builtin_tensor %23342 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %23343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_20876 = torch.constant.int 5
%23344 = torch.prims.convert_element_type %23343, %int5_20876 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_20877 = torch.constant.int 131072
%none_20878 = torch.constant.none
%none_20879 = torch.constant.none
%cpu_20880 = torch.constant.device "cpu"
%false_20881 = torch.constant.bool false
%23345 = torch.aten.arange %int131072_20877, %none_20878, %none_20879, %cpu_20880, %false_20881 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_20882 = torch.constant.int 0
%int128_20883 = torch.constant.int 128
%int2_20884 = torch.constant.int 2
%none_20885 = torch.constant.none
%none_20886 = torch.constant.none
%cpu_20887 = torch.constant.device "cpu"
%false_20888 = torch.constant.bool false
%23346 = torch.aten.arange.start_step %int0_20882, %int128_20883, %int2_20884, %none_20885, %none_20886, %cpu_20887, %false_20888 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_20889 = torch.constant.int 0
%int0_20890 = torch.constant.int 0
%int64_20891 = torch.constant.int 64
%int1_20892 = torch.constant.int 1
%23347 = torch.aten.slice.Tensor %23346, %int0_20889, %int0_20890, %int64_20891, %int1_20892 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_20893 = torch.constant.int 6
%23348 = torch.prims.convert_element_type %23347, %int6_20893 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_20894 = torch.constant.int 128
%23349 = torch.aten.div.Scalar %23348, %int128_20894 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_20895 = torch.constant.float 5.000000e+05
%23350 = torch.aten.pow.Scalar %float5.000000e05_20895, %23349 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%23351 = torch.aten.reciprocal %23350 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_20896 = torch.constant.float 1.000000e+00
%23352 = torch.aten.mul.Scalar %23351, %float1.000000e00_20896 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_20897 = torch.constant.int 131072
%int1_20898 = torch.constant.int 1
%23353 = torch.prim.ListConstruct %int131072_20897, %int1_20898 : (!torch.int, !torch.int) -> !torch.list<int>
%23354 = torch.aten.view %23345, %23353 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%23355 = torch.aten.mul.Tensor %23354, %23352 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%23356 = torch.aten.cos %23355 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%23357 = torch.aten.sin %23355 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%23358 = torch.aten.complex %23356, %23357 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%23359 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23360 = flow.tensor.transfer %23359 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%23361 = torch_c.from_builtin_tensor %23360 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23362 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23363 = flow.tensor.transfer %23362 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%23364 = torch_c.from_builtin_tensor %23363 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23365 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23366 = flow.tensor.transfer %23365 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%23367 = torch_c.from_builtin_tensor %23366 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23368 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23369 = flow.tensor.transfer %23368 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%23370 = torch_c.from_builtin_tensor %23369 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23371 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23372 = flow.tensor.transfer %23371 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%23373 = torch_c.from_builtin_tensor %23372 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23374 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23375 = flow.tensor.transfer %23374 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%23376 = torch_c.from_builtin_tensor %23375 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23377 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23378 = flow.tensor.transfer %23377 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%23379 = torch_c.from_builtin_tensor %23378 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%23380 = torch_c.to_builtin_tensor %23358 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%23381 = flow.tensor.transfer %23380 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%23382 = torch_c.from_builtin_tensor %23381 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_20899 = torch.constant.int 1
%23383 = torch.aten.size.int %23032, %int1_20899 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_20900 = torch.constant.int 0
%23384 = torch.aten.add.int %int0_20900, %23383 : !torch.int, !torch.int -> !torch.int
%int0_20901 = torch.constant.int 0
%int0_20902 = torch.constant.int 0
%int1_20903 = torch.constant.int 1
%23385 = torch.aten.slice.Tensor %23361, %int0_20901, %int0_20902, %23384, %int1_20903 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23385, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20904 = torch.constant.int 1
%int0_20905 = torch.constant.int 0
%int9223372036854775807_20906 = torch.constant.int 9223372036854775807
%int1_20907 = torch.constant.int 1
%23386 = torch.aten.slice.Tensor %23385, %int1_20904, %int0_20905, %int9223372036854775807_20906, %int1_20907 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23386, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20908 = torch.constant.int 0
%23387 = torch.aten.unsqueeze %23386, %int0_20908 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23387, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20909 = torch.constant.int 2
%23388 = torch.aten.unsqueeze %23387, %int2_20909 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23388, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20910 = torch.constant.int 3
%int0_20911 = torch.constant.int 0
%int9223372036854775807_20912 = torch.constant.int 9223372036854775807
%int1_20913 = torch.constant.int 1
%23389 = torch.aten.slice.Tensor %23388, %int3_20910, %int0_20911, %int9223372036854775807_20912, %int1_20913 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23389, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23390 = torch_c.to_builtin_tensor %23156 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_20914 = arith.constant 1 : index
%dim_20915 = tensor.dim %23390, %c1_20914 : tensor<4x?x1x128xf16>
%23391 = flow.tensor.bitcast %23390 : tensor<4x?x1x128xf16>{%dim_20915} -> tensor<4x?x1x64xcomplex<f16>>{%dim_20915}
%23392 = torch_c.from_builtin_tensor %23391 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23393 = torch.aten.mul.Tensor %23392, %23389 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23394 = torch_c.to_builtin_tensor %23393 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_20916 = arith.constant 1 : index
%dim_20917 = tensor.dim %23394, %c1_20916 : tensor<4x?x1x64xcomplex<f32>>
%23395 = flow.tensor.bitcast %23394 : tensor<4x?x1x64xcomplex<f32>>{%dim_20917} -> tensor<4x?x1x128xf32>{%dim_20917}
%23396 = torch_c.from_builtin_tensor %23395 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_20918 = torch.constant.int 5
%23397 = torch.prims.convert_element_type %23396, %int5_20918 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_20919 = torch.constant.int 1
%23398 = torch.aten.size.int %23038, %int1_20919 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_20920 = torch.constant.int 0
%23399 = torch.aten.add.int %int0_20920, %23398 : !torch.int, !torch.int -> !torch.int
%int0_20921 = torch.constant.int 0
%int0_20922 = torch.constant.int 0
%int1_20923 = torch.constant.int 1
%23400 = torch.aten.slice.Tensor %23364, %int0_20921, %int0_20922, %23399, %int1_20923 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23400, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20924 = torch.constant.int 1
%int0_20925 = torch.constant.int 0
%int9223372036854775807_20926 = torch.constant.int 9223372036854775807
%int1_20927 = torch.constant.int 1
%23401 = torch.aten.slice.Tensor %23400, %int1_20924, %int0_20925, %int9223372036854775807_20926, %int1_20927 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23401, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20928 = torch.constant.int 0
%23402 = torch.aten.unsqueeze %23401, %int0_20928 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23402, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20929 = torch.constant.int 2
%23403 = torch.aten.unsqueeze %23402, %int2_20929 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23403, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20930 = torch.constant.int 3
%int0_20931 = torch.constant.int 0
%int9223372036854775807_20932 = torch.constant.int 9223372036854775807
%int1_20933 = torch.constant.int 1
%23404 = torch.aten.slice.Tensor %23403, %int3_20930, %int0_20931, %int9223372036854775807_20932, %int1_20933 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23404, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23405 = torch_c.to_builtin_tensor %23158 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_20934 = arith.constant 1 : index
%dim_20935 = tensor.dim %23405, %c1_20934 : tensor<4x?x1x128xf16>
%23406 = flow.tensor.bitcast %23405 : tensor<4x?x1x128xf16>{%dim_20935} -> tensor<4x?x1x64xcomplex<f16>>{%dim_20935}
%23407 = torch_c.from_builtin_tensor %23406 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23408 = torch.aten.mul.Tensor %23407, %23404 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23409 = torch_c.to_builtin_tensor %23408 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_20936 = arith.constant 1 : index
%dim_20937 = tensor.dim %23409, %c1_20936 : tensor<4x?x1x64xcomplex<f32>>
%23410 = flow.tensor.bitcast %23409 : tensor<4x?x1x64xcomplex<f32>>{%dim_20937} -> tensor<4x?x1x128xf32>{%dim_20937}
%23411 = torch_c.from_builtin_tensor %23410 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_20938 = torch.constant.int 5
%23412 = torch.prims.convert_element_type %23411, %int5_20938 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_20939 = torch.constant.int 1
%23413 = torch.aten.size.int %23044, %int1_20939 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_20940 = torch.constant.int 0
%23414 = torch.aten.add.int %int0_20940, %23413 : !torch.int, !torch.int -> !torch.int
%int0_20941 = torch.constant.int 0
%int0_20942 = torch.constant.int 0
%int1_20943 = torch.constant.int 1
%23415 = torch.aten.slice.Tensor %23367, %int0_20941, %int0_20942, %23414, %int1_20943 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23415, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20944 = torch.constant.int 1
%int0_20945 = torch.constant.int 0
%int9223372036854775807_20946 = torch.constant.int 9223372036854775807
%int1_20947 = torch.constant.int 1
%23416 = torch.aten.slice.Tensor %23415, %int1_20944, %int0_20945, %int9223372036854775807_20946, %int1_20947 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23416, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20948 = torch.constant.int 0
%23417 = torch.aten.unsqueeze %23416, %int0_20948 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23417, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20949 = torch.constant.int 2
%23418 = torch.aten.unsqueeze %23417, %int2_20949 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23418, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20950 = torch.constant.int 3
%int0_20951 = torch.constant.int 0
%int9223372036854775807_20952 = torch.constant.int 9223372036854775807
%int1_20953 = torch.constant.int 1
%23419 = torch.aten.slice.Tensor %23418, %int3_20950, %int0_20951, %int9223372036854775807_20952, %int1_20953 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23419, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23420 = torch_c.to_builtin_tensor %23160 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_20954 = arith.constant 1 : index
%dim_20955 = tensor.dim %23420, %c1_20954 : tensor<4x?x1x128xf16>
%23421 = flow.tensor.bitcast %23420 : tensor<4x?x1x128xf16>{%dim_20955} -> tensor<4x?x1x64xcomplex<f16>>{%dim_20955}
%23422 = torch_c.from_builtin_tensor %23421 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23423 = torch.aten.mul.Tensor %23422, %23419 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23424 = torch_c.to_builtin_tensor %23423 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_20956 = arith.constant 1 : index
%dim_20957 = tensor.dim %23424, %c1_20956 : tensor<4x?x1x64xcomplex<f32>>
%23425 = flow.tensor.bitcast %23424 : tensor<4x?x1x64xcomplex<f32>>{%dim_20957} -> tensor<4x?x1x128xf32>{%dim_20957}
%23426 = torch_c.from_builtin_tensor %23425 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_20958 = torch.constant.int 5
%23427 = torch.prims.convert_element_type %23426, %int5_20958 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_20959 = torch.constant.int 1
%23428 = torch.aten.size.int %23050, %int1_20959 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_20960 = torch.constant.int 0
%23429 = torch.aten.add.int %int0_20960, %23428 : !torch.int, !torch.int -> !torch.int
%int0_20961 = torch.constant.int 0
%int0_20962 = torch.constant.int 0
%int1_20963 = torch.constant.int 1
%23430 = torch.aten.slice.Tensor %23370, %int0_20961, %int0_20962, %23429, %int1_20963 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23430, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20964 = torch.constant.int 1
%int0_20965 = torch.constant.int 0
%int9223372036854775807_20966 = torch.constant.int 9223372036854775807
%int1_20967 = torch.constant.int 1
%23431 = torch.aten.slice.Tensor %23430, %int1_20964, %int0_20965, %int9223372036854775807_20966, %int1_20967 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23431, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20968 = torch.constant.int 0
%23432 = torch.aten.unsqueeze %23431, %int0_20968 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23432, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20969 = torch.constant.int 2
%23433 = torch.aten.unsqueeze %23432, %int2_20969 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23433, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20970 = torch.constant.int 3
%int0_20971 = torch.constant.int 0
%int9223372036854775807_20972 = torch.constant.int 9223372036854775807
%int1_20973 = torch.constant.int 1
%23434 = torch.aten.slice.Tensor %23433, %int3_20970, %int0_20971, %int9223372036854775807_20972, %int1_20973 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23434, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23435 = torch_c.to_builtin_tensor %23162 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_20974 = arith.constant 1 : index
%dim_20975 = tensor.dim %23435, %c1_20974 : tensor<4x?x1x128xf16>
%23436 = flow.tensor.bitcast %23435 : tensor<4x?x1x128xf16>{%dim_20975} -> tensor<4x?x1x64xcomplex<f16>>{%dim_20975}
%23437 = torch_c.from_builtin_tensor %23436 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23438 = torch.aten.mul.Tensor %23437, %23434 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23439 = torch_c.to_builtin_tensor %23438 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_20976 = arith.constant 1 : index
%dim_20977 = tensor.dim %23439, %c1_20976 : tensor<4x?x1x64xcomplex<f32>>
%23440 = flow.tensor.bitcast %23439 : tensor<4x?x1x64xcomplex<f32>>{%dim_20977} -> tensor<4x?x1x128xf32>{%dim_20977}
%23441 = torch_c.from_builtin_tensor %23440 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_20978 = torch.constant.int 5
%23442 = torch.prims.convert_element_type %23441, %int5_20978 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_20979 = torch.constant.int 1
%23443 = torch.aten.size.int %23056, %int1_20979 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_20980 = torch.constant.int 0
%23444 = torch.aten.add.int %int0_20980, %23443 : !torch.int, !torch.int -> !torch.int
%int0_20981 = torch.constant.int 0
%int0_20982 = torch.constant.int 0
%int1_20983 = torch.constant.int 1
%23445 = torch.aten.slice.Tensor %23373, %int0_20981, %int0_20982, %23444, %int1_20983 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23445, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_20984 = torch.constant.int 1
%int0_20985 = torch.constant.int 0
%int9223372036854775807_20986 = torch.constant.int 9223372036854775807
%int1_20987 = torch.constant.int 1
%23446 = torch.aten.slice.Tensor %23445, %int1_20984, %int0_20985, %int9223372036854775807_20986, %int1_20987 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23446, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_20988 = torch.constant.int 0
%23447 = torch.aten.unsqueeze %23446, %int0_20988 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23447, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_20989 = torch.constant.int 2
%23448 = torch.aten.unsqueeze %23447, %int2_20989 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23448, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_20990 = torch.constant.int 3
%int0_20991 = torch.constant.int 0
%int9223372036854775807_20992 = torch.constant.int 9223372036854775807
%int1_20993 = torch.constant.int 1
%23449 = torch.aten.slice.Tensor %23448, %int3_20990, %int0_20991, %int9223372036854775807_20992, %int1_20993 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23449, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23450 = torch_c.to_builtin_tensor %23164 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_20994 = arith.constant 1 : index
%dim_20995 = tensor.dim %23450, %c1_20994 : tensor<4x?x1x128xf16>
%23451 = flow.tensor.bitcast %23450 : tensor<4x?x1x128xf16>{%dim_20995} -> tensor<4x?x1x64xcomplex<f16>>{%dim_20995}
%23452 = torch_c.from_builtin_tensor %23451 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23453 = torch.aten.mul.Tensor %23452, %23449 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23454 = torch_c.to_builtin_tensor %23453 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_20996 = arith.constant 1 : index
%dim_20997 = tensor.dim %23454, %c1_20996 : tensor<4x?x1x64xcomplex<f32>>
%23455 = flow.tensor.bitcast %23454 : tensor<4x?x1x64xcomplex<f32>>{%dim_20997} -> tensor<4x?x1x128xf32>{%dim_20997}
%23456 = torch_c.from_builtin_tensor %23455 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_20998 = torch.constant.int 5
%23457 = torch.prims.convert_element_type %23456, %int5_20998 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_20999 = torch.constant.int 1
%23458 = torch.aten.size.int %23062, %int1_20999 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_21000 = torch.constant.int 0
%23459 = torch.aten.add.int %int0_21000, %23458 : !torch.int, !torch.int -> !torch.int
%int0_21001 = torch.constant.int 0
%int0_21002 = torch.constant.int 0
%int1_21003 = torch.constant.int 1
%23460 = torch.aten.slice.Tensor %23376, %int0_21001, %int0_21002, %23459, %int1_21003 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23460, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_21004 = torch.constant.int 1
%int0_21005 = torch.constant.int 0
%int9223372036854775807_21006 = torch.constant.int 9223372036854775807
%int1_21007 = torch.constant.int 1
%23461 = torch.aten.slice.Tensor %23460, %int1_21004, %int0_21005, %int9223372036854775807_21006, %int1_21007 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23461, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_21008 = torch.constant.int 0
%23462 = torch.aten.unsqueeze %23461, %int0_21008 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23462, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_21009 = torch.constant.int 2
%23463 = torch.aten.unsqueeze %23462, %int2_21009 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23463, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_21010 = torch.constant.int 3
%int0_21011 = torch.constant.int 0
%int9223372036854775807_21012 = torch.constant.int 9223372036854775807
%int1_21013 = torch.constant.int 1
%23464 = torch.aten.slice.Tensor %23463, %int3_21010, %int0_21011, %int9223372036854775807_21012, %int1_21013 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23464, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23465 = torch_c.to_builtin_tensor %23166 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_21014 = arith.constant 1 : index
%dim_21015 = tensor.dim %23465, %c1_21014 : tensor<4x?x1x128xf16>
%23466 = flow.tensor.bitcast %23465 : tensor<4x?x1x128xf16>{%dim_21015} -> tensor<4x?x1x64xcomplex<f16>>{%dim_21015}
%23467 = torch_c.from_builtin_tensor %23466 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23468 = torch.aten.mul.Tensor %23467, %23464 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23469 = torch_c.to_builtin_tensor %23468 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_21016 = arith.constant 1 : index
%dim_21017 = tensor.dim %23469, %c1_21016 : tensor<4x?x1x64xcomplex<f32>>
%23470 = flow.tensor.bitcast %23469 : tensor<4x?x1x64xcomplex<f32>>{%dim_21017} -> tensor<4x?x1x128xf32>{%dim_21017}
%23471 = torch_c.from_builtin_tensor %23470 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_21018 = torch.constant.int 5
%23472 = torch.prims.convert_element_type %23471, %int5_21018 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_21019 = torch.constant.int 1
%23473 = torch.aten.size.int %23068, %int1_21019 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_21020 = torch.constant.int 0
%23474 = torch.aten.add.int %int0_21020, %23473 : !torch.int, !torch.int -> !torch.int
%int0_21021 = torch.constant.int 0
%int0_21022 = torch.constant.int 0
%int1_21023 = torch.constant.int 1
%23475 = torch.aten.slice.Tensor %23379, %int0_21021, %int0_21022, %23474, %int1_21023 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23475, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_21024 = torch.constant.int 1
%int0_21025 = torch.constant.int 0
%int9223372036854775807_21026 = torch.constant.int 9223372036854775807
%int1_21027 = torch.constant.int 1
%23476 = torch.aten.slice.Tensor %23475, %int1_21024, %int0_21025, %int9223372036854775807_21026, %int1_21027 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23476, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_21028 = torch.constant.int 0
%23477 = torch.aten.unsqueeze %23476, %int0_21028 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23477, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_21029 = torch.constant.int 2
%23478 = torch.aten.unsqueeze %23477, %int2_21029 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23478, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_21030 = torch.constant.int 3
%int0_21031 = torch.constant.int 0
%int9223372036854775807_21032 = torch.constant.int 9223372036854775807
%int1_21033 = torch.constant.int 1
%23479 = torch.aten.slice.Tensor %23478, %int3_21030, %int0_21031, %int9223372036854775807_21032, %int1_21033 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23479, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23480 = torch_c.to_builtin_tensor %23168 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_21034 = arith.constant 1 : index
%dim_21035 = tensor.dim %23480, %c1_21034 : tensor<4x?x1x128xf16>
%23481 = flow.tensor.bitcast %23480 : tensor<4x?x1x128xf16>{%dim_21035} -> tensor<4x?x1x64xcomplex<f16>>{%dim_21035}
%23482 = torch_c.from_builtin_tensor %23481 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23483 = torch.aten.mul.Tensor %23482, %23479 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23484 = torch_c.to_builtin_tensor %23483 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_21036 = arith.constant 1 : index
%dim_21037 = tensor.dim %23484, %c1_21036 : tensor<4x?x1x64xcomplex<f32>>
%23485 = flow.tensor.bitcast %23484 : tensor<4x?x1x64xcomplex<f32>>{%dim_21037} -> tensor<4x?x1x128xf32>{%dim_21037}
%23486 = torch_c.from_builtin_tensor %23485 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_21038 = torch.constant.int 5
%23487 = torch.prims.convert_element_type %23486, %int5_21038 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_21039 = torch.constant.int 1
%23488 = torch.aten.size.int %23074, %int1_21039 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_21040 = torch.constant.int 0
%23489 = torch.aten.add.int %int0_21040, %23488 : !torch.int, !torch.int -> !torch.int
%int0_21041 = torch.constant.int 0
%int0_21042 = torch.constant.int 0
%int1_21043 = torch.constant.int 1
%23490 = torch.aten.slice.Tensor %23382, %int0_21041, %int0_21042, %23489, %int1_21043 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23490, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_21044 = torch.constant.int 1
%int0_21045 = torch.constant.int 0
%int9223372036854775807_21046 = torch.constant.int 9223372036854775807
%int1_21047 = torch.constant.int 1
%23491 = torch.aten.slice.Tensor %23490, %int1_21044, %int0_21045, %int9223372036854775807_21046, %int1_21047 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %23491, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_21048 = torch.constant.int 0
%23492 = torch.aten.unsqueeze %23491, %int0_21048 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %23492, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_21049 = torch.constant.int 2
%23493 = torch.aten.unsqueeze %23492, %int2_21049 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23493, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_21050 = torch.constant.int 3
%int0_21051 = torch.constant.int 0
%int9223372036854775807_21052 = torch.constant.int 9223372036854775807
%int1_21053 = torch.constant.int 1
%23494 = torch.aten.slice.Tensor %23493, %int3_21050, %int0_21051, %int9223372036854775807_21052, %int1_21053 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23494, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%23495 = torch_c.to_builtin_tensor %23170 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_21054 = arith.constant 1 : index
%dim_21055 = tensor.dim %23495, %c1_21054 : tensor<4x?x1x128xf16>
%23496 = flow.tensor.bitcast %23495 : tensor<4x?x1x128xf16>{%dim_21055} -> tensor<4x?x1x64xcomplex<f16>>{%dim_21055}
%23497 = torch_c.from_builtin_tensor %23496 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %23497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%23498 = torch.aten.mul.Tensor %23497, %23494 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %23498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%23499 = torch_c.to_builtin_tensor %23498 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_21056 = arith.constant 1 : index
%dim_21057 = tensor.dim %23499, %c1_21056 : tensor<4x?x1x64xcomplex<f32>>
%23500 = flow.tensor.bitcast %23499 : tensor<4x?x1x64xcomplex<f32>>{%dim_21057} -> tensor<4x?x1x128xf32>{%dim_21057}
%23501 = torch_c.from_builtin_tensor %23500 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %23501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_21058 = torch.constant.int 5
%23502 = torch.prims.convert_element_type %23501, %int5_21058 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %23502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_21059 = torch.constant.int 64
%23503 = torch.aten.mul.Scalar %2364, %int64_21059 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23503, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21060 = torch.constant.int 64
%23504 = torch.aten.mul.Scalar %2367, %int64_21060 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23504, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21061 = torch.constant.int 64
%23505 = torch.aten.mul.Scalar %2370, %int64_21061 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23505, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21062 = torch.constant.int 64
%23506 = torch.aten.mul.Scalar %2373, %int64_21062 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23506, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21063 = torch.constant.int 64
%23507 = torch.aten.mul.Scalar %2376, %int64_21063 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23507, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21064 = torch.constant.int 64
%23508 = torch.aten.mul.Scalar %2379, %int64_21064 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23508, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21065 = torch.constant.int 64
%23509 = torch.aten.mul.Scalar %2382, %int64_21065 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23509, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_21066 = torch.constant.int 64
%23510 = torch.aten.mul.Scalar %2385, %int64_21066 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23510, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22 = torch.constant.int 22
%int1_21067 = torch.constant.int 1
%23511 = torch.aten.add.Scalar %23503, %int22, %int1_21067 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23511, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21068 = torch.constant.int 22
%int1_21069 = torch.constant.int 1
%23512 = torch.aten.add.Scalar %23504, %int22_21068, %int1_21069 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23512, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21070 = torch.constant.int 22
%int1_21071 = torch.constant.int 1
%23513 = torch.aten.add.Scalar %23505, %int22_21070, %int1_21071 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23513, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21072 = torch.constant.int 22
%int1_21073 = torch.constant.int 1
%23514 = torch.aten.add.Scalar %23506, %int22_21072, %int1_21073 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23514, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21074 = torch.constant.int 22
%int1_21075 = torch.constant.int 1
%23515 = torch.aten.add.Scalar %23507, %int22_21074, %int1_21075 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23515, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21076 = torch.constant.int 22
%int1_21077 = torch.constant.int 1
%23516 = torch.aten.add.Scalar %23508, %int22_21076, %int1_21077 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23516, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21078 = torch.constant.int 22
%int1_21079 = torch.constant.int 1
%23517 = torch.aten.add.Scalar %23509, %int22_21078, %int1_21079 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23517, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int22_21080 = torch.constant.int 22
%int1_21081 = torch.constant.int 1
%23518 = torch.aten.add.Scalar %23510, %int22_21080, %int1_21081 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23518, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_21082 = torch.constant.int 4
%int16_21083 = torch.constant.int 16
%int1_21084 = torch.constant.int 1
%int128_21085 = torch.constant.int 128
%23519 = torch.prim.ListConstruct %int4_21082, %3095, %int16_21083, %int1_21084, %int128_21085 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23520 = torch.aten.view %23397, %23519 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23520, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21086 = torch.constant.int 4
%int16_21087 = torch.constant.int 16
%int1_21088 = torch.constant.int 1
%int128_21089 = torch.constant.int 128
%23521 = torch.prim.ListConstruct %int4_21086, %3095, %int16_21087, %int1_21088, %int128_21089 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23522 = torch.aten.view %23412, %23521 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23522, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21090 = torch.constant.int 4
%int16_21091 = torch.constant.int 16
%int1_21092 = torch.constant.int 1
%int128_21093 = torch.constant.int 128
%23523 = torch.prim.ListConstruct %int4_21090, %3095, %int16_21091, %int1_21092, %int128_21093 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23524 = torch.aten.view %23427, %23523 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23524, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21094 = torch.constant.int 4
%int16_21095 = torch.constant.int 16
%int1_21096 = torch.constant.int 1
%int128_21097 = torch.constant.int 128
%23525 = torch.prim.ListConstruct %int4_21094, %3095, %int16_21095, %int1_21096, %int128_21097 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23526 = torch.aten.view %23442, %23525 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23526, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21098 = torch.constant.int 4
%int16_21099 = torch.constant.int 16
%int1_21100 = torch.constant.int 1
%int128_21101 = torch.constant.int 128
%23527 = torch.prim.ListConstruct %int4_21098, %3095, %int16_21099, %int1_21100, %int128_21101 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23528 = torch.aten.view %23457, %23527 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23528, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21102 = torch.constant.int 4
%int16_21103 = torch.constant.int 16
%int1_21104 = torch.constant.int 1
%int128_21105 = torch.constant.int 128
%23529 = torch.prim.ListConstruct %int4_21102, %3095, %int16_21103, %int1_21104, %int128_21105 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23530 = torch.aten.view %23472, %23529 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23530, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21106 = torch.constant.int 4
%int16_21107 = torch.constant.int 16
%int1_21108 = torch.constant.int 1
%int128_21109 = torch.constant.int 128
%23531 = torch.prim.ListConstruct %int4_21106, %3095, %int16_21107, %int1_21108, %int128_21109 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23532 = torch.aten.view %23487, %23531 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23532, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21110 = torch.constant.int 4
%int16_21111 = torch.constant.int 16
%int1_21112 = torch.constant.int 1
%int128_21113 = torch.constant.int 128
%23533 = torch.prim.ListConstruct %int4_21110, %3095, %int16_21111, %int1_21112, %int128_21113 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23534 = torch.aten.view %23502, %23533 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23534, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21114 = torch.constant.int 4
%23535 = torch.aten.mul.int %int4_21114, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21115 = torch.constant.int 16
%int1_21116 = torch.constant.int 1
%int128_21117 = torch.constant.int 128
%23536 = torch.prim.ListConstruct %23535, %int16_21115, %int1_21116, %int128_21117 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23537 = torch.aten.view %23520, %23536 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23537, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21118 = torch.constant.int 4
%23538 = torch.aten.mul.int %int4_21118, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21119 = torch.constant.int 16
%int1_21120 = torch.constant.int 1
%int128_21121 = torch.constant.int 128
%23539 = torch.prim.ListConstruct %23538, %int16_21119, %int1_21120, %int128_21121 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23540 = torch.aten.view %23522, %23539 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23540, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21122 = torch.constant.int 4
%23541 = torch.aten.mul.int %int4_21122, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21123 = torch.constant.int 16
%int1_21124 = torch.constant.int 1
%int128_21125 = torch.constant.int 128
%23542 = torch.prim.ListConstruct %23541, %int16_21123, %int1_21124, %int128_21125 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23543 = torch.aten.view %23524, %23542 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23543, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21126 = torch.constant.int 4
%23544 = torch.aten.mul.int %int4_21126, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21127 = torch.constant.int 16
%int1_21128 = torch.constant.int 1
%int128_21129 = torch.constant.int 128
%23545 = torch.prim.ListConstruct %23544, %int16_21127, %int1_21128, %int128_21129 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23546 = torch.aten.view %23526, %23545 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23546, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21130 = torch.constant.int 4
%23547 = torch.aten.mul.int %int4_21130, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21131 = torch.constant.int 16
%int1_21132 = torch.constant.int 1
%int128_21133 = torch.constant.int 128
%23548 = torch.prim.ListConstruct %23547, %int16_21131, %int1_21132, %int128_21133 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23549 = torch.aten.view %23528, %23548 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23549, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21134 = torch.constant.int 4
%23550 = torch.aten.mul.int %int4_21134, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21135 = torch.constant.int 16
%int1_21136 = torch.constant.int 1
%int128_21137 = torch.constant.int 128
%23551 = torch.prim.ListConstruct %23550, %int16_21135, %int1_21136, %int128_21137 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23552 = torch.aten.view %23530, %23551 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23552, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21138 = torch.constant.int 4
%23553 = torch.aten.mul.int %int4_21138, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21139 = torch.constant.int 16
%int1_21140 = torch.constant.int 1
%int128_21141 = torch.constant.int 128
%23554 = torch.prim.ListConstruct %23553, %int16_21139, %int1_21140, %int128_21141 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23555 = torch.aten.view %23532, %23554 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23555, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21142 = torch.constant.int 4
%23556 = torch.aten.mul.int %int4_21142, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21143 = torch.constant.int 16
%int1_21144 = torch.constant.int 1
%int128_21145 = torch.constant.int 128
%23557 = torch.prim.ListConstruct %23556, %int16_21143, %int1_21144, %int128_21145 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23558 = torch.aten.view %23534, %23557 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23558, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21146 = torch.constant.int 4
%23559 = torch.aten.mul.int %int4_21146, %3095 : !torch.int, !torch.int -> !torch.int
%23560 = torch.prim.ListConstruct %23559 : (!torch.int) -> !torch.list<int>
%23561 = torch.aten.view %23511, %23560 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23561, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21147 = torch.constant.int 4
%23562 = torch.aten.mul.int %int4_21147, %3095 : !torch.int, !torch.int -> !torch.int
%23563 = torch.prim.ListConstruct %23562 : (!torch.int) -> !torch.list<int>
%23564 = torch.aten.view %23512, %23563 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23564, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21148 = torch.constant.int 4
%23565 = torch.aten.mul.int %int4_21148, %3095 : !torch.int, !torch.int -> !torch.int
%23566 = torch.prim.ListConstruct %23565 : (!torch.int) -> !torch.list<int>
%23567 = torch.aten.view %23513, %23566 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23567, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21149 = torch.constant.int 4
%23568 = torch.aten.mul.int %int4_21149, %3095 : !torch.int, !torch.int -> !torch.int
%23569 = torch.prim.ListConstruct %23568 : (!torch.int) -> !torch.list<int>
%23570 = torch.aten.view %23514, %23569 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23570, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21150 = torch.constant.int 4
%23571 = torch.aten.mul.int %int4_21150, %3095 : !torch.int, !torch.int -> !torch.int
%23572 = torch.prim.ListConstruct %23571 : (!torch.int) -> !torch.list<int>
%23573 = torch.aten.view %23515, %23572 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23573, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21151 = torch.constant.int 4
%23574 = torch.aten.mul.int %int4_21151, %3095 : !torch.int, !torch.int -> !torch.int
%23575 = torch.prim.ListConstruct %23574 : (!torch.int) -> !torch.list<int>
%23576 = torch.aten.view %23516, %23575 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23576, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21152 = torch.constant.int 4
%23577 = torch.aten.mul.int %int4_21152, %3095 : !torch.int, !torch.int -> !torch.int
%23578 = torch.prim.ListConstruct %23577 : (!torch.int) -> !torch.list<int>
%23579 = torch.aten.view %23517, %23578 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23579, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21153 = torch.constant.int 4
%23580 = torch.aten.mul.int %int4_21153, %3095 : !torch.int, !torch.int -> !torch.int
%23581 = torch.prim.ListConstruct %23580 : (!torch.int) -> !torch.list<int>
%23582 = torch.aten.view %23518, %23581 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23582, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21154 = torch.constant.int 4
%int16_21155 = torch.constant.int 16
%int1_21156 = torch.constant.int 1
%int128_21157 = torch.constant.int 128
%23583 = torch.prim.ListConstruct %int4_21154, %3095, %int16_21155, %int1_21156, %int128_21157 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23584 = torch.aten.view %23172, %23583 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23584, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21158 = torch.constant.int 4
%int16_21159 = torch.constant.int 16
%int1_21160 = torch.constant.int 1
%int128_21161 = torch.constant.int 128
%23585 = torch.prim.ListConstruct %int4_21158, %3095, %int16_21159, %int1_21160, %int128_21161 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23586 = torch.aten.view %23174, %23585 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23586, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21162 = torch.constant.int 4
%int16_21163 = torch.constant.int 16
%int1_21164 = torch.constant.int 1
%int128_21165 = torch.constant.int 128
%23587 = torch.prim.ListConstruct %int4_21162, %3095, %int16_21163, %int1_21164, %int128_21165 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23588 = torch.aten.view %23176, %23587 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23588, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21166 = torch.constant.int 4
%int16_21167 = torch.constant.int 16
%int1_21168 = torch.constant.int 1
%int128_21169 = torch.constant.int 128
%23589 = torch.prim.ListConstruct %int4_21166, %3095, %int16_21167, %int1_21168, %int128_21169 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23590 = torch.aten.view %23178, %23589 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23590, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21170 = torch.constant.int 4
%int16_21171 = torch.constant.int 16
%int1_21172 = torch.constant.int 1
%int128_21173 = torch.constant.int 128
%23591 = torch.prim.ListConstruct %int4_21170, %3095, %int16_21171, %int1_21172, %int128_21173 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23592 = torch.aten.view %23180, %23591 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23592, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21174 = torch.constant.int 4
%int16_21175 = torch.constant.int 16
%int1_21176 = torch.constant.int 1
%int128_21177 = torch.constant.int 128
%23593 = torch.prim.ListConstruct %int4_21174, %3095, %int16_21175, %int1_21176, %int128_21177 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23594 = torch.aten.view %23182, %23593 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23594, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21178 = torch.constant.int 4
%int16_21179 = torch.constant.int 16
%int1_21180 = torch.constant.int 1
%int128_21181 = torch.constant.int 128
%23595 = torch.prim.ListConstruct %int4_21178, %3095, %int16_21179, %int1_21180, %int128_21181 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23596 = torch.aten.view %23184, %23595 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23596, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21182 = torch.constant.int 4
%int16_21183 = torch.constant.int 16
%int1_21184 = torch.constant.int 1
%int128_21185 = torch.constant.int 128
%23597 = torch.prim.ListConstruct %int4_21182, %3095, %int16_21183, %int1_21184, %int128_21185 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23598 = torch.aten.view %23186, %23597 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %23598, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_21186 = torch.constant.int 4
%23599 = torch.aten.mul.int %int4_21186, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21187 = torch.constant.int 16
%int1_21188 = torch.constant.int 1
%int128_21189 = torch.constant.int 128
%23600 = torch.prim.ListConstruct %23599, %int16_21187, %int1_21188, %int128_21189 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23601 = torch.aten.view %23584, %23600 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23601, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21190 = torch.constant.int 4
%23602 = torch.aten.mul.int %int4_21190, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21191 = torch.constant.int 16
%int1_21192 = torch.constant.int 1
%int128_21193 = torch.constant.int 128
%23603 = torch.prim.ListConstruct %23602, %int16_21191, %int1_21192, %int128_21193 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23604 = torch.aten.view %23586, %23603 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23604, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21194 = torch.constant.int 4
%23605 = torch.aten.mul.int %int4_21194, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21195 = torch.constant.int 16
%int1_21196 = torch.constant.int 1
%int128_21197 = torch.constant.int 128
%23606 = torch.prim.ListConstruct %23605, %int16_21195, %int1_21196, %int128_21197 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23607 = torch.aten.view %23588, %23606 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23607, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21198 = torch.constant.int 4
%23608 = torch.aten.mul.int %int4_21198, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21199 = torch.constant.int 16
%int1_21200 = torch.constant.int 1
%int128_21201 = torch.constant.int 128
%23609 = torch.prim.ListConstruct %23608, %int16_21199, %int1_21200, %int128_21201 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23610 = torch.aten.view %23590, %23609 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23610, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21202 = torch.constant.int 4
%23611 = torch.aten.mul.int %int4_21202, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21203 = torch.constant.int 16
%int1_21204 = torch.constant.int 1
%int128_21205 = torch.constant.int 128
%23612 = torch.prim.ListConstruct %23611, %int16_21203, %int1_21204, %int128_21205 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23613 = torch.aten.view %23592, %23612 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23613, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21206 = torch.constant.int 4
%23614 = torch.aten.mul.int %int4_21206, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21207 = torch.constant.int 16
%int1_21208 = torch.constant.int 1
%int128_21209 = torch.constant.int 128
%23615 = torch.prim.ListConstruct %23614, %int16_21207, %int1_21208, %int128_21209 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23616 = torch.aten.view %23594, %23615 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23616, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21210 = torch.constant.int 4
%23617 = torch.aten.mul.int %int4_21210, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21211 = torch.constant.int 16
%int1_21212 = torch.constant.int 1
%int128_21213 = torch.constant.int 128
%23618 = torch.prim.ListConstruct %23617, %int16_21211, %int1_21212, %int128_21213 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23619 = torch.aten.view %23596, %23618 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23619, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_21214 = torch.constant.int 4
%23620 = torch.aten.mul.int %int4_21214, %3095 : !torch.int, !torch.int -> !torch.int
%int16_21215 = torch.constant.int 16
%int1_21216 = torch.constant.int 1
%int128_21217 = torch.constant.int 128
%23621 = torch.prim.ListConstruct %23620, %int16_21215, %int1_21216, %int128_21217 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23622 = torch.aten.view %23598, %23621 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23622, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_21218 = torch.constant.int 1
%int1_21219 = torch.constant.int 1
%23623 = torch.aten.add.Scalar %23511, %int1_21218, %int1_21219 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23623, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21220 = torch.constant.int 1
%int1_21221 = torch.constant.int 1
%23624 = torch.aten.add.Scalar %23512, %int1_21220, %int1_21221 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23624, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21222 = torch.constant.int 1
%int1_21223 = torch.constant.int 1
%23625 = torch.aten.add.Scalar %23513, %int1_21222, %int1_21223 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23625, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21224 = torch.constant.int 1
%int1_21225 = torch.constant.int 1
%23626 = torch.aten.add.Scalar %23514, %int1_21224, %int1_21225 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23626, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21226 = torch.constant.int 1
%int1_21227 = torch.constant.int 1
%23627 = torch.aten.add.Scalar %23515, %int1_21226, %int1_21227 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23627, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21228 = torch.constant.int 1
%int1_21229 = torch.constant.int 1
%23628 = torch.aten.add.Scalar %23516, %int1_21228, %int1_21229 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23628, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21230 = torch.constant.int 1
%int1_21231 = torch.constant.int 1
%23629 = torch.aten.add.Scalar %23517, %int1_21230, %int1_21231 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23629, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_21232 = torch.constant.int 1
%int1_21233 = torch.constant.int 1
%23630 = torch.aten.add.Scalar %23518, %int1_21232, %int1_21233 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %23630, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_21234 = torch.constant.int 4
%23631 = torch.aten.mul.int %int4_21234, %3095 : !torch.int, !torch.int -> !torch.int
%23632 = torch.prim.ListConstruct %23631 : (!torch.int) -> !torch.list<int>
%23633 = torch.aten.view %23623, %23632 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23633, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21235 = torch.constant.int 4
%23634 = torch.aten.mul.int %int4_21235, %3095 : !torch.int, !torch.int -> !torch.int
%23635 = torch.prim.ListConstruct %23634 : (!torch.int) -> !torch.list<int>
%23636 = torch.aten.view %23624, %23635 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23636, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21236 = torch.constant.int 4
%23637 = torch.aten.mul.int %int4_21236, %3095 : !torch.int, !torch.int -> !torch.int
%23638 = torch.prim.ListConstruct %23637 : (!torch.int) -> !torch.list<int>
%23639 = torch.aten.view %23625, %23638 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23639, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21237 = torch.constant.int 4
%23640 = torch.aten.mul.int %int4_21237, %3095 : !torch.int, !torch.int -> !torch.int
%23641 = torch.prim.ListConstruct %23640 : (!torch.int) -> !torch.list<int>
%23642 = torch.aten.view %23626, %23641 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23642, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21238 = torch.constant.int 4
%23643 = torch.aten.mul.int %int4_21238, %3095 : !torch.int, !torch.int -> !torch.int
%23644 = torch.prim.ListConstruct %23643 : (!torch.int) -> !torch.list<int>
%23645 = torch.aten.view %23627, %23644 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23645, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21239 = torch.constant.int 4
%23646 = torch.aten.mul.int %int4_21239, %3095 : !torch.int, !torch.int -> !torch.int
%23647 = torch.prim.ListConstruct %23646 : (!torch.int) -> !torch.list<int>
%23648 = torch.aten.view %23628, %23647 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23648, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21240 = torch.constant.int 4
%23649 = torch.aten.mul.int %int4_21240, %3095 : !torch.int, !torch.int -> !torch.int
%23650 = torch.prim.ListConstruct %23649 : (!torch.int) -> !torch.list<int>
%23651 = torch.aten.view %23629, %23650 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23651, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_21241 = torch.constant.int 4
%23652 = torch.aten.mul.int %int4_21241, %3095 : !torch.int, !torch.int -> !torch.int
%23653 = torch.prim.ListConstruct %23652 : (!torch.int) -> !torch.list<int>
%23654 = torch.aten.view %23630, %23653 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23654, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%23655 = torch.prim.ListConstruct %23561, %23633 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21242 = torch.constant.int 0
%23656 = torch.aten.cat %23655, %int0_21242 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23656, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23657 = torch.prim.ListConstruct %23564, %23636 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21243 = torch.constant.int 0
%23658 = torch.aten.cat %23657, %int0_21243 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23658, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23659 = torch.prim.ListConstruct %23567, %23639 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21244 = torch.constant.int 0
%23660 = torch.aten.cat %23659, %int0_21244 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23660, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23661 = torch.prim.ListConstruct %23570, %23642 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21245 = torch.constant.int 0
%23662 = torch.aten.cat %23661, %int0_21245 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23662, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23663 = torch.prim.ListConstruct %23573, %23645 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21246 = torch.constant.int 0
%23664 = torch.aten.cat %23663, %int0_21246 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23664, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23665 = torch.prim.ListConstruct %23576, %23648 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21247 = torch.constant.int 0
%23666 = torch.aten.cat %23665, %int0_21247 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23666, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23667 = torch.prim.ListConstruct %23579, %23651 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21248 = torch.constant.int 0
%23668 = torch.aten.cat %23667, %int0_21248 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23668, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23669 = torch.prim.ListConstruct %23582, %23654 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_21249 = torch.constant.int 0
%23670 = torch.aten.cat %23669, %int0_21249 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %23670, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%23671 = torch.prim.ListConstruct %23537, %23601 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21250 = torch.constant.int 0
%23672 = torch.aten.cat %23671, %int0_21250 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23672, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23673 = torch.prim.ListConstruct %23540, %23604 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21251 = torch.constant.int 0
%23674 = torch.aten.cat %23673, %int0_21251 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23674, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23675 = torch.prim.ListConstruct %23543, %23607 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21252 = torch.constant.int 0
%23676 = torch.aten.cat %23675, %int0_21252 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23676, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23677 = torch.prim.ListConstruct %23546, %23610 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21253 = torch.constant.int 0
%23678 = torch.aten.cat %23677, %int0_21253 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23678, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23679 = torch.prim.ListConstruct %23549, %23613 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21254 = torch.constant.int 0
%23680 = torch.aten.cat %23679, %int0_21254 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23680, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23681 = torch.prim.ListConstruct %23552, %23616 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21255 = torch.constant.int 0
%23682 = torch.aten.cat %23681, %int0_21255 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23682, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23683 = torch.prim.ListConstruct %23555, %23619 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21256 = torch.constant.int 0
%23684 = torch.aten.cat %23683, %int0_21256 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23684, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23685 = torch.prim.ListConstruct %23558, %23622 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_21257 = torch.constant.int 0
%23686 = torch.aten.cat %23685, %int0_21257 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23686, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21258 = torch.constant.int 32
%int2_21259 = torch.constant.int 2
%int16_21260 = torch.constant.int 16
%int1_21261 = torch.constant.int 1
%int128_21262 = torch.constant.int 128
%23687 = torch.prim.ListConstruct %3023, %int32_21258, %int2_21259, %int16_21260, %int1_21261, %int128_21262 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23688 = torch.aten.view %21837, %23687 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23688, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21263 = torch.constant.int 32
%23689 = torch.aten.mul.int %3023, %int32_21263 : !torch.int, !torch.int -> !torch.int
%int2_21264 = torch.constant.int 2
%23690 = torch.aten.mul.int %23689, %int2_21264 : !torch.int, !torch.int -> !torch.int
%int16_21265 = torch.constant.int 16
%int1_21266 = torch.constant.int 1
%int128_21267 = torch.constant.int 128
%23691 = torch.prim.ListConstruct %23690, %int16_21265, %int1_21266, %int128_21267 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23692 = torch.aten.view %23688, %23691 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23692, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23693 = torch.prim.ListConstruct %23656 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21268 = torch.constant.bool false
%23694 = torch.aten.index_put %23692, %23693, %23672, %false_21268 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23694, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21269 = torch.constant.int 32
%int2_21270 = torch.constant.int 2
%int16_21271 = torch.constant.int 16
%int1_21272 = torch.constant.int 1
%int128_21273 = torch.constant.int 128
%23695 = torch.prim.ListConstruct %3023, %int32_21269, %int2_21270, %int16_21271, %int1_21272, %int128_21273 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23696 = torch.aten.view %23694, %23695 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23696, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21274 = torch.constant.int 131072
%23697 = torch.prim.ListConstruct %3023, %int131072_21274 : (!torch.int, !torch.int) -> !torch.list<int>
%23698 = torch.aten.view %23696, %23697 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23698, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21275 = torch.constant.int 32
%int2_21276 = torch.constant.int 2
%int16_21277 = torch.constant.int 16
%int1_21278 = torch.constant.int 1
%int128_21279 = torch.constant.int 128
%23699 = torch.prim.ListConstruct %3026, %int32_21275, %int2_21276, %int16_21277, %int1_21278, %int128_21279 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23700 = torch.aten.view %21849, %23699 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23700, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21280 = torch.constant.int 32
%23701 = torch.aten.mul.int %3026, %int32_21280 : !torch.int, !torch.int -> !torch.int
%int2_21281 = torch.constant.int 2
%23702 = torch.aten.mul.int %23701, %int2_21281 : !torch.int, !torch.int -> !torch.int
%int16_21282 = torch.constant.int 16
%int1_21283 = torch.constant.int 1
%int128_21284 = torch.constant.int 128
%23703 = torch.prim.ListConstruct %23702, %int16_21282, %int1_21283, %int128_21284 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23704 = torch.aten.view %23700, %23703 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23704, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23705 = torch.prim.ListConstruct %23658 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21285 = torch.constant.bool false
%23706 = torch.aten.index_put %23704, %23705, %23674, %false_21285 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23706, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21286 = torch.constant.int 32
%int2_21287 = torch.constant.int 2
%int16_21288 = torch.constant.int 16
%int1_21289 = torch.constant.int 1
%int128_21290 = torch.constant.int 128
%23707 = torch.prim.ListConstruct %3026, %int32_21286, %int2_21287, %int16_21288, %int1_21289, %int128_21290 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23708 = torch.aten.view %23706, %23707 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23708, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21291 = torch.constant.int 131072
%23709 = torch.prim.ListConstruct %3026, %int131072_21291 : (!torch.int, !torch.int) -> !torch.list<int>
%23710 = torch.aten.view %23708, %23709 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23710, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21292 = torch.constant.int 32
%int2_21293 = torch.constant.int 2
%int16_21294 = torch.constant.int 16
%int1_21295 = torch.constant.int 1
%int128_21296 = torch.constant.int 128
%23711 = torch.prim.ListConstruct %3029, %int32_21292, %int2_21293, %int16_21294, %int1_21295, %int128_21296 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23712 = torch.aten.view %21861, %23711 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23712, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21297 = torch.constant.int 32
%23713 = torch.aten.mul.int %3029, %int32_21297 : !torch.int, !torch.int -> !torch.int
%int2_21298 = torch.constant.int 2
%23714 = torch.aten.mul.int %23713, %int2_21298 : !torch.int, !torch.int -> !torch.int
%int16_21299 = torch.constant.int 16
%int1_21300 = torch.constant.int 1
%int128_21301 = torch.constant.int 128
%23715 = torch.prim.ListConstruct %23714, %int16_21299, %int1_21300, %int128_21301 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23716 = torch.aten.view %23712, %23715 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23716, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23717 = torch.prim.ListConstruct %23660 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21302 = torch.constant.bool false
%23718 = torch.aten.index_put %23716, %23717, %23676, %false_21302 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23718, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21303 = torch.constant.int 32
%int2_21304 = torch.constant.int 2
%int16_21305 = torch.constant.int 16
%int1_21306 = torch.constant.int 1
%int128_21307 = torch.constant.int 128
%23719 = torch.prim.ListConstruct %3029, %int32_21303, %int2_21304, %int16_21305, %int1_21306, %int128_21307 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23720 = torch.aten.view %23718, %23719 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23720, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21308 = torch.constant.int 131072
%23721 = torch.prim.ListConstruct %3029, %int131072_21308 : (!torch.int, !torch.int) -> !torch.list<int>
%23722 = torch.aten.view %23720, %23721 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23722, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21309 = torch.constant.int 32
%int2_21310 = torch.constant.int 2
%int16_21311 = torch.constant.int 16
%int1_21312 = torch.constant.int 1
%int128_21313 = torch.constant.int 128
%23723 = torch.prim.ListConstruct %3032, %int32_21309, %int2_21310, %int16_21311, %int1_21312, %int128_21313 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23724 = torch.aten.view %21873, %23723 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23724, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21314 = torch.constant.int 32
%23725 = torch.aten.mul.int %3032, %int32_21314 : !torch.int, !torch.int -> !torch.int
%int2_21315 = torch.constant.int 2
%23726 = torch.aten.mul.int %23725, %int2_21315 : !torch.int, !torch.int -> !torch.int
%int16_21316 = torch.constant.int 16
%int1_21317 = torch.constant.int 1
%int128_21318 = torch.constant.int 128
%23727 = torch.prim.ListConstruct %23726, %int16_21316, %int1_21317, %int128_21318 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23728 = torch.aten.view %23724, %23727 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23728, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23729 = torch.prim.ListConstruct %23662 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21319 = torch.constant.bool false
%23730 = torch.aten.index_put %23728, %23729, %23678, %false_21319 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23730, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21320 = torch.constant.int 32
%int2_21321 = torch.constant.int 2
%int16_21322 = torch.constant.int 16
%int1_21323 = torch.constant.int 1
%int128_21324 = torch.constant.int 128
%23731 = torch.prim.ListConstruct %3032, %int32_21320, %int2_21321, %int16_21322, %int1_21323, %int128_21324 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23732 = torch.aten.view %23730, %23731 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23732, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21325 = torch.constant.int 131072
%23733 = torch.prim.ListConstruct %3032, %int131072_21325 : (!torch.int, !torch.int) -> !torch.list<int>
%23734 = torch.aten.view %23732, %23733 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23734, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21326 = torch.constant.int 32
%int2_21327 = torch.constant.int 2
%int16_21328 = torch.constant.int 16
%int1_21329 = torch.constant.int 1
%int128_21330 = torch.constant.int 128
%23735 = torch.prim.ListConstruct %3035, %int32_21326, %int2_21327, %int16_21328, %int1_21329, %int128_21330 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23736 = torch.aten.view %21885, %23735 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23736, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21331 = torch.constant.int 32
%23737 = torch.aten.mul.int %3035, %int32_21331 : !torch.int, !torch.int -> !torch.int
%int2_21332 = torch.constant.int 2
%23738 = torch.aten.mul.int %23737, %int2_21332 : !torch.int, !torch.int -> !torch.int
%int16_21333 = torch.constant.int 16
%int1_21334 = torch.constant.int 1
%int128_21335 = torch.constant.int 128
%23739 = torch.prim.ListConstruct %23738, %int16_21333, %int1_21334, %int128_21335 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23740 = torch.aten.view %23736, %23739 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23740, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23741 = torch.prim.ListConstruct %23664 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21336 = torch.constant.bool false
%23742 = torch.aten.index_put %23740, %23741, %23680, %false_21336 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23742, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21337 = torch.constant.int 32
%int2_21338 = torch.constant.int 2
%int16_21339 = torch.constant.int 16
%int1_21340 = torch.constant.int 1
%int128_21341 = torch.constant.int 128
%23743 = torch.prim.ListConstruct %3035, %int32_21337, %int2_21338, %int16_21339, %int1_21340, %int128_21341 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23744 = torch.aten.view %23742, %23743 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23744, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21342 = torch.constant.int 131072
%23745 = torch.prim.ListConstruct %3035, %int131072_21342 : (!torch.int, !torch.int) -> !torch.list<int>
%23746 = torch.aten.view %23744, %23745 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23746, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21343 = torch.constant.int 32
%int2_21344 = torch.constant.int 2
%int16_21345 = torch.constant.int 16
%int1_21346 = torch.constant.int 1
%int128_21347 = torch.constant.int 128
%23747 = torch.prim.ListConstruct %3038, %int32_21343, %int2_21344, %int16_21345, %int1_21346, %int128_21347 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23748 = torch.aten.view %21897, %23747 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23748, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21348 = torch.constant.int 32
%23749 = torch.aten.mul.int %3038, %int32_21348 : !torch.int, !torch.int -> !torch.int
%int2_21349 = torch.constant.int 2
%23750 = torch.aten.mul.int %23749, %int2_21349 : !torch.int, !torch.int -> !torch.int
%int16_21350 = torch.constant.int 16
%int1_21351 = torch.constant.int 1
%int128_21352 = torch.constant.int 128
%23751 = torch.prim.ListConstruct %23750, %int16_21350, %int1_21351, %int128_21352 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23752 = torch.aten.view %23748, %23751 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23752, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23753 = torch.prim.ListConstruct %23666 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21353 = torch.constant.bool false
%23754 = torch.aten.index_put %23752, %23753, %23682, %false_21353 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23754, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21354 = torch.constant.int 32
%int2_21355 = torch.constant.int 2
%int16_21356 = torch.constant.int 16
%int1_21357 = torch.constant.int 1
%int128_21358 = torch.constant.int 128
%23755 = torch.prim.ListConstruct %3038, %int32_21354, %int2_21355, %int16_21356, %int1_21357, %int128_21358 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23756 = torch.aten.view %23754, %23755 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23756, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21359 = torch.constant.int 131072
%23757 = torch.prim.ListConstruct %3038, %int131072_21359 : (!torch.int, !torch.int) -> !torch.list<int>
%23758 = torch.aten.view %23756, %23757 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23758, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21360 = torch.constant.int 32
%int2_21361 = torch.constant.int 2
%int16_21362 = torch.constant.int 16
%int1_21363 = torch.constant.int 1
%int128_21364 = torch.constant.int 128
%23759 = torch.prim.ListConstruct %3041, %int32_21360, %int2_21361, %int16_21362, %int1_21363, %int128_21364 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23760 = torch.aten.view %21909, %23759 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23760, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21365 = torch.constant.int 32
%23761 = torch.aten.mul.int %3041, %int32_21365 : !torch.int, !torch.int -> !torch.int
%int2_21366 = torch.constant.int 2
%23762 = torch.aten.mul.int %23761, %int2_21366 : !torch.int, !torch.int -> !torch.int
%int16_21367 = torch.constant.int 16
%int1_21368 = torch.constant.int 1
%int128_21369 = torch.constant.int 128
%23763 = torch.prim.ListConstruct %23762, %int16_21367, %int1_21368, %int128_21369 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23764 = torch.aten.view %23760, %23763 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23764, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23765 = torch.prim.ListConstruct %23668 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21370 = torch.constant.bool false
%23766 = torch.aten.index_put %23764, %23765, %23684, %false_21370 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23766, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21371 = torch.constant.int 32
%int2_21372 = torch.constant.int 2
%int16_21373 = torch.constant.int 16
%int1_21374 = torch.constant.int 1
%int128_21375 = torch.constant.int 128
%23767 = torch.prim.ListConstruct %3041, %int32_21371, %int2_21372, %int16_21373, %int1_21374, %int128_21375 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23768 = torch.aten.view %23766, %23767 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23768, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21376 = torch.constant.int 131072
%23769 = torch.prim.ListConstruct %3041, %int131072_21376 : (!torch.int, !torch.int) -> !torch.list<int>
%23770 = torch.aten.view %23768, %23769 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23770, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_21377 = torch.constant.int 32
%int2_21378 = torch.constant.int 2
%int16_21379 = torch.constant.int 16
%int1_21380 = torch.constant.int 1
%int128_21381 = torch.constant.int 128
%23771 = torch.prim.ListConstruct %3044, %int32_21377, %int2_21378, %int16_21379, %int1_21380, %int128_21381 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23772 = torch.aten.view %21921, %23771 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23772, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_21382 = torch.constant.int 32
%23773 = torch.aten.mul.int %3044, %int32_21382 : !torch.int, !torch.int -> !torch.int
%int2_21383 = torch.constant.int 2
%23774 = torch.aten.mul.int %23773, %int2_21383 : !torch.int, !torch.int -> !torch.int
%int16_21384 = torch.constant.int 16
%int1_21385 = torch.constant.int 1
%int128_21386 = torch.constant.int 128
%23775 = torch.prim.ListConstruct %23774, %int16_21384, %int1_21385, %int128_21386 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23776 = torch.aten.view %23772, %23775 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23776, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%23777 = torch.prim.ListConstruct %23670 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_21387 = torch.constant.bool false
%23778 = torch.aten.index_put %23776, %23777, %23686, %false_21387 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %23778, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_21388 = torch.constant.int 32
%int2_21389 = torch.constant.int 2
%int16_21390 = torch.constant.int 16
%int1_21391 = torch.constant.int 1
%int128_21392 = torch.constant.int 128
%23779 = torch.prim.ListConstruct %3044, %int32_21388, %int2_21389, %int16_21390, %int1_21391, %int128_21392 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23780 = torch.aten.view %23778, %23779 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %23780, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_21393 = torch.constant.int 131072
%23781 = torch.prim.ListConstruct %3044, %int131072_21393 : (!torch.int, !torch.int) -> !torch.list<int>
%23782 = torch.aten.view %23780, %23781 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %23782, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_21394 = torch.constant.int -2
%23783 = torch.aten.unsqueeze %23397, %int-2_21394 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21395 = torch.constant.int -2
%23784 = torch.aten.unsqueeze %23412, %int-2_21395 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21396 = torch.constant.int -2
%23785 = torch.aten.unsqueeze %23427, %int-2_21396 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21397 = torch.constant.int -2
%23786 = torch.aten.unsqueeze %23442, %int-2_21397 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21398 = torch.constant.int -2
%23787 = torch.aten.unsqueeze %23457, %int-2_21398 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21399 = torch.constant.int -2
%23788 = torch.aten.unsqueeze %23472, %int-2_21399 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21400 = torch.constant.int -2
%23789 = torch.aten.unsqueeze %23487, %int-2_21400 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21401 = torch.constant.int -2
%23790 = torch.aten.unsqueeze %23502, %int-2_21401 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_21402 = torch.constant.int 4
%int1_21403 = torch.constant.int 1
%int4_21404 = torch.constant.int 4
%int128_21405 = torch.constant.int 128
%23791 = torch.prim.ListConstruct %int4_21402, %23383, %int1_21403, %int4_21404, %int128_21405 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21406 = torch.constant.bool false
%23792 = torch.aten.expand %23783, %23791, %false_21406 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21407 = torch.constant.int 4
%int1_21408 = torch.constant.int 1
%int4_21409 = torch.constant.int 4
%int128_21410 = torch.constant.int 128
%23793 = torch.prim.ListConstruct %int4_21407, %23383, %int1_21408, %int4_21409, %int128_21410 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21411 = torch.constant.bool false
%23794 = torch.aten.expand %23784, %23793, %false_21411 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21412 = torch.constant.int 4
%int1_21413 = torch.constant.int 1
%int4_21414 = torch.constant.int 4
%int128_21415 = torch.constant.int 128
%23795 = torch.prim.ListConstruct %int4_21412, %23383, %int1_21413, %int4_21414, %int128_21415 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21416 = torch.constant.bool false
%23796 = torch.aten.expand %23785, %23795, %false_21416 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21417 = torch.constant.int 4
%int1_21418 = torch.constant.int 1
%int4_21419 = torch.constant.int 4
%int128_21420 = torch.constant.int 128
%23797 = torch.prim.ListConstruct %int4_21417, %23383, %int1_21418, %int4_21419, %int128_21420 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21421 = torch.constant.bool false
%23798 = torch.aten.expand %23786, %23797, %false_21421 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21422 = torch.constant.int 4
%int1_21423 = torch.constant.int 1
%int4_21424 = torch.constant.int 4
%int128_21425 = torch.constant.int 128
%23799 = torch.prim.ListConstruct %int4_21422, %23383, %int1_21423, %int4_21424, %int128_21425 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21426 = torch.constant.bool false
%23800 = torch.aten.expand %23787, %23799, %false_21426 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21427 = torch.constant.int 4
%int1_21428 = torch.constant.int 1
%int4_21429 = torch.constant.int 4
%int128_21430 = torch.constant.int 128
%23801 = torch.prim.ListConstruct %int4_21427, %23383, %int1_21428, %int4_21429, %int128_21430 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21431 = torch.constant.bool false
%23802 = torch.aten.expand %23788, %23801, %false_21431 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21432 = torch.constant.int 4
%int1_21433 = torch.constant.int 1
%int4_21434 = torch.constant.int 4
%int128_21435 = torch.constant.int 128
%23803 = torch.prim.ListConstruct %int4_21432, %23383, %int1_21433, %int4_21434, %int128_21435 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21436 = torch.constant.bool false
%23804 = torch.aten.expand %23789, %23803, %false_21436 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21437 = torch.constant.int 4
%int1_21438 = torch.constant.int 1
%int4_21439 = torch.constant.int 4
%int128_21440 = torch.constant.int 128
%23805 = torch.prim.ListConstruct %int4_21437, %23383, %int1_21438, %int4_21439, %int128_21440 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21441 = torch.constant.bool false
%23806 = torch.aten.expand %23790, %23805, %false_21441 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21442 = torch.constant.int 4
%int4_21443 = torch.constant.int 4
%int128_21444 = torch.constant.int 128
%23807 = torch.prim.ListConstruct %int4_21442, %23383, %int4_21443, %int128_21444 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23808 = torch.aten.view %23792, %23807 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21445 = torch.constant.int 4
%int4_21446 = torch.constant.int 4
%int128_21447 = torch.constant.int 128
%23809 = torch.prim.ListConstruct %int4_21445, %23383, %int4_21446, %int128_21447 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23810 = torch.aten.view %23794, %23809 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21448 = torch.constant.int 4
%int4_21449 = torch.constant.int 4
%int128_21450 = torch.constant.int 128
%23811 = torch.prim.ListConstruct %int4_21448, %23383, %int4_21449, %int128_21450 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23812 = torch.aten.view %23796, %23811 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21451 = torch.constant.int 4
%int4_21452 = torch.constant.int 4
%int128_21453 = torch.constant.int 128
%23813 = torch.prim.ListConstruct %int4_21451, %23383, %int4_21452, %int128_21453 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23814 = torch.aten.view %23798, %23813 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21454 = torch.constant.int 4
%int4_21455 = torch.constant.int 4
%int128_21456 = torch.constant.int 128
%23815 = torch.prim.ListConstruct %int4_21454, %23383, %int4_21455, %int128_21456 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23816 = torch.aten.view %23800, %23815 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21457 = torch.constant.int 4
%int4_21458 = torch.constant.int 4
%int128_21459 = torch.constant.int 128
%23817 = torch.prim.ListConstruct %int4_21457, %23383, %int4_21458, %int128_21459 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23818 = torch.aten.view %23802, %23817 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21460 = torch.constant.int 4
%int4_21461 = torch.constant.int 4
%int128_21462 = torch.constant.int 128
%23819 = torch.prim.ListConstruct %int4_21460, %23383, %int4_21461, %int128_21462 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23820 = torch.aten.view %23804, %23819 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21463 = torch.constant.int 4
%int4_21464 = torch.constant.int 4
%int128_21465 = torch.constant.int 128
%23821 = torch.prim.ListConstruct %int4_21463, %23383, %int4_21464, %int128_21465 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23822 = torch.aten.view %23806, %23821 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_21466 = torch.constant.int -2
%23823 = torch.aten.unsqueeze %23172, %int-2_21466 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21467 = torch.constant.int -2
%23824 = torch.aten.unsqueeze %23174, %int-2_21467 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21468 = torch.constant.int -2
%23825 = torch.aten.unsqueeze %23176, %int-2_21468 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21469 = torch.constant.int -2
%23826 = torch.aten.unsqueeze %23178, %int-2_21469 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21470 = torch.constant.int -2
%23827 = torch.aten.unsqueeze %23180, %int-2_21470 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21471 = torch.constant.int -2
%23828 = torch.aten.unsqueeze %23182, %int-2_21471 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21472 = torch.constant.int -2
%23829 = torch.aten.unsqueeze %23184, %int-2_21472 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_21473 = torch.constant.int -2
%23830 = torch.aten.unsqueeze %23186, %int-2_21473 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %23830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_21474 = torch.constant.int 1
%23831 = torch.aten.size.int %23096, %int1_21474 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_21475 = torch.constant.int 4
%int1_21476 = torch.constant.int 1
%int4_21477 = torch.constant.int 4
%int128_21478 = torch.constant.int 128
%23832 = torch.prim.ListConstruct %int4_21475, %23831, %int1_21476, %int4_21477, %int128_21478 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21479 = torch.constant.bool false
%23833 = torch.aten.expand %23823, %23832, %false_21479 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21480 = torch.constant.int 4
%int1_21481 = torch.constant.int 1
%int4_21482 = torch.constant.int 4
%int128_21483 = torch.constant.int 128
%23834 = torch.prim.ListConstruct %int4_21480, %23831, %int1_21481, %int4_21482, %int128_21483 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21484 = torch.constant.bool false
%23835 = torch.aten.expand %23824, %23834, %false_21484 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21485 = torch.constant.int 4
%int1_21486 = torch.constant.int 1
%int4_21487 = torch.constant.int 4
%int128_21488 = torch.constant.int 128
%23836 = torch.prim.ListConstruct %int4_21485, %23831, %int1_21486, %int4_21487, %int128_21488 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21489 = torch.constant.bool false
%23837 = torch.aten.expand %23825, %23836, %false_21489 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21490 = torch.constant.int 4
%int1_21491 = torch.constant.int 1
%int4_21492 = torch.constant.int 4
%int128_21493 = torch.constant.int 128
%23838 = torch.prim.ListConstruct %int4_21490, %23831, %int1_21491, %int4_21492, %int128_21493 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21494 = torch.constant.bool false
%23839 = torch.aten.expand %23826, %23838, %false_21494 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21495 = torch.constant.int 4
%int1_21496 = torch.constant.int 1
%int4_21497 = torch.constant.int 4
%int128_21498 = torch.constant.int 128
%23840 = torch.prim.ListConstruct %int4_21495, %23831, %int1_21496, %int4_21497, %int128_21498 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21499 = torch.constant.bool false
%23841 = torch.aten.expand %23827, %23840, %false_21499 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21500 = torch.constant.int 4
%int1_21501 = torch.constant.int 1
%int4_21502 = torch.constant.int 4
%int128_21503 = torch.constant.int 128
%23842 = torch.prim.ListConstruct %int4_21500, %23831, %int1_21501, %int4_21502, %int128_21503 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21504 = torch.constant.bool false
%23843 = torch.aten.expand %23828, %23842, %false_21504 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21505 = torch.constant.int 4
%int1_21506 = torch.constant.int 1
%int4_21507 = torch.constant.int 4
%int128_21508 = torch.constant.int 128
%23844 = torch.prim.ListConstruct %int4_21505, %23831, %int1_21506, %int4_21507, %int128_21508 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21509 = torch.constant.bool false
%23845 = torch.aten.expand %23829, %23844, %false_21509 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21510 = torch.constant.int 4
%int1_21511 = torch.constant.int 1
%int4_21512 = torch.constant.int 4
%int128_21513 = torch.constant.int 128
%23846 = torch.prim.ListConstruct %int4_21510, %23831, %int1_21511, %int4_21512, %int128_21513 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_21514 = torch.constant.bool false
%23847 = torch.aten.expand %23830, %23846, %false_21514 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %23847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_21515 = torch.constant.int 4
%int4_21516 = torch.constant.int 4
%int128_21517 = torch.constant.int 128
%23848 = torch.prim.ListConstruct %int4_21515, %23831, %int4_21516, %int128_21517 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23849 = torch.aten.view %23833, %23848 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21518 = torch.constant.int 4
%int4_21519 = torch.constant.int 4
%int128_21520 = torch.constant.int 128
%23850 = torch.prim.ListConstruct %int4_21518, %23831, %int4_21519, %int128_21520 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23851 = torch.aten.view %23835, %23850 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21521 = torch.constant.int 4
%int4_21522 = torch.constant.int 4
%int128_21523 = torch.constant.int 128
%23852 = torch.prim.ListConstruct %int4_21521, %23831, %int4_21522, %int128_21523 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23853 = torch.aten.view %23837, %23852 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21524 = torch.constant.int 4
%int4_21525 = torch.constant.int 4
%int128_21526 = torch.constant.int 128
%23854 = torch.prim.ListConstruct %int4_21524, %23831, %int4_21525, %int128_21526 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23855 = torch.aten.view %23839, %23854 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21527 = torch.constant.int 4
%int4_21528 = torch.constant.int 4
%int128_21529 = torch.constant.int 128
%23856 = torch.prim.ListConstruct %int4_21527, %23831, %int4_21528, %int128_21529 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23857 = torch.aten.view %23841, %23856 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21530 = torch.constant.int 4
%int4_21531 = torch.constant.int 4
%int128_21532 = torch.constant.int 128
%23858 = torch.prim.ListConstruct %int4_21530, %23831, %int4_21531, %int128_21532 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23859 = torch.aten.view %23843, %23858 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21533 = torch.constant.int 4
%int4_21534 = torch.constant.int 4
%int128_21535 = torch.constant.int 128
%23860 = torch.prim.ListConstruct %int4_21533, %23831, %int4_21534, %int128_21535 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23861 = torch.aten.view %23845, %23860 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21536 = torch.constant.int 4
%int4_21537 = torch.constant.int 4
%int128_21538 = torch.constant.int 128
%23862 = torch.prim.ListConstruct %int4_21536, %23831, %int4_21537, %int128_21538 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23863 = torch.aten.view %23847, %23862 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21539 = torch.constant.int 1
%int2_21540 = torch.constant.int 2
%23864 = torch.aten.transpose.int %23239, %int1_21539, %int2_21540 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23864, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21541 = torch.constant.int 1
%int2_21542 = torch.constant.int 2
%23865 = torch.aten.transpose.int %23254, %int1_21541, %int2_21542 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23865, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21543 = torch.constant.int 1
%int2_21544 = torch.constant.int 2
%23866 = torch.aten.transpose.int %23269, %int1_21543, %int2_21544 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23866, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21545 = torch.constant.int 1
%int2_21546 = torch.constant.int 2
%23867 = torch.aten.transpose.int %23284, %int1_21545, %int2_21546 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23867, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21547 = torch.constant.int 1
%int2_21548 = torch.constant.int 2
%23868 = torch.aten.transpose.int %23299, %int1_21547, %int2_21548 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23868, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21549 = torch.constant.int 1
%int2_21550 = torch.constant.int 2
%23869 = torch.aten.transpose.int %23314, %int1_21549, %int2_21550 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23869, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21551 = torch.constant.int 1
%int2_21552 = torch.constant.int 2
%23870 = torch.aten.transpose.int %23329, %int1_21551, %int2_21552 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23870, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21553 = torch.constant.int 1
%int2_21554 = torch.constant.int 2
%23871 = torch.aten.transpose.int %23344, %int1_21553, %int2_21554 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23871, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21555 = torch.constant.int 1
%int2_21556 = torch.constant.int 2
%23872 = torch.aten.transpose.int %23808, %int1_21555, %int2_21556 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23872, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21557 = torch.constant.int 1
%int2_21558 = torch.constant.int 2
%23873 = torch.aten.transpose.int %23810, %int1_21557, %int2_21558 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23873, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21559 = torch.constant.int 1
%int2_21560 = torch.constant.int 2
%23874 = torch.aten.transpose.int %23812, %int1_21559, %int2_21560 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23874, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21561 = torch.constant.int 1
%int2_21562 = torch.constant.int 2
%23875 = torch.aten.transpose.int %23814, %int1_21561, %int2_21562 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23875, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21563 = torch.constant.int 1
%int2_21564 = torch.constant.int 2
%23876 = torch.aten.transpose.int %23816, %int1_21563, %int2_21564 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23876, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21565 = torch.constant.int 1
%int2_21566 = torch.constant.int 2
%23877 = torch.aten.transpose.int %23818, %int1_21565, %int2_21566 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23877, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21567 = torch.constant.int 1
%int2_21568 = torch.constant.int 2
%23878 = torch.aten.transpose.int %23820, %int1_21567, %int2_21568 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23878, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21569 = torch.constant.int 1
%int2_21570 = torch.constant.int 2
%23879 = torch.aten.transpose.int %23822, %int1_21569, %int2_21570 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23879, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21571 = torch.constant.int 1
%int2_21572 = torch.constant.int 2
%23880 = torch.aten.transpose.int %23849, %int1_21571, %int2_21572 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23880, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21573 = torch.constant.int 1
%int2_21574 = torch.constant.int 2
%23881 = torch.aten.transpose.int %23851, %int1_21573, %int2_21574 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23881, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21575 = torch.constant.int 1
%int2_21576 = torch.constant.int 2
%23882 = torch.aten.transpose.int %23853, %int1_21575, %int2_21576 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23882, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21577 = torch.constant.int 1
%int2_21578 = torch.constant.int 2
%23883 = torch.aten.transpose.int %23855, %int1_21577, %int2_21578 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23883, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21579 = torch.constant.int 1
%int2_21580 = torch.constant.int 2
%23884 = torch.aten.transpose.int %23857, %int1_21579, %int2_21580 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23884, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21581 = torch.constant.int 1
%int2_21582 = torch.constant.int 2
%23885 = torch.aten.transpose.int %23859, %int1_21581, %int2_21582 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23885, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21583 = torch.constant.int 1
%int2_21584 = torch.constant.int 2
%23886 = torch.aten.transpose.int %23861, %int1_21583, %int2_21584 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23886, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21585 = torch.constant.int 1
%int2_21586 = torch.constant.int 2
%23887 = torch.aten.transpose.int %23863, %int1_21585, %int2_21586 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %23887, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21587 = torch.constant.float 0.000000e+00
%true_21588 = torch.constant.bool true
%none_21589 = torch.constant.none
%none_21590 = torch.constant.none
%23888:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23864, %23872, %23880, %float0.000000e00_21587, %true_21588, %none_21589, %none_21590) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23888#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21591 = torch.constant.float 0.000000e+00
%true_21592 = torch.constant.bool true
%none_21593 = torch.constant.none
%none_21594 = torch.constant.none
%23889:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23865, %23873, %23881, %float0.000000e00_21591, %true_21592, %none_21593, %none_21594) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23889#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21595 = torch.constant.float 0.000000e+00
%true_21596 = torch.constant.bool true
%none_21597 = torch.constant.none
%none_21598 = torch.constant.none
%23890:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23866, %23874, %23882, %float0.000000e00_21595, %true_21596, %none_21597, %none_21598) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23890#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21599 = torch.constant.float 0.000000e+00
%true_21600 = torch.constant.bool true
%none_21601 = torch.constant.none
%none_21602 = torch.constant.none
%23891:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23867, %23875, %23883, %float0.000000e00_21599, %true_21600, %none_21601, %none_21602) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23891#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21603 = torch.constant.float 0.000000e+00
%true_21604 = torch.constant.bool true
%none_21605 = torch.constant.none
%none_21606 = torch.constant.none
%23892:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23868, %23876, %23884, %float0.000000e00_21603, %true_21604, %none_21605, %none_21606) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23892#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21607 = torch.constant.float 0.000000e+00
%true_21608 = torch.constant.bool true
%none_21609 = torch.constant.none
%none_21610 = torch.constant.none
%23893:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23869, %23877, %23885, %float0.000000e00_21607, %true_21608, %none_21609, %none_21610) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23893#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21611 = torch.constant.float 0.000000e+00
%true_21612 = torch.constant.bool true
%none_21613 = torch.constant.none
%none_21614 = torch.constant.none
%23894:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23870, %23878, %23886, %float0.000000e00_21611, %true_21612, %none_21613, %none_21614) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23894#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_21615 = torch.constant.float 0.000000e+00
%true_21616 = torch.constant.bool true
%none_21617 = torch.constant.none
%none_21618 = torch.constant.none
%23895:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%23871, %23879, %23887, %float0.000000e00_21615, %true_21616, %none_21617, %none_21618) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %23895#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_21619 = torch.constant.int 1
%int2_21620 = torch.constant.int 2
%23896 = torch.aten.transpose.int %23888#0, %int1_21619, %int2_21620 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21621 = torch.constant.int 1
%int2_21622 = torch.constant.int 2
%23897 = torch.aten.transpose.int %23889#0, %int1_21621, %int2_21622 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21623 = torch.constant.int 1
%int2_21624 = torch.constant.int 2
%23898 = torch.aten.transpose.int %23890#0, %int1_21623, %int2_21624 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21625 = torch.constant.int 1
%int2_21626 = torch.constant.int 2
%23899 = torch.aten.transpose.int %23891#0, %int1_21625, %int2_21626 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21627 = torch.constant.int 1
%int2_21628 = torch.constant.int 2
%23900 = torch.aten.transpose.int %23892#0, %int1_21627, %int2_21628 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21629 = torch.constant.int 1
%int2_21630 = torch.constant.int 2
%23901 = torch.aten.transpose.int %23893#0, %int1_21629, %int2_21630 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21631 = torch.constant.int 1
%int2_21632 = torch.constant.int 2
%23902 = torch.aten.transpose.int %23894#0, %int1_21631, %int2_21632 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_21633 = torch.constant.int 1
%int2_21634 = torch.constant.int 2
%23903 = torch.aten.transpose.int %23895#0, %int1_21633, %int2_21634 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %23903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_21635 = torch.constant.int 4
%int512_21636 = torch.constant.int 512
%23904 = torch.prim.ListConstruct %int4_21635, %23225, %int512_21636 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23905 = torch.aten.view %23896, %23904 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21637 = torch.constant.int 4
%int512_21638 = torch.constant.int 512
%23906 = torch.prim.ListConstruct %int4_21637, %23240, %int512_21638 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23907 = torch.aten.view %23897, %23906 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21639 = torch.constant.int 4
%int512_21640 = torch.constant.int 512
%23908 = torch.prim.ListConstruct %int4_21639, %23255, %int512_21640 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23909 = torch.aten.view %23898, %23908 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21641 = torch.constant.int 4
%int512_21642 = torch.constant.int 512
%23910 = torch.prim.ListConstruct %int4_21641, %23270, %int512_21642 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23911 = torch.aten.view %23899, %23910 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21643 = torch.constant.int 4
%int512_21644 = torch.constant.int 512
%23912 = torch.prim.ListConstruct %int4_21643, %23285, %int512_21644 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23913 = torch.aten.view %23900, %23912 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21645 = torch.constant.int 4
%int512_21646 = torch.constant.int 512
%23914 = torch.prim.ListConstruct %int4_21645, %23300, %int512_21646 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23915 = torch.aten.view %23901, %23914 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21647 = torch.constant.int 4
%int512_21648 = torch.constant.int 512
%23916 = torch.prim.ListConstruct %int4_21647, %23315, %int512_21648 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23917 = torch.aten.view %23902, %23916 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_21649 = torch.constant.int 4
%int512_21650 = torch.constant.int 512
%23918 = torch.prim.ListConstruct %int4_21649, %23330, %int512_21650 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23919 = torch.aten.view %23903, %23918 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %23919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_21651 = torch.constant.int 1
%int0_21652 = torch.constant.int 0
%23920 = torch.prim.ListConstruct %int1_21651, %int0_21652 : (!torch.int, !torch.int) -> !torch.list<int>
%23921 = torch.aten.permute %832, %23920 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21653 = torch.constant.int 1
%int0_21654 = torch.constant.int 0
%23922 = torch.prim.ListConstruct %int1_21653, %int0_21654 : (!torch.int, !torch.int) -> !torch.list<int>
%23923 = torch.aten.permute %833, %23922 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21655 = torch.constant.int 1
%int0_21656 = torch.constant.int 0
%23924 = torch.prim.ListConstruct %int1_21655, %int0_21656 : (!torch.int, !torch.int) -> !torch.list<int>
%23925 = torch.aten.permute %834, %23924 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21657 = torch.constant.int 1
%int0_21658 = torch.constant.int 0
%23926 = torch.prim.ListConstruct %int1_21657, %int0_21658 : (!torch.int, !torch.int) -> !torch.list<int>
%23927 = torch.aten.permute %835, %23926 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21659 = torch.constant.int 1
%int0_21660 = torch.constant.int 0
%23928 = torch.prim.ListConstruct %int1_21659, %int0_21660 : (!torch.int, !torch.int) -> !torch.list<int>
%23929 = torch.aten.permute %836, %23928 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21661 = torch.constant.int 1
%int0_21662 = torch.constant.int 0
%23930 = torch.prim.ListConstruct %int1_21661, %int0_21662 : (!torch.int, !torch.int) -> !torch.list<int>
%23931 = torch.aten.permute %837, %23930 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21663 = torch.constant.int 1
%int0_21664 = torch.constant.int 0
%23932 = torch.prim.ListConstruct %int1_21663, %int0_21664 : (!torch.int, !torch.int) -> !torch.list<int>
%23933 = torch.aten.permute %838, %23932 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_21665 = torch.constant.int 1
%int0_21666 = torch.constant.int 0
%23934 = torch.prim.ListConstruct %int1_21665, %int0_21666 : (!torch.int, !torch.int) -> !torch.list<int>
%23935 = torch.aten.permute %839, %23934 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_21667 = torch.constant.int 4
%23936 = torch.aten.mul.int %int4_21667, %23225 : !torch.int, !torch.int -> !torch.int
%int512_21668 = torch.constant.int 512
%23937 = torch.prim.ListConstruct %23936, %int512_21668 : (!torch.int, !torch.int) -> !torch.list<int>
%23938 = torch.aten.view %23905, %23937 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23938, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23939 = torch.aten.mm %23938, %23921 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23939, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21669 = torch.constant.int 4
%int4096_21670 = torch.constant.int 4096
%23940 = torch.prim.ListConstruct %int4_21669, %23225, %int4096_21670 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23941 = torch.aten.view %23939, %23940 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21671 = torch.constant.int 4
%23942 = torch.aten.mul.int %int4_21671, %23240 : !torch.int, !torch.int -> !torch.int
%int512_21672 = torch.constant.int 512
%23943 = torch.prim.ListConstruct %23942, %int512_21672 : (!torch.int, !torch.int) -> !torch.list<int>
%23944 = torch.aten.view %23907, %23943 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23944, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23945 = torch.aten.mm %23944, %23923 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23945, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21673 = torch.constant.int 4
%int4096_21674 = torch.constant.int 4096
%23946 = torch.prim.ListConstruct %int4_21673, %23240, %int4096_21674 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23947 = torch.aten.view %23945, %23946 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21675 = torch.constant.int 4
%23948 = torch.aten.mul.int %int4_21675, %23255 : !torch.int, !torch.int -> !torch.int
%int512_21676 = torch.constant.int 512
%23949 = torch.prim.ListConstruct %23948, %int512_21676 : (!torch.int, !torch.int) -> !torch.list<int>
%23950 = torch.aten.view %23909, %23949 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23950, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23951 = torch.aten.mm %23950, %23925 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23951, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21677 = torch.constant.int 4
%int4096_21678 = torch.constant.int 4096
%23952 = torch.prim.ListConstruct %int4_21677, %23255, %int4096_21678 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23953 = torch.aten.view %23951, %23952 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21679 = torch.constant.int 4
%23954 = torch.aten.mul.int %int4_21679, %23270 : !torch.int, !torch.int -> !torch.int
%int512_21680 = torch.constant.int 512
%23955 = torch.prim.ListConstruct %23954, %int512_21680 : (!torch.int, !torch.int) -> !torch.list<int>
%23956 = torch.aten.view %23911, %23955 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23956, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23957 = torch.aten.mm %23956, %23927 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23957, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21681 = torch.constant.int 4
%int4096_21682 = torch.constant.int 4096
%23958 = torch.prim.ListConstruct %int4_21681, %23270, %int4096_21682 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23959 = torch.aten.view %23957, %23958 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21683 = torch.constant.int 4
%23960 = torch.aten.mul.int %int4_21683, %23285 : !torch.int, !torch.int -> !torch.int
%int512_21684 = torch.constant.int 512
%23961 = torch.prim.ListConstruct %23960, %int512_21684 : (!torch.int, !torch.int) -> !torch.list<int>
%23962 = torch.aten.view %23913, %23961 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23962, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23963 = torch.aten.mm %23962, %23929 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23963, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21685 = torch.constant.int 4
%int4096_21686 = torch.constant.int 4096
%23964 = torch.prim.ListConstruct %int4_21685, %23285, %int4096_21686 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23965 = torch.aten.view %23963, %23964 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21687 = torch.constant.int 4
%23966 = torch.aten.mul.int %int4_21687, %23300 : !torch.int, !torch.int -> !torch.int
%int512_21688 = torch.constant.int 512
%23967 = torch.prim.ListConstruct %23966, %int512_21688 : (!torch.int, !torch.int) -> !torch.list<int>
%23968 = torch.aten.view %23915, %23967 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23968, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23969 = torch.aten.mm %23968, %23931 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23969, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21689 = torch.constant.int 4
%int4096_21690 = torch.constant.int 4096
%23970 = torch.prim.ListConstruct %int4_21689, %23300, %int4096_21690 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23971 = torch.aten.view %23969, %23970 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21691 = torch.constant.int 4
%23972 = torch.aten.mul.int %int4_21691, %23315 : !torch.int, !torch.int -> !torch.int
%int512_21692 = torch.constant.int 512
%23973 = torch.prim.ListConstruct %23972, %int512_21692 : (!torch.int, !torch.int) -> !torch.list<int>
%23974 = torch.aten.view %23917, %23973 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23974, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23975 = torch.aten.mm %23974, %23933 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23975, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21693 = torch.constant.int 4
%int4096_21694 = torch.constant.int 4096
%23976 = torch.prim.ListConstruct %int4_21693, %23315, %int4096_21694 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23977 = torch.aten.view %23975, %23976 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_21695 = torch.constant.int 4
%23978 = torch.aten.mul.int %int4_21695, %23330 : !torch.int, !torch.int -> !torch.int
%int512_21696 = torch.constant.int 512
%23979 = torch.prim.ListConstruct %23978, %int512_21696 : (!torch.int, !torch.int) -> !torch.list<int>
%23980 = torch.aten.view %23919, %23979 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %23980, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%23981 = torch.aten.mm %23980, %23935 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %23981, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_21697 = torch.constant.int 4
%int4096_21698 = torch.constant.int 4096
%23982 = torch.prim.ListConstruct %int4_21697, %23330, %int4096_21698 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%23983 = torch.aten.view %23981, %23982 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%23984 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21699 = arith.constant 1 : index
%dim_21700 = tensor.dim %23984, %c1_21699 : tensor<4x?x4096xf16>
%23985 = flow.tensor.transfer %23984 : tensor<4x?x4096xf16>{%dim_21700} to #hal.device.promise<@__device_0>
%23986 = torch_c.from_builtin_tensor %23985 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%23987 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21701 = arith.constant 1 : index
%dim_21702 = tensor.dim %23987, %c1_21701 : tensor<4x?x4096xf16>
%23988 = flow.tensor.transfer %23987 : tensor<4x?x4096xf16>{%dim_21702} to #hal.device.promise<@__device_0>
%23989 = torch_c.from_builtin_tensor %23988 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%23990 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21703 = arith.constant 1 : index
%dim_21704 = tensor.dim %23990, %c1_21703 : tensor<4x?x4096xf16>
%23991 = flow.tensor.transfer %23990 : tensor<4x?x4096xf16>{%dim_21704} to #hal.device.promise<@__device_0>
%23992 = torch_c.from_builtin_tensor %23991 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%23993 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21705 = arith.constant 1 : index
%dim_21706 = tensor.dim %23993, %c1_21705 : tensor<4x?x4096xf16>
%23994 = flow.tensor.transfer %23993 : tensor<4x?x4096xf16>{%dim_21706} to #hal.device.promise<@__device_0>
%23995 = torch_c.from_builtin_tensor %23994 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%23996 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21707 = arith.constant 1 : index
%dim_21708 = tensor.dim %23996, %c1_21707 : tensor<4x?x4096xf16>
%23997 = flow.tensor.transfer %23996 : tensor<4x?x4096xf16>{%dim_21708} to #hal.device.promise<@__device_0>
%23998 = torch_c.from_builtin_tensor %23997 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %23998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%23999 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21709 = arith.constant 1 : index
%dim_21710 = tensor.dim %23999, %c1_21709 : tensor<4x?x4096xf16>
%24000 = flow.tensor.transfer %23999 : tensor<4x?x4096xf16>{%dim_21710} to #hal.device.promise<@__device_0>
%24001 = torch_c.from_builtin_tensor %24000 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24002 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21711 = arith.constant 1 : index
%dim_21712 = tensor.dim %24002, %c1_21711 : tensor<4x?x4096xf16>
%24003 = flow.tensor.transfer %24002 : tensor<4x?x4096xf16>{%dim_21712} to #hal.device.promise<@__device_0>
%24004 = torch_c.from_builtin_tensor %24003 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21713 = torch.constant.int 1
%24005 = torch.aten.add.Tensor %23941, %23986, %int1_21713 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21714 = torch.constant.int 1
%24006 = torch.aten.add.Tensor %24005, %23989, %int1_21714 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21715 = torch.constant.int 1
%24007 = torch.aten.add.Tensor %24006, %23992, %int1_21715 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21716 = torch.constant.int 1
%24008 = torch.aten.add.Tensor %24007, %23995, %int1_21716 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21717 = torch.constant.int 1
%24009 = torch.aten.add.Tensor %24008, %23998, %int1_21717 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21718 = torch.constant.int 1
%24010 = torch.aten.add.Tensor %24009, %24001, %int1_21718 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21719 = torch.constant.int 1
%24011 = torch.aten.add.Tensor %24010, %24004, %int1_21719 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24012 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21720 = arith.constant 1 : index
%dim_21721 = tensor.dim %24012, %c1_21720 : tensor<4x?x4096xf16>
%24013 = flow.tensor.transfer %24012 : tensor<4x?x4096xf16>{%dim_21721} to #hal.device.promise<@__device_1>
%24014 = torch_c.from_builtin_tensor %24013 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24015 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21722 = arith.constant 1 : index
%dim_21723 = tensor.dim %24015, %c1_21722 : tensor<4x?x4096xf16>
%24016 = flow.tensor.transfer %24015 : tensor<4x?x4096xf16>{%dim_21723} to #hal.device.promise<@__device_1>
%24017 = torch_c.from_builtin_tensor %24016 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24018 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21724 = arith.constant 1 : index
%dim_21725 = tensor.dim %24018, %c1_21724 : tensor<4x?x4096xf16>
%24019 = flow.tensor.transfer %24018 : tensor<4x?x4096xf16>{%dim_21725} to #hal.device.promise<@__device_1>
%24020 = torch_c.from_builtin_tensor %24019 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24021 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21726 = arith.constant 1 : index
%dim_21727 = tensor.dim %24021, %c1_21726 : tensor<4x?x4096xf16>
%24022 = flow.tensor.transfer %24021 : tensor<4x?x4096xf16>{%dim_21727} to #hal.device.promise<@__device_1>
%24023 = torch_c.from_builtin_tensor %24022 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24024 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21728 = arith.constant 1 : index
%dim_21729 = tensor.dim %24024, %c1_21728 : tensor<4x?x4096xf16>
%24025 = flow.tensor.transfer %24024 : tensor<4x?x4096xf16>{%dim_21729} to #hal.device.promise<@__device_1>
%24026 = torch_c.from_builtin_tensor %24025 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24027 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21730 = arith.constant 1 : index
%dim_21731 = tensor.dim %24027, %c1_21730 : tensor<4x?x4096xf16>
%24028 = flow.tensor.transfer %24027 : tensor<4x?x4096xf16>{%dim_21731} to #hal.device.promise<@__device_1>
%24029 = torch_c.from_builtin_tensor %24028 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24030 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21732 = arith.constant 1 : index
%dim_21733 = tensor.dim %24030, %c1_21732 : tensor<4x?x4096xf16>
%24031 = flow.tensor.transfer %24030 : tensor<4x?x4096xf16>{%dim_21733} to #hal.device.promise<@__device_1>
%24032 = torch_c.from_builtin_tensor %24031 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21734 = torch.constant.int 1
%24033 = torch.aten.add.Tensor %24014, %23947, %int1_21734 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21735 = torch.constant.int 1
%24034 = torch.aten.add.Tensor %24033, %24017, %int1_21735 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21736 = torch.constant.int 1
%24035 = torch.aten.add.Tensor %24034, %24020, %int1_21736 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21737 = torch.constant.int 1
%24036 = torch.aten.add.Tensor %24035, %24023, %int1_21737 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21738 = torch.constant.int 1
%24037 = torch.aten.add.Tensor %24036, %24026, %int1_21738 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21739 = torch.constant.int 1
%24038 = torch.aten.add.Tensor %24037, %24029, %int1_21739 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21740 = torch.constant.int 1
%24039 = torch.aten.add.Tensor %24038, %24032, %int1_21740 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24040 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21741 = arith.constant 1 : index
%dim_21742 = tensor.dim %24040, %c1_21741 : tensor<4x?x4096xf16>
%24041 = flow.tensor.transfer %24040 : tensor<4x?x4096xf16>{%dim_21742} to #hal.device.promise<@__device_2>
%24042 = torch_c.from_builtin_tensor %24041 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24043 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21743 = arith.constant 1 : index
%dim_21744 = tensor.dim %24043, %c1_21743 : tensor<4x?x4096xf16>
%24044 = flow.tensor.transfer %24043 : tensor<4x?x4096xf16>{%dim_21744} to #hal.device.promise<@__device_2>
%24045 = torch_c.from_builtin_tensor %24044 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24046 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21745 = arith.constant 1 : index
%dim_21746 = tensor.dim %24046, %c1_21745 : tensor<4x?x4096xf16>
%24047 = flow.tensor.transfer %24046 : tensor<4x?x4096xf16>{%dim_21746} to #hal.device.promise<@__device_2>
%24048 = torch_c.from_builtin_tensor %24047 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24049 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21747 = arith.constant 1 : index
%dim_21748 = tensor.dim %24049, %c1_21747 : tensor<4x?x4096xf16>
%24050 = flow.tensor.transfer %24049 : tensor<4x?x4096xf16>{%dim_21748} to #hal.device.promise<@__device_2>
%24051 = torch_c.from_builtin_tensor %24050 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24052 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21749 = arith.constant 1 : index
%dim_21750 = tensor.dim %24052, %c1_21749 : tensor<4x?x4096xf16>
%24053 = flow.tensor.transfer %24052 : tensor<4x?x4096xf16>{%dim_21750} to #hal.device.promise<@__device_2>
%24054 = torch_c.from_builtin_tensor %24053 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24055 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21751 = arith.constant 1 : index
%dim_21752 = tensor.dim %24055, %c1_21751 : tensor<4x?x4096xf16>
%24056 = flow.tensor.transfer %24055 : tensor<4x?x4096xf16>{%dim_21752} to #hal.device.promise<@__device_2>
%24057 = torch_c.from_builtin_tensor %24056 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24058 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21753 = arith.constant 1 : index
%dim_21754 = tensor.dim %24058, %c1_21753 : tensor<4x?x4096xf16>
%24059 = flow.tensor.transfer %24058 : tensor<4x?x4096xf16>{%dim_21754} to #hal.device.promise<@__device_2>
%24060 = torch_c.from_builtin_tensor %24059 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21755 = torch.constant.int 1
%24061 = torch.aten.add.Tensor %24042, %24045, %int1_21755 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21756 = torch.constant.int 1
%24062 = torch.aten.add.Tensor %24061, %23953, %int1_21756 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21757 = torch.constant.int 1
%24063 = torch.aten.add.Tensor %24062, %24048, %int1_21757 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21758 = torch.constant.int 1
%24064 = torch.aten.add.Tensor %24063, %24051, %int1_21758 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21759 = torch.constant.int 1
%24065 = torch.aten.add.Tensor %24064, %24054, %int1_21759 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21760 = torch.constant.int 1
%24066 = torch.aten.add.Tensor %24065, %24057, %int1_21760 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21761 = torch.constant.int 1
%24067 = torch.aten.add.Tensor %24066, %24060, %int1_21761 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24068 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21762 = arith.constant 1 : index
%dim_21763 = tensor.dim %24068, %c1_21762 : tensor<4x?x4096xf16>
%24069 = flow.tensor.transfer %24068 : tensor<4x?x4096xf16>{%dim_21763} to #hal.device.promise<@__device_3>
%24070 = torch_c.from_builtin_tensor %24069 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24071 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21764 = arith.constant 1 : index
%dim_21765 = tensor.dim %24071, %c1_21764 : tensor<4x?x4096xf16>
%24072 = flow.tensor.transfer %24071 : tensor<4x?x4096xf16>{%dim_21765} to #hal.device.promise<@__device_3>
%24073 = torch_c.from_builtin_tensor %24072 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24074 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21766 = arith.constant 1 : index
%dim_21767 = tensor.dim %24074, %c1_21766 : tensor<4x?x4096xf16>
%24075 = flow.tensor.transfer %24074 : tensor<4x?x4096xf16>{%dim_21767} to #hal.device.promise<@__device_3>
%24076 = torch_c.from_builtin_tensor %24075 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24077 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21768 = arith.constant 1 : index
%dim_21769 = tensor.dim %24077, %c1_21768 : tensor<4x?x4096xf16>
%24078 = flow.tensor.transfer %24077 : tensor<4x?x4096xf16>{%dim_21769} to #hal.device.promise<@__device_3>
%24079 = torch_c.from_builtin_tensor %24078 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24080 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21770 = arith.constant 1 : index
%dim_21771 = tensor.dim %24080, %c1_21770 : tensor<4x?x4096xf16>
%24081 = flow.tensor.transfer %24080 : tensor<4x?x4096xf16>{%dim_21771} to #hal.device.promise<@__device_3>
%24082 = torch_c.from_builtin_tensor %24081 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24083 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21772 = arith.constant 1 : index
%dim_21773 = tensor.dim %24083, %c1_21772 : tensor<4x?x4096xf16>
%24084 = flow.tensor.transfer %24083 : tensor<4x?x4096xf16>{%dim_21773} to #hal.device.promise<@__device_3>
%24085 = torch_c.from_builtin_tensor %24084 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24086 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21774 = arith.constant 1 : index
%dim_21775 = tensor.dim %24086, %c1_21774 : tensor<4x?x4096xf16>
%24087 = flow.tensor.transfer %24086 : tensor<4x?x4096xf16>{%dim_21775} to #hal.device.promise<@__device_3>
%24088 = torch_c.from_builtin_tensor %24087 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21776 = torch.constant.int 1
%24089 = torch.aten.add.Tensor %24070, %24073, %int1_21776 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21777 = torch.constant.int 1
%24090 = torch.aten.add.Tensor %24089, %24076, %int1_21777 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21778 = torch.constant.int 1
%24091 = torch.aten.add.Tensor %24090, %23959, %int1_21778 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21779 = torch.constant.int 1
%24092 = torch.aten.add.Tensor %24091, %24079, %int1_21779 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21780 = torch.constant.int 1
%24093 = torch.aten.add.Tensor %24092, %24082, %int1_21780 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21781 = torch.constant.int 1
%24094 = torch.aten.add.Tensor %24093, %24085, %int1_21781 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21782 = torch.constant.int 1
%24095 = torch.aten.add.Tensor %24094, %24088, %int1_21782 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24096 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21783 = arith.constant 1 : index
%dim_21784 = tensor.dim %24096, %c1_21783 : tensor<4x?x4096xf16>
%24097 = flow.tensor.transfer %24096 : tensor<4x?x4096xf16>{%dim_21784} to #hal.device.promise<@__device_4>
%24098 = torch_c.from_builtin_tensor %24097 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24099 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21785 = arith.constant 1 : index
%dim_21786 = tensor.dim %24099, %c1_21785 : tensor<4x?x4096xf16>
%24100 = flow.tensor.transfer %24099 : tensor<4x?x4096xf16>{%dim_21786} to #hal.device.promise<@__device_4>
%24101 = torch_c.from_builtin_tensor %24100 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24102 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21787 = arith.constant 1 : index
%dim_21788 = tensor.dim %24102, %c1_21787 : tensor<4x?x4096xf16>
%24103 = flow.tensor.transfer %24102 : tensor<4x?x4096xf16>{%dim_21788} to #hal.device.promise<@__device_4>
%24104 = torch_c.from_builtin_tensor %24103 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24105 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21789 = arith.constant 1 : index
%dim_21790 = tensor.dim %24105, %c1_21789 : tensor<4x?x4096xf16>
%24106 = flow.tensor.transfer %24105 : tensor<4x?x4096xf16>{%dim_21790} to #hal.device.promise<@__device_4>
%24107 = torch_c.from_builtin_tensor %24106 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24108 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21791 = arith.constant 1 : index
%dim_21792 = tensor.dim %24108, %c1_21791 : tensor<4x?x4096xf16>
%24109 = flow.tensor.transfer %24108 : tensor<4x?x4096xf16>{%dim_21792} to #hal.device.promise<@__device_4>
%24110 = torch_c.from_builtin_tensor %24109 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24111 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21793 = arith.constant 1 : index
%dim_21794 = tensor.dim %24111, %c1_21793 : tensor<4x?x4096xf16>
%24112 = flow.tensor.transfer %24111 : tensor<4x?x4096xf16>{%dim_21794} to #hal.device.promise<@__device_4>
%24113 = torch_c.from_builtin_tensor %24112 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24114 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21795 = arith.constant 1 : index
%dim_21796 = tensor.dim %24114, %c1_21795 : tensor<4x?x4096xf16>
%24115 = flow.tensor.transfer %24114 : tensor<4x?x4096xf16>{%dim_21796} to #hal.device.promise<@__device_4>
%24116 = torch_c.from_builtin_tensor %24115 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21797 = torch.constant.int 1
%24117 = torch.aten.add.Tensor %24098, %24101, %int1_21797 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21798 = torch.constant.int 1
%24118 = torch.aten.add.Tensor %24117, %24104, %int1_21798 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21799 = torch.constant.int 1
%24119 = torch.aten.add.Tensor %24118, %24107, %int1_21799 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21800 = torch.constant.int 1
%24120 = torch.aten.add.Tensor %24119, %23965, %int1_21800 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21801 = torch.constant.int 1
%24121 = torch.aten.add.Tensor %24120, %24110, %int1_21801 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21802 = torch.constant.int 1
%24122 = torch.aten.add.Tensor %24121, %24113, %int1_21802 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21803 = torch.constant.int 1
%24123 = torch.aten.add.Tensor %24122, %24116, %int1_21803 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24124 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21804 = arith.constant 1 : index
%dim_21805 = tensor.dim %24124, %c1_21804 : tensor<4x?x4096xf16>
%24125 = flow.tensor.transfer %24124 : tensor<4x?x4096xf16>{%dim_21805} to #hal.device.promise<@__device_5>
%24126 = torch_c.from_builtin_tensor %24125 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24127 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21806 = arith.constant 1 : index
%dim_21807 = tensor.dim %24127, %c1_21806 : tensor<4x?x4096xf16>
%24128 = flow.tensor.transfer %24127 : tensor<4x?x4096xf16>{%dim_21807} to #hal.device.promise<@__device_5>
%24129 = torch_c.from_builtin_tensor %24128 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24130 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21808 = arith.constant 1 : index
%dim_21809 = tensor.dim %24130, %c1_21808 : tensor<4x?x4096xf16>
%24131 = flow.tensor.transfer %24130 : tensor<4x?x4096xf16>{%dim_21809} to #hal.device.promise<@__device_5>
%24132 = torch_c.from_builtin_tensor %24131 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24133 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21810 = arith.constant 1 : index
%dim_21811 = tensor.dim %24133, %c1_21810 : tensor<4x?x4096xf16>
%24134 = flow.tensor.transfer %24133 : tensor<4x?x4096xf16>{%dim_21811} to #hal.device.promise<@__device_5>
%24135 = torch_c.from_builtin_tensor %24134 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24136 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21812 = arith.constant 1 : index
%dim_21813 = tensor.dim %24136, %c1_21812 : tensor<4x?x4096xf16>
%24137 = flow.tensor.transfer %24136 : tensor<4x?x4096xf16>{%dim_21813} to #hal.device.promise<@__device_5>
%24138 = torch_c.from_builtin_tensor %24137 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24139 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21814 = arith.constant 1 : index
%dim_21815 = tensor.dim %24139, %c1_21814 : tensor<4x?x4096xf16>
%24140 = flow.tensor.transfer %24139 : tensor<4x?x4096xf16>{%dim_21815} to #hal.device.promise<@__device_5>
%24141 = torch_c.from_builtin_tensor %24140 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24142 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21816 = arith.constant 1 : index
%dim_21817 = tensor.dim %24142, %c1_21816 : tensor<4x?x4096xf16>
%24143 = flow.tensor.transfer %24142 : tensor<4x?x4096xf16>{%dim_21817} to #hal.device.promise<@__device_5>
%24144 = torch_c.from_builtin_tensor %24143 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21818 = torch.constant.int 1
%24145 = torch.aten.add.Tensor %24126, %24129, %int1_21818 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21819 = torch.constant.int 1
%24146 = torch.aten.add.Tensor %24145, %24132, %int1_21819 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21820 = torch.constant.int 1
%24147 = torch.aten.add.Tensor %24146, %24135, %int1_21820 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21821 = torch.constant.int 1
%24148 = torch.aten.add.Tensor %24147, %24138, %int1_21821 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21822 = torch.constant.int 1
%24149 = torch.aten.add.Tensor %24148, %23971, %int1_21822 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21823 = torch.constant.int 1
%24150 = torch.aten.add.Tensor %24149, %24141, %int1_21823 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21824 = torch.constant.int 1
%24151 = torch.aten.add.Tensor %24150, %24144, %int1_21824 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24152 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21825 = arith.constant 1 : index
%dim_21826 = tensor.dim %24152, %c1_21825 : tensor<4x?x4096xf16>
%24153 = flow.tensor.transfer %24152 : tensor<4x?x4096xf16>{%dim_21826} to #hal.device.promise<@__device_6>
%24154 = torch_c.from_builtin_tensor %24153 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24155 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21827 = arith.constant 1 : index
%dim_21828 = tensor.dim %24155, %c1_21827 : tensor<4x?x4096xf16>
%24156 = flow.tensor.transfer %24155 : tensor<4x?x4096xf16>{%dim_21828} to #hal.device.promise<@__device_6>
%24157 = torch_c.from_builtin_tensor %24156 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24158 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21829 = arith.constant 1 : index
%dim_21830 = tensor.dim %24158, %c1_21829 : tensor<4x?x4096xf16>
%24159 = flow.tensor.transfer %24158 : tensor<4x?x4096xf16>{%dim_21830} to #hal.device.promise<@__device_6>
%24160 = torch_c.from_builtin_tensor %24159 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24161 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21831 = arith.constant 1 : index
%dim_21832 = tensor.dim %24161, %c1_21831 : tensor<4x?x4096xf16>
%24162 = flow.tensor.transfer %24161 : tensor<4x?x4096xf16>{%dim_21832} to #hal.device.promise<@__device_6>
%24163 = torch_c.from_builtin_tensor %24162 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24164 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21833 = arith.constant 1 : index
%dim_21834 = tensor.dim %24164, %c1_21833 : tensor<4x?x4096xf16>
%24165 = flow.tensor.transfer %24164 : tensor<4x?x4096xf16>{%dim_21834} to #hal.device.promise<@__device_6>
%24166 = torch_c.from_builtin_tensor %24165 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24167 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21835 = arith.constant 1 : index
%dim_21836 = tensor.dim %24167, %c1_21835 : tensor<4x?x4096xf16>
%24168 = flow.tensor.transfer %24167 : tensor<4x?x4096xf16>{%dim_21836} to #hal.device.promise<@__device_6>
%24169 = torch_c.from_builtin_tensor %24168 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24170 = torch_c.to_builtin_tensor %23983 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21837 = arith.constant 1 : index
%dim_21838 = tensor.dim %24170, %c1_21837 : tensor<4x?x4096xf16>
%24171 = flow.tensor.transfer %24170 : tensor<4x?x4096xf16>{%dim_21838} to #hal.device.promise<@__device_6>
%24172 = torch_c.from_builtin_tensor %24171 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21839 = torch.constant.int 1
%24173 = torch.aten.add.Tensor %24154, %24157, %int1_21839 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21840 = torch.constant.int 1
%24174 = torch.aten.add.Tensor %24173, %24160, %int1_21840 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21841 = torch.constant.int 1
%24175 = torch.aten.add.Tensor %24174, %24163, %int1_21841 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21842 = torch.constant.int 1
%24176 = torch.aten.add.Tensor %24175, %24166, %int1_21842 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21843 = torch.constant.int 1
%24177 = torch.aten.add.Tensor %24176, %24169, %int1_21843 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21844 = torch.constant.int 1
%24178 = torch.aten.add.Tensor %24177, %23977, %int1_21844 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21845 = torch.constant.int 1
%24179 = torch.aten.add.Tensor %24178, %24172, %int1_21845 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24180 = torch_c.to_builtin_tensor %23941 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21846 = arith.constant 1 : index
%dim_21847 = tensor.dim %24180, %c1_21846 : tensor<4x?x4096xf16>
%24181 = flow.tensor.transfer %24180 : tensor<4x?x4096xf16>{%dim_21847} to #hal.device.promise<@__device_7>
%24182 = torch_c.from_builtin_tensor %24181 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24183 = torch_c.to_builtin_tensor %23947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21848 = arith.constant 1 : index
%dim_21849 = tensor.dim %24183, %c1_21848 : tensor<4x?x4096xf16>
%24184 = flow.tensor.transfer %24183 : tensor<4x?x4096xf16>{%dim_21849} to #hal.device.promise<@__device_7>
%24185 = torch_c.from_builtin_tensor %24184 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24186 = torch_c.to_builtin_tensor %23953 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21850 = arith.constant 1 : index
%dim_21851 = tensor.dim %24186, %c1_21850 : tensor<4x?x4096xf16>
%24187 = flow.tensor.transfer %24186 : tensor<4x?x4096xf16>{%dim_21851} to #hal.device.promise<@__device_7>
%24188 = torch_c.from_builtin_tensor %24187 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24189 = torch_c.to_builtin_tensor %23959 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21852 = arith.constant 1 : index
%dim_21853 = tensor.dim %24189, %c1_21852 : tensor<4x?x4096xf16>
%24190 = flow.tensor.transfer %24189 : tensor<4x?x4096xf16>{%dim_21853} to #hal.device.promise<@__device_7>
%24191 = torch_c.from_builtin_tensor %24190 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24192 = torch_c.to_builtin_tensor %23965 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21854 = arith.constant 1 : index
%dim_21855 = tensor.dim %24192, %c1_21854 : tensor<4x?x4096xf16>
%24193 = flow.tensor.transfer %24192 : tensor<4x?x4096xf16>{%dim_21855} to #hal.device.promise<@__device_7>
%24194 = torch_c.from_builtin_tensor %24193 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24195 = torch_c.to_builtin_tensor %23971 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21856 = arith.constant 1 : index
%dim_21857 = tensor.dim %24195, %c1_21856 : tensor<4x?x4096xf16>
%24196 = flow.tensor.transfer %24195 : tensor<4x?x4096xf16>{%dim_21857} to #hal.device.promise<@__device_7>
%24197 = torch_c.from_builtin_tensor %24196 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24198 = torch_c.to_builtin_tensor %23977 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_21858 = arith.constant 1 : index
%dim_21859 = tensor.dim %24198, %c1_21858 : tensor<4x?x4096xf16>
%24199 = flow.tensor.transfer %24198 : tensor<4x?x4096xf16>{%dim_21859} to #hal.device.promise<@__device_7>
%24200 = torch_c.from_builtin_tensor %24199 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21860 = torch.constant.int 1
%24201 = torch.aten.add.Tensor %24182, %24185, %int1_21860 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21861 = torch.constant.int 1
%24202 = torch.aten.add.Tensor %24201, %24188, %int1_21861 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21862 = torch.constant.int 1
%24203 = torch.aten.add.Tensor %24202, %24191, %int1_21862 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21863 = torch.constant.int 1
%24204 = torch.aten.add.Tensor %24203, %24194, %int1_21863 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21864 = torch.constant.int 1
%24205 = torch.aten.add.Tensor %24204, %24197, %int1_21864 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21865 = torch.constant.int 1
%24206 = torch.aten.add.Tensor %24205, %24200, %int1_21865 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21866 = torch.constant.int 1
%24207 = torch.aten.add.Tensor %24206, %23983, %int1_21866 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21867 = torch.constant.int 1
%24208 = torch.aten.add.Tensor %22867, %24011, %int1_21867 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21868 = torch.constant.int 1
%24209 = torch.aten.add.Tensor %22868, %24039, %int1_21868 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21869 = torch.constant.int 1
%24210 = torch.aten.add.Tensor %22869, %24067, %int1_21869 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21870 = torch.constant.int 1
%24211 = torch.aten.add.Tensor %22870, %24095, %int1_21870 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21871 = torch.constant.int 1
%24212 = torch.aten.add.Tensor %22871, %24123, %int1_21871 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21872 = torch.constant.int 1
%24213 = torch.aten.add.Tensor %22872, %24151, %int1_21872 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21873 = torch.constant.int 1
%24214 = torch.aten.add.Tensor %22873, %24179, %int1_21873 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21874 = torch.constant.int 1
%24215 = torch.aten.add.Tensor %22874, %24207, %int1_21874 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_21875 = torch.constant.int 6
%24216 = torch.prims.convert_element_type %24208, %int6_21875 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21876 = torch.constant.int 6
%24217 = torch.prims.convert_element_type %24209, %int6_21876 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21877 = torch.constant.int 6
%24218 = torch.prims.convert_element_type %24210, %int6_21877 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21878 = torch.constant.int 6
%24219 = torch.prims.convert_element_type %24211, %int6_21878 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21879 = torch.constant.int 6
%24220 = torch.prims.convert_element_type %24212, %int6_21879 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21880 = torch.constant.int 6
%24221 = torch.prims.convert_element_type %24213, %int6_21880 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21881 = torch.constant.int 6
%24222 = torch.prims.convert_element_type %24214, %int6_21881 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_21882 = torch.constant.int 6
%24223 = torch.prims.convert_element_type %24215, %int6_21882 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21883 = torch.constant.int 2
%24224 = torch.aten.pow.Tensor_Scalar %24216, %int2_21883 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21884 = torch.constant.int 2
%24225 = torch.aten.pow.Tensor_Scalar %24217, %int2_21884 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21885 = torch.constant.int 2
%24226 = torch.aten.pow.Tensor_Scalar %24218, %int2_21885 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21886 = torch.constant.int 2
%24227 = torch.aten.pow.Tensor_Scalar %24219, %int2_21886 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21887 = torch.constant.int 2
%24228 = torch.aten.pow.Tensor_Scalar %24220, %int2_21887 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21888 = torch.constant.int 2
%24229 = torch.aten.pow.Tensor_Scalar %24221, %int2_21888 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21889 = torch.constant.int 2
%24230 = torch.aten.pow.Tensor_Scalar %24222, %int2_21889 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_21890 = torch.constant.int 2
%24231 = torch.aten.pow.Tensor_Scalar %24223, %int2_21890 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_21891 = torch.constant.int -1
%24232 = torch.prim.ListConstruct %int-1_21891 : (!torch.int) -> !torch.list<int>
%true_21892 = torch.constant.bool true
%none_21893 = torch.constant.none
%24233 = torch.aten.mean.dim %24224, %24232, %true_21892, %none_21893 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21894 = torch.constant.int -1
%24234 = torch.prim.ListConstruct %int-1_21894 : (!torch.int) -> !torch.list<int>
%true_21895 = torch.constant.bool true
%none_21896 = torch.constant.none
%24235 = torch.aten.mean.dim %24225, %24234, %true_21895, %none_21896 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21897 = torch.constant.int -1
%24236 = torch.prim.ListConstruct %int-1_21897 : (!torch.int) -> !torch.list<int>
%true_21898 = torch.constant.bool true
%none_21899 = torch.constant.none
%24237 = torch.aten.mean.dim %24226, %24236, %true_21898, %none_21899 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21900 = torch.constant.int -1
%24238 = torch.prim.ListConstruct %int-1_21900 : (!torch.int) -> !torch.list<int>
%true_21901 = torch.constant.bool true
%none_21902 = torch.constant.none
%24239 = torch.aten.mean.dim %24227, %24238, %true_21901, %none_21902 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21903 = torch.constant.int -1
%24240 = torch.prim.ListConstruct %int-1_21903 : (!torch.int) -> !torch.list<int>
%true_21904 = torch.constant.bool true
%none_21905 = torch.constant.none
%24241 = torch.aten.mean.dim %24228, %24240, %true_21904, %none_21905 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21906 = torch.constant.int -1
%24242 = torch.prim.ListConstruct %int-1_21906 : (!torch.int) -> !torch.list<int>
%true_21907 = torch.constant.bool true
%none_21908 = torch.constant.none
%24243 = torch.aten.mean.dim %24229, %24242, %true_21907, %none_21908 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21909 = torch.constant.int -1
%24244 = torch.prim.ListConstruct %int-1_21909 : (!torch.int) -> !torch.list<int>
%true_21910 = torch.constant.bool true
%none_21911 = torch.constant.none
%24245 = torch.aten.mean.dim %24230, %24244, %true_21910, %none_21911 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_21912 = torch.constant.int -1
%24246 = torch.prim.ListConstruct %int-1_21912 : (!torch.int) -> !torch.list<int>
%true_21913 = torch.constant.bool true
%none_21914 = torch.constant.none
%24247 = torch.aten.mean.dim %24231, %24246, %true_21913, %none_21914 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21915 = torch.constant.float 9.9999997473787516E-6
%int1_21916 = torch.constant.int 1
%24248 = torch.aten.add.Scalar %24233, %float9.999990e-06_21915, %int1_21916 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21917 = torch.constant.float 9.9999997473787516E-6
%int1_21918 = torch.constant.int 1
%24249 = torch.aten.add.Scalar %24235, %float9.999990e-06_21917, %int1_21918 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21919 = torch.constant.float 9.9999997473787516E-6
%int1_21920 = torch.constant.int 1
%24250 = torch.aten.add.Scalar %24237, %float9.999990e-06_21919, %int1_21920 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21921 = torch.constant.float 9.9999997473787516E-6
%int1_21922 = torch.constant.int 1
%24251 = torch.aten.add.Scalar %24239, %float9.999990e-06_21921, %int1_21922 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21923 = torch.constant.float 9.9999997473787516E-6
%int1_21924 = torch.constant.int 1
%24252 = torch.aten.add.Scalar %24241, %float9.999990e-06_21923, %int1_21924 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21925 = torch.constant.float 9.9999997473787516E-6
%int1_21926 = torch.constant.int 1
%24253 = torch.aten.add.Scalar %24243, %float9.999990e-06_21925, %int1_21926 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21927 = torch.constant.float 9.9999997473787516E-6
%int1_21928 = torch.constant.int 1
%24254 = torch.aten.add.Scalar %24245, %float9.999990e-06_21927, %int1_21928 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_21929 = torch.constant.float 9.9999997473787516E-6
%int1_21930 = torch.constant.int 1
%24255 = torch.aten.add.Scalar %24247, %float9.999990e-06_21929, %int1_21930 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24256 = torch.aten.rsqrt %24248 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24257 = torch.aten.rsqrt %24249 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24258 = torch.aten.rsqrt %24250 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24259 = torch.aten.rsqrt %24251 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24260 = torch.aten.rsqrt %24252 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24261 = torch.aten.rsqrt %24253 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24262 = torch.aten.rsqrt %24254 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24263 = torch.aten.rsqrt %24255 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24263, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24264 = torch.aten.mul.Tensor %24216, %24256 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24265 = torch.aten.mul.Tensor %24217, %24257 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24266 = torch.aten.mul.Tensor %24218, %24258 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24267 = torch.aten.mul.Tensor %24219, %24259 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24268 = torch.aten.mul.Tensor %24220, %24260 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24269 = torch.aten.mul.Tensor %24221, %24261 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24270 = torch.aten.mul.Tensor %24222, %24262 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24271 = torch.aten.mul.Tensor %24223, %24263 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24272 = torch.aten.mul.Tensor %840, %24264 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24273 = torch.aten.mul.Tensor %841, %24265 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24274 = torch.aten.mul.Tensor %842, %24266 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24275 = torch.aten.mul.Tensor %843, %24267 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24276 = torch.aten.mul.Tensor %844, %24268 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24277 = torch.aten.mul.Tensor %845, %24269 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24278 = torch.aten.mul.Tensor %846, %24270 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24279 = torch.aten.mul.Tensor %847, %24271 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_21931 = torch.constant.int 5
%24280 = torch.prims.convert_element_type %24272, %int5_21931 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21932 = torch.constant.int 5
%24281 = torch.prims.convert_element_type %24273, %int5_21932 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21933 = torch.constant.int 5
%24282 = torch.prims.convert_element_type %24274, %int5_21933 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21934 = torch.constant.int 5
%24283 = torch.prims.convert_element_type %24275, %int5_21934 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21935 = torch.constant.int 5
%24284 = torch.prims.convert_element_type %24276, %int5_21935 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21936 = torch.constant.int 5
%24285 = torch.prims.convert_element_type %24277, %int5_21936 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21937 = torch.constant.int 5
%24286 = torch.prims.convert_element_type %24278, %int5_21937 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_21938 = torch.constant.int 5
%24287 = torch.prims.convert_element_type %24279, %int5_21938 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_21939 = torch.constant.int 1
%int0_21940 = torch.constant.int 0
%24288 = torch.prim.ListConstruct %int1_21939, %int0_21940 : (!torch.int, !torch.int) -> !torch.list<int>
%24289 = torch.aten.permute %848, %24288 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21941 = torch.constant.int 1
%int0_21942 = torch.constant.int 0
%24290 = torch.prim.ListConstruct %int1_21941, %int0_21942 : (!torch.int, !torch.int) -> !torch.list<int>
%24291 = torch.aten.permute %849, %24290 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21943 = torch.constant.int 1
%int0_21944 = torch.constant.int 0
%24292 = torch.prim.ListConstruct %int1_21943, %int0_21944 : (!torch.int, !torch.int) -> !torch.list<int>
%24293 = torch.aten.permute %850, %24292 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21945 = torch.constant.int 1
%int0_21946 = torch.constant.int 0
%24294 = torch.prim.ListConstruct %int1_21945, %int0_21946 : (!torch.int, !torch.int) -> !torch.list<int>
%24295 = torch.aten.permute %851, %24294 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21947 = torch.constant.int 1
%int0_21948 = torch.constant.int 0
%24296 = torch.prim.ListConstruct %int1_21947, %int0_21948 : (!torch.int, !torch.int) -> !torch.list<int>
%24297 = torch.aten.permute %852, %24296 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21949 = torch.constant.int 1
%int0_21950 = torch.constant.int 0
%24298 = torch.prim.ListConstruct %int1_21949, %int0_21950 : (!torch.int, !torch.int) -> !torch.list<int>
%24299 = torch.aten.permute %853, %24298 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21951 = torch.constant.int 1
%int0_21952 = torch.constant.int 0
%24300 = torch.prim.ListConstruct %int1_21951, %int0_21952 : (!torch.int, !torch.int) -> !torch.list<int>
%24301 = torch.aten.permute %854, %24300 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21953 = torch.constant.int 1
%int0_21954 = torch.constant.int 0
%24302 = torch.prim.ListConstruct %int1_21953, %int0_21954 : (!torch.int, !torch.int) -> !torch.list<int>
%24303 = torch.aten.permute %855, %24302 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_21955 = torch.constant.int 4
%24304 = torch.aten.mul.int %int4_21955, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21956 = torch.constant.int 4096
%24305 = torch.prim.ListConstruct %24304, %int4096_21956 : (!torch.int, !torch.int) -> !torch.list<int>
%24306 = torch.aten.view %24280, %24305 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24306, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24307 = torch.aten.mm %24306, %24289 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24307, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21957 = torch.constant.int 4
%int1792_21958 = torch.constant.int 1792
%24308 = torch.prim.ListConstruct %int4_21957, %2482, %int1792_21958 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24309 = torch.aten.view %24307, %24308 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21959 = torch.constant.int 4
%24310 = torch.aten.mul.int %int4_21959, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21960 = torch.constant.int 4096
%24311 = torch.prim.ListConstruct %24310, %int4096_21960 : (!torch.int, !torch.int) -> !torch.list<int>
%24312 = torch.aten.view %24281, %24311 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24312, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24313 = torch.aten.mm %24312, %24291 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24313, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21961 = torch.constant.int 4
%int1792_21962 = torch.constant.int 1792
%24314 = torch.prim.ListConstruct %int4_21961, %2482, %int1792_21962 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24315 = torch.aten.view %24313, %24314 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21963 = torch.constant.int 4
%24316 = torch.aten.mul.int %int4_21963, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21964 = torch.constant.int 4096
%24317 = torch.prim.ListConstruct %24316, %int4096_21964 : (!torch.int, !torch.int) -> !torch.list<int>
%24318 = torch.aten.view %24282, %24317 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24318, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24319 = torch.aten.mm %24318, %24293 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24319, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21965 = torch.constant.int 4
%int1792_21966 = torch.constant.int 1792
%24320 = torch.prim.ListConstruct %int4_21965, %2482, %int1792_21966 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24321 = torch.aten.view %24319, %24320 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21967 = torch.constant.int 4
%24322 = torch.aten.mul.int %int4_21967, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21968 = torch.constant.int 4096
%24323 = torch.prim.ListConstruct %24322, %int4096_21968 : (!torch.int, !torch.int) -> !torch.list<int>
%24324 = torch.aten.view %24283, %24323 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24324, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24325 = torch.aten.mm %24324, %24295 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24325, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21969 = torch.constant.int 4
%int1792_21970 = torch.constant.int 1792
%24326 = torch.prim.ListConstruct %int4_21969, %2482, %int1792_21970 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24327 = torch.aten.view %24325, %24326 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21971 = torch.constant.int 4
%24328 = torch.aten.mul.int %int4_21971, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21972 = torch.constant.int 4096
%24329 = torch.prim.ListConstruct %24328, %int4096_21972 : (!torch.int, !torch.int) -> !torch.list<int>
%24330 = torch.aten.view %24284, %24329 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24330, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24331 = torch.aten.mm %24330, %24297 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24331, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21973 = torch.constant.int 4
%int1792_21974 = torch.constant.int 1792
%24332 = torch.prim.ListConstruct %int4_21973, %2482, %int1792_21974 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24333 = torch.aten.view %24331, %24332 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21975 = torch.constant.int 4
%24334 = torch.aten.mul.int %int4_21975, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21976 = torch.constant.int 4096
%24335 = torch.prim.ListConstruct %24334, %int4096_21976 : (!torch.int, !torch.int) -> !torch.list<int>
%24336 = torch.aten.view %24285, %24335 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24336, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24337 = torch.aten.mm %24336, %24299 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24337, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21977 = torch.constant.int 4
%int1792_21978 = torch.constant.int 1792
%24338 = torch.prim.ListConstruct %int4_21977, %2482, %int1792_21978 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24339 = torch.aten.view %24337, %24338 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21979 = torch.constant.int 4
%24340 = torch.aten.mul.int %int4_21979, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21980 = torch.constant.int 4096
%24341 = torch.prim.ListConstruct %24340, %int4096_21980 : (!torch.int, !torch.int) -> !torch.list<int>
%24342 = torch.aten.view %24286, %24341 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24342, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24343 = torch.aten.mm %24342, %24301 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24343, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21981 = torch.constant.int 4
%int1792_21982 = torch.constant.int 1792
%24344 = torch.prim.ListConstruct %int4_21981, %2482, %int1792_21982 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24345 = torch.aten.view %24343, %24344 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_21983 = torch.constant.int 4
%24346 = torch.aten.mul.int %int4_21983, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_21984 = torch.constant.int 4096
%24347 = torch.prim.ListConstruct %24346, %int4096_21984 : (!torch.int, !torch.int) -> !torch.list<int>
%24348 = torch.aten.view %24287, %24347 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24348, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24349 = torch.aten.mm %24348, %24303 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24349, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_21985 = torch.constant.int 4
%int1792_21986 = torch.constant.int 1792
%24350 = torch.prim.ListConstruct %int4_21985, %2482, %int1792_21986 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24351 = torch.aten.view %24349, %24350 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24352 = torch.aten.silu %24309 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24353 = torch.aten.silu %24315 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24354 = torch.aten.silu %24321 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24355 = torch.aten.silu %24327 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24356 = torch.aten.silu %24333 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24357 = torch.aten.silu %24339 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24358 = torch.aten.silu %24345 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24359 = torch.aten.silu %24351 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_21987 = torch.constant.int 1
%int0_21988 = torch.constant.int 0
%24360 = torch.prim.ListConstruct %int1_21987, %int0_21988 : (!torch.int, !torch.int) -> !torch.list<int>
%24361 = torch.aten.permute %856, %24360 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21989 = torch.constant.int 1
%int0_21990 = torch.constant.int 0
%24362 = torch.prim.ListConstruct %int1_21989, %int0_21990 : (!torch.int, !torch.int) -> !torch.list<int>
%24363 = torch.aten.permute %857, %24362 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21991 = torch.constant.int 1
%int0_21992 = torch.constant.int 0
%24364 = torch.prim.ListConstruct %int1_21991, %int0_21992 : (!torch.int, !torch.int) -> !torch.list<int>
%24365 = torch.aten.permute %858, %24364 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21993 = torch.constant.int 1
%int0_21994 = torch.constant.int 0
%24366 = torch.prim.ListConstruct %int1_21993, %int0_21994 : (!torch.int, !torch.int) -> !torch.list<int>
%24367 = torch.aten.permute %859, %24366 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21995 = torch.constant.int 1
%int0_21996 = torch.constant.int 0
%24368 = torch.prim.ListConstruct %int1_21995, %int0_21996 : (!torch.int, !torch.int) -> !torch.list<int>
%24369 = torch.aten.permute %860, %24368 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21997 = torch.constant.int 1
%int0_21998 = torch.constant.int 0
%24370 = torch.prim.ListConstruct %int1_21997, %int0_21998 : (!torch.int, !torch.int) -> !torch.list<int>
%24371 = torch.aten.permute %861, %24370 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_21999 = torch.constant.int 1
%int0_22000 = torch.constant.int 0
%24372 = torch.prim.ListConstruct %int1_21999, %int0_22000 : (!torch.int, !torch.int) -> !torch.list<int>
%24373 = torch.aten.permute %862, %24372 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_22001 = torch.constant.int 1
%int0_22002 = torch.constant.int 0
%24374 = torch.prim.ListConstruct %int1_22001, %int0_22002 : (!torch.int, !torch.int) -> !torch.list<int>
%24375 = torch.aten.permute %863, %24374 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_22003 = torch.constant.int 4
%24376 = torch.aten.mul.int %int4_22003, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22004 = torch.constant.int 4096
%24377 = torch.prim.ListConstruct %24376, %int4096_22004 : (!torch.int, !torch.int) -> !torch.list<int>
%24378 = torch.aten.view %24280, %24377 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24378, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24379 = torch.aten.mm %24378, %24361 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24379, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22005 = torch.constant.int 4
%int1792_22006 = torch.constant.int 1792
%24380 = torch.prim.ListConstruct %int4_22005, %2482, %int1792_22006 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24381 = torch.aten.view %24379, %24380 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22007 = torch.constant.int 4
%24382 = torch.aten.mul.int %int4_22007, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22008 = torch.constant.int 4096
%24383 = torch.prim.ListConstruct %24382, %int4096_22008 : (!torch.int, !torch.int) -> !torch.list<int>
%24384 = torch.aten.view %24281, %24383 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24384, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24385 = torch.aten.mm %24384, %24363 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24385, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22009 = torch.constant.int 4
%int1792_22010 = torch.constant.int 1792
%24386 = torch.prim.ListConstruct %int4_22009, %2482, %int1792_22010 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24387 = torch.aten.view %24385, %24386 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22011 = torch.constant.int 4
%24388 = torch.aten.mul.int %int4_22011, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22012 = torch.constant.int 4096
%24389 = torch.prim.ListConstruct %24388, %int4096_22012 : (!torch.int, !torch.int) -> !torch.list<int>
%24390 = torch.aten.view %24282, %24389 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24390, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24391 = torch.aten.mm %24390, %24365 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24391, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22013 = torch.constant.int 4
%int1792_22014 = torch.constant.int 1792
%24392 = torch.prim.ListConstruct %int4_22013, %2482, %int1792_22014 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24393 = torch.aten.view %24391, %24392 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22015 = torch.constant.int 4
%24394 = torch.aten.mul.int %int4_22015, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22016 = torch.constant.int 4096
%24395 = torch.prim.ListConstruct %24394, %int4096_22016 : (!torch.int, !torch.int) -> !torch.list<int>
%24396 = torch.aten.view %24283, %24395 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24396, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24397 = torch.aten.mm %24396, %24367 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24397, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22017 = torch.constant.int 4
%int1792_22018 = torch.constant.int 1792
%24398 = torch.prim.ListConstruct %int4_22017, %2482, %int1792_22018 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24399 = torch.aten.view %24397, %24398 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22019 = torch.constant.int 4
%24400 = torch.aten.mul.int %int4_22019, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22020 = torch.constant.int 4096
%24401 = torch.prim.ListConstruct %24400, %int4096_22020 : (!torch.int, !torch.int) -> !torch.list<int>
%24402 = torch.aten.view %24284, %24401 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24402, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24403 = torch.aten.mm %24402, %24369 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24403, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22021 = torch.constant.int 4
%int1792_22022 = torch.constant.int 1792
%24404 = torch.prim.ListConstruct %int4_22021, %2482, %int1792_22022 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24405 = torch.aten.view %24403, %24404 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22023 = torch.constant.int 4
%24406 = torch.aten.mul.int %int4_22023, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22024 = torch.constant.int 4096
%24407 = torch.prim.ListConstruct %24406, %int4096_22024 : (!torch.int, !torch.int) -> !torch.list<int>
%24408 = torch.aten.view %24285, %24407 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24408, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24409 = torch.aten.mm %24408, %24371 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24409, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22025 = torch.constant.int 4
%int1792_22026 = torch.constant.int 1792
%24410 = torch.prim.ListConstruct %int4_22025, %2482, %int1792_22026 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24411 = torch.aten.view %24409, %24410 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22027 = torch.constant.int 4
%24412 = torch.aten.mul.int %int4_22027, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22028 = torch.constant.int 4096
%24413 = torch.prim.ListConstruct %24412, %int4096_22028 : (!torch.int, !torch.int) -> !torch.list<int>
%24414 = torch.aten.view %24286, %24413 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24414, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24415 = torch.aten.mm %24414, %24373 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24415, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22029 = torch.constant.int 4
%int1792_22030 = torch.constant.int 1792
%24416 = torch.prim.ListConstruct %int4_22029, %2482, %int1792_22030 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24417 = torch.aten.view %24415, %24416 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_22031 = torch.constant.int 4
%24418 = torch.aten.mul.int %int4_22031, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22032 = torch.constant.int 4096
%24419 = torch.prim.ListConstruct %24418, %int4096_22032 : (!torch.int, !torch.int) -> !torch.list<int>
%24420 = torch.aten.view %24287, %24419 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24420, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24421 = torch.aten.mm %24420, %24375 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24421, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_22033 = torch.constant.int 4
%int1792_22034 = torch.constant.int 1792
%24422 = torch.prim.ListConstruct %int4_22033, %2482, %int1792_22034 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24423 = torch.aten.view %24421, %24422 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24424 = torch.aten.mul.Tensor %24352, %24381 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24425 = torch.aten.mul.Tensor %24353, %24387 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24426 = torch.aten.mul.Tensor %24354, %24393 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24427 = torch.aten.mul.Tensor %24355, %24399 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24428 = torch.aten.mul.Tensor %24356, %24405 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24429 = torch.aten.mul.Tensor %24357, %24411 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24430 = torch.aten.mul.Tensor %24358, %24417 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%24431 = torch.aten.mul.Tensor %24359, %24423 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %24431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_22035 = torch.constant.int 1
%int0_22036 = torch.constant.int 0
%24432 = torch.prim.ListConstruct %int1_22035, %int0_22036 : (!torch.int, !torch.int) -> !torch.list<int>
%24433 = torch.aten.permute %864, %24432 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22037 = torch.constant.int 1
%int0_22038 = torch.constant.int 0
%24434 = torch.prim.ListConstruct %int1_22037, %int0_22038 : (!torch.int, !torch.int) -> !torch.list<int>
%24435 = torch.aten.permute %865, %24434 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22039 = torch.constant.int 1
%int0_22040 = torch.constant.int 0
%24436 = torch.prim.ListConstruct %int1_22039, %int0_22040 : (!torch.int, !torch.int) -> !torch.list<int>
%24437 = torch.aten.permute %866, %24436 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22041 = torch.constant.int 1
%int0_22042 = torch.constant.int 0
%24438 = torch.prim.ListConstruct %int1_22041, %int0_22042 : (!torch.int, !torch.int) -> !torch.list<int>
%24439 = torch.aten.permute %867, %24438 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22043 = torch.constant.int 1
%int0_22044 = torch.constant.int 0
%24440 = torch.prim.ListConstruct %int1_22043, %int0_22044 : (!torch.int, !torch.int) -> !torch.list<int>
%24441 = torch.aten.permute %868, %24440 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22045 = torch.constant.int 1
%int0_22046 = torch.constant.int 0
%24442 = torch.prim.ListConstruct %int1_22045, %int0_22046 : (!torch.int, !torch.int) -> !torch.list<int>
%24443 = torch.aten.permute %869, %24442 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22047 = torch.constant.int 1
%int0_22048 = torch.constant.int 0
%24444 = torch.prim.ListConstruct %int1_22047, %int0_22048 : (!torch.int, !torch.int) -> !torch.list<int>
%24445 = torch.aten.permute %870, %24444 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22049 = torch.constant.int 1
%int0_22050 = torch.constant.int 0
%24446 = torch.prim.ListConstruct %int1_22049, %int0_22050 : (!torch.int, !torch.int) -> !torch.list<int>
%24447 = torch.aten.permute %871, %24446 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_22051 = torch.constant.int 1
%24448 = torch.aten.size.int %24309, %int1_22051 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22052 = torch.constant.int 4
%24449 = torch.aten.mul.int %int4_22052, %24448 : !torch.int, !torch.int -> !torch.int
%int1792_22053 = torch.constant.int 1792
%24450 = torch.prim.ListConstruct %24449, %int1792_22053 : (!torch.int, !torch.int) -> !torch.list<int>
%24451 = torch.aten.view %24424, %24450 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24451, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24452 = torch.aten.mm %24451, %24433 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24452, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22054 = torch.constant.int 4
%int4096_22055 = torch.constant.int 4096
%24453 = torch.prim.ListConstruct %int4_22054, %24448, %int4096_22055 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24454 = torch.aten.view %24452, %24453 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22056 = torch.constant.int 1
%24455 = torch.aten.size.int %24315, %int1_22056 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22057 = torch.constant.int 4
%24456 = torch.aten.mul.int %int4_22057, %24455 : !torch.int, !torch.int -> !torch.int
%int1792_22058 = torch.constant.int 1792
%24457 = torch.prim.ListConstruct %24456, %int1792_22058 : (!torch.int, !torch.int) -> !torch.list<int>
%24458 = torch.aten.view %24425, %24457 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24458, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24459 = torch.aten.mm %24458, %24435 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24459, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22059 = torch.constant.int 4
%int4096_22060 = torch.constant.int 4096
%24460 = torch.prim.ListConstruct %int4_22059, %24455, %int4096_22060 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24461 = torch.aten.view %24459, %24460 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22061 = torch.constant.int 1
%24462 = torch.aten.size.int %24321, %int1_22061 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22062 = torch.constant.int 4
%24463 = torch.aten.mul.int %int4_22062, %24462 : !torch.int, !torch.int -> !torch.int
%int1792_22063 = torch.constant.int 1792
%24464 = torch.prim.ListConstruct %24463, %int1792_22063 : (!torch.int, !torch.int) -> !torch.list<int>
%24465 = torch.aten.view %24426, %24464 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24465, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24466 = torch.aten.mm %24465, %24437 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24466, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22064 = torch.constant.int 4
%int4096_22065 = torch.constant.int 4096
%24467 = torch.prim.ListConstruct %int4_22064, %24462, %int4096_22065 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24468 = torch.aten.view %24466, %24467 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22066 = torch.constant.int 1
%24469 = torch.aten.size.int %24327, %int1_22066 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22067 = torch.constant.int 4
%24470 = torch.aten.mul.int %int4_22067, %24469 : !torch.int, !torch.int -> !torch.int
%int1792_22068 = torch.constant.int 1792
%24471 = torch.prim.ListConstruct %24470, %int1792_22068 : (!torch.int, !torch.int) -> !torch.list<int>
%24472 = torch.aten.view %24427, %24471 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24472, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24473 = torch.aten.mm %24472, %24439 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24473, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22069 = torch.constant.int 4
%int4096_22070 = torch.constant.int 4096
%24474 = torch.prim.ListConstruct %int4_22069, %24469, %int4096_22070 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24475 = torch.aten.view %24473, %24474 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22071 = torch.constant.int 1
%24476 = torch.aten.size.int %24333, %int1_22071 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22072 = torch.constant.int 4
%24477 = torch.aten.mul.int %int4_22072, %24476 : !torch.int, !torch.int -> !torch.int
%int1792_22073 = torch.constant.int 1792
%24478 = torch.prim.ListConstruct %24477, %int1792_22073 : (!torch.int, !torch.int) -> !torch.list<int>
%24479 = torch.aten.view %24428, %24478 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24479, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24480 = torch.aten.mm %24479, %24441 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24480, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22074 = torch.constant.int 4
%int4096_22075 = torch.constant.int 4096
%24481 = torch.prim.ListConstruct %int4_22074, %24476, %int4096_22075 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24482 = torch.aten.view %24480, %24481 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22076 = torch.constant.int 1
%24483 = torch.aten.size.int %24339, %int1_22076 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22077 = torch.constant.int 4
%24484 = torch.aten.mul.int %int4_22077, %24483 : !torch.int, !torch.int -> !torch.int
%int1792_22078 = torch.constant.int 1792
%24485 = torch.prim.ListConstruct %24484, %int1792_22078 : (!torch.int, !torch.int) -> !torch.list<int>
%24486 = torch.aten.view %24429, %24485 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24486, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24487 = torch.aten.mm %24486, %24443 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24487, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22079 = torch.constant.int 4
%int4096_22080 = torch.constant.int 4096
%24488 = torch.prim.ListConstruct %int4_22079, %24483, %int4096_22080 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24489 = torch.aten.view %24487, %24488 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22081 = torch.constant.int 1
%24490 = torch.aten.size.int %24345, %int1_22081 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22082 = torch.constant.int 4
%24491 = torch.aten.mul.int %int4_22082, %24490 : !torch.int, !torch.int -> !torch.int
%int1792_22083 = torch.constant.int 1792
%24492 = torch.prim.ListConstruct %24491, %int1792_22083 : (!torch.int, !torch.int) -> !torch.list<int>
%24493 = torch.aten.view %24430, %24492 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24493, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24494 = torch.aten.mm %24493, %24445 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24494, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22084 = torch.constant.int 4
%int4096_22085 = torch.constant.int 4096
%24495 = torch.prim.ListConstruct %int4_22084, %24490, %int4096_22085 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24496 = torch.aten.view %24494, %24495 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22086 = torch.constant.int 1
%24497 = torch.aten.size.int %24351, %int1_22086 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_22087 = torch.constant.int 4
%24498 = torch.aten.mul.int %int4_22087, %24497 : !torch.int, !torch.int -> !torch.int
%int1792_22088 = torch.constant.int 1792
%24499 = torch.prim.ListConstruct %24498, %int1792_22088 : (!torch.int, !torch.int) -> !torch.list<int>
%24500 = torch.aten.view %24431, %24499 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %24500, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%24501 = torch.aten.mm %24500, %24447 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24501, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_22089 = torch.constant.int 4
%int4096_22090 = torch.constant.int 4096
%24502 = torch.prim.ListConstruct %int4_22089, %24497, %int4096_22090 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24503 = torch.aten.view %24501, %24502 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24504 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22091 = arith.constant 1 : index
%dim_22092 = tensor.dim %24504, %c1_22091 : tensor<4x?x4096xf16>
%24505 = flow.tensor.transfer %24504 : tensor<4x?x4096xf16>{%dim_22092} to #hal.device.promise<@__device_0>
%24506 = torch_c.from_builtin_tensor %24505 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24507 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22093 = arith.constant 1 : index
%dim_22094 = tensor.dim %24507, %c1_22093 : tensor<4x?x4096xf16>
%24508 = flow.tensor.transfer %24507 : tensor<4x?x4096xf16>{%dim_22094} to #hal.device.promise<@__device_0>
%24509 = torch_c.from_builtin_tensor %24508 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24510 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22095 = arith.constant 1 : index
%dim_22096 = tensor.dim %24510, %c1_22095 : tensor<4x?x4096xf16>
%24511 = flow.tensor.transfer %24510 : tensor<4x?x4096xf16>{%dim_22096} to #hal.device.promise<@__device_0>
%24512 = torch_c.from_builtin_tensor %24511 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24513 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22097 = arith.constant 1 : index
%dim_22098 = tensor.dim %24513, %c1_22097 : tensor<4x?x4096xf16>
%24514 = flow.tensor.transfer %24513 : tensor<4x?x4096xf16>{%dim_22098} to #hal.device.promise<@__device_0>
%24515 = torch_c.from_builtin_tensor %24514 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24516 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22099 = arith.constant 1 : index
%dim_22100 = tensor.dim %24516, %c1_22099 : tensor<4x?x4096xf16>
%24517 = flow.tensor.transfer %24516 : tensor<4x?x4096xf16>{%dim_22100} to #hal.device.promise<@__device_0>
%24518 = torch_c.from_builtin_tensor %24517 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24519 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22101 = arith.constant 1 : index
%dim_22102 = tensor.dim %24519, %c1_22101 : tensor<4x?x4096xf16>
%24520 = flow.tensor.transfer %24519 : tensor<4x?x4096xf16>{%dim_22102} to #hal.device.promise<@__device_0>
%24521 = torch_c.from_builtin_tensor %24520 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24522 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22103 = arith.constant 1 : index
%dim_22104 = tensor.dim %24522, %c1_22103 : tensor<4x?x4096xf16>
%24523 = flow.tensor.transfer %24522 : tensor<4x?x4096xf16>{%dim_22104} to #hal.device.promise<@__device_0>
%24524 = torch_c.from_builtin_tensor %24523 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22105 = torch.constant.int 1
%24525 = torch.aten.add.Tensor %24454, %24506, %int1_22105 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22106 = torch.constant.int 1
%24526 = torch.aten.add.Tensor %24525, %24509, %int1_22106 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22107 = torch.constant.int 1
%24527 = torch.aten.add.Tensor %24526, %24512, %int1_22107 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22108 = torch.constant.int 1
%24528 = torch.aten.add.Tensor %24527, %24515, %int1_22108 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22109 = torch.constant.int 1
%24529 = torch.aten.add.Tensor %24528, %24518, %int1_22109 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22110 = torch.constant.int 1
%24530 = torch.aten.add.Tensor %24529, %24521, %int1_22110 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22111 = torch.constant.int 1
%24531 = torch.aten.add.Tensor %24530, %24524, %int1_22111 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24532 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22112 = arith.constant 1 : index
%dim_22113 = tensor.dim %24532, %c1_22112 : tensor<4x?x4096xf16>
%24533 = flow.tensor.transfer %24532 : tensor<4x?x4096xf16>{%dim_22113} to #hal.device.promise<@__device_1>
%24534 = torch_c.from_builtin_tensor %24533 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24535 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22114 = arith.constant 1 : index
%dim_22115 = tensor.dim %24535, %c1_22114 : tensor<4x?x4096xf16>
%24536 = flow.tensor.transfer %24535 : tensor<4x?x4096xf16>{%dim_22115} to #hal.device.promise<@__device_1>
%24537 = torch_c.from_builtin_tensor %24536 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24538 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22116 = arith.constant 1 : index
%dim_22117 = tensor.dim %24538, %c1_22116 : tensor<4x?x4096xf16>
%24539 = flow.tensor.transfer %24538 : tensor<4x?x4096xf16>{%dim_22117} to #hal.device.promise<@__device_1>
%24540 = torch_c.from_builtin_tensor %24539 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24541 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22118 = arith.constant 1 : index
%dim_22119 = tensor.dim %24541, %c1_22118 : tensor<4x?x4096xf16>
%24542 = flow.tensor.transfer %24541 : tensor<4x?x4096xf16>{%dim_22119} to #hal.device.promise<@__device_1>
%24543 = torch_c.from_builtin_tensor %24542 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24544 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22120 = arith.constant 1 : index
%dim_22121 = tensor.dim %24544, %c1_22120 : tensor<4x?x4096xf16>
%24545 = flow.tensor.transfer %24544 : tensor<4x?x4096xf16>{%dim_22121} to #hal.device.promise<@__device_1>
%24546 = torch_c.from_builtin_tensor %24545 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24547 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22122 = arith.constant 1 : index
%dim_22123 = tensor.dim %24547, %c1_22122 : tensor<4x?x4096xf16>
%24548 = flow.tensor.transfer %24547 : tensor<4x?x4096xf16>{%dim_22123} to #hal.device.promise<@__device_1>
%24549 = torch_c.from_builtin_tensor %24548 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24550 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22124 = arith.constant 1 : index
%dim_22125 = tensor.dim %24550, %c1_22124 : tensor<4x?x4096xf16>
%24551 = flow.tensor.transfer %24550 : tensor<4x?x4096xf16>{%dim_22125} to #hal.device.promise<@__device_1>
%24552 = torch_c.from_builtin_tensor %24551 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22126 = torch.constant.int 1
%24553 = torch.aten.add.Tensor %24534, %24461, %int1_22126 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22127 = torch.constant.int 1
%24554 = torch.aten.add.Tensor %24553, %24537, %int1_22127 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22128 = torch.constant.int 1
%24555 = torch.aten.add.Tensor %24554, %24540, %int1_22128 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22129 = torch.constant.int 1
%24556 = torch.aten.add.Tensor %24555, %24543, %int1_22129 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22130 = torch.constant.int 1
%24557 = torch.aten.add.Tensor %24556, %24546, %int1_22130 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22131 = torch.constant.int 1
%24558 = torch.aten.add.Tensor %24557, %24549, %int1_22131 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22132 = torch.constant.int 1
%24559 = torch.aten.add.Tensor %24558, %24552, %int1_22132 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24560 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22133 = arith.constant 1 : index
%dim_22134 = tensor.dim %24560, %c1_22133 : tensor<4x?x4096xf16>
%24561 = flow.tensor.transfer %24560 : tensor<4x?x4096xf16>{%dim_22134} to #hal.device.promise<@__device_2>
%24562 = torch_c.from_builtin_tensor %24561 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24563 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22135 = arith.constant 1 : index
%dim_22136 = tensor.dim %24563, %c1_22135 : tensor<4x?x4096xf16>
%24564 = flow.tensor.transfer %24563 : tensor<4x?x4096xf16>{%dim_22136} to #hal.device.promise<@__device_2>
%24565 = torch_c.from_builtin_tensor %24564 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24566 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22137 = arith.constant 1 : index
%dim_22138 = tensor.dim %24566, %c1_22137 : tensor<4x?x4096xf16>
%24567 = flow.tensor.transfer %24566 : tensor<4x?x4096xf16>{%dim_22138} to #hal.device.promise<@__device_2>
%24568 = torch_c.from_builtin_tensor %24567 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24569 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22139 = arith.constant 1 : index
%dim_22140 = tensor.dim %24569, %c1_22139 : tensor<4x?x4096xf16>
%24570 = flow.tensor.transfer %24569 : tensor<4x?x4096xf16>{%dim_22140} to #hal.device.promise<@__device_2>
%24571 = torch_c.from_builtin_tensor %24570 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24572 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22141 = arith.constant 1 : index
%dim_22142 = tensor.dim %24572, %c1_22141 : tensor<4x?x4096xf16>
%24573 = flow.tensor.transfer %24572 : tensor<4x?x4096xf16>{%dim_22142} to #hal.device.promise<@__device_2>
%24574 = torch_c.from_builtin_tensor %24573 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24575 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22143 = arith.constant 1 : index
%dim_22144 = tensor.dim %24575, %c1_22143 : tensor<4x?x4096xf16>
%24576 = flow.tensor.transfer %24575 : tensor<4x?x4096xf16>{%dim_22144} to #hal.device.promise<@__device_2>
%24577 = torch_c.from_builtin_tensor %24576 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24578 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22145 = arith.constant 1 : index
%dim_22146 = tensor.dim %24578, %c1_22145 : tensor<4x?x4096xf16>
%24579 = flow.tensor.transfer %24578 : tensor<4x?x4096xf16>{%dim_22146} to #hal.device.promise<@__device_2>
%24580 = torch_c.from_builtin_tensor %24579 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22147 = torch.constant.int 1
%24581 = torch.aten.add.Tensor %24562, %24565, %int1_22147 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22148 = torch.constant.int 1
%24582 = torch.aten.add.Tensor %24581, %24468, %int1_22148 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22149 = torch.constant.int 1
%24583 = torch.aten.add.Tensor %24582, %24568, %int1_22149 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22150 = torch.constant.int 1
%24584 = torch.aten.add.Tensor %24583, %24571, %int1_22150 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22151 = torch.constant.int 1
%24585 = torch.aten.add.Tensor %24584, %24574, %int1_22151 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22152 = torch.constant.int 1
%24586 = torch.aten.add.Tensor %24585, %24577, %int1_22152 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22153 = torch.constant.int 1
%24587 = torch.aten.add.Tensor %24586, %24580, %int1_22153 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24588 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22154 = arith.constant 1 : index
%dim_22155 = tensor.dim %24588, %c1_22154 : tensor<4x?x4096xf16>
%24589 = flow.tensor.transfer %24588 : tensor<4x?x4096xf16>{%dim_22155} to #hal.device.promise<@__device_3>
%24590 = torch_c.from_builtin_tensor %24589 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24591 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22156 = arith.constant 1 : index
%dim_22157 = tensor.dim %24591, %c1_22156 : tensor<4x?x4096xf16>
%24592 = flow.tensor.transfer %24591 : tensor<4x?x4096xf16>{%dim_22157} to #hal.device.promise<@__device_3>
%24593 = torch_c.from_builtin_tensor %24592 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24594 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22158 = arith.constant 1 : index
%dim_22159 = tensor.dim %24594, %c1_22158 : tensor<4x?x4096xf16>
%24595 = flow.tensor.transfer %24594 : tensor<4x?x4096xf16>{%dim_22159} to #hal.device.promise<@__device_3>
%24596 = torch_c.from_builtin_tensor %24595 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24597 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22160 = arith.constant 1 : index
%dim_22161 = tensor.dim %24597, %c1_22160 : tensor<4x?x4096xf16>
%24598 = flow.tensor.transfer %24597 : tensor<4x?x4096xf16>{%dim_22161} to #hal.device.promise<@__device_3>
%24599 = torch_c.from_builtin_tensor %24598 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24600 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22162 = arith.constant 1 : index
%dim_22163 = tensor.dim %24600, %c1_22162 : tensor<4x?x4096xf16>
%24601 = flow.tensor.transfer %24600 : tensor<4x?x4096xf16>{%dim_22163} to #hal.device.promise<@__device_3>
%24602 = torch_c.from_builtin_tensor %24601 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24603 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22164 = arith.constant 1 : index
%dim_22165 = tensor.dim %24603, %c1_22164 : tensor<4x?x4096xf16>
%24604 = flow.tensor.transfer %24603 : tensor<4x?x4096xf16>{%dim_22165} to #hal.device.promise<@__device_3>
%24605 = torch_c.from_builtin_tensor %24604 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24606 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22166 = arith.constant 1 : index
%dim_22167 = tensor.dim %24606, %c1_22166 : tensor<4x?x4096xf16>
%24607 = flow.tensor.transfer %24606 : tensor<4x?x4096xf16>{%dim_22167} to #hal.device.promise<@__device_3>
%24608 = torch_c.from_builtin_tensor %24607 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22168 = torch.constant.int 1
%24609 = torch.aten.add.Tensor %24590, %24593, %int1_22168 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22169 = torch.constant.int 1
%24610 = torch.aten.add.Tensor %24609, %24596, %int1_22169 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22170 = torch.constant.int 1
%24611 = torch.aten.add.Tensor %24610, %24475, %int1_22170 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22171 = torch.constant.int 1
%24612 = torch.aten.add.Tensor %24611, %24599, %int1_22171 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22172 = torch.constant.int 1
%24613 = torch.aten.add.Tensor %24612, %24602, %int1_22172 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22173 = torch.constant.int 1
%24614 = torch.aten.add.Tensor %24613, %24605, %int1_22173 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22174 = torch.constant.int 1
%24615 = torch.aten.add.Tensor %24614, %24608, %int1_22174 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24616 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22175 = arith.constant 1 : index
%dim_22176 = tensor.dim %24616, %c1_22175 : tensor<4x?x4096xf16>
%24617 = flow.tensor.transfer %24616 : tensor<4x?x4096xf16>{%dim_22176} to #hal.device.promise<@__device_4>
%24618 = torch_c.from_builtin_tensor %24617 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24619 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22177 = arith.constant 1 : index
%dim_22178 = tensor.dim %24619, %c1_22177 : tensor<4x?x4096xf16>
%24620 = flow.tensor.transfer %24619 : tensor<4x?x4096xf16>{%dim_22178} to #hal.device.promise<@__device_4>
%24621 = torch_c.from_builtin_tensor %24620 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24622 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22179 = arith.constant 1 : index
%dim_22180 = tensor.dim %24622, %c1_22179 : tensor<4x?x4096xf16>
%24623 = flow.tensor.transfer %24622 : tensor<4x?x4096xf16>{%dim_22180} to #hal.device.promise<@__device_4>
%24624 = torch_c.from_builtin_tensor %24623 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24625 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22181 = arith.constant 1 : index
%dim_22182 = tensor.dim %24625, %c1_22181 : tensor<4x?x4096xf16>
%24626 = flow.tensor.transfer %24625 : tensor<4x?x4096xf16>{%dim_22182} to #hal.device.promise<@__device_4>
%24627 = torch_c.from_builtin_tensor %24626 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24628 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22183 = arith.constant 1 : index
%dim_22184 = tensor.dim %24628, %c1_22183 : tensor<4x?x4096xf16>
%24629 = flow.tensor.transfer %24628 : tensor<4x?x4096xf16>{%dim_22184} to #hal.device.promise<@__device_4>
%24630 = torch_c.from_builtin_tensor %24629 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24631 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22185 = arith.constant 1 : index
%dim_22186 = tensor.dim %24631, %c1_22185 : tensor<4x?x4096xf16>
%24632 = flow.tensor.transfer %24631 : tensor<4x?x4096xf16>{%dim_22186} to #hal.device.promise<@__device_4>
%24633 = torch_c.from_builtin_tensor %24632 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24634 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22187 = arith.constant 1 : index
%dim_22188 = tensor.dim %24634, %c1_22187 : tensor<4x?x4096xf16>
%24635 = flow.tensor.transfer %24634 : tensor<4x?x4096xf16>{%dim_22188} to #hal.device.promise<@__device_4>
%24636 = torch_c.from_builtin_tensor %24635 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22189 = torch.constant.int 1
%24637 = torch.aten.add.Tensor %24618, %24621, %int1_22189 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22190 = torch.constant.int 1
%24638 = torch.aten.add.Tensor %24637, %24624, %int1_22190 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22191 = torch.constant.int 1
%24639 = torch.aten.add.Tensor %24638, %24627, %int1_22191 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22192 = torch.constant.int 1
%24640 = torch.aten.add.Tensor %24639, %24482, %int1_22192 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22193 = torch.constant.int 1
%24641 = torch.aten.add.Tensor %24640, %24630, %int1_22193 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22194 = torch.constant.int 1
%24642 = torch.aten.add.Tensor %24641, %24633, %int1_22194 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22195 = torch.constant.int 1
%24643 = torch.aten.add.Tensor %24642, %24636, %int1_22195 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24644 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22196 = arith.constant 1 : index
%dim_22197 = tensor.dim %24644, %c1_22196 : tensor<4x?x4096xf16>
%24645 = flow.tensor.transfer %24644 : tensor<4x?x4096xf16>{%dim_22197} to #hal.device.promise<@__device_5>
%24646 = torch_c.from_builtin_tensor %24645 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24647 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22198 = arith.constant 1 : index
%dim_22199 = tensor.dim %24647, %c1_22198 : tensor<4x?x4096xf16>
%24648 = flow.tensor.transfer %24647 : tensor<4x?x4096xf16>{%dim_22199} to #hal.device.promise<@__device_5>
%24649 = torch_c.from_builtin_tensor %24648 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24650 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22200 = arith.constant 1 : index
%dim_22201 = tensor.dim %24650, %c1_22200 : tensor<4x?x4096xf16>
%24651 = flow.tensor.transfer %24650 : tensor<4x?x4096xf16>{%dim_22201} to #hal.device.promise<@__device_5>
%24652 = torch_c.from_builtin_tensor %24651 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24653 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22202 = arith.constant 1 : index
%dim_22203 = tensor.dim %24653, %c1_22202 : tensor<4x?x4096xf16>
%24654 = flow.tensor.transfer %24653 : tensor<4x?x4096xf16>{%dim_22203} to #hal.device.promise<@__device_5>
%24655 = torch_c.from_builtin_tensor %24654 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24656 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22204 = arith.constant 1 : index
%dim_22205 = tensor.dim %24656, %c1_22204 : tensor<4x?x4096xf16>
%24657 = flow.tensor.transfer %24656 : tensor<4x?x4096xf16>{%dim_22205} to #hal.device.promise<@__device_5>
%24658 = torch_c.from_builtin_tensor %24657 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24659 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22206 = arith.constant 1 : index
%dim_22207 = tensor.dim %24659, %c1_22206 : tensor<4x?x4096xf16>
%24660 = flow.tensor.transfer %24659 : tensor<4x?x4096xf16>{%dim_22207} to #hal.device.promise<@__device_5>
%24661 = torch_c.from_builtin_tensor %24660 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24662 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22208 = arith.constant 1 : index
%dim_22209 = tensor.dim %24662, %c1_22208 : tensor<4x?x4096xf16>
%24663 = flow.tensor.transfer %24662 : tensor<4x?x4096xf16>{%dim_22209} to #hal.device.promise<@__device_5>
%24664 = torch_c.from_builtin_tensor %24663 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22210 = torch.constant.int 1
%24665 = torch.aten.add.Tensor %24646, %24649, %int1_22210 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22211 = torch.constant.int 1
%24666 = torch.aten.add.Tensor %24665, %24652, %int1_22211 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22212 = torch.constant.int 1
%24667 = torch.aten.add.Tensor %24666, %24655, %int1_22212 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22213 = torch.constant.int 1
%24668 = torch.aten.add.Tensor %24667, %24658, %int1_22213 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22214 = torch.constant.int 1
%24669 = torch.aten.add.Tensor %24668, %24489, %int1_22214 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22215 = torch.constant.int 1
%24670 = torch.aten.add.Tensor %24669, %24661, %int1_22215 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22216 = torch.constant.int 1
%24671 = torch.aten.add.Tensor %24670, %24664, %int1_22216 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24672 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22217 = arith.constant 1 : index
%dim_22218 = tensor.dim %24672, %c1_22217 : tensor<4x?x4096xf16>
%24673 = flow.tensor.transfer %24672 : tensor<4x?x4096xf16>{%dim_22218} to #hal.device.promise<@__device_6>
%24674 = torch_c.from_builtin_tensor %24673 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24675 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22219 = arith.constant 1 : index
%dim_22220 = tensor.dim %24675, %c1_22219 : tensor<4x?x4096xf16>
%24676 = flow.tensor.transfer %24675 : tensor<4x?x4096xf16>{%dim_22220} to #hal.device.promise<@__device_6>
%24677 = torch_c.from_builtin_tensor %24676 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24678 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22221 = arith.constant 1 : index
%dim_22222 = tensor.dim %24678, %c1_22221 : tensor<4x?x4096xf16>
%24679 = flow.tensor.transfer %24678 : tensor<4x?x4096xf16>{%dim_22222} to #hal.device.promise<@__device_6>
%24680 = torch_c.from_builtin_tensor %24679 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24681 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22223 = arith.constant 1 : index
%dim_22224 = tensor.dim %24681, %c1_22223 : tensor<4x?x4096xf16>
%24682 = flow.tensor.transfer %24681 : tensor<4x?x4096xf16>{%dim_22224} to #hal.device.promise<@__device_6>
%24683 = torch_c.from_builtin_tensor %24682 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24684 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22225 = arith.constant 1 : index
%dim_22226 = tensor.dim %24684, %c1_22225 : tensor<4x?x4096xf16>
%24685 = flow.tensor.transfer %24684 : tensor<4x?x4096xf16>{%dim_22226} to #hal.device.promise<@__device_6>
%24686 = torch_c.from_builtin_tensor %24685 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24687 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22227 = arith.constant 1 : index
%dim_22228 = tensor.dim %24687, %c1_22227 : tensor<4x?x4096xf16>
%24688 = flow.tensor.transfer %24687 : tensor<4x?x4096xf16>{%dim_22228} to #hal.device.promise<@__device_6>
%24689 = torch_c.from_builtin_tensor %24688 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24690 = torch_c.to_builtin_tensor %24503 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22229 = arith.constant 1 : index
%dim_22230 = tensor.dim %24690, %c1_22229 : tensor<4x?x4096xf16>
%24691 = flow.tensor.transfer %24690 : tensor<4x?x4096xf16>{%dim_22230} to #hal.device.promise<@__device_6>
%24692 = torch_c.from_builtin_tensor %24691 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22231 = torch.constant.int 1
%24693 = torch.aten.add.Tensor %24674, %24677, %int1_22231 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22232 = torch.constant.int 1
%24694 = torch.aten.add.Tensor %24693, %24680, %int1_22232 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22233 = torch.constant.int 1
%24695 = torch.aten.add.Tensor %24694, %24683, %int1_22233 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22234 = torch.constant.int 1
%24696 = torch.aten.add.Tensor %24695, %24686, %int1_22234 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22235 = torch.constant.int 1
%24697 = torch.aten.add.Tensor %24696, %24689, %int1_22235 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22236 = torch.constant.int 1
%24698 = torch.aten.add.Tensor %24697, %24496, %int1_22236 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22237 = torch.constant.int 1
%24699 = torch.aten.add.Tensor %24698, %24692, %int1_22237 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24700 = torch_c.to_builtin_tensor %24454 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22238 = arith.constant 1 : index
%dim_22239 = tensor.dim %24700, %c1_22238 : tensor<4x?x4096xf16>
%24701 = flow.tensor.transfer %24700 : tensor<4x?x4096xf16>{%dim_22239} to #hal.device.promise<@__device_7>
%24702 = torch_c.from_builtin_tensor %24701 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24703 = torch_c.to_builtin_tensor %24461 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22240 = arith.constant 1 : index
%dim_22241 = tensor.dim %24703, %c1_22240 : tensor<4x?x4096xf16>
%24704 = flow.tensor.transfer %24703 : tensor<4x?x4096xf16>{%dim_22241} to #hal.device.promise<@__device_7>
%24705 = torch_c.from_builtin_tensor %24704 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24706 = torch_c.to_builtin_tensor %24468 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22242 = arith.constant 1 : index
%dim_22243 = tensor.dim %24706, %c1_22242 : tensor<4x?x4096xf16>
%24707 = flow.tensor.transfer %24706 : tensor<4x?x4096xf16>{%dim_22243} to #hal.device.promise<@__device_7>
%24708 = torch_c.from_builtin_tensor %24707 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24709 = torch_c.to_builtin_tensor %24475 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22244 = arith.constant 1 : index
%dim_22245 = tensor.dim %24709, %c1_22244 : tensor<4x?x4096xf16>
%24710 = flow.tensor.transfer %24709 : tensor<4x?x4096xf16>{%dim_22245} to #hal.device.promise<@__device_7>
%24711 = torch_c.from_builtin_tensor %24710 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24712 = torch_c.to_builtin_tensor %24482 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22246 = arith.constant 1 : index
%dim_22247 = tensor.dim %24712, %c1_22246 : tensor<4x?x4096xf16>
%24713 = flow.tensor.transfer %24712 : tensor<4x?x4096xf16>{%dim_22247} to #hal.device.promise<@__device_7>
%24714 = torch_c.from_builtin_tensor %24713 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24715 = torch_c.to_builtin_tensor %24489 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22248 = arith.constant 1 : index
%dim_22249 = tensor.dim %24715, %c1_22248 : tensor<4x?x4096xf16>
%24716 = flow.tensor.transfer %24715 : tensor<4x?x4096xf16>{%dim_22249} to #hal.device.promise<@__device_7>
%24717 = torch_c.from_builtin_tensor %24716 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%24718 = torch_c.to_builtin_tensor %24496 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_22250 = arith.constant 1 : index
%dim_22251 = tensor.dim %24718, %c1_22250 : tensor<4x?x4096xf16>
%24719 = flow.tensor.transfer %24718 : tensor<4x?x4096xf16>{%dim_22251} to #hal.device.promise<@__device_7>
%24720 = torch_c.from_builtin_tensor %24719 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22252 = torch.constant.int 1
%24721 = torch.aten.add.Tensor %24702, %24705, %int1_22252 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22253 = torch.constant.int 1
%24722 = torch.aten.add.Tensor %24721, %24708, %int1_22253 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22254 = torch.constant.int 1
%24723 = torch.aten.add.Tensor %24722, %24711, %int1_22254 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22255 = torch.constant.int 1
%24724 = torch.aten.add.Tensor %24723, %24714, %int1_22255 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22256 = torch.constant.int 1
%24725 = torch.aten.add.Tensor %24724, %24717, %int1_22256 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22257 = torch.constant.int 1
%24726 = torch.aten.add.Tensor %24725, %24720, %int1_22257 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22258 = torch.constant.int 1
%24727 = torch.aten.add.Tensor %24726, %24503, %int1_22258 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22259 = torch.constant.int 1
%24728 = torch.aten.add.Tensor %24208, %24531, %int1_22259 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22260 = torch.constant.int 1
%24729 = torch.aten.add.Tensor %24209, %24559, %int1_22260 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22261 = torch.constant.int 1
%24730 = torch.aten.add.Tensor %24210, %24587, %int1_22261 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22262 = torch.constant.int 1
%24731 = torch.aten.add.Tensor %24211, %24615, %int1_22262 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22263 = torch.constant.int 1
%24732 = torch.aten.add.Tensor %24212, %24643, %int1_22263 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22264 = torch.constant.int 1
%24733 = torch.aten.add.Tensor %24213, %24671, %int1_22264 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22265 = torch.constant.int 1
%24734 = torch.aten.add.Tensor %24214, %24699, %int1_22265 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22266 = torch.constant.int 1
%24735 = torch.aten.add.Tensor %24215, %24727, %int1_22266 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_22267 = torch.constant.int 6
%24736 = torch.prims.convert_element_type %24728, %int6_22267 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22268 = torch.constant.int 6
%24737 = torch.prims.convert_element_type %24729, %int6_22268 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22269 = torch.constant.int 6
%24738 = torch.prims.convert_element_type %24730, %int6_22269 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22270 = torch.constant.int 6
%24739 = torch.prims.convert_element_type %24731, %int6_22270 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22271 = torch.constant.int 6
%24740 = torch.prims.convert_element_type %24732, %int6_22271 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22272 = torch.constant.int 6
%24741 = torch.prims.convert_element_type %24733, %int6_22272 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22273 = torch.constant.int 6
%24742 = torch.prims.convert_element_type %24734, %int6_22273 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_22274 = torch.constant.int 6
%24743 = torch.prims.convert_element_type %24735, %int6_22274 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22275 = torch.constant.int 2
%24744 = torch.aten.pow.Tensor_Scalar %24736, %int2_22275 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22276 = torch.constant.int 2
%24745 = torch.aten.pow.Tensor_Scalar %24737, %int2_22276 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22277 = torch.constant.int 2
%24746 = torch.aten.pow.Tensor_Scalar %24738, %int2_22277 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22278 = torch.constant.int 2
%24747 = torch.aten.pow.Tensor_Scalar %24739, %int2_22278 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22279 = torch.constant.int 2
%24748 = torch.aten.pow.Tensor_Scalar %24740, %int2_22279 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22280 = torch.constant.int 2
%24749 = torch.aten.pow.Tensor_Scalar %24741, %int2_22280 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22281 = torch.constant.int 2
%24750 = torch.aten.pow.Tensor_Scalar %24742, %int2_22281 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_22282 = torch.constant.int 2
%24751 = torch.aten.pow.Tensor_Scalar %24743, %int2_22282 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_22283 = torch.constant.int -1
%24752 = torch.prim.ListConstruct %int-1_22283 : (!torch.int) -> !torch.list<int>
%true_22284 = torch.constant.bool true
%none_22285 = torch.constant.none
%24753 = torch.aten.mean.dim %24744, %24752, %true_22284, %none_22285 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22286 = torch.constant.int -1
%24754 = torch.prim.ListConstruct %int-1_22286 : (!torch.int) -> !torch.list<int>
%true_22287 = torch.constant.bool true
%none_22288 = torch.constant.none
%24755 = torch.aten.mean.dim %24745, %24754, %true_22287, %none_22288 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22289 = torch.constant.int -1
%24756 = torch.prim.ListConstruct %int-1_22289 : (!torch.int) -> !torch.list<int>
%true_22290 = torch.constant.bool true
%none_22291 = torch.constant.none
%24757 = torch.aten.mean.dim %24746, %24756, %true_22290, %none_22291 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22292 = torch.constant.int -1
%24758 = torch.prim.ListConstruct %int-1_22292 : (!torch.int) -> !torch.list<int>
%true_22293 = torch.constant.bool true
%none_22294 = torch.constant.none
%24759 = torch.aten.mean.dim %24747, %24758, %true_22293, %none_22294 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22295 = torch.constant.int -1
%24760 = torch.prim.ListConstruct %int-1_22295 : (!torch.int) -> !torch.list<int>
%true_22296 = torch.constant.bool true
%none_22297 = torch.constant.none
%24761 = torch.aten.mean.dim %24748, %24760, %true_22296, %none_22297 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22298 = torch.constant.int -1
%24762 = torch.prim.ListConstruct %int-1_22298 : (!torch.int) -> !torch.list<int>
%true_22299 = torch.constant.bool true
%none_22300 = torch.constant.none
%24763 = torch.aten.mean.dim %24749, %24762, %true_22299, %none_22300 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22301 = torch.constant.int -1
%24764 = torch.prim.ListConstruct %int-1_22301 : (!torch.int) -> !torch.list<int>
%true_22302 = torch.constant.bool true
%none_22303 = torch.constant.none
%24765 = torch.aten.mean.dim %24750, %24764, %true_22302, %none_22303 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_22304 = torch.constant.int -1
%24766 = torch.prim.ListConstruct %int-1_22304 : (!torch.int) -> !torch.list<int>
%true_22305 = torch.constant.bool true
%none_22306 = torch.constant.none
%24767 = torch.aten.mean.dim %24751, %24766, %true_22305, %none_22306 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22307 = torch.constant.float 9.9999997473787516E-6
%int1_22308 = torch.constant.int 1
%24768 = torch.aten.add.Scalar %24753, %float9.999990e-06_22307, %int1_22308 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22309 = torch.constant.float 9.9999997473787516E-6
%int1_22310 = torch.constant.int 1
%24769 = torch.aten.add.Scalar %24755, %float9.999990e-06_22309, %int1_22310 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22311 = torch.constant.float 9.9999997473787516E-6
%int1_22312 = torch.constant.int 1
%24770 = torch.aten.add.Scalar %24757, %float9.999990e-06_22311, %int1_22312 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22313 = torch.constant.float 9.9999997473787516E-6
%int1_22314 = torch.constant.int 1
%24771 = torch.aten.add.Scalar %24759, %float9.999990e-06_22313, %int1_22314 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22315 = torch.constant.float 9.9999997473787516E-6
%int1_22316 = torch.constant.int 1
%24772 = torch.aten.add.Scalar %24761, %float9.999990e-06_22315, %int1_22316 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22317 = torch.constant.float 9.9999997473787516E-6
%int1_22318 = torch.constant.int 1
%24773 = torch.aten.add.Scalar %24763, %float9.999990e-06_22317, %int1_22318 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22319 = torch.constant.float 9.9999997473787516E-6
%int1_22320 = torch.constant.int 1
%24774 = torch.aten.add.Scalar %24765, %float9.999990e-06_22319, %int1_22320 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_22321 = torch.constant.float 9.9999997473787516E-6
%int1_22322 = torch.constant.int 1
%24775 = torch.aten.add.Scalar %24767, %float9.999990e-06_22321, %int1_22322 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24776 = torch.aten.rsqrt %24768 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24777 = torch.aten.rsqrt %24769 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24778 = torch.aten.rsqrt %24770 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24779 = torch.aten.rsqrt %24771 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24780 = torch.aten.rsqrt %24772 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24781 = torch.aten.rsqrt %24773 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24782 = torch.aten.rsqrt %24774 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24783 = torch.aten.rsqrt %24775 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %24783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%24784 = torch.aten.mul.Tensor %24736, %24776 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24785 = torch.aten.mul.Tensor %24737, %24777 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24786 = torch.aten.mul.Tensor %24738, %24778 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24787 = torch.aten.mul.Tensor %24739, %24779 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24788 = torch.aten.mul.Tensor %24740, %24780 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24789 = torch.aten.mul.Tensor %24741, %24781 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24790 = torch.aten.mul.Tensor %24742, %24782 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24791 = torch.aten.mul.Tensor %24743, %24783 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24792 = torch.aten.mul.Tensor %872, %24784 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24793 = torch.aten.mul.Tensor %873, %24785 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24794 = torch.aten.mul.Tensor %874, %24786 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24795 = torch.aten.mul.Tensor %875, %24787 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24796 = torch.aten.mul.Tensor %876, %24788 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24797 = torch.aten.mul.Tensor %877, %24789 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24798 = torch.aten.mul.Tensor %878, %24790 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%24799 = torch.aten.mul.Tensor %879, %24791 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %24799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_22323 = torch.constant.int 5
%24800 = torch.prims.convert_element_type %24792, %int5_22323 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22324 = torch.constant.int 5
%24801 = torch.prims.convert_element_type %24793, %int5_22324 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22325 = torch.constant.int 5
%24802 = torch.prims.convert_element_type %24794, %int5_22325 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22326 = torch.constant.int 5
%24803 = torch.prims.convert_element_type %24795, %int5_22326 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22327 = torch.constant.int 5
%24804 = torch.prims.convert_element_type %24796, %int5_22327 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22328 = torch.constant.int 5
%24805 = torch.prims.convert_element_type %24797, %int5_22328 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22329 = torch.constant.int 5
%24806 = torch.prims.convert_element_type %24798, %int5_22329 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_22330 = torch.constant.int 5
%24807 = torch.prims.convert_element_type %24799, %int5_22330 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %24807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_22331 = torch.constant.int 1
%int0_22332 = torch.constant.int 0
%24808 = torch.prim.ListConstruct %int1_22331, %int0_22332 : (!torch.int, !torch.int) -> !torch.list<int>
%24809 = torch.aten.permute %880, %24808 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22333 = torch.constant.int 1
%int0_22334 = torch.constant.int 0
%24810 = torch.prim.ListConstruct %int1_22333, %int0_22334 : (!torch.int, !torch.int) -> !torch.list<int>
%24811 = torch.aten.permute %881, %24810 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22335 = torch.constant.int 1
%int0_22336 = torch.constant.int 0
%24812 = torch.prim.ListConstruct %int1_22335, %int0_22336 : (!torch.int, !torch.int) -> !torch.list<int>
%24813 = torch.aten.permute %882, %24812 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22337 = torch.constant.int 1
%int0_22338 = torch.constant.int 0
%24814 = torch.prim.ListConstruct %int1_22337, %int0_22338 : (!torch.int, !torch.int) -> !torch.list<int>
%24815 = torch.aten.permute %883, %24814 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22339 = torch.constant.int 1
%int0_22340 = torch.constant.int 0
%24816 = torch.prim.ListConstruct %int1_22339, %int0_22340 : (!torch.int, !torch.int) -> !torch.list<int>
%24817 = torch.aten.permute %884, %24816 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22341 = torch.constant.int 1
%int0_22342 = torch.constant.int 0
%24818 = torch.prim.ListConstruct %int1_22341, %int0_22342 : (!torch.int, !torch.int) -> !torch.list<int>
%24819 = torch.aten.permute %885, %24818 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22343 = torch.constant.int 1
%int0_22344 = torch.constant.int 0
%24820 = torch.prim.ListConstruct %int1_22343, %int0_22344 : (!torch.int, !torch.int) -> !torch.list<int>
%24821 = torch.aten.permute %886, %24820 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_22345 = torch.constant.int 1
%int0_22346 = torch.constant.int 0
%24822 = torch.prim.ListConstruct %int1_22345, %int0_22346 : (!torch.int, !torch.int) -> !torch.list<int>
%24823 = torch.aten.permute %887, %24822 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_22347 = torch.constant.int 4
%24824 = torch.aten.mul.int %int4_22347, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22348 = torch.constant.int 4096
%24825 = torch.prim.ListConstruct %24824, %int4096_22348 : (!torch.int, !torch.int) -> !torch.list<int>
%24826 = torch.aten.view %24800, %24825 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24826, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24827 = torch.aten.mm %24826, %24809 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24827, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22349 = torch.constant.int 4
%int512_22350 = torch.constant.int 512
%24828 = torch.prim.ListConstruct %int4_22349, %2482, %int512_22350 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24829 = torch.aten.view %24827, %24828 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22351 = torch.constant.int 4
%24830 = torch.aten.mul.int %int4_22351, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22352 = torch.constant.int 4096
%24831 = torch.prim.ListConstruct %24830, %int4096_22352 : (!torch.int, !torch.int) -> !torch.list<int>
%24832 = torch.aten.view %24801, %24831 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24832, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24833 = torch.aten.mm %24832, %24811 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24833, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22353 = torch.constant.int 4
%int512_22354 = torch.constant.int 512
%24834 = torch.prim.ListConstruct %int4_22353, %2482, %int512_22354 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24835 = torch.aten.view %24833, %24834 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22355 = torch.constant.int 4
%24836 = torch.aten.mul.int %int4_22355, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22356 = torch.constant.int 4096
%24837 = torch.prim.ListConstruct %24836, %int4096_22356 : (!torch.int, !torch.int) -> !torch.list<int>
%24838 = torch.aten.view %24802, %24837 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24838, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24839 = torch.aten.mm %24838, %24813 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24839, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22357 = torch.constant.int 4
%int512_22358 = torch.constant.int 512
%24840 = torch.prim.ListConstruct %int4_22357, %2482, %int512_22358 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24841 = torch.aten.view %24839, %24840 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22359 = torch.constant.int 4
%24842 = torch.aten.mul.int %int4_22359, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22360 = torch.constant.int 4096
%24843 = torch.prim.ListConstruct %24842, %int4096_22360 : (!torch.int, !torch.int) -> !torch.list<int>
%24844 = torch.aten.view %24803, %24843 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24844, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24845 = torch.aten.mm %24844, %24815 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24845, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22361 = torch.constant.int 4
%int512_22362 = torch.constant.int 512
%24846 = torch.prim.ListConstruct %int4_22361, %2482, %int512_22362 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24847 = torch.aten.view %24845, %24846 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22363 = torch.constant.int 4
%24848 = torch.aten.mul.int %int4_22363, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22364 = torch.constant.int 4096
%24849 = torch.prim.ListConstruct %24848, %int4096_22364 : (!torch.int, !torch.int) -> !torch.list<int>
%24850 = torch.aten.view %24804, %24849 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24850, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24851 = torch.aten.mm %24850, %24817 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24851, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22365 = torch.constant.int 4
%int512_22366 = torch.constant.int 512
%24852 = torch.prim.ListConstruct %int4_22365, %2482, %int512_22366 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24853 = torch.aten.view %24851, %24852 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22367 = torch.constant.int 4
%24854 = torch.aten.mul.int %int4_22367, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22368 = torch.constant.int 4096
%24855 = torch.prim.ListConstruct %24854, %int4096_22368 : (!torch.int, !torch.int) -> !torch.list<int>
%24856 = torch.aten.view %24805, %24855 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24856, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24857 = torch.aten.mm %24856, %24819 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24857, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22369 = torch.constant.int 4
%int512_22370 = torch.constant.int 512
%24858 = torch.prim.ListConstruct %int4_22369, %2482, %int512_22370 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24859 = torch.aten.view %24857, %24858 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22371 = torch.constant.int 4
%24860 = torch.aten.mul.int %int4_22371, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22372 = torch.constant.int 4096
%24861 = torch.prim.ListConstruct %24860, %int4096_22372 : (!torch.int, !torch.int) -> !torch.list<int>
%24862 = torch.aten.view %24806, %24861 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24862, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24863 = torch.aten.mm %24862, %24821 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24863, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22373 = torch.constant.int 4
%int512_22374 = torch.constant.int 512
%24864 = torch.prim.ListConstruct %int4_22373, %2482, %int512_22374 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24865 = torch.aten.view %24863, %24864 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_22375 = torch.constant.int 4
%24866 = torch.aten.mul.int %int4_22375, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22376 = torch.constant.int 4096
%24867 = torch.prim.ListConstruct %24866, %int4096_22376 : (!torch.int, !torch.int) -> !torch.list<int>
%24868 = torch.aten.view %24807, %24867 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24868, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24869 = torch.aten.mm %24868, %24823 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %24869, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_22377 = torch.constant.int 4
%int512_22378 = torch.constant.int 512
%24870 = torch.prim.ListConstruct %int4_22377, %2482, %int512_22378 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24871 = torch.aten.view %24869, %24870 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %24871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_22379 = torch.constant.int 1
%int0_22380 = torch.constant.int 0
%24872 = torch.prim.ListConstruct %int1_22379, %int0_22380 : (!torch.int, !torch.int) -> !torch.list<int>
%24873 = torch.aten.permute %888, %24872 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22381 = torch.constant.int 1
%int0_22382 = torch.constant.int 0
%24874 = torch.prim.ListConstruct %int1_22381, %int0_22382 : (!torch.int, !torch.int) -> !torch.list<int>
%24875 = torch.aten.permute %889, %24874 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22383 = torch.constant.int 1
%int0_22384 = torch.constant.int 0
%24876 = torch.prim.ListConstruct %int1_22383, %int0_22384 : (!torch.int, !torch.int) -> !torch.list<int>
%24877 = torch.aten.permute %890, %24876 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22385 = torch.constant.int 1
%int0_22386 = torch.constant.int 0
%24878 = torch.prim.ListConstruct %int1_22385, %int0_22386 : (!torch.int, !torch.int) -> !torch.list<int>
%24879 = torch.aten.permute %891, %24878 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22387 = torch.constant.int 1
%int0_22388 = torch.constant.int 0
%24880 = torch.prim.ListConstruct %int1_22387, %int0_22388 : (!torch.int, !torch.int) -> !torch.list<int>
%24881 = torch.aten.permute %892, %24880 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22389 = torch.constant.int 1
%int0_22390 = torch.constant.int 0
%24882 = torch.prim.ListConstruct %int1_22389, %int0_22390 : (!torch.int, !torch.int) -> !torch.list<int>
%24883 = torch.aten.permute %893, %24882 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22391 = torch.constant.int 1
%int0_22392 = torch.constant.int 0
%24884 = torch.prim.ListConstruct %int1_22391, %int0_22392 : (!torch.int, !torch.int) -> !torch.list<int>
%24885 = torch.aten.permute %894, %24884 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22393 = torch.constant.int 1
%int0_22394 = torch.constant.int 0
%24886 = torch.prim.ListConstruct %int1_22393, %int0_22394 : (!torch.int, !torch.int) -> !torch.list<int>
%24887 = torch.aten.permute %895, %24886 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_22395 = torch.constant.int 4
%24888 = torch.aten.mul.int %int4_22395, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22396 = torch.constant.int 4096
%24889 = torch.prim.ListConstruct %24888, %int4096_22396 : (!torch.int, !torch.int) -> !torch.list<int>
%24890 = torch.aten.view %24800, %24889 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24890, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24891 = torch.aten.mm %24890, %24873 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24891, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22397 = torch.constant.int 4
%int128_22398 = torch.constant.int 128
%24892 = torch.prim.ListConstruct %int4_22397, %2482, %int128_22398 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24893 = torch.aten.view %24891, %24892 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22399 = torch.constant.int 4
%24894 = torch.aten.mul.int %int4_22399, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22400 = torch.constant.int 4096
%24895 = torch.prim.ListConstruct %24894, %int4096_22400 : (!torch.int, !torch.int) -> !torch.list<int>
%24896 = torch.aten.view %24801, %24895 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24896, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24897 = torch.aten.mm %24896, %24875 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24897, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22401 = torch.constant.int 4
%int128_22402 = torch.constant.int 128
%24898 = torch.prim.ListConstruct %int4_22401, %2482, %int128_22402 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24899 = torch.aten.view %24897, %24898 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22403 = torch.constant.int 4
%24900 = torch.aten.mul.int %int4_22403, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22404 = torch.constant.int 4096
%24901 = torch.prim.ListConstruct %24900, %int4096_22404 : (!torch.int, !torch.int) -> !torch.list<int>
%24902 = torch.aten.view %24802, %24901 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24902, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24903 = torch.aten.mm %24902, %24877 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24903, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22405 = torch.constant.int 4
%int128_22406 = torch.constant.int 128
%24904 = torch.prim.ListConstruct %int4_22405, %2482, %int128_22406 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24905 = torch.aten.view %24903, %24904 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22407 = torch.constant.int 4
%24906 = torch.aten.mul.int %int4_22407, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22408 = torch.constant.int 4096
%24907 = torch.prim.ListConstruct %24906, %int4096_22408 : (!torch.int, !torch.int) -> !torch.list<int>
%24908 = torch.aten.view %24803, %24907 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24908, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24909 = torch.aten.mm %24908, %24879 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24909, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22409 = torch.constant.int 4
%int128_22410 = torch.constant.int 128
%24910 = torch.prim.ListConstruct %int4_22409, %2482, %int128_22410 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24911 = torch.aten.view %24909, %24910 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22411 = torch.constant.int 4
%24912 = torch.aten.mul.int %int4_22411, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22412 = torch.constant.int 4096
%24913 = torch.prim.ListConstruct %24912, %int4096_22412 : (!torch.int, !torch.int) -> !torch.list<int>
%24914 = torch.aten.view %24804, %24913 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24914, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24915 = torch.aten.mm %24914, %24881 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24915, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22413 = torch.constant.int 4
%int128_22414 = torch.constant.int 128
%24916 = torch.prim.ListConstruct %int4_22413, %2482, %int128_22414 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24917 = torch.aten.view %24915, %24916 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22415 = torch.constant.int 4
%24918 = torch.aten.mul.int %int4_22415, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22416 = torch.constant.int 4096
%24919 = torch.prim.ListConstruct %24918, %int4096_22416 : (!torch.int, !torch.int) -> !torch.list<int>
%24920 = torch.aten.view %24805, %24919 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24920, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24921 = torch.aten.mm %24920, %24883 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24921, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22417 = torch.constant.int 4
%int128_22418 = torch.constant.int 128
%24922 = torch.prim.ListConstruct %int4_22417, %2482, %int128_22418 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24923 = torch.aten.view %24921, %24922 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22419 = torch.constant.int 4
%24924 = torch.aten.mul.int %int4_22419, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22420 = torch.constant.int 4096
%24925 = torch.prim.ListConstruct %24924, %int4096_22420 : (!torch.int, !torch.int) -> !torch.list<int>
%24926 = torch.aten.view %24806, %24925 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24926, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24927 = torch.aten.mm %24926, %24885 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24927, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22421 = torch.constant.int 4
%int128_22422 = torch.constant.int 128
%24928 = torch.prim.ListConstruct %int4_22421, %2482, %int128_22422 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24929 = torch.aten.view %24927, %24928 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22423 = torch.constant.int 4
%24930 = torch.aten.mul.int %int4_22423, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22424 = torch.constant.int 4096
%24931 = torch.prim.ListConstruct %24930, %int4096_22424 : (!torch.int, !torch.int) -> !torch.list<int>
%24932 = torch.aten.view %24807, %24931 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24932, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24933 = torch.aten.mm %24932, %24887 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24933, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22425 = torch.constant.int 4
%int128_22426 = torch.constant.int 128
%24934 = torch.prim.ListConstruct %int4_22425, %2482, %int128_22426 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24935 = torch.aten.view %24933, %24934 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_22427 = torch.constant.int 1
%int0_22428 = torch.constant.int 0
%24936 = torch.prim.ListConstruct %int1_22427, %int0_22428 : (!torch.int, !torch.int) -> !torch.list<int>
%24937 = torch.aten.permute %896, %24936 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22429 = torch.constant.int 1
%int0_22430 = torch.constant.int 0
%24938 = torch.prim.ListConstruct %int1_22429, %int0_22430 : (!torch.int, !torch.int) -> !torch.list<int>
%24939 = torch.aten.permute %897, %24938 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22431 = torch.constant.int 1
%int0_22432 = torch.constant.int 0
%24940 = torch.prim.ListConstruct %int1_22431, %int0_22432 : (!torch.int, !torch.int) -> !torch.list<int>
%24941 = torch.aten.permute %898, %24940 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22433 = torch.constant.int 1
%int0_22434 = torch.constant.int 0
%24942 = torch.prim.ListConstruct %int1_22433, %int0_22434 : (!torch.int, !torch.int) -> !torch.list<int>
%24943 = torch.aten.permute %899, %24942 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22435 = torch.constant.int 1
%int0_22436 = torch.constant.int 0
%24944 = torch.prim.ListConstruct %int1_22435, %int0_22436 : (!torch.int, !torch.int) -> !torch.list<int>
%24945 = torch.aten.permute %900, %24944 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22437 = torch.constant.int 1
%int0_22438 = torch.constant.int 0
%24946 = torch.prim.ListConstruct %int1_22437, %int0_22438 : (!torch.int, !torch.int) -> !torch.list<int>
%24947 = torch.aten.permute %901, %24946 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22439 = torch.constant.int 1
%int0_22440 = torch.constant.int 0
%24948 = torch.prim.ListConstruct %int1_22439, %int0_22440 : (!torch.int, !torch.int) -> !torch.list<int>
%24949 = torch.aten.permute %902, %24948 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_22441 = torch.constant.int 1
%int0_22442 = torch.constant.int 0
%24950 = torch.prim.ListConstruct %int1_22441, %int0_22442 : (!torch.int, !torch.int) -> !torch.list<int>
%24951 = torch.aten.permute %903, %24950 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_22443 = torch.constant.int 4
%24952 = torch.aten.mul.int %int4_22443, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22444 = torch.constant.int 4096
%24953 = torch.prim.ListConstruct %24952, %int4096_22444 : (!torch.int, !torch.int) -> !torch.list<int>
%24954 = torch.aten.view %24800, %24953 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24954, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24955 = torch.aten.mm %24954, %24937 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24955, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22445 = torch.constant.int 4
%int128_22446 = torch.constant.int 128
%24956 = torch.prim.ListConstruct %int4_22445, %2482, %int128_22446 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24957 = torch.aten.view %24955, %24956 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22447 = torch.constant.int 4
%24958 = torch.aten.mul.int %int4_22447, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22448 = torch.constant.int 4096
%24959 = torch.prim.ListConstruct %24958, %int4096_22448 : (!torch.int, !torch.int) -> !torch.list<int>
%24960 = torch.aten.view %24801, %24959 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24960, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24961 = torch.aten.mm %24960, %24939 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24961, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22449 = torch.constant.int 4
%int128_22450 = torch.constant.int 128
%24962 = torch.prim.ListConstruct %int4_22449, %2482, %int128_22450 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24963 = torch.aten.view %24961, %24962 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22451 = torch.constant.int 4
%24964 = torch.aten.mul.int %int4_22451, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22452 = torch.constant.int 4096
%24965 = torch.prim.ListConstruct %24964, %int4096_22452 : (!torch.int, !torch.int) -> !torch.list<int>
%24966 = torch.aten.view %24802, %24965 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24966, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24967 = torch.aten.mm %24966, %24941 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24967, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22453 = torch.constant.int 4
%int128_22454 = torch.constant.int 128
%24968 = torch.prim.ListConstruct %int4_22453, %2482, %int128_22454 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24969 = torch.aten.view %24967, %24968 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22455 = torch.constant.int 4
%24970 = torch.aten.mul.int %int4_22455, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22456 = torch.constant.int 4096
%24971 = torch.prim.ListConstruct %24970, %int4096_22456 : (!torch.int, !torch.int) -> !torch.list<int>
%24972 = torch.aten.view %24803, %24971 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24972, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24973 = torch.aten.mm %24972, %24943 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24973, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22457 = torch.constant.int 4
%int128_22458 = torch.constant.int 128
%24974 = torch.prim.ListConstruct %int4_22457, %2482, %int128_22458 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24975 = torch.aten.view %24973, %24974 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22459 = torch.constant.int 4
%24976 = torch.aten.mul.int %int4_22459, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22460 = torch.constant.int 4096
%24977 = torch.prim.ListConstruct %24976, %int4096_22460 : (!torch.int, !torch.int) -> !torch.list<int>
%24978 = torch.aten.view %24804, %24977 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24978, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24979 = torch.aten.mm %24978, %24945 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24979, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22461 = torch.constant.int 4
%int128_22462 = torch.constant.int 128
%24980 = torch.prim.ListConstruct %int4_22461, %2482, %int128_22462 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24981 = torch.aten.view %24979, %24980 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22463 = torch.constant.int 4
%24982 = torch.aten.mul.int %int4_22463, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22464 = torch.constant.int 4096
%24983 = torch.prim.ListConstruct %24982, %int4096_22464 : (!torch.int, !torch.int) -> !torch.list<int>
%24984 = torch.aten.view %24805, %24983 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24984, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24985 = torch.aten.mm %24984, %24947 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24985, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22465 = torch.constant.int 4
%int128_22466 = torch.constant.int 128
%24986 = torch.prim.ListConstruct %int4_22465, %2482, %int128_22466 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24987 = torch.aten.view %24985, %24986 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22467 = torch.constant.int 4
%24988 = torch.aten.mul.int %int4_22467, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22468 = torch.constant.int 4096
%24989 = torch.prim.ListConstruct %24988, %int4096_22468 : (!torch.int, !torch.int) -> !torch.list<int>
%24990 = torch.aten.view %24806, %24989 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24990, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24991 = torch.aten.mm %24990, %24949 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24991, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22469 = torch.constant.int 4
%int128_22470 = torch.constant.int 128
%24992 = torch.prim.ListConstruct %int4_22469, %2482, %int128_22470 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24993 = torch.aten.view %24991, %24992 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22471 = torch.constant.int 4
%24994 = torch.aten.mul.int %int4_22471, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_22472 = torch.constant.int 4096
%24995 = torch.prim.ListConstruct %24994, %int4096_22472 : (!torch.int, !torch.int) -> !torch.list<int>
%24996 = torch.aten.view %24807, %24995 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %24996, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%24997 = torch.aten.mm %24996, %24951 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %24997, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_22473 = torch.constant.int 4
%int128_22474 = torch.constant.int 128
%24998 = torch.prim.ListConstruct %int4_22473, %2482, %int128_22474 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%24999 = torch.aten.view %24997, %24998 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %24999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_22475 = torch.constant.int 4
%int4_22476 = torch.constant.int 4
%int128_22477 = torch.constant.int 128
%25000 = torch.prim.ListConstruct %int4_22475, %2482, %int4_22476, %int128_22477 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25001 = torch.aten.view %24829, %25000 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22478 = torch.constant.int 4
%int4_22479 = torch.constant.int 4
%int128_22480 = torch.constant.int 128
%25002 = torch.prim.ListConstruct %int4_22478, %2482, %int4_22479, %int128_22480 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25003 = torch.aten.view %24835, %25002 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22481 = torch.constant.int 4
%int4_22482 = torch.constant.int 4
%int128_22483 = torch.constant.int 128
%25004 = torch.prim.ListConstruct %int4_22481, %2482, %int4_22482, %int128_22483 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25005 = torch.aten.view %24841, %25004 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22484 = torch.constant.int 4
%int4_22485 = torch.constant.int 4
%int128_22486 = torch.constant.int 128
%25006 = torch.prim.ListConstruct %int4_22484, %2482, %int4_22485, %int128_22486 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25007 = torch.aten.view %24847, %25006 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22487 = torch.constant.int 4
%int4_22488 = torch.constant.int 4
%int128_22489 = torch.constant.int 128
%25008 = torch.prim.ListConstruct %int4_22487, %2482, %int4_22488, %int128_22489 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25009 = torch.aten.view %24853, %25008 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22490 = torch.constant.int 4
%int4_22491 = torch.constant.int 4
%int128_22492 = torch.constant.int 128
%25010 = torch.prim.ListConstruct %int4_22490, %2482, %int4_22491, %int128_22492 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25011 = torch.aten.view %24859, %25010 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22493 = torch.constant.int 4
%int4_22494 = torch.constant.int 4
%int128_22495 = torch.constant.int 128
%25012 = torch.prim.ListConstruct %int4_22493, %2482, %int4_22494, %int128_22495 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25013 = torch.aten.view %24865, %25012 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22496 = torch.constant.int 4
%int4_22497 = torch.constant.int 4
%int128_22498 = torch.constant.int 128
%25014 = torch.prim.ListConstruct %int4_22496, %2482, %int4_22497, %int128_22498 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25015 = torch.aten.view %24871, %25014 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_22499 = torch.constant.int 4
%int1_22500 = torch.constant.int 1
%int128_22501 = torch.constant.int 128
%25016 = torch.prim.ListConstruct %int4_22499, %2482, %int1_22500, %int128_22501 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25017 = torch.aten.view %24893, %25016 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22502 = torch.constant.int 4
%int1_22503 = torch.constant.int 1
%int128_22504 = torch.constant.int 128
%25018 = torch.prim.ListConstruct %int4_22502, %2482, %int1_22503, %int128_22504 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25019 = torch.aten.view %24899, %25018 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22505 = torch.constant.int 4
%int1_22506 = torch.constant.int 1
%int128_22507 = torch.constant.int 128
%25020 = torch.prim.ListConstruct %int4_22505, %2482, %int1_22506, %int128_22507 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25021 = torch.aten.view %24905, %25020 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22508 = torch.constant.int 4
%int1_22509 = torch.constant.int 1
%int128_22510 = torch.constant.int 128
%25022 = torch.prim.ListConstruct %int4_22508, %2482, %int1_22509, %int128_22510 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25023 = torch.aten.view %24911, %25022 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22511 = torch.constant.int 4
%int1_22512 = torch.constant.int 1
%int128_22513 = torch.constant.int 128
%25024 = torch.prim.ListConstruct %int4_22511, %2482, %int1_22512, %int128_22513 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25025 = torch.aten.view %24917, %25024 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22514 = torch.constant.int 4
%int1_22515 = torch.constant.int 1
%int128_22516 = torch.constant.int 128
%25026 = torch.prim.ListConstruct %int4_22514, %2482, %int1_22515, %int128_22516 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25027 = torch.aten.view %24923, %25026 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22517 = torch.constant.int 4
%int1_22518 = torch.constant.int 1
%int128_22519 = torch.constant.int 128
%25028 = torch.prim.ListConstruct %int4_22517, %2482, %int1_22518, %int128_22519 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25029 = torch.aten.view %24929, %25028 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22520 = torch.constant.int 4
%int1_22521 = torch.constant.int 1
%int128_22522 = torch.constant.int 128
%25030 = torch.prim.ListConstruct %int4_22520, %2482, %int1_22521, %int128_22522 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25031 = torch.aten.view %24935, %25030 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22523 = torch.constant.int 4
%int1_22524 = torch.constant.int 1
%int128_22525 = torch.constant.int 128
%25032 = torch.prim.ListConstruct %int4_22523, %2482, %int1_22524, %int128_22525 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25033 = torch.aten.view %24957, %25032 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22526 = torch.constant.int 4
%int1_22527 = torch.constant.int 1
%int128_22528 = torch.constant.int 128
%25034 = torch.prim.ListConstruct %int4_22526, %2482, %int1_22527, %int128_22528 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25035 = torch.aten.view %24963, %25034 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22529 = torch.constant.int 4
%int1_22530 = torch.constant.int 1
%int128_22531 = torch.constant.int 128
%25036 = torch.prim.ListConstruct %int4_22529, %2482, %int1_22530, %int128_22531 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25037 = torch.aten.view %24969, %25036 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22532 = torch.constant.int 4
%int1_22533 = torch.constant.int 1
%int128_22534 = torch.constant.int 128
%25038 = torch.prim.ListConstruct %int4_22532, %2482, %int1_22533, %int128_22534 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25039 = torch.aten.view %24975, %25038 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22535 = torch.constant.int 4
%int1_22536 = torch.constant.int 1
%int128_22537 = torch.constant.int 128
%25040 = torch.prim.ListConstruct %int4_22535, %2482, %int1_22536, %int128_22537 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25041 = torch.aten.view %24981, %25040 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22538 = torch.constant.int 4
%int1_22539 = torch.constant.int 1
%int128_22540 = torch.constant.int 128
%25042 = torch.prim.ListConstruct %int4_22538, %2482, %int1_22539, %int128_22540 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25043 = torch.aten.view %24987, %25042 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22541 = torch.constant.int 4
%int1_22542 = torch.constant.int 1
%int128_22543 = torch.constant.int 128
%25044 = torch.prim.ListConstruct %int4_22541, %2482, %int1_22542, %int128_22543 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25045 = torch.aten.view %24993, %25044 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_22544 = torch.constant.int 4
%int1_22545 = torch.constant.int 1
%int128_22546 = torch.constant.int 128
%25046 = torch.prim.ListConstruct %int4_22544, %2482, %int1_22545, %int128_22546 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25047 = torch.aten.view %24999, %25046 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_22547 = torch.constant.int 131072
%none_22548 = torch.constant.none
%none_22549 = torch.constant.none
%cpu_22550 = torch.constant.device "cpu"
%false_22551 = torch.constant.bool false
%25048 = torch.aten.arange %int131072_22547, %none_22548, %none_22549, %cpu_22550, %false_22551 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_22552 = torch.constant.int 0
%int128_22553 = torch.constant.int 128
%int2_22554 = torch.constant.int 2
%none_22555 = torch.constant.none
%none_22556 = torch.constant.none
%cpu_22557 = torch.constant.device "cpu"
%false_22558 = torch.constant.bool false
%25049 = torch.aten.arange.start_step %int0_22552, %int128_22553, %int2_22554, %none_22555, %none_22556, %cpu_22557, %false_22558 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_22559 = torch.constant.int 0
%int0_22560 = torch.constant.int 0
%int64_22561 = torch.constant.int 64
%int1_22562 = torch.constant.int 1
%25050 = torch.aten.slice.Tensor %25049, %int0_22559, %int0_22560, %int64_22561, %int1_22562 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_22563 = torch.constant.int 6
%25051 = torch.prims.convert_element_type %25050, %int6_22563 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_22564 = torch.constant.int 128
%25052 = torch.aten.div.Scalar %25051, %int128_22564 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_22565 = torch.constant.float 5.000000e+05
%25053 = torch.aten.pow.Scalar %float5.000000e05_22565, %25052 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%25054 = torch.aten.reciprocal %25053 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_22566 = torch.constant.float 1.000000e+00
%25055 = torch.aten.mul.Scalar %25054, %float1.000000e00_22566 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_22567 = torch.constant.int 131072
%int1_22568 = torch.constant.int 1
%25056 = torch.prim.ListConstruct %int131072_22567, %int1_22568 : (!torch.int, !torch.int) -> !torch.list<int>
%25057 = torch.aten.view %25048, %25056 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%25058 = torch.aten.mul.Tensor %25057, %25055 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%25059 = torch.aten.cos %25058 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%25060 = torch.aten.sin %25058 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%25061 = torch.aten.complex %25059, %25060 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%25062 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25063 = flow.tensor.transfer %25062 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%25064 = torch_c.from_builtin_tensor %25063 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25065 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25066 = flow.tensor.transfer %25065 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%25067 = torch_c.from_builtin_tensor %25066 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25068 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25069 = flow.tensor.transfer %25068 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%25070 = torch_c.from_builtin_tensor %25069 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25071 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25072 = flow.tensor.transfer %25071 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%25073 = torch_c.from_builtin_tensor %25072 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25074 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25075 = flow.tensor.transfer %25074 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%25076 = torch_c.from_builtin_tensor %25075 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25077 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25078 = flow.tensor.transfer %25077 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%25079 = torch_c.from_builtin_tensor %25078 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25080 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25081 = flow.tensor.transfer %25080 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%25082 = torch_c.from_builtin_tensor %25081 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25083 = torch_c.to_builtin_tensor %25061 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25084 = flow.tensor.transfer %25083 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%25085 = torch_c.from_builtin_tensor %25084 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_22569 = torch.constant.int 1
%25086 = torch.aten.size.int %24829, %int1_22569 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22570 = torch.constant.int 0
%25087 = torch.aten.add.int %int0_22570, %25086 : !torch.int, !torch.int -> !torch.int
%int0_22571 = torch.constant.int 0
%int0_22572 = torch.constant.int 0
%int1_22573 = torch.constant.int 1
%25088 = torch.aten.slice.Tensor %25064, %int0_22571, %int0_22572, %25087, %int1_22573 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25088, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22574 = torch.constant.int 1
%int0_22575 = torch.constant.int 0
%int9223372036854775807_22576 = torch.constant.int 9223372036854775807
%int1_22577 = torch.constant.int 1
%25089 = torch.aten.slice.Tensor %25088, %int1_22574, %int0_22575, %int9223372036854775807_22576, %int1_22577 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25089, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22578 = torch.constant.int 0
%25090 = torch.aten.unsqueeze %25089, %int0_22578 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25090, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22579 = torch.constant.int 2
%25091 = torch.aten.unsqueeze %25090, %int2_22579 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25091, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22580 = torch.constant.int 3
%int0_22581 = torch.constant.int 0
%int9223372036854775807_22582 = torch.constant.int 9223372036854775807
%int1_22583 = torch.constant.int 1
%25092 = torch.aten.slice.Tensor %25091, %int3_22580, %int0_22581, %int9223372036854775807_22582, %int1_22583 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25092, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25093 = torch_c.to_builtin_tensor %25001 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22584 = arith.constant 1 : index
%dim_22585 = tensor.dim %25093, %c1_22584 : tensor<4x?x4x128xf16>
%25094 = flow.tensor.bitcast %25093 : tensor<4x?x4x128xf16>{%dim_22585} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22585}
%25095 = torch_c.from_builtin_tensor %25094 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25096 = torch.aten.mul.Tensor %25095, %25092 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25097 = torch_c.to_builtin_tensor %25096 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22586 = arith.constant 1 : index
%dim_22587 = tensor.dim %25097, %c1_22586 : tensor<4x?x4x64xcomplex<f32>>
%25098 = flow.tensor.bitcast %25097 : tensor<4x?x4x64xcomplex<f32>>{%dim_22587} -> tensor<4x?x4x128xf32>{%dim_22587}
%25099 = torch_c.from_builtin_tensor %25098 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22588 = torch.constant.int 5
%25100 = torch.prims.convert_element_type %25099, %int5_22588 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22589 = torch.constant.int 1
%25101 = torch.aten.size.int %24835, %int1_22589 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22590 = torch.constant.int 0
%25102 = torch.aten.add.int %int0_22590, %25101 : !torch.int, !torch.int -> !torch.int
%int0_22591 = torch.constant.int 0
%int0_22592 = torch.constant.int 0
%int1_22593 = torch.constant.int 1
%25103 = torch.aten.slice.Tensor %25067, %int0_22591, %int0_22592, %25102, %int1_22593 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25103, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22594 = torch.constant.int 1
%int0_22595 = torch.constant.int 0
%int9223372036854775807_22596 = torch.constant.int 9223372036854775807
%int1_22597 = torch.constant.int 1
%25104 = torch.aten.slice.Tensor %25103, %int1_22594, %int0_22595, %int9223372036854775807_22596, %int1_22597 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25104, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22598 = torch.constant.int 0
%25105 = torch.aten.unsqueeze %25104, %int0_22598 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25105, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22599 = torch.constant.int 2
%25106 = torch.aten.unsqueeze %25105, %int2_22599 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25106, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22600 = torch.constant.int 3
%int0_22601 = torch.constant.int 0
%int9223372036854775807_22602 = torch.constant.int 9223372036854775807
%int1_22603 = torch.constant.int 1
%25107 = torch.aten.slice.Tensor %25106, %int3_22600, %int0_22601, %int9223372036854775807_22602, %int1_22603 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25107, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25108 = torch_c.to_builtin_tensor %25003 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22604 = arith.constant 1 : index
%dim_22605 = tensor.dim %25108, %c1_22604 : tensor<4x?x4x128xf16>
%25109 = flow.tensor.bitcast %25108 : tensor<4x?x4x128xf16>{%dim_22605} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22605}
%25110 = torch_c.from_builtin_tensor %25109 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25111 = torch.aten.mul.Tensor %25110, %25107 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25112 = torch_c.to_builtin_tensor %25111 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22606 = arith.constant 1 : index
%dim_22607 = tensor.dim %25112, %c1_22606 : tensor<4x?x4x64xcomplex<f32>>
%25113 = flow.tensor.bitcast %25112 : tensor<4x?x4x64xcomplex<f32>>{%dim_22607} -> tensor<4x?x4x128xf32>{%dim_22607}
%25114 = torch_c.from_builtin_tensor %25113 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22608 = torch.constant.int 5
%25115 = torch.prims.convert_element_type %25114, %int5_22608 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22609 = torch.constant.int 1
%25116 = torch.aten.size.int %24841, %int1_22609 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22610 = torch.constant.int 0
%25117 = torch.aten.add.int %int0_22610, %25116 : !torch.int, !torch.int -> !torch.int
%int0_22611 = torch.constant.int 0
%int0_22612 = torch.constant.int 0
%int1_22613 = torch.constant.int 1
%25118 = torch.aten.slice.Tensor %25070, %int0_22611, %int0_22612, %25117, %int1_22613 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25118, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22614 = torch.constant.int 1
%int0_22615 = torch.constant.int 0
%int9223372036854775807_22616 = torch.constant.int 9223372036854775807
%int1_22617 = torch.constant.int 1
%25119 = torch.aten.slice.Tensor %25118, %int1_22614, %int0_22615, %int9223372036854775807_22616, %int1_22617 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25119, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22618 = torch.constant.int 0
%25120 = torch.aten.unsqueeze %25119, %int0_22618 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25120, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22619 = torch.constant.int 2
%25121 = torch.aten.unsqueeze %25120, %int2_22619 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25121, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22620 = torch.constant.int 3
%int0_22621 = torch.constant.int 0
%int9223372036854775807_22622 = torch.constant.int 9223372036854775807
%int1_22623 = torch.constant.int 1
%25122 = torch.aten.slice.Tensor %25121, %int3_22620, %int0_22621, %int9223372036854775807_22622, %int1_22623 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25122, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25123 = torch_c.to_builtin_tensor %25005 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22624 = arith.constant 1 : index
%dim_22625 = tensor.dim %25123, %c1_22624 : tensor<4x?x4x128xf16>
%25124 = flow.tensor.bitcast %25123 : tensor<4x?x4x128xf16>{%dim_22625} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22625}
%25125 = torch_c.from_builtin_tensor %25124 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25126 = torch.aten.mul.Tensor %25125, %25122 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25127 = torch_c.to_builtin_tensor %25126 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22626 = arith.constant 1 : index
%dim_22627 = tensor.dim %25127, %c1_22626 : tensor<4x?x4x64xcomplex<f32>>
%25128 = flow.tensor.bitcast %25127 : tensor<4x?x4x64xcomplex<f32>>{%dim_22627} -> tensor<4x?x4x128xf32>{%dim_22627}
%25129 = torch_c.from_builtin_tensor %25128 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22628 = torch.constant.int 5
%25130 = torch.prims.convert_element_type %25129, %int5_22628 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22629 = torch.constant.int 1
%25131 = torch.aten.size.int %24847, %int1_22629 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22630 = torch.constant.int 0
%25132 = torch.aten.add.int %int0_22630, %25131 : !torch.int, !torch.int -> !torch.int
%int0_22631 = torch.constant.int 0
%int0_22632 = torch.constant.int 0
%int1_22633 = torch.constant.int 1
%25133 = torch.aten.slice.Tensor %25073, %int0_22631, %int0_22632, %25132, %int1_22633 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25133, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22634 = torch.constant.int 1
%int0_22635 = torch.constant.int 0
%int9223372036854775807_22636 = torch.constant.int 9223372036854775807
%int1_22637 = torch.constant.int 1
%25134 = torch.aten.slice.Tensor %25133, %int1_22634, %int0_22635, %int9223372036854775807_22636, %int1_22637 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25134, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22638 = torch.constant.int 0
%25135 = torch.aten.unsqueeze %25134, %int0_22638 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25135, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22639 = torch.constant.int 2
%25136 = torch.aten.unsqueeze %25135, %int2_22639 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25136, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22640 = torch.constant.int 3
%int0_22641 = torch.constant.int 0
%int9223372036854775807_22642 = torch.constant.int 9223372036854775807
%int1_22643 = torch.constant.int 1
%25137 = torch.aten.slice.Tensor %25136, %int3_22640, %int0_22641, %int9223372036854775807_22642, %int1_22643 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25137, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25138 = torch_c.to_builtin_tensor %25007 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22644 = arith.constant 1 : index
%dim_22645 = tensor.dim %25138, %c1_22644 : tensor<4x?x4x128xf16>
%25139 = flow.tensor.bitcast %25138 : tensor<4x?x4x128xf16>{%dim_22645} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22645}
%25140 = torch_c.from_builtin_tensor %25139 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25141 = torch.aten.mul.Tensor %25140, %25137 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25142 = torch_c.to_builtin_tensor %25141 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22646 = arith.constant 1 : index
%dim_22647 = tensor.dim %25142, %c1_22646 : tensor<4x?x4x64xcomplex<f32>>
%25143 = flow.tensor.bitcast %25142 : tensor<4x?x4x64xcomplex<f32>>{%dim_22647} -> tensor<4x?x4x128xf32>{%dim_22647}
%25144 = torch_c.from_builtin_tensor %25143 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22648 = torch.constant.int 5
%25145 = torch.prims.convert_element_type %25144, %int5_22648 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22649 = torch.constant.int 1
%25146 = torch.aten.size.int %24853, %int1_22649 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22650 = torch.constant.int 0
%25147 = torch.aten.add.int %int0_22650, %25146 : !torch.int, !torch.int -> !torch.int
%int0_22651 = torch.constant.int 0
%int0_22652 = torch.constant.int 0
%int1_22653 = torch.constant.int 1
%25148 = torch.aten.slice.Tensor %25076, %int0_22651, %int0_22652, %25147, %int1_22653 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25148, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22654 = torch.constant.int 1
%int0_22655 = torch.constant.int 0
%int9223372036854775807_22656 = torch.constant.int 9223372036854775807
%int1_22657 = torch.constant.int 1
%25149 = torch.aten.slice.Tensor %25148, %int1_22654, %int0_22655, %int9223372036854775807_22656, %int1_22657 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25149, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22658 = torch.constant.int 0
%25150 = torch.aten.unsqueeze %25149, %int0_22658 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25150, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22659 = torch.constant.int 2
%25151 = torch.aten.unsqueeze %25150, %int2_22659 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25151, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22660 = torch.constant.int 3
%int0_22661 = torch.constant.int 0
%int9223372036854775807_22662 = torch.constant.int 9223372036854775807
%int1_22663 = torch.constant.int 1
%25152 = torch.aten.slice.Tensor %25151, %int3_22660, %int0_22661, %int9223372036854775807_22662, %int1_22663 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25152, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25153 = torch_c.to_builtin_tensor %25009 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22664 = arith.constant 1 : index
%dim_22665 = tensor.dim %25153, %c1_22664 : tensor<4x?x4x128xf16>
%25154 = flow.tensor.bitcast %25153 : tensor<4x?x4x128xf16>{%dim_22665} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22665}
%25155 = torch_c.from_builtin_tensor %25154 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25156 = torch.aten.mul.Tensor %25155, %25152 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25157 = torch_c.to_builtin_tensor %25156 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22666 = arith.constant 1 : index
%dim_22667 = tensor.dim %25157, %c1_22666 : tensor<4x?x4x64xcomplex<f32>>
%25158 = flow.tensor.bitcast %25157 : tensor<4x?x4x64xcomplex<f32>>{%dim_22667} -> tensor<4x?x4x128xf32>{%dim_22667}
%25159 = torch_c.from_builtin_tensor %25158 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22668 = torch.constant.int 5
%25160 = torch.prims.convert_element_type %25159, %int5_22668 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22669 = torch.constant.int 1
%25161 = torch.aten.size.int %24859, %int1_22669 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22670 = torch.constant.int 0
%25162 = torch.aten.add.int %int0_22670, %25161 : !torch.int, !torch.int -> !torch.int
%int0_22671 = torch.constant.int 0
%int0_22672 = torch.constant.int 0
%int1_22673 = torch.constant.int 1
%25163 = torch.aten.slice.Tensor %25079, %int0_22671, %int0_22672, %25162, %int1_22673 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25163, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22674 = torch.constant.int 1
%int0_22675 = torch.constant.int 0
%int9223372036854775807_22676 = torch.constant.int 9223372036854775807
%int1_22677 = torch.constant.int 1
%25164 = torch.aten.slice.Tensor %25163, %int1_22674, %int0_22675, %int9223372036854775807_22676, %int1_22677 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25164, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22678 = torch.constant.int 0
%25165 = torch.aten.unsqueeze %25164, %int0_22678 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25165, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22679 = torch.constant.int 2
%25166 = torch.aten.unsqueeze %25165, %int2_22679 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25166, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22680 = torch.constant.int 3
%int0_22681 = torch.constant.int 0
%int9223372036854775807_22682 = torch.constant.int 9223372036854775807
%int1_22683 = torch.constant.int 1
%25167 = torch.aten.slice.Tensor %25166, %int3_22680, %int0_22681, %int9223372036854775807_22682, %int1_22683 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25167, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25168 = torch_c.to_builtin_tensor %25011 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22684 = arith.constant 1 : index
%dim_22685 = tensor.dim %25168, %c1_22684 : tensor<4x?x4x128xf16>
%25169 = flow.tensor.bitcast %25168 : tensor<4x?x4x128xf16>{%dim_22685} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22685}
%25170 = torch_c.from_builtin_tensor %25169 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25171 = torch.aten.mul.Tensor %25170, %25167 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25172 = torch_c.to_builtin_tensor %25171 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22686 = arith.constant 1 : index
%dim_22687 = tensor.dim %25172, %c1_22686 : tensor<4x?x4x64xcomplex<f32>>
%25173 = flow.tensor.bitcast %25172 : tensor<4x?x4x64xcomplex<f32>>{%dim_22687} -> tensor<4x?x4x128xf32>{%dim_22687}
%25174 = torch_c.from_builtin_tensor %25173 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22688 = torch.constant.int 5
%25175 = torch.prims.convert_element_type %25174, %int5_22688 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22689 = torch.constant.int 1
%25176 = torch.aten.size.int %24865, %int1_22689 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22690 = torch.constant.int 0
%25177 = torch.aten.add.int %int0_22690, %25176 : !torch.int, !torch.int -> !torch.int
%int0_22691 = torch.constant.int 0
%int0_22692 = torch.constant.int 0
%int1_22693 = torch.constant.int 1
%25178 = torch.aten.slice.Tensor %25082, %int0_22691, %int0_22692, %25177, %int1_22693 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25178, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22694 = torch.constant.int 1
%int0_22695 = torch.constant.int 0
%int9223372036854775807_22696 = torch.constant.int 9223372036854775807
%int1_22697 = torch.constant.int 1
%25179 = torch.aten.slice.Tensor %25178, %int1_22694, %int0_22695, %int9223372036854775807_22696, %int1_22697 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25179, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22698 = torch.constant.int 0
%25180 = torch.aten.unsqueeze %25179, %int0_22698 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25180, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22699 = torch.constant.int 2
%25181 = torch.aten.unsqueeze %25180, %int2_22699 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25181, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22700 = torch.constant.int 3
%int0_22701 = torch.constant.int 0
%int9223372036854775807_22702 = torch.constant.int 9223372036854775807
%int1_22703 = torch.constant.int 1
%25182 = torch.aten.slice.Tensor %25181, %int3_22700, %int0_22701, %int9223372036854775807_22702, %int1_22703 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25182, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25183 = torch_c.to_builtin_tensor %25013 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22704 = arith.constant 1 : index
%dim_22705 = tensor.dim %25183, %c1_22704 : tensor<4x?x4x128xf16>
%25184 = flow.tensor.bitcast %25183 : tensor<4x?x4x128xf16>{%dim_22705} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22705}
%25185 = torch_c.from_builtin_tensor %25184 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25186 = torch.aten.mul.Tensor %25185, %25182 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25187 = torch_c.to_builtin_tensor %25186 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22706 = arith.constant 1 : index
%dim_22707 = tensor.dim %25187, %c1_22706 : tensor<4x?x4x64xcomplex<f32>>
%25188 = flow.tensor.bitcast %25187 : tensor<4x?x4x64xcomplex<f32>>{%dim_22707} -> tensor<4x?x4x128xf32>{%dim_22707}
%25189 = torch_c.from_builtin_tensor %25188 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22708 = torch.constant.int 5
%25190 = torch.prims.convert_element_type %25189, %int5_22708 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_22709 = torch.constant.int 1
%25191 = torch.aten.size.int %24871, %int1_22709 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_22710 = torch.constant.int 0
%25192 = torch.aten.add.int %int0_22710, %25191 : !torch.int, !torch.int -> !torch.int
%int0_22711 = torch.constant.int 0
%int0_22712 = torch.constant.int 0
%int1_22713 = torch.constant.int 1
%25193 = torch.aten.slice.Tensor %25085, %int0_22711, %int0_22712, %25192, %int1_22713 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25193, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22714 = torch.constant.int 1
%int0_22715 = torch.constant.int 0
%int9223372036854775807_22716 = torch.constant.int 9223372036854775807
%int1_22717 = torch.constant.int 1
%25194 = torch.aten.slice.Tensor %25193, %int1_22714, %int0_22715, %int9223372036854775807_22716, %int1_22717 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25194, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22718 = torch.constant.int 0
%25195 = torch.aten.unsqueeze %25194, %int0_22718 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25195, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22719 = torch.constant.int 2
%25196 = torch.aten.unsqueeze %25195, %int2_22719 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25196, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22720 = torch.constant.int 3
%int0_22721 = torch.constant.int 0
%int9223372036854775807_22722 = torch.constant.int 9223372036854775807
%int1_22723 = torch.constant.int 1
%25197 = torch.aten.slice.Tensor %25196, %int3_22720, %int0_22721, %int9223372036854775807_22722, %int1_22723 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25197, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25198 = torch_c.to_builtin_tensor %25015 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_22724 = arith.constant 1 : index
%dim_22725 = tensor.dim %25198, %c1_22724 : tensor<4x?x4x128xf16>
%25199 = flow.tensor.bitcast %25198 : tensor<4x?x4x128xf16>{%dim_22725} -> tensor<4x?x4x64xcomplex<f16>>{%dim_22725}
%25200 = torch_c.from_builtin_tensor %25199 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %25200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%25201 = torch.aten.mul.Tensor %25200, %25197 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %25201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%25202 = torch_c.to_builtin_tensor %25201 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_22726 = arith.constant 1 : index
%dim_22727 = tensor.dim %25202, %c1_22726 : tensor<4x?x4x64xcomplex<f32>>
%25203 = flow.tensor.bitcast %25202 : tensor<4x?x4x64xcomplex<f32>>{%dim_22727} -> tensor<4x?x4x128xf32>{%dim_22727}
%25204 = torch_c.from_builtin_tensor %25203 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %25204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_22728 = torch.constant.int 5
%25205 = torch.prims.convert_element_type %25204, %int5_22728 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_22729 = torch.constant.int 131072
%none_22730 = torch.constant.none
%none_22731 = torch.constant.none
%cpu_22732 = torch.constant.device "cpu"
%false_22733 = torch.constant.bool false
%25206 = torch.aten.arange %int131072_22729, %none_22730, %none_22731, %cpu_22732, %false_22733 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_22734 = torch.constant.int 0
%int128_22735 = torch.constant.int 128
%int2_22736 = torch.constant.int 2
%none_22737 = torch.constant.none
%none_22738 = torch.constant.none
%cpu_22739 = torch.constant.device "cpu"
%false_22740 = torch.constant.bool false
%25207 = torch.aten.arange.start_step %int0_22734, %int128_22735, %int2_22736, %none_22737, %none_22738, %cpu_22739, %false_22740 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_22741 = torch.constant.int 0
%int0_22742 = torch.constant.int 0
%int64_22743 = torch.constant.int 64
%int1_22744 = torch.constant.int 1
%25208 = torch.aten.slice.Tensor %25207, %int0_22741, %int0_22742, %int64_22743, %int1_22744 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_22745 = torch.constant.int 6
%25209 = torch.prims.convert_element_type %25208, %int6_22745 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_22746 = torch.constant.int 128
%25210 = torch.aten.div.Scalar %25209, %int128_22746 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_22747 = torch.constant.float 5.000000e+05
%25211 = torch.aten.pow.Scalar %float5.000000e05_22747, %25210 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%25212 = torch.aten.reciprocal %25211 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_22748 = torch.constant.float 1.000000e+00
%25213 = torch.aten.mul.Scalar %25212, %float1.000000e00_22748 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_22749 = torch.constant.int 131072
%int1_22750 = torch.constant.int 1
%25214 = torch.prim.ListConstruct %int131072_22749, %int1_22750 : (!torch.int, !torch.int) -> !torch.list<int>
%25215 = torch.aten.view %25206, %25214 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%25216 = torch.aten.mul.Tensor %25215, %25213 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%25217 = torch.aten.cos %25216 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%25218 = torch.aten.sin %25216 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%25219 = torch.aten.complex %25217, %25218 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%25220 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25221 = flow.tensor.transfer %25220 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%25222 = torch_c.from_builtin_tensor %25221 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25223 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25224 = flow.tensor.transfer %25223 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%25225 = torch_c.from_builtin_tensor %25224 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25226 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25227 = flow.tensor.transfer %25226 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%25228 = torch_c.from_builtin_tensor %25227 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25229 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25230 = flow.tensor.transfer %25229 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%25231 = torch_c.from_builtin_tensor %25230 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25232 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25233 = flow.tensor.transfer %25232 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%25234 = torch_c.from_builtin_tensor %25233 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25235 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25236 = flow.tensor.transfer %25235 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%25237 = torch_c.from_builtin_tensor %25236 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25238 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25239 = flow.tensor.transfer %25238 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%25240 = torch_c.from_builtin_tensor %25239 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%25241 = torch_c.to_builtin_tensor %25219 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%25242 = flow.tensor.transfer %25241 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%25243 = torch_c.from_builtin_tensor %25242 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_22751 = torch.constant.int 1
%25244 = torch.aten.size.int %24893, %int1_22751 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22752 = torch.constant.int 0
%25245 = torch.aten.add.int %int0_22752, %25244 : !torch.int, !torch.int -> !torch.int
%int0_22753 = torch.constant.int 0
%int0_22754 = torch.constant.int 0
%int1_22755 = torch.constant.int 1
%25246 = torch.aten.slice.Tensor %25222, %int0_22753, %int0_22754, %25245, %int1_22755 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25246, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22756 = torch.constant.int 1
%int0_22757 = torch.constant.int 0
%int9223372036854775807_22758 = torch.constant.int 9223372036854775807
%int1_22759 = torch.constant.int 1
%25247 = torch.aten.slice.Tensor %25246, %int1_22756, %int0_22757, %int9223372036854775807_22758, %int1_22759 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25247, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22760 = torch.constant.int 0
%25248 = torch.aten.unsqueeze %25247, %int0_22760 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25248, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22761 = torch.constant.int 2
%25249 = torch.aten.unsqueeze %25248, %int2_22761 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25249, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22762 = torch.constant.int 3
%int0_22763 = torch.constant.int 0
%int9223372036854775807_22764 = torch.constant.int 9223372036854775807
%int1_22765 = torch.constant.int 1
%25250 = torch.aten.slice.Tensor %25249, %int3_22762, %int0_22763, %int9223372036854775807_22764, %int1_22765 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25250, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25251 = torch_c.to_builtin_tensor %25017 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22766 = arith.constant 1 : index
%dim_22767 = tensor.dim %25251, %c1_22766 : tensor<4x?x1x128xf16>
%25252 = flow.tensor.bitcast %25251 : tensor<4x?x1x128xf16>{%dim_22767} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22767}
%25253 = torch_c.from_builtin_tensor %25252 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25254 = torch.aten.mul.Tensor %25253, %25250 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25255 = torch_c.to_builtin_tensor %25254 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22768 = arith.constant 1 : index
%dim_22769 = tensor.dim %25255, %c1_22768 : tensor<4x?x1x64xcomplex<f32>>
%25256 = flow.tensor.bitcast %25255 : tensor<4x?x1x64xcomplex<f32>>{%dim_22769} -> tensor<4x?x1x128xf32>{%dim_22769}
%25257 = torch_c.from_builtin_tensor %25256 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22770 = torch.constant.int 5
%25258 = torch.prims.convert_element_type %25257, %int5_22770 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22771 = torch.constant.int 1
%25259 = torch.aten.size.int %24899, %int1_22771 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22772 = torch.constant.int 0
%25260 = torch.aten.add.int %int0_22772, %25259 : !torch.int, !torch.int -> !torch.int
%int0_22773 = torch.constant.int 0
%int0_22774 = torch.constant.int 0
%int1_22775 = torch.constant.int 1
%25261 = torch.aten.slice.Tensor %25225, %int0_22773, %int0_22774, %25260, %int1_22775 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25261, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22776 = torch.constant.int 1
%int0_22777 = torch.constant.int 0
%int9223372036854775807_22778 = torch.constant.int 9223372036854775807
%int1_22779 = torch.constant.int 1
%25262 = torch.aten.slice.Tensor %25261, %int1_22776, %int0_22777, %int9223372036854775807_22778, %int1_22779 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25262, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22780 = torch.constant.int 0
%25263 = torch.aten.unsqueeze %25262, %int0_22780 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25263, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22781 = torch.constant.int 2
%25264 = torch.aten.unsqueeze %25263, %int2_22781 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25264, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22782 = torch.constant.int 3
%int0_22783 = torch.constant.int 0
%int9223372036854775807_22784 = torch.constant.int 9223372036854775807
%int1_22785 = torch.constant.int 1
%25265 = torch.aten.slice.Tensor %25264, %int3_22782, %int0_22783, %int9223372036854775807_22784, %int1_22785 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25265, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25266 = torch_c.to_builtin_tensor %25019 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22786 = arith.constant 1 : index
%dim_22787 = tensor.dim %25266, %c1_22786 : tensor<4x?x1x128xf16>
%25267 = flow.tensor.bitcast %25266 : tensor<4x?x1x128xf16>{%dim_22787} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22787}
%25268 = torch_c.from_builtin_tensor %25267 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25269 = torch.aten.mul.Tensor %25268, %25265 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25270 = torch_c.to_builtin_tensor %25269 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22788 = arith.constant 1 : index
%dim_22789 = tensor.dim %25270, %c1_22788 : tensor<4x?x1x64xcomplex<f32>>
%25271 = flow.tensor.bitcast %25270 : tensor<4x?x1x64xcomplex<f32>>{%dim_22789} -> tensor<4x?x1x128xf32>{%dim_22789}
%25272 = torch_c.from_builtin_tensor %25271 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22790 = torch.constant.int 5
%25273 = torch.prims.convert_element_type %25272, %int5_22790 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22791 = torch.constant.int 1
%25274 = torch.aten.size.int %24905, %int1_22791 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22792 = torch.constant.int 0
%25275 = torch.aten.add.int %int0_22792, %25274 : !torch.int, !torch.int -> !torch.int
%int0_22793 = torch.constant.int 0
%int0_22794 = torch.constant.int 0
%int1_22795 = torch.constant.int 1
%25276 = torch.aten.slice.Tensor %25228, %int0_22793, %int0_22794, %25275, %int1_22795 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25276, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22796 = torch.constant.int 1
%int0_22797 = torch.constant.int 0
%int9223372036854775807_22798 = torch.constant.int 9223372036854775807
%int1_22799 = torch.constant.int 1
%25277 = torch.aten.slice.Tensor %25276, %int1_22796, %int0_22797, %int9223372036854775807_22798, %int1_22799 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25277, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22800 = torch.constant.int 0
%25278 = torch.aten.unsqueeze %25277, %int0_22800 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25278, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22801 = torch.constant.int 2
%25279 = torch.aten.unsqueeze %25278, %int2_22801 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25279, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22802 = torch.constant.int 3
%int0_22803 = torch.constant.int 0
%int9223372036854775807_22804 = torch.constant.int 9223372036854775807
%int1_22805 = torch.constant.int 1
%25280 = torch.aten.slice.Tensor %25279, %int3_22802, %int0_22803, %int9223372036854775807_22804, %int1_22805 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25280, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25281 = torch_c.to_builtin_tensor %25021 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22806 = arith.constant 1 : index
%dim_22807 = tensor.dim %25281, %c1_22806 : tensor<4x?x1x128xf16>
%25282 = flow.tensor.bitcast %25281 : tensor<4x?x1x128xf16>{%dim_22807} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22807}
%25283 = torch_c.from_builtin_tensor %25282 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25284 = torch.aten.mul.Tensor %25283, %25280 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25285 = torch_c.to_builtin_tensor %25284 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22808 = arith.constant 1 : index
%dim_22809 = tensor.dim %25285, %c1_22808 : tensor<4x?x1x64xcomplex<f32>>
%25286 = flow.tensor.bitcast %25285 : tensor<4x?x1x64xcomplex<f32>>{%dim_22809} -> tensor<4x?x1x128xf32>{%dim_22809}
%25287 = torch_c.from_builtin_tensor %25286 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22810 = torch.constant.int 5
%25288 = torch.prims.convert_element_type %25287, %int5_22810 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22811 = torch.constant.int 1
%25289 = torch.aten.size.int %24911, %int1_22811 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22812 = torch.constant.int 0
%25290 = torch.aten.add.int %int0_22812, %25289 : !torch.int, !torch.int -> !torch.int
%int0_22813 = torch.constant.int 0
%int0_22814 = torch.constant.int 0
%int1_22815 = torch.constant.int 1
%25291 = torch.aten.slice.Tensor %25231, %int0_22813, %int0_22814, %25290, %int1_22815 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25291, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22816 = torch.constant.int 1
%int0_22817 = torch.constant.int 0
%int9223372036854775807_22818 = torch.constant.int 9223372036854775807
%int1_22819 = torch.constant.int 1
%25292 = torch.aten.slice.Tensor %25291, %int1_22816, %int0_22817, %int9223372036854775807_22818, %int1_22819 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25292, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22820 = torch.constant.int 0
%25293 = torch.aten.unsqueeze %25292, %int0_22820 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25293, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22821 = torch.constant.int 2
%25294 = torch.aten.unsqueeze %25293, %int2_22821 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25294, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22822 = torch.constant.int 3
%int0_22823 = torch.constant.int 0
%int9223372036854775807_22824 = torch.constant.int 9223372036854775807
%int1_22825 = torch.constant.int 1
%25295 = torch.aten.slice.Tensor %25294, %int3_22822, %int0_22823, %int9223372036854775807_22824, %int1_22825 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25295, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25296 = torch_c.to_builtin_tensor %25023 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22826 = arith.constant 1 : index
%dim_22827 = tensor.dim %25296, %c1_22826 : tensor<4x?x1x128xf16>
%25297 = flow.tensor.bitcast %25296 : tensor<4x?x1x128xf16>{%dim_22827} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22827}
%25298 = torch_c.from_builtin_tensor %25297 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25299 = torch.aten.mul.Tensor %25298, %25295 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25300 = torch_c.to_builtin_tensor %25299 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22828 = arith.constant 1 : index
%dim_22829 = tensor.dim %25300, %c1_22828 : tensor<4x?x1x64xcomplex<f32>>
%25301 = flow.tensor.bitcast %25300 : tensor<4x?x1x64xcomplex<f32>>{%dim_22829} -> tensor<4x?x1x128xf32>{%dim_22829}
%25302 = torch_c.from_builtin_tensor %25301 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22830 = torch.constant.int 5
%25303 = torch.prims.convert_element_type %25302, %int5_22830 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22831 = torch.constant.int 1
%25304 = torch.aten.size.int %24917, %int1_22831 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22832 = torch.constant.int 0
%25305 = torch.aten.add.int %int0_22832, %25304 : !torch.int, !torch.int -> !torch.int
%int0_22833 = torch.constant.int 0
%int0_22834 = torch.constant.int 0
%int1_22835 = torch.constant.int 1
%25306 = torch.aten.slice.Tensor %25234, %int0_22833, %int0_22834, %25305, %int1_22835 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25306, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22836 = torch.constant.int 1
%int0_22837 = torch.constant.int 0
%int9223372036854775807_22838 = torch.constant.int 9223372036854775807
%int1_22839 = torch.constant.int 1
%25307 = torch.aten.slice.Tensor %25306, %int1_22836, %int0_22837, %int9223372036854775807_22838, %int1_22839 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25307, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22840 = torch.constant.int 0
%25308 = torch.aten.unsqueeze %25307, %int0_22840 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25308, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22841 = torch.constant.int 2
%25309 = torch.aten.unsqueeze %25308, %int2_22841 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25309, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22842 = torch.constant.int 3
%int0_22843 = torch.constant.int 0
%int9223372036854775807_22844 = torch.constant.int 9223372036854775807
%int1_22845 = torch.constant.int 1
%25310 = torch.aten.slice.Tensor %25309, %int3_22842, %int0_22843, %int9223372036854775807_22844, %int1_22845 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25310, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25311 = torch_c.to_builtin_tensor %25025 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22846 = arith.constant 1 : index
%dim_22847 = tensor.dim %25311, %c1_22846 : tensor<4x?x1x128xf16>
%25312 = flow.tensor.bitcast %25311 : tensor<4x?x1x128xf16>{%dim_22847} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22847}
%25313 = torch_c.from_builtin_tensor %25312 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25314 = torch.aten.mul.Tensor %25313, %25310 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25315 = torch_c.to_builtin_tensor %25314 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22848 = arith.constant 1 : index
%dim_22849 = tensor.dim %25315, %c1_22848 : tensor<4x?x1x64xcomplex<f32>>
%25316 = flow.tensor.bitcast %25315 : tensor<4x?x1x64xcomplex<f32>>{%dim_22849} -> tensor<4x?x1x128xf32>{%dim_22849}
%25317 = torch_c.from_builtin_tensor %25316 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22850 = torch.constant.int 5
%25318 = torch.prims.convert_element_type %25317, %int5_22850 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22851 = torch.constant.int 1
%25319 = torch.aten.size.int %24923, %int1_22851 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22852 = torch.constant.int 0
%25320 = torch.aten.add.int %int0_22852, %25319 : !torch.int, !torch.int -> !torch.int
%int0_22853 = torch.constant.int 0
%int0_22854 = torch.constant.int 0
%int1_22855 = torch.constant.int 1
%25321 = torch.aten.slice.Tensor %25237, %int0_22853, %int0_22854, %25320, %int1_22855 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25321, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22856 = torch.constant.int 1
%int0_22857 = torch.constant.int 0
%int9223372036854775807_22858 = torch.constant.int 9223372036854775807
%int1_22859 = torch.constant.int 1
%25322 = torch.aten.slice.Tensor %25321, %int1_22856, %int0_22857, %int9223372036854775807_22858, %int1_22859 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25322, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22860 = torch.constant.int 0
%25323 = torch.aten.unsqueeze %25322, %int0_22860 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25323, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22861 = torch.constant.int 2
%25324 = torch.aten.unsqueeze %25323, %int2_22861 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25324, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22862 = torch.constant.int 3
%int0_22863 = torch.constant.int 0
%int9223372036854775807_22864 = torch.constant.int 9223372036854775807
%int1_22865 = torch.constant.int 1
%25325 = torch.aten.slice.Tensor %25324, %int3_22862, %int0_22863, %int9223372036854775807_22864, %int1_22865 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25325, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25326 = torch_c.to_builtin_tensor %25027 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22866 = arith.constant 1 : index
%dim_22867 = tensor.dim %25326, %c1_22866 : tensor<4x?x1x128xf16>
%25327 = flow.tensor.bitcast %25326 : tensor<4x?x1x128xf16>{%dim_22867} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22867}
%25328 = torch_c.from_builtin_tensor %25327 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25329 = torch.aten.mul.Tensor %25328, %25325 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25330 = torch_c.to_builtin_tensor %25329 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22868 = arith.constant 1 : index
%dim_22869 = tensor.dim %25330, %c1_22868 : tensor<4x?x1x64xcomplex<f32>>
%25331 = flow.tensor.bitcast %25330 : tensor<4x?x1x64xcomplex<f32>>{%dim_22869} -> tensor<4x?x1x128xf32>{%dim_22869}
%25332 = torch_c.from_builtin_tensor %25331 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22870 = torch.constant.int 5
%25333 = torch.prims.convert_element_type %25332, %int5_22870 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22871 = torch.constant.int 1
%25334 = torch.aten.size.int %24929, %int1_22871 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22872 = torch.constant.int 0
%25335 = torch.aten.add.int %int0_22872, %25334 : !torch.int, !torch.int -> !torch.int
%int0_22873 = torch.constant.int 0
%int0_22874 = torch.constant.int 0
%int1_22875 = torch.constant.int 1
%25336 = torch.aten.slice.Tensor %25240, %int0_22873, %int0_22874, %25335, %int1_22875 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25336, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22876 = torch.constant.int 1
%int0_22877 = torch.constant.int 0
%int9223372036854775807_22878 = torch.constant.int 9223372036854775807
%int1_22879 = torch.constant.int 1
%25337 = torch.aten.slice.Tensor %25336, %int1_22876, %int0_22877, %int9223372036854775807_22878, %int1_22879 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25337, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22880 = torch.constant.int 0
%25338 = torch.aten.unsqueeze %25337, %int0_22880 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25338, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22881 = torch.constant.int 2
%25339 = torch.aten.unsqueeze %25338, %int2_22881 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25339, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22882 = torch.constant.int 3
%int0_22883 = torch.constant.int 0
%int9223372036854775807_22884 = torch.constant.int 9223372036854775807
%int1_22885 = torch.constant.int 1
%25340 = torch.aten.slice.Tensor %25339, %int3_22882, %int0_22883, %int9223372036854775807_22884, %int1_22885 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25340, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25341 = torch_c.to_builtin_tensor %25029 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22886 = arith.constant 1 : index
%dim_22887 = tensor.dim %25341, %c1_22886 : tensor<4x?x1x128xf16>
%25342 = flow.tensor.bitcast %25341 : tensor<4x?x1x128xf16>{%dim_22887} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22887}
%25343 = torch_c.from_builtin_tensor %25342 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25344 = torch.aten.mul.Tensor %25343, %25340 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25345 = torch_c.to_builtin_tensor %25344 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22888 = arith.constant 1 : index
%dim_22889 = tensor.dim %25345, %c1_22888 : tensor<4x?x1x64xcomplex<f32>>
%25346 = flow.tensor.bitcast %25345 : tensor<4x?x1x64xcomplex<f32>>{%dim_22889} -> tensor<4x?x1x128xf32>{%dim_22889}
%25347 = torch_c.from_builtin_tensor %25346 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22890 = torch.constant.int 5
%25348 = torch.prims.convert_element_type %25347, %int5_22890 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_22891 = torch.constant.int 1
%25349 = torch.aten.size.int %24935, %int1_22891 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_22892 = torch.constant.int 0
%25350 = torch.aten.add.int %int0_22892, %25349 : !torch.int, !torch.int -> !torch.int
%int0_22893 = torch.constant.int 0
%int0_22894 = torch.constant.int 0
%int1_22895 = torch.constant.int 1
%25351 = torch.aten.slice.Tensor %25243, %int0_22893, %int0_22894, %25350, %int1_22895 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25351, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_22896 = torch.constant.int 1
%int0_22897 = torch.constant.int 0
%int9223372036854775807_22898 = torch.constant.int 9223372036854775807
%int1_22899 = torch.constant.int 1
%25352 = torch.aten.slice.Tensor %25351, %int1_22896, %int0_22897, %int9223372036854775807_22898, %int1_22899 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %25352, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_22900 = torch.constant.int 0
%25353 = torch.aten.unsqueeze %25352, %int0_22900 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %25353, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_22901 = torch.constant.int 2
%25354 = torch.aten.unsqueeze %25353, %int2_22901 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25354, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_22902 = torch.constant.int 3
%int0_22903 = torch.constant.int 0
%int9223372036854775807_22904 = torch.constant.int 9223372036854775807
%int1_22905 = torch.constant.int 1
%25355 = torch.aten.slice.Tensor %25354, %int3_22902, %int0_22903, %int9223372036854775807_22904, %int1_22905 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25355, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%25356 = torch_c.to_builtin_tensor %25031 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_22906 = arith.constant 1 : index
%dim_22907 = tensor.dim %25356, %c1_22906 : tensor<4x?x1x128xf16>
%25357 = flow.tensor.bitcast %25356 : tensor<4x?x1x128xf16>{%dim_22907} -> tensor<4x?x1x64xcomplex<f16>>{%dim_22907}
%25358 = torch_c.from_builtin_tensor %25357 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %25358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%25359 = torch.aten.mul.Tensor %25358, %25355 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %25359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%25360 = torch_c.to_builtin_tensor %25359 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_22908 = arith.constant 1 : index
%dim_22909 = tensor.dim %25360, %c1_22908 : tensor<4x?x1x64xcomplex<f32>>
%25361 = flow.tensor.bitcast %25360 : tensor<4x?x1x64xcomplex<f32>>{%dim_22909} -> tensor<4x?x1x128xf32>{%dim_22909}
%25362 = torch_c.from_builtin_tensor %25361 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %25362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_22910 = torch.constant.int 5
%25363 = torch.prims.convert_element_type %25362, %int5_22910 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %25363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_22911 = torch.constant.int 64
%25364 = torch.aten.mul.Scalar %2364, %int64_22911 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25364, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22912 = torch.constant.int 64
%25365 = torch.aten.mul.Scalar %2367, %int64_22912 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25365, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22913 = torch.constant.int 64
%25366 = torch.aten.mul.Scalar %2370, %int64_22913 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25366, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22914 = torch.constant.int 64
%25367 = torch.aten.mul.Scalar %2373, %int64_22914 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25367, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22915 = torch.constant.int 64
%25368 = torch.aten.mul.Scalar %2376, %int64_22915 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25368, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22916 = torch.constant.int 64
%25369 = torch.aten.mul.Scalar %2379, %int64_22916 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25369, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22917 = torch.constant.int 64
%25370 = torch.aten.mul.Scalar %2382, %int64_22917 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25370, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_22918 = torch.constant.int 64
%25371 = torch.aten.mul.Scalar %2385, %int64_22918 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25371, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24 = torch.constant.int 24
%int1_22919 = torch.constant.int 1
%25372 = torch.aten.add.Scalar %25364, %int24, %int1_22919 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25372, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22920 = torch.constant.int 24
%int1_22921 = torch.constant.int 1
%25373 = torch.aten.add.Scalar %25365, %int24_22920, %int1_22921 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25373, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22922 = torch.constant.int 24
%int1_22923 = torch.constant.int 1
%25374 = torch.aten.add.Scalar %25366, %int24_22922, %int1_22923 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25374, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22924 = torch.constant.int 24
%int1_22925 = torch.constant.int 1
%25375 = torch.aten.add.Scalar %25367, %int24_22924, %int1_22925 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25375, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22926 = torch.constant.int 24
%int1_22927 = torch.constant.int 1
%25376 = torch.aten.add.Scalar %25368, %int24_22926, %int1_22927 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25376, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22928 = torch.constant.int 24
%int1_22929 = torch.constant.int 1
%25377 = torch.aten.add.Scalar %25369, %int24_22928, %int1_22929 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25377, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22930 = torch.constant.int 24
%int1_22931 = torch.constant.int 1
%25378 = torch.aten.add.Scalar %25370, %int24_22930, %int1_22931 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25378, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int24_22932 = torch.constant.int 24
%int1_22933 = torch.constant.int 1
%25379 = torch.aten.add.Scalar %25371, %int24_22932, %int1_22933 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25379, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_22934 = torch.constant.int 4
%int16_22935 = torch.constant.int 16
%int1_22936 = torch.constant.int 1
%int128_22937 = torch.constant.int 128
%25380 = torch.prim.ListConstruct %int4_22934, %3095, %int16_22935, %int1_22936, %int128_22937 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25381 = torch.aten.view %25258, %25380 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25381, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22938 = torch.constant.int 4
%int16_22939 = torch.constant.int 16
%int1_22940 = torch.constant.int 1
%int128_22941 = torch.constant.int 128
%25382 = torch.prim.ListConstruct %int4_22938, %3095, %int16_22939, %int1_22940, %int128_22941 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25383 = torch.aten.view %25273, %25382 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25383, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22942 = torch.constant.int 4
%int16_22943 = torch.constant.int 16
%int1_22944 = torch.constant.int 1
%int128_22945 = torch.constant.int 128
%25384 = torch.prim.ListConstruct %int4_22942, %3095, %int16_22943, %int1_22944, %int128_22945 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25385 = torch.aten.view %25288, %25384 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25385, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22946 = torch.constant.int 4
%int16_22947 = torch.constant.int 16
%int1_22948 = torch.constant.int 1
%int128_22949 = torch.constant.int 128
%25386 = torch.prim.ListConstruct %int4_22946, %3095, %int16_22947, %int1_22948, %int128_22949 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25387 = torch.aten.view %25303, %25386 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25387, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22950 = torch.constant.int 4
%int16_22951 = torch.constant.int 16
%int1_22952 = torch.constant.int 1
%int128_22953 = torch.constant.int 128
%25388 = torch.prim.ListConstruct %int4_22950, %3095, %int16_22951, %int1_22952, %int128_22953 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25389 = torch.aten.view %25318, %25388 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25389, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22954 = torch.constant.int 4
%int16_22955 = torch.constant.int 16
%int1_22956 = torch.constant.int 1
%int128_22957 = torch.constant.int 128
%25390 = torch.prim.ListConstruct %int4_22954, %3095, %int16_22955, %int1_22956, %int128_22957 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25391 = torch.aten.view %25333, %25390 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25391, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22958 = torch.constant.int 4
%int16_22959 = torch.constant.int 16
%int1_22960 = torch.constant.int 1
%int128_22961 = torch.constant.int 128
%25392 = torch.prim.ListConstruct %int4_22958, %3095, %int16_22959, %int1_22960, %int128_22961 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25393 = torch.aten.view %25348, %25392 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25393, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22962 = torch.constant.int 4
%int16_22963 = torch.constant.int 16
%int1_22964 = torch.constant.int 1
%int128_22965 = torch.constant.int 128
%25394 = torch.prim.ListConstruct %int4_22962, %3095, %int16_22963, %int1_22964, %int128_22965 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25395 = torch.aten.view %25363, %25394 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25395, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_22966 = torch.constant.int 4
%25396 = torch.aten.mul.int %int4_22966, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22967 = torch.constant.int 16
%int1_22968 = torch.constant.int 1
%int128_22969 = torch.constant.int 128
%25397 = torch.prim.ListConstruct %25396, %int16_22967, %int1_22968, %int128_22969 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25398 = torch.aten.view %25381, %25397 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25398, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22970 = torch.constant.int 4
%25399 = torch.aten.mul.int %int4_22970, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22971 = torch.constant.int 16
%int1_22972 = torch.constant.int 1
%int128_22973 = torch.constant.int 128
%25400 = torch.prim.ListConstruct %25399, %int16_22971, %int1_22972, %int128_22973 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25401 = torch.aten.view %25383, %25400 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25401, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22974 = torch.constant.int 4
%25402 = torch.aten.mul.int %int4_22974, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22975 = torch.constant.int 16
%int1_22976 = torch.constant.int 1
%int128_22977 = torch.constant.int 128
%25403 = torch.prim.ListConstruct %25402, %int16_22975, %int1_22976, %int128_22977 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25404 = torch.aten.view %25385, %25403 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25404, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22978 = torch.constant.int 4
%25405 = torch.aten.mul.int %int4_22978, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22979 = torch.constant.int 16
%int1_22980 = torch.constant.int 1
%int128_22981 = torch.constant.int 128
%25406 = torch.prim.ListConstruct %25405, %int16_22979, %int1_22980, %int128_22981 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25407 = torch.aten.view %25387, %25406 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25407, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22982 = torch.constant.int 4
%25408 = torch.aten.mul.int %int4_22982, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22983 = torch.constant.int 16
%int1_22984 = torch.constant.int 1
%int128_22985 = torch.constant.int 128
%25409 = torch.prim.ListConstruct %25408, %int16_22983, %int1_22984, %int128_22985 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25410 = torch.aten.view %25389, %25409 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25410, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22986 = torch.constant.int 4
%25411 = torch.aten.mul.int %int4_22986, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22987 = torch.constant.int 16
%int1_22988 = torch.constant.int 1
%int128_22989 = torch.constant.int 128
%25412 = torch.prim.ListConstruct %25411, %int16_22987, %int1_22988, %int128_22989 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25413 = torch.aten.view %25391, %25412 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25413, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22990 = torch.constant.int 4
%25414 = torch.aten.mul.int %int4_22990, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22991 = torch.constant.int 16
%int1_22992 = torch.constant.int 1
%int128_22993 = torch.constant.int 128
%25415 = torch.prim.ListConstruct %25414, %int16_22991, %int1_22992, %int128_22993 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25416 = torch.aten.view %25393, %25415 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25416, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22994 = torch.constant.int 4
%25417 = torch.aten.mul.int %int4_22994, %3095 : !torch.int, !torch.int -> !torch.int
%int16_22995 = torch.constant.int 16
%int1_22996 = torch.constant.int 1
%int128_22997 = torch.constant.int 128
%25418 = torch.prim.ListConstruct %25417, %int16_22995, %int1_22996, %int128_22997 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25419 = torch.aten.view %25395, %25418 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25419, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_22998 = torch.constant.int 4
%25420 = torch.aten.mul.int %int4_22998, %3095 : !torch.int, !torch.int -> !torch.int
%25421 = torch.prim.ListConstruct %25420 : (!torch.int) -> !torch.list<int>
%25422 = torch.aten.view %25372, %25421 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25422, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_22999 = torch.constant.int 4
%25423 = torch.aten.mul.int %int4_22999, %3095 : !torch.int, !torch.int -> !torch.int
%25424 = torch.prim.ListConstruct %25423 : (!torch.int) -> !torch.list<int>
%25425 = torch.aten.view %25373, %25424 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25425, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23000 = torch.constant.int 4
%25426 = torch.aten.mul.int %int4_23000, %3095 : !torch.int, !torch.int -> !torch.int
%25427 = torch.prim.ListConstruct %25426 : (!torch.int) -> !torch.list<int>
%25428 = torch.aten.view %25374, %25427 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25428, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23001 = torch.constant.int 4
%25429 = torch.aten.mul.int %int4_23001, %3095 : !torch.int, !torch.int -> !torch.int
%25430 = torch.prim.ListConstruct %25429 : (!torch.int) -> !torch.list<int>
%25431 = torch.aten.view %25375, %25430 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25431, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23002 = torch.constant.int 4
%25432 = torch.aten.mul.int %int4_23002, %3095 : !torch.int, !torch.int -> !torch.int
%25433 = torch.prim.ListConstruct %25432 : (!torch.int) -> !torch.list<int>
%25434 = torch.aten.view %25376, %25433 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25434, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23003 = torch.constant.int 4
%25435 = torch.aten.mul.int %int4_23003, %3095 : !torch.int, !torch.int -> !torch.int
%25436 = torch.prim.ListConstruct %25435 : (!torch.int) -> !torch.list<int>
%25437 = torch.aten.view %25377, %25436 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25437, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23004 = torch.constant.int 4
%25438 = torch.aten.mul.int %int4_23004, %3095 : !torch.int, !torch.int -> !torch.int
%25439 = torch.prim.ListConstruct %25438 : (!torch.int) -> !torch.list<int>
%25440 = torch.aten.view %25378, %25439 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25440, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23005 = torch.constant.int 4
%25441 = torch.aten.mul.int %int4_23005, %3095 : !torch.int, !torch.int -> !torch.int
%25442 = torch.prim.ListConstruct %25441 : (!torch.int) -> !torch.list<int>
%25443 = torch.aten.view %25379, %25442 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25443, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23006 = torch.constant.int 4
%int16_23007 = torch.constant.int 16
%int1_23008 = torch.constant.int 1
%int128_23009 = torch.constant.int 128
%25444 = torch.prim.ListConstruct %int4_23006, %3095, %int16_23007, %int1_23008, %int128_23009 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25445 = torch.aten.view %25033, %25444 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25445, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23010 = torch.constant.int 4
%int16_23011 = torch.constant.int 16
%int1_23012 = torch.constant.int 1
%int128_23013 = torch.constant.int 128
%25446 = torch.prim.ListConstruct %int4_23010, %3095, %int16_23011, %int1_23012, %int128_23013 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25447 = torch.aten.view %25035, %25446 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25447, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23014 = torch.constant.int 4
%int16_23015 = torch.constant.int 16
%int1_23016 = torch.constant.int 1
%int128_23017 = torch.constant.int 128
%25448 = torch.prim.ListConstruct %int4_23014, %3095, %int16_23015, %int1_23016, %int128_23017 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25449 = torch.aten.view %25037, %25448 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25449, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23018 = torch.constant.int 4
%int16_23019 = torch.constant.int 16
%int1_23020 = torch.constant.int 1
%int128_23021 = torch.constant.int 128
%25450 = torch.prim.ListConstruct %int4_23018, %3095, %int16_23019, %int1_23020, %int128_23021 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25451 = torch.aten.view %25039, %25450 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25451, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23022 = torch.constant.int 4
%int16_23023 = torch.constant.int 16
%int1_23024 = torch.constant.int 1
%int128_23025 = torch.constant.int 128
%25452 = torch.prim.ListConstruct %int4_23022, %3095, %int16_23023, %int1_23024, %int128_23025 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25453 = torch.aten.view %25041, %25452 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25453, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23026 = torch.constant.int 4
%int16_23027 = torch.constant.int 16
%int1_23028 = torch.constant.int 1
%int128_23029 = torch.constant.int 128
%25454 = torch.prim.ListConstruct %int4_23026, %3095, %int16_23027, %int1_23028, %int128_23029 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25455 = torch.aten.view %25043, %25454 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25455, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23030 = torch.constant.int 4
%int16_23031 = torch.constant.int 16
%int1_23032 = torch.constant.int 1
%int128_23033 = torch.constant.int 128
%25456 = torch.prim.ListConstruct %int4_23030, %3095, %int16_23031, %int1_23032, %int128_23033 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25457 = torch.aten.view %25045, %25456 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25457, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23034 = torch.constant.int 4
%int16_23035 = torch.constant.int 16
%int1_23036 = torch.constant.int 1
%int128_23037 = torch.constant.int 128
%25458 = torch.prim.ListConstruct %int4_23034, %3095, %int16_23035, %int1_23036, %int128_23037 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25459 = torch.aten.view %25047, %25458 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %25459, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_23038 = torch.constant.int 4
%25460 = torch.aten.mul.int %int4_23038, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23039 = torch.constant.int 16
%int1_23040 = torch.constant.int 1
%int128_23041 = torch.constant.int 128
%25461 = torch.prim.ListConstruct %25460, %int16_23039, %int1_23040, %int128_23041 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25462 = torch.aten.view %25445, %25461 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25462, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23042 = torch.constant.int 4
%25463 = torch.aten.mul.int %int4_23042, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23043 = torch.constant.int 16
%int1_23044 = torch.constant.int 1
%int128_23045 = torch.constant.int 128
%25464 = torch.prim.ListConstruct %25463, %int16_23043, %int1_23044, %int128_23045 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25465 = torch.aten.view %25447, %25464 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25465, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23046 = torch.constant.int 4
%25466 = torch.aten.mul.int %int4_23046, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23047 = torch.constant.int 16
%int1_23048 = torch.constant.int 1
%int128_23049 = torch.constant.int 128
%25467 = torch.prim.ListConstruct %25466, %int16_23047, %int1_23048, %int128_23049 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25468 = torch.aten.view %25449, %25467 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25468, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23050 = torch.constant.int 4
%25469 = torch.aten.mul.int %int4_23050, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23051 = torch.constant.int 16
%int1_23052 = torch.constant.int 1
%int128_23053 = torch.constant.int 128
%25470 = torch.prim.ListConstruct %25469, %int16_23051, %int1_23052, %int128_23053 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25471 = torch.aten.view %25451, %25470 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25471, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23054 = torch.constant.int 4
%25472 = torch.aten.mul.int %int4_23054, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23055 = torch.constant.int 16
%int1_23056 = torch.constant.int 1
%int128_23057 = torch.constant.int 128
%25473 = torch.prim.ListConstruct %25472, %int16_23055, %int1_23056, %int128_23057 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25474 = torch.aten.view %25453, %25473 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25474, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23058 = torch.constant.int 4
%25475 = torch.aten.mul.int %int4_23058, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23059 = torch.constant.int 16
%int1_23060 = torch.constant.int 1
%int128_23061 = torch.constant.int 128
%25476 = torch.prim.ListConstruct %25475, %int16_23059, %int1_23060, %int128_23061 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25477 = torch.aten.view %25455, %25476 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25477, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23062 = torch.constant.int 4
%25478 = torch.aten.mul.int %int4_23062, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23063 = torch.constant.int 16
%int1_23064 = torch.constant.int 1
%int128_23065 = torch.constant.int 128
%25479 = torch.prim.ListConstruct %25478, %int16_23063, %int1_23064, %int128_23065 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25480 = torch.aten.view %25457, %25479 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25480, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_23066 = torch.constant.int 4
%25481 = torch.aten.mul.int %int4_23066, %3095 : !torch.int, !torch.int -> !torch.int
%int16_23067 = torch.constant.int 16
%int1_23068 = torch.constant.int 1
%int128_23069 = torch.constant.int 128
%25482 = torch.prim.ListConstruct %25481, %int16_23067, %int1_23068, %int128_23069 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25483 = torch.aten.view %25459, %25482 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25483, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_23070 = torch.constant.int 1
%int1_23071 = torch.constant.int 1
%25484 = torch.aten.add.Scalar %25372, %int1_23070, %int1_23071 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25484, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23072 = torch.constant.int 1
%int1_23073 = torch.constant.int 1
%25485 = torch.aten.add.Scalar %25373, %int1_23072, %int1_23073 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25485, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23074 = torch.constant.int 1
%int1_23075 = torch.constant.int 1
%25486 = torch.aten.add.Scalar %25374, %int1_23074, %int1_23075 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25486, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23076 = torch.constant.int 1
%int1_23077 = torch.constant.int 1
%25487 = torch.aten.add.Scalar %25375, %int1_23076, %int1_23077 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25487, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23078 = torch.constant.int 1
%int1_23079 = torch.constant.int 1
%25488 = torch.aten.add.Scalar %25376, %int1_23078, %int1_23079 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25488, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23080 = torch.constant.int 1
%int1_23081 = torch.constant.int 1
%25489 = torch.aten.add.Scalar %25377, %int1_23080, %int1_23081 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25489, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23082 = torch.constant.int 1
%int1_23083 = torch.constant.int 1
%25490 = torch.aten.add.Scalar %25378, %int1_23082, %int1_23083 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25490, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_23084 = torch.constant.int 1
%int1_23085 = torch.constant.int 1
%25491 = torch.aten.add.Scalar %25379, %int1_23084, %int1_23085 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %25491, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_23086 = torch.constant.int 4
%25492 = torch.aten.mul.int %int4_23086, %3095 : !torch.int, !torch.int -> !torch.int
%25493 = torch.prim.ListConstruct %25492 : (!torch.int) -> !torch.list<int>
%25494 = torch.aten.view %25484, %25493 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25494, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23087 = torch.constant.int 4
%25495 = torch.aten.mul.int %int4_23087, %3095 : !torch.int, !torch.int -> !torch.int
%25496 = torch.prim.ListConstruct %25495 : (!torch.int) -> !torch.list<int>
%25497 = torch.aten.view %25485, %25496 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25497, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23088 = torch.constant.int 4
%25498 = torch.aten.mul.int %int4_23088, %3095 : !torch.int, !torch.int -> !torch.int
%25499 = torch.prim.ListConstruct %25498 : (!torch.int) -> !torch.list<int>
%25500 = torch.aten.view %25486, %25499 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25500, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23089 = torch.constant.int 4
%25501 = torch.aten.mul.int %int4_23089, %3095 : !torch.int, !torch.int -> !torch.int
%25502 = torch.prim.ListConstruct %25501 : (!torch.int) -> !torch.list<int>
%25503 = torch.aten.view %25487, %25502 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25503, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23090 = torch.constant.int 4
%25504 = torch.aten.mul.int %int4_23090, %3095 : !torch.int, !torch.int -> !torch.int
%25505 = torch.prim.ListConstruct %25504 : (!torch.int) -> !torch.list<int>
%25506 = torch.aten.view %25488, %25505 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25506, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23091 = torch.constant.int 4
%25507 = torch.aten.mul.int %int4_23091, %3095 : !torch.int, !torch.int -> !torch.int
%25508 = torch.prim.ListConstruct %25507 : (!torch.int) -> !torch.list<int>
%25509 = torch.aten.view %25489, %25508 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25509, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23092 = torch.constant.int 4
%25510 = torch.aten.mul.int %int4_23092, %3095 : !torch.int, !torch.int -> !torch.int
%25511 = torch.prim.ListConstruct %25510 : (!torch.int) -> !torch.list<int>
%25512 = torch.aten.view %25490, %25511 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25512, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_23093 = torch.constant.int 4
%25513 = torch.aten.mul.int %int4_23093, %3095 : !torch.int, !torch.int -> !torch.int
%25514 = torch.prim.ListConstruct %25513 : (!torch.int) -> !torch.list<int>
%25515 = torch.aten.view %25491, %25514 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25515, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%25516 = torch.prim.ListConstruct %25422, %25494 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23094 = torch.constant.int 0
%25517 = torch.aten.cat %25516, %int0_23094 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25517, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25518 = torch.prim.ListConstruct %25425, %25497 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23095 = torch.constant.int 0
%25519 = torch.aten.cat %25518, %int0_23095 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25519, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25520 = torch.prim.ListConstruct %25428, %25500 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23096 = torch.constant.int 0
%25521 = torch.aten.cat %25520, %int0_23096 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25521, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25522 = torch.prim.ListConstruct %25431, %25503 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23097 = torch.constant.int 0
%25523 = torch.aten.cat %25522, %int0_23097 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25523, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25524 = torch.prim.ListConstruct %25434, %25506 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23098 = torch.constant.int 0
%25525 = torch.aten.cat %25524, %int0_23098 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25525, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25526 = torch.prim.ListConstruct %25437, %25509 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23099 = torch.constant.int 0
%25527 = torch.aten.cat %25526, %int0_23099 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25527, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25528 = torch.prim.ListConstruct %25440, %25512 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23100 = torch.constant.int 0
%25529 = torch.aten.cat %25528, %int0_23100 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25529, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25530 = torch.prim.ListConstruct %25443, %25515 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_23101 = torch.constant.int 0
%25531 = torch.aten.cat %25530, %int0_23101 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %25531, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%25532 = torch.prim.ListConstruct %25398, %25462 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23102 = torch.constant.int 0
%25533 = torch.aten.cat %25532, %int0_23102 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25533, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25534 = torch.prim.ListConstruct %25401, %25465 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23103 = torch.constant.int 0
%25535 = torch.aten.cat %25534, %int0_23103 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25535, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25536 = torch.prim.ListConstruct %25404, %25468 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23104 = torch.constant.int 0
%25537 = torch.aten.cat %25536, %int0_23104 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25537, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25538 = torch.prim.ListConstruct %25407, %25471 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23105 = torch.constant.int 0
%25539 = torch.aten.cat %25538, %int0_23105 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25539, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25540 = torch.prim.ListConstruct %25410, %25474 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23106 = torch.constant.int 0
%25541 = torch.aten.cat %25540, %int0_23106 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25541, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25542 = torch.prim.ListConstruct %25413, %25477 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23107 = torch.constant.int 0
%25543 = torch.aten.cat %25542, %int0_23107 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25543, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25544 = torch.prim.ListConstruct %25416, %25480 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23108 = torch.constant.int 0
%25545 = torch.aten.cat %25544, %int0_23108 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25545, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25546 = torch.prim.ListConstruct %25419, %25483 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_23109 = torch.constant.int 0
%25547 = torch.aten.cat %25546, %int0_23109 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25547, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23110 = torch.constant.int 32
%int2_23111 = torch.constant.int 2
%int16_23112 = torch.constant.int 16
%int1_23113 = torch.constant.int 1
%int128_23114 = torch.constant.int 128
%25548 = torch.prim.ListConstruct %3023, %int32_23110, %int2_23111, %int16_23112, %int1_23113, %int128_23114 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25549 = torch.aten.view %23698, %25548 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25549, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23115 = torch.constant.int 32
%25550 = torch.aten.mul.int %3023, %int32_23115 : !torch.int, !torch.int -> !torch.int
%int2_23116 = torch.constant.int 2
%25551 = torch.aten.mul.int %25550, %int2_23116 : !torch.int, !torch.int -> !torch.int
%int16_23117 = torch.constant.int 16
%int1_23118 = torch.constant.int 1
%int128_23119 = torch.constant.int 128
%25552 = torch.prim.ListConstruct %25551, %int16_23117, %int1_23118, %int128_23119 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25553 = torch.aten.view %25549, %25552 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25553, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25554 = torch.prim.ListConstruct %25517 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23120 = torch.constant.bool false
%25555 = torch.aten.index_put %25553, %25554, %25533, %false_23120 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25555, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23121 = torch.constant.int 32
%int2_23122 = torch.constant.int 2
%int16_23123 = torch.constant.int 16
%int1_23124 = torch.constant.int 1
%int128_23125 = torch.constant.int 128
%25556 = torch.prim.ListConstruct %3023, %int32_23121, %int2_23122, %int16_23123, %int1_23124, %int128_23125 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25557 = torch.aten.view %25555, %25556 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25557, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23126 = torch.constant.int 131072
%25558 = torch.prim.ListConstruct %3023, %int131072_23126 : (!torch.int, !torch.int) -> !torch.list<int>
%25559 = torch.aten.view %25557, %25558 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25559, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23127 = torch.constant.int 32
%int2_23128 = torch.constant.int 2
%int16_23129 = torch.constant.int 16
%int1_23130 = torch.constant.int 1
%int128_23131 = torch.constant.int 128
%25560 = torch.prim.ListConstruct %3026, %int32_23127, %int2_23128, %int16_23129, %int1_23130, %int128_23131 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25561 = torch.aten.view %23710, %25560 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25561, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23132 = torch.constant.int 32
%25562 = torch.aten.mul.int %3026, %int32_23132 : !torch.int, !torch.int -> !torch.int
%int2_23133 = torch.constant.int 2
%25563 = torch.aten.mul.int %25562, %int2_23133 : !torch.int, !torch.int -> !torch.int
%int16_23134 = torch.constant.int 16
%int1_23135 = torch.constant.int 1
%int128_23136 = torch.constant.int 128
%25564 = torch.prim.ListConstruct %25563, %int16_23134, %int1_23135, %int128_23136 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25565 = torch.aten.view %25561, %25564 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25565, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25566 = torch.prim.ListConstruct %25519 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23137 = torch.constant.bool false
%25567 = torch.aten.index_put %25565, %25566, %25535, %false_23137 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25567, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23138 = torch.constant.int 32
%int2_23139 = torch.constant.int 2
%int16_23140 = torch.constant.int 16
%int1_23141 = torch.constant.int 1
%int128_23142 = torch.constant.int 128
%25568 = torch.prim.ListConstruct %3026, %int32_23138, %int2_23139, %int16_23140, %int1_23141, %int128_23142 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25569 = torch.aten.view %25567, %25568 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25569, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23143 = torch.constant.int 131072
%25570 = torch.prim.ListConstruct %3026, %int131072_23143 : (!torch.int, !torch.int) -> !torch.list<int>
%25571 = torch.aten.view %25569, %25570 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25571, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23144 = torch.constant.int 32
%int2_23145 = torch.constant.int 2
%int16_23146 = torch.constant.int 16
%int1_23147 = torch.constant.int 1
%int128_23148 = torch.constant.int 128
%25572 = torch.prim.ListConstruct %3029, %int32_23144, %int2_23145, %int16_23146, %int1_23147, %int128_23148 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25573 = torch.aten.view %23722, %25572 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25573, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23149 = torch.constant.int 32
%25574 = torch.aten.mul.int %3029, %int32_23149 : !torch.int, !torch.int -> !torch.int
%int2_23150 = torch.constant.int 2
%25575 = torch.aten.mul.int %25574, %int2_23150 : !torch.int, !torch.int -> !torch.int
%int16_23151 = torch.constant.int 16
%int1_23152 = torch.constant.int 1
%int128_23153 = torch.constant.int 128
%25576 = torch.prim.ListConstruct %25575, %int16_23151, %int1_23152, %int128_23153 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25577 = torch.aten.view %25573, %25576 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25577, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25578 = torch.prim.ListConstruct %25521 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23154 = torch.constant.bool false
%25579 = torch.aten.index_put %25577, %25578, %25537, %false_23154 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25579, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23155 = torch.constant.int 32
%int2_23156 = torch.constant.int 2
%int16_23157 = torch.constant.int 16
%int1_23158 = torch.constant.int 1
%int128_23159 = torch.constant.int 128
%25580 = torch.prim.ListConstruct %3029, %int32_23155, %int2_23156, %int16_23157, %int1_23158, %int128_23159 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25581 = torch.aten.view %25579, %25580 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25581, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23160 = torch.constant.int 131072
%25582 = torch.prim.ListConstruct %3029, %int131072_23160 : (!torch.int, !torch.int) -> !torch.list<int>
%25583 = torch.aten.view %25581, %25582 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25583, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23161 = torch.constant.int 32
%int2_23162 = torch.constant.int 2
%int16_23163 = torch.constant.int 16
%int1_23164 = torch.constant.int 1
%int128_23165 = torch.constant.int 128
%25584 = torch.prim.ListConstruct %3032, %int32_23161, %int2_23162, %int16_23163, %int1_23164, %int128_23165 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25585 = torch.aten.view %23734, %25584 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25585, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23166 = torch.constant.int 32
%25586 = torch.aten.mul.int %3032, %int32_23166 : !torch.int, !torch.int -> !torch.int
%int2_23167 = torch.constant.int 2
%25587 = torch.aten.mul.int %25586, %int2_23167 : !torch.int, !torch.int -> !torch.int
%int16_23168 = torch.constant.int 16
%int1_23169 = torch.constant.int 1
%int128_23170 = torch.constant.int 128
%25588 = torch.prim.ListConstruct %25587, %int16_23168, %int1_23169, %int128_23170 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25589 = torch.aten.view %25585, %25588 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25589, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25590 = torch.prim.ListConstruct %25523 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23171 = torch.constant.bool false
%25591 = torch.aten.index_put %25589, %25590, %25539, %false_23171 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25591, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23172 = torch.constant.int 32
%int2_23173 = torch.constant.int 2
%int16_23174 = torch.constant.int 16
%int1_23175 = torch.constant.int 1
%int128_23176 = torch.constant.int 128
%25592 = torch.prim.ListConstruct %3032, %int32_23172, %int2_23173, %int16_23174, %int1_23175, %int128_23176 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25593 = torch.aten.view %25591, %25592 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25593, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23177 = torch.constant.int 131072
%25594 = torch.prim.ListConstruct %3032, %int131072_23177 : (!torch.int, !torch.int) -> !torch.list<int>
%25595 = torch.aten.view %25593, %25594 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25595, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23178 = torch.constant.int 32
%int2_23179 = torch.constant.int 2
%int16_23180 = torch.constant.int 16
%int1_23181 = torch.constant.int 1
%int128_23182 = torch.constant.int 128
%25596 = torch.prim.ListConstruct %3035, %int32_23178, %int2_23179, %int16_23180, %int1_23181, %int128_23182 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25597 = torch.aten.view %23746, %25596 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25597, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23183 = torch.constant.int 32
%25598 = torch.aten.mul.int %3035, %int32_23183 : !torch.int, !torch.int -> !torch.int
%int2_23184 = torch.constant.int 2
%25599 = torch.aten.mul.int %25598, %int2_23184 : !torch.int, !torch.int -> !torch.int
%int16_23185 = torch.constant.int 16
%int1_23186 = torch.constant.int 1
%int128_23187 = torch.constant.int 128
%25600 = torch.prim.ListConstruct %25599, %int16_23185, %int1_23186, %int128_23187 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25601 = torch.aten.view %25597, %25600 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25601, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25602 = torch.prim.ListConstruct %25525 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23188 = torch.constant.bool false
%25603 = torch.aten.index_put %25601, %25602, %25541, %false_23188 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25603, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23189 = torch.constant.int 32
%int2_23190 = torch.constant.int 2
%int16_23191 = torch.constant.int 16
%int1_23192 = torch.constant.int 1
%int128_23193 = torch.constant.int 128
%25604 = torch.prim.ListConstruct %3035, %int32_23189, %int2_23190, %int16_23191, %int1_23192, %int128_23193 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25605 = torch.aten.view %25603, %25604 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25605, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23194 = torch.constant.int 131072
%25606 = torch.prim.ListConstruct %3035, %int131072_23194 : (!torch.int, !torch.int) -> !torch.list<int>
%25607 = torch.aten.view %25605, %25606 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25607, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23195 = torch.constant.int 32
%int2_23196 = torch.constant.int 2
%int16_23197 = torch.constant.int 16
%int1_23198 = torch.constant.int 1
%int128_23199 = torch.constant.int 128
%25608 = torch.prim.ListConstruct %3038, %int32_23195, %int2_23196, %int16_23197, %int1_23198, %int128_23199 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25609 = torch.aten.view %23758, %25608 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25609, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23200 = torch.constant.int 32
%25610 = torch.aten.mul.int %3038, %int32_23200 : !torch.int, !torch.int -> !torch.int
%int2_23201 = torch.constant.int 2
%25611 = torch.aten.mul.int %25610, %int2_23201 : !torch.int, !torch.int -> !torch.int
%int16_23202 = torch.constant.int 16
%int1_23203 = torch.constant.int 1
%int128_23204 = torch.constant.int 128
%25612 = torch.prim.ListConstruct %25611, %int16_23202, %int1_23203, %int128_23204 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25613 = torch.aten.view %25609, %25612 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25613, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25614 = torch.prim.ListConstruct %25527 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23205 = torch.constant.bool false
%25615 = torch.aten.index_put %25613, %25614, %25543, %false_23205 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25615, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23206 = torch.constant.int 32
%int2_23207 = torch.constant.int 2
%int16_23208 = torch.constant.int 16
%int1_23209 = torch.constant.int 1
%int128_23210 = torch.constant.int 128
%25616 = torch.prim.ListConstruct %3038, %int32_23206, %int2_23207, %int16_23208, %int1_23209, %int128_23210 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25617 = torch.aten.view %25615, %25616 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25617, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23211 = torch.constant.int 131072
%25618 = torch.prim.ListConstruct %3038, %int131072_23211 : (!torch.int, !torch.int) -> !torch.list<int>
%25619 = torch.aten.view %25617, %25618 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25619, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23212 = torch.constant.int 32
%int2_23213 = torch.constant.int 2
%int16_23214 = torch.constant.int 16
%int1_23215 = torch.constant.int 1
%int128_23216 = torch.constant.int 128
%25620 = torch.prim.ListConstruct %3041, %int32_23212, %int2_23213, %int16_23214, %int1_23215, %int128_23216 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25621 = torch.aten.view %23770, %25620 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25621, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23217 = torch.constant.int 32
%25622 = torch.aten.mul.int %3041, %int32_23217 : !torch.int, !torch.int -> !torch.int
%int2_23218 = torch.constant.int 2
%25623 = torch.aten.mul.int %25622, %int2_23218 : !torch.int, !torch.int -> !torch.int
%int16_23219 = torch.constant.int 16
%int1_23220 = torch.constant.int 1
%int128_23221 = torch.constant.int 128
%25624 = torch.prim.ListConstruct %25623, %int16_23219, %int1_23220, %int128_23221 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25625 = torch.aten.view %25621, %25624 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25625, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25626 = torch.prim.ListConstruct %25529 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23222 = torch.constant.bool false
%25627 = torch.aten.index_put %25625, %25626, %25545, %false_23222 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25627, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23223 = torch.constant.int 32
%int2_23224 = torch.constant.int 2
%int16_23225 = torch.constant.int 16
%int1_23226 = torch.constant.int 1
%int128_23227 = torch.constant.int 128
%25628 = torch.prim.ListConstruct %3041, %int32_23223, %int2_23224, %int16_23225, %int1_23226, %int128_23227 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25629 = torch.aten.view %25627, %25628 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25629, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23228 = torch.constant.int 131072
%25630 = torch.prim.ListConstruct %3041, %int131072_23228 : (!torch.int, !torch.int) -> !torch.list<int>
%25631 = torch.aten.view %25629, %25630 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25631, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_23229 = torch.constant.int 32
%int2_23230 = torch.constant.int 2
%int16_23231 = torch.constant.int 16
%int1_23232 = torch.constant.int 1
%int128_23233 = torch.constant.int 128
%25632 = torch.prim.ListConstruct %3044, %int32_23229, %int2_23230, %int16_23231, %int1_23232, %int128_23233 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25633 = torch.aten.view %23782, %25632 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25633, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_23234 = torch.constant.int 32
%25634 = torch.aten.mul.int %3044, %int32_23234 : !torch.int, !torch.int -> !torch.int
%int2_23235 = torch.constant.int 2
%25635 = torch.aten.mul.int %25634, %int2_23235 : !torch.int, !torch.int -> !torch.int
%int16_23236 = torch.constant.int 16
%int1_23237 = torch.constant.int 1
%int128_23238 = torch.constant.int 128
%25636 = torch.prim.ListConstruct %25635, %int16_23236, %int1_23237, %int128_23238 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25637 = torch.aten.view %25633, %25636 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25637, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%25638 = torch.prim.ListConstruct %25531 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_23239 = torch.constant.bool false
%25639 = torch.aten.index_put %25637, %25638, %25547, %false_23239 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %25639, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_23240 = torch.constant.int 32
%int2_23241 = torch.constant.int 2
%int16_23242 = torch.constant.int 16
%int1_23243 = torch.constant.int 1
%int128_23244 = torch.constant.int 128
%25640 = torch.prim.ListConstruct %3044, %int32_23240, %int2_23241, %int16_23242, %int1_23243, %int128_23244 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25641 = torch.aten.view %25639, %25640 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %25641, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_23245 = torch.constant.int 131072
%25642 = torch.prim.ListConstruct %3044, %int131072_23245 : (!torch.int, !torch.int) -> !torch.list<int>
%25643 = torch.aten.view %25641, %25642 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %25643, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_23246 = torch.constant.int -2
%25644 = torch.aten.unsqueeze %25258, %int-2_23246 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23247 = torch.constant.int -2
%25645 = torch.aten.unsqueeze %25273, %int-2_23247 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23248 = torch.constant.int -2
%25646 = torch.aten.unsqueeze %25288, %int-2_23248 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23249 = torch.constant.int -2
%25647 = torch.aten.unsqueeze %25303, %int-2_23249 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23250 = torch.constant.int -2
%25648 = torch.aten.unsqueeze %25318, %int-2_23250 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23251 = torch.constant.int -2
%25649 = torch.aten.unsqueeze %25333, %int-2_23251 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23252 = torch.constant.int -2
%25650 = torch.aten.unsqueeze %25348, %int-2_23252 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23253 = torch.constant.int -2
%25651 = torch.aten.unsqueeze %25363, %int-2_23253 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_23254 = torch.constant.int 4
%int1_23255 = torch.constant.int 1
%int4_23256 = torch.constant.int 4
%int128_23257 = torch.constant.int 128
%25652 = torch.prim.ListConstruct %int4_23254, %25244, %int1_23255, %int4_23256, %int128_23257 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23258 = torch.constant.bool false
%25653 = torch.aten.expand %25644, %25652, %false_23258 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23259 = torch.constant.int 4
%int1_23260 = torch.constant.int 1
%int4_23261 = torch.constant.int 4
%int128_23262 = torch.constant.int 128
%25654 = torch.prim.ListConstruct %int4_23259, %25244, %int1_23260, %int4_23261, %int128_23262 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23263 = torch.constant.bool false
%25655 = torch.aten.expand %25645, %25654, %false_23263 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23264 = torch.constant.int 4
%int1_23265 = torch.constant.int 1
%int4_23266 = torch.constant.int 4
%int128_23267 = torch.constant.int 128
%25656 = torch.prim.ListConstruct %int4_23264, %25244, %int1_23265, %int4_23266, %int128_23267 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23268 = torch.constant.bool false
%25657 = torch.aten.expand %25646, %25656, %false_23268 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23269 = torch.constant.int 4
%int1_23270 = torch.constant.int 1
%int4_23271 = torch.constant.int 4
%int128_23272 = torch.constant.int 128
%25658 = torch.prim.ListConstruct %int4_23269, %25244, %int1_23270, %int4_23271, %int128_23272 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23273 = torch.constant.bool false
%25659 = torch.aten.expand %25647, %25658, %false_23273 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23274 = torch.constant.int 4
%int1_23275 = torch.constant.int 1
%int4_23276 = torch.constant.int 4
%int128_23277 = torch.constant.int 128
%25660 = torch.prim.ListConstruct %int4_23274, %25244, %int1_23275, %int4_23276, %int128_23277 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23278 = torch.constant.bool false
%25661 = torch.aten.expand %25648, %25660, %false_23278 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23279 = torch.constant.int 4
%int1_23280 = torch.constant.int 1
%int4_23281 = torch.constant.int 4
%int128_23282 = torch.constant.int 128
%25662 = torch.prim.ListConstruct %int4_23279, %25244, %int1_23280, %int4_23281, %int128_23282 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23283 = torch.constant.bool false
%25663 = torch.aten.expand %25649, %25662, %false_23283 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23284 = torch.constant.int 4
%int1_23285 = torch.constant.int 1
%int4_23286 = torch.constant.int 4
%int128_23287 = torch.constant.int 128
%25664 = torch.prim.ListConstruct %int4_23284, %25244, %int1_23285, %int4_23286, %int128_23287 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23288 = torch.constant.bool false
%25665 = torch.aten.expand %25650, %25664, %false_23288 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23289 = torch.constant.int 4
%int1_23290 = torch.constant.int 1
%int4_23291 = torch.constant.int 4
%int128_23292 = torch.constant.int 128
%25666 = torch.prim.ListConstruct %int4_23289, %25244, %int1_23290, %int4_23291, %int128_23292 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23293 = torch.constant.bool false
%25667 = torch.aten.expand %25651, %25666, %false_23293 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23294 = torch.constant.int 4
%int4_23295 = torch.constant.int 4
%int128_23296 = torch.constant.int 128
%25668 = torch.prim.ListConstruct %int4_23294, %25244, %int4_23295, %int128_23296 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25669 = torch.aten.view %25653, %25668 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23297 = torch.constant.int 4
%int4_23298 = torch.constant.int 4
%int128_23299 = torch.constant.int 128
%25670 = torch.prim.ListConstruct %int4_23297, %25244, %int4_23298, %int128_23299 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25671 = torch.aten.view %25655, %25670 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23300 = torch.constant.int 4
%int4_23301 = torch.constant.int 4
%int128_23302 = torch.constant.int 128
%25672 = torch.prim.ListConstruct %int4_23300, %25244, %int4_23301, %int128_23302 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25673 = torch.aten.view %25657, %25672 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23303 = torch.constant.int 4
%int4_23304 = torch.constant.int 4
%int128_23305 = torch.constant.int 128
%25674 = torch.prim.ListConstruct %int4_23303, %25244, %int4_23304, %int128_23305 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25675 = torch.aten.view %25659, %25674 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23306 = torch.constant.int 4
%int4_23307 = torch.constant.int 4
%int128_23308 = torch.constant.int 128
%25676 = torch.prim.ListConstruct %int4_23306, %25244, %int4_23307, %int128_23308 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25677 = torch.aten.view %25661, %25676 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23309 = torch.constant.int 4
%int4_23310 = torch.constant.int 4
%int128_23311 = torch.constant.int 128
%25678 = torch.prim.ListConstruct %int4_23309, %25244, %int4_23310, %int128_23311 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25679 = torch.aten.view %25663, %25678 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23312 = torch.constant.int 4
%int4_23313 = torch.constant.int 4
%int128_23314 = torch.constant.int 128
%25680 = torch.prim.ListConstruct %int4_23312, %25244, %int4_23313, %int128_23314 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25681 = torch.aten.view %25665, %25680 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23315 = torch.constant.int 4
%int4_23316 = torch.constant.int 4
%int128_23317 = torch.constant.int 128
%25682 = torch.prim.ListConstruct %int4_23315, %25244, %int4_23316, %int128_23317 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25683 = torch.aten.view %25667, %25682 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_23318 = torch.constant.int -2
%25684 = torch.aten.unsqueeze %25033, %int-2_23318 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23319 = torch.constant.int -2
%25685 = torch.aten.unsqueeze %25035, %int-2_23319 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23320 = torch.constant.int -2
%25686 = torch.aten.unsqueeze %25037, %int-2_23320 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23321 = torch.constant.int -2
%25687 = torch.aten.unsqueeze %25039, %int-2_23321 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23322 = torch.constant.int -2
%25688 = torch.aten.unsqueeze %25041, %int-2_23322 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23323 = torch.constant.int -2
%25689 = torch.aten.unsqueeze %25043, %int-2_23323 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23324 = torch.constant.int -2
%25690 = torch.aten.unsqueeze %25045, %int-2_23324 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_23325 = torch.constant.int -2
%25691 = torch.aten.unsqueeze %25047, %int-2_23325 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %25691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_23326 = torch.constant.int 1
%25692 = torch.aten.size.int %24957, %int1_23326 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_23327 = torch.constant.int 4
%int1_23328 = torch.constant.int 1
%int4_23329 = torch.constant.int 4
%int128_23330 = torch.constant.int 128
%25693 = torch.prim.ListConstruct %int4_23327, %25692, %int1_23328, %int4_23329, %int128_23330 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23331 = torch.constant.bool false
%25694 = torch.aten.expand %25684, %25693, %false_23331 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23332 = torch.constant.int 4
%int1_23333 = torch.constant.int 1
%int4_23334 = torch.constant.int 4
%int128_23335 = torch.constant.int 128
%25695 = torch.prim.ListConstruct %int4_23332, %25692, %int1_23333, %int4_23334, %int128_23335 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23336 = torch.constant.bool false
%25696 = torch.aten.expand %25685, %25695, %false_23336 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23337 = torch.constant.int 4
%int1_23338 = torch.constant.int 1
%int4_23339 = torch.constant.int 4
%int128_23340 = torch.constant.int 128
%25697 = torch.prim.ListConstruct %int4_23337, %25692, %int1_23338, %int4_23339, %int128_23340 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23341 = torch.constant.bool false
%25698 = torch.aten.expand %25686, %25697, %false_23341 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23342 = torch.constant.int 4
%int1_23343 = torch.constant.int 1
%int4_23344 = torch.constant.int 4
%int128_23345 = torch.constant.int 128
%25699 = torch.prim.ListConstruct %int4_23342, %25692, %int1_23343, %int4_23344, %int128_23345 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23346 = torch.constant.bool false
%25700 = torch.aten.expand %25687, %25699, %false_23346 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23347 = torch.constant.int 4
%int1_23348 = torch.constant.int 1
%int4_23349 = torch.constant.int 4
%int128_23350 = torch.constant.int 128
%25701 = torch.prim.ListConstruct %int4_23347, %25692, %int1_23348, %int4_23349, %int128_23350 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23351 = torch.constant.bool false
%25702 = torch.aten.expand %25688, %25701, %false_23351 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23352 = torch.constant.int 4
%int1_23353 = torch.constant.int 1
%int4_23354 = torch.constant.int 4
%int128_23355 = torch.constant.int 128
%25703 = torch.prim.ListConstruct %int4_23352, %25692, %int1_23353, %int4_23354, %int128_23355 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23356 = torch.constant.bool false
%25704 = torch.aten.expand %25689, %25703, %false_23356 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23357 = torch.constant.int 4
%int1_23358 = torch.constant.int 1
%int4_23359 = torch.constant.int 4
%int128_23360 = torch.constant.int 128
%25705 = torch.prim.ListConstruct %int4_23357, %25692, %int1_23358, %int4_23359, %int128_23360 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23361 = torch.constant.bool false
%25706 = torch.aten.expand %25690, %25705, %false_23361 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23362 = torch.constant.int 4
%int1_23363 = torch.constant.int 1
%int4_23364 = torch.constant.int 4
%int128_23365 = torch.constant.int 128
%25707 = torch.prim.ListConstruct %int4_23362, %25692, %int1_23363, %int4_23364, %int128_23365 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_23366 = torch.constant.bool false
%25708 = torch.aten.expand %25691, %25707, %false_23366 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %25708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_23367 = torch.constant.int 4
%int4_23368 = torch.constant.int 4
%int128_23369 = torch.constant.int 128
%25709 = torch.prim.ListConstruct %int4_23367, %25692, %int4_23368, %int128_23369 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25710 = torch.aten.view %25694, %25709 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23370 = torch.constant.int 4
%int4_23371 = torch.constant.int 4
%int128_23372 = torch.constant.int 128
%25711 = torch.prim.ListConstruct %int4_23370, %25692, %int4_23371, %int128_23372 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25712 = torch.aten.view %25696, %25711 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23373 = torch.constant.int 4
%int4_23374 = torch.constant.int 4
%int128_23375 = torch.constant.int 128
%25713 = torch.prim.ListConstruct %int4_23373, %25692, %int4_23374, %int128_23375 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25714 = torch.aten.view %25698, %25713 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23376 = torch.constant.int 4
%int4_23377 = torch.constant.int 4
%int128_23378 = torch.constant.int 128
%25715 = torch.prim.ListConstruct %int4_23376, %25692, %int4_23377, %int128_23378 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25716 = torch.aten.view %25700, %25715 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23379 = torch.constant.int 4
%int4_23380 = torch.constant.int 4
%int128_23381 = torch.constant.int 128
%25717 = torch.prim.ListConstruct %int4_23379, %25692, %int4_23380, %int128_23381 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25718 = torch.aten.view %25702, %25717 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23382 = torch.constant.int 4
%int4_23383 = torch.constant.int 4
%int128_23384 = torch.constant.int 128
%25719 = torch.prim.ListConstruct %int4_23382, %25692, %int4_23383, %int128_23384 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25720 = torch.aten.view %25704, %25719 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23385 = torch.constant.int 4
%int4_23386 = torch.constant.int 4
%int128_23387 = torch.constant.int 128
%25721 = torch.prim.ListConstruct %int4_23385, %25692, %int4_23386, %int128_23387 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25722 = torch.aten.view %25706, %25721 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23388 = torch.constant.int 4
%int4_23389 = torch.constant.int 4
%int128_23390 = torch.constant.int 128
%25723 = torch.prim.ListConstruct %int4_23388, %25692, %int4_23389, %int128_23390 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25724 = torch.aten.view %25708, %25723 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23391 = torch.constant.int 1
%int2_23392 = torch.constant.int 2
%25725 = torch.aten.transpose.int %25100, %int1_23391, %int2_23392 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25725, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23393 = torch.constant.int 1
%int2_23394 = torch.constant.int 2
%25726 = torch.aten.transpose.int %25115, %int1_23393, %int2_23394 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25726, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23395 = torch.constant.int 1
%int2_23396 = torch.constant.int 2
%25727 = torch.aten.transpose.int %25130, %int1_23395, %int2_23396 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25727, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23397 = torch.constant.int 1
%int2_23398 = torch.constant.int 2
%25728 = torch.aten.transpose.int %25145, %int1_23397, %int2_23398 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25728, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23399 = torch.constant.int 1
%int2_23400 = torch.constant.int 2
%25729 = torch.aten.transpose.int %25160, %int1_23399, %int2_23400 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25729, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23401 = torch.constant.int 1
%int2_23402 = torch.constant.int 2
%25730 = torch.aten.transpose.int %25175, %int1_23401, %int2_23402 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25730, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23403 = torch.constant.int 1
%int2_23404 = torch.constant.int 2
%25731 = torch.aten.transpose.int %25190, %int1_23403, %int2_23404 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25731, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23405 = torch.constant.int 1
%int2_23406 = torch.constant.int 2
%25732 = torch.aten.transpose.int %25205, %int1_23405, %int2_23406 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25732, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23407 = torch.constant.int 1
%int2_23408 = torch.constant.int 2
%25733 = torch.aten.transpose.int %25669, %int1_23407, %int2_23408 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25733, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23409 = torch.constant.int 1
%int2_23410 = torch.constant.int 2
%25734 = torch.aten.transpose.int %25671, %int1_23409, %int2_23410 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25734, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23411 = torch.constant.int 1
%int2_23412 = torch.constant.int 2
%25735 = torch.aten.transpose.int %25673, %int1_23411, %int2_23412 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25735, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23413 = torch.constant.int 1
%int2_23414 = torch.constant.int 2
%25736 = torch.aten.transpose.int %25675, %int1_23413, %int2_23414 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25736, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23415 = torch.constant.int 1
%int2_23416 = torch.constant.int 2
%25737 = torch.aten.transpose.int %25677, %int1_23415, %int2_23416 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25737, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23417 = torch.constant.int 1
%int2_23418 = torch.constant.int 2
%25738 = torch.aten.transpose.int %25679, %int1_23417, %int2_23418 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25738, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23419 = torch.constant.int 1
%int2_23420 = torch.constant.int 2
%25739 = torch.aten.transpose.int %25681, %int1_23419, %int2_23420 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25739, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23421 = torch.constant.int 1
%int2_23422 = torch.constant.int 2
%25740 = torch.aten.transpose.int %25683, %int1_23421, %int2_23422 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25740, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23423 = torch.constant.int 1
%int2_23424 = torch.constant.int 2
%25741 = torch.aten.transpose.int %25710, %int1_23423, %int2_23424 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25741, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23425 = torch.constant.int 1
%int2_23426 = torch.constant.int 2
%25742 = torch.aten.transpose.int %25712, %int1_23425, %int2_23426 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25742, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23427 = torch.constant.int 1
%int2_23428 = torch.constant.int 2
%25743 = torch.aten.transpose.int %25714, %int1_23427, %int2_23428 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25743, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23429 = torch.constant.int 1
%int2_23430 = torch.constant.int 2
%25744 = torch.aten.transpose.int %25716, %int1_23429, %int2_23430 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25744, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23431 = torch.constant.int 1
%int2_23432 = torch.constant.int 2
%25745 = torch.aten.transpose.int %25718, %int1_23431, %int2_23432 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25745, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23433 = torch.constant.int 1
%int2_23434 = torch.constant.int 2
%25746 = torch.aten.transpose.int %25720, %int1_23433, %int2_23434 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25746, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23435 = torch.constant.int 1
%int2_23436 = torch.constant.int 2
%25747 = torch.aten.transpose.int %25722, %int1_23435, %int2_23436 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25747, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23437 = torch.constant.int 1
%int2_23438 = torch.constant.int 2
%25748 = torch.aten.transpose.int %25724, %int1_23437, %int2_23438 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %25748, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23439 = torch.constant.float 0.000000e+00
%true_23440 = torch.constant.bool true
%none_23441 = torch.constant.none
%none_23442 = torch.constant.none
%25749:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25725, %25733, %25741, %float0.000000e00_23439, %true_23440, %none_23441, %none_23442) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25749#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23443 = torch.constant.float 0.000000e+00
%true_23444 = torch.constant.bool true
%none_23445 = torch.constant.none
%none_23446 = torch.constant.none
%25750:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25726, %25734, %25742, %float0.000000e00_23443, %true_23444, %none_23445, %none_23446) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25750#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23447 = torch.constant.float 0.000000e+00
%true_23448 = torch.constant.bool true
%none_23449 = torch.constant.none
%none_23450 = torch.constant.none
%25751:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25727, %25735, %25743, %float0.000000e00_23447, %true_23448, %none_23449, %none_23450) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25751#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23451 = torch.constant.float 0.000000e+00
%true_23452 = torch.constant.bool true
%none_23453 = torch.constant.none
%none_23454 = torch.constant.none
%25752:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25728, %25736, %25744, %float0.000000e00_23451, %true_23452, %none_23453, %none_23454) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25752#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23455 = torch.constant.float 0.000000e+00
%true_23456 = torch.constant.bool true
%none_23457 = torch.constant.none
%none_23458 = torch.constant.none
%25753:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25729, %25737, %25745, %float0.000000e00_23455, %true_23456, %none_23457, %none_23458) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25753#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23459 = torch.constant.float 0.000000e+00
%true_23460 = torch.constant.bool true
%none_23461 = torch.constant.none
%none_23462 = torch.constant.none
%25754:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25730, %25738, %25746, %float0.000000e00_23459, %true_23460, %none_23461, %none_23462) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25754#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23463 = torch.constant.float 0.000000e+00
%true_23464 = torch.constant.bool true
%none_23465 = torch.constant.none
%none_23466 = torch.constant.none
%25755:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25731, %25739, %25747, %float0.000000e00_23463, %true_23464, %none_23465, %none_23466) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25755#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_23467 = torch.constant.float 0.000000e+00
%true_23468 = torch.constant.bool true
%none_23469 = torch.constant.none
%none_23470 = torch.constant.none
%25756:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%25732, %25740, %25748, %float0.000000e00_23467, %true_23468, %none_23469, %none_23470) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %25756#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_23471 = torch.constant.int 1
%int2_23472 = torch.constant.int 2
%25757 = torch.aten.transpose.int %25749#0, %int1_23471, %int2_23472 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23473 = torch.constant.int 1
%int2_23474 = torch.constant.int 2
%25758 = torch.aten.transpose.int %25750#0, %int1_23473, %int2_23474 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23475 = torch.constant.int 1
%int2_23476 = torch.constant.int 2
%25759 = torch.aten.transpose.int %25751#0, %int1_23475, %int2_23476 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23477 = torch.constant.int 1
%int2_23478 = torch.constant.int 2
%25760 = torch.aten.transpose.int %25752#0, %int1_23477, %int2_23478 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23479 = torch.constant.int 1
%int2_23480 = torch.constant.int 2
%25761 = torch.aten.transpose.int %25753#0, %int1_23479, %int2_23480 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23481 = torch.constant.int 1
%int2_23482 = torch.constant.int 2
%25762 = torch.aten.transpose.int %25754#0, %int1_23481, %int2_23482 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23483 = torch.constant.int 1
%int2_23484 = torch.constant.int 2
%25763 = torch.aten.transpose.int %25755#0, %int1_23483, %int2_23484 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_23485 = torch.constant.int 1
%int2_23486 = torch.constant.int 2
%25764 = torch.aten.transpose.int %25756#0, %int1_23485, %int2_23486 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %25764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_23487 = torch.constant.int 4
%int512_23488 = torch.constant.int 512
%25765 = torch.prim.ListConstruct %int4_23487, %25086, %int512_23488 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25766 = torch.aten.view %25757, %25765 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23489 = torch.constant.int 4
%int512_23490 = torch.constant.int 512
%25767 = torch.prim.ListConstruct %int4_23489, %25101, %int512_23490 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25768 = torch.aten.view %25758, %25767 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23491 = torch.constant.int 4
%int512_23492 = torch.constant.int 512
%25769 = torch.prim.ListConstruct %int4_23491, %25116, %int512_23492 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25770 = torch.aten.view %25759, %25769 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23493 = torch.constant.int 4
%int512_23494 = torch.constant.int 512
%25771 = torch.prim.ListConstruct %int4_23493, %25131, %int512_23494 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25772 = torch.aten.view %25760, %25771 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23495 = torch.constant.int 4
%int512_23496 = torch.constant.int 512
%25773 = torch.prim.ListConstruct %int4_23495, %25146, %int512_23496 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25774 = torch.aten.view %25761, %25773 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23497 = torch.constant.int 4
%int512_23498 = torch.constant.int 512
%25775 = torch.prim.ListConstruct %int4_23497, %25161, %int512_23498 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25776 = torch.aten.view %25762, %25775 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23499 = torch.constant.int 4
%int512_23500 = torch.constant.int 512
%25777 = torch.prim.ListConstruct %int4_23499, %25176, %int512_23500 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25778 = torch.aten.view %25763, %25777 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_23501 = torch.constant.int 4
%int512_23502 = torch.constant.int 512
%25779 = torch.prim.ListConstruct %int4_23501, %25191, %int512_23502 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25780 = torch.aten.view %25764, %25779 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %25780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_23503 = torch.constant.int 1
%int0_23504 = torch.constant.int 0
%25781 = torch.prim.ListConstruct %int1_23503, %int0_23504 : (!torch.int, !torch.int) -> !torch.list<int>
%25782 = torch.aten.permute %904, %25781 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23505 = torch.constant.int 1
%int0_23506 = torch.constant.int 0
%25783 = torch.prim.ListConstruct %int1_23505, %int0_23506 : (!torch.int, !torch.int) -> !torch.list<int>
%25784 = torch.aten.permute %905, %25783 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23507 = torch.constant.int 1
%int0_23508 = torch.constant.int 0
%25785 = torch.prim.ListConstruct %int1_23507, %int0_23508 : (!torch.int, !torch.int) -> !torch.list<int>
%25786 = torch.aten.permute %906, %25785 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23509 = torch.constant.int 1
%int0_23510 = torch.constant.int 0
%25787 = torch.prim.ListConstruct %int1_23509, %int0_23510 : (!torch.int, !torch.int) -> !torch.list<int>
%25788 = torch.aten.permute %907, %25787 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23511 = torch.constant.int 1
%int0_23512 = torch.constant.int 0
%25789 = torch.prim.ListConstruct %int1_23511, %int0_23512 : (!torch.int, !torch.int) -> !torch.list<int>
%25790 = torch.aten.permute %908, %25789 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23513 = torch.constant.int 1
%int0_23514 = torch.constant.int 0
%25791 = torch.prim.ListConstruct %int1_23513, %int0_23514 : (!torch.int, !torch.int) -> !torch.list<int>
%25792 = torch.aten.permute %909, %25791 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23515 = torch.constant.int 1
%int0_23516 = torch.constant.int 0
%25793 = torch.prim.ListConstruct %int1_23515, %int0_23516 : (!torch.int, !torch.int) -> !torch.list<int>
%25794 = torch.aten.permute %910, %25793 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_23517 = torch.constant.int 1
%int0_23518 = torch.constant.int 0
%25795 = torch.prim.ListConstruct %int1_23517, %int0_23518 : (!torch.int, !torch.int) -> !torch.list<int>
%25796 = torch.aten.permute %911, %25795 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_23519 = torch.constant.int 4
%25797 = torch.aten.mul.int %int4_23519, %25086 : !torch.int, !torch.int -> !torch.int
%int512_23520 = torch.constant.int 512
%25798 = torch.prim.ListConstruct %25797, %int512_23520 : (!torch.int, !torch.int) -> !torch.list<int>
%25799 = torch.aten.view %25766, %25798 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25799, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25800 = torch.aten.mm %25799, %25782 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25800, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23521 = torch.constant.int 4
%int4096_23522 = torch.constant.int 4096
%25801 = torch.prim.ListConstruct %int4_23521, %25086, %int4096_23522 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25802 = torch.aten.view %25800, %25801 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23523 = torch.constant.int 4
%25803 = torch.aten.mul.int %int4_23523, %25101 : !torch.int, !torch.int -> !torch.int
%int512_23524 = torch.constant.int 512
%25804 = torch.prim.ListConstruct %25803, %int512_23524 : (!torch.int, !torch.int) -> !torch.list<int>
%25805 = torch.aten.view %25768, %25804 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25805, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25806 = torch.aten.mm %25805, %25784 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25806, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23525 = torch.constant.int 4
%int4096_23526 = torch.constant.int 4096
%25807 = torch.prim.ListConstruct %int4_23525, %25101, %int4096_23526 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25808 = torch.aten.view %25806, %25807 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23527 = torch.constant.int 4
%25809 = torch.aten.mul.int %int4_23527, %25116 : !torch.int, !torch.int -> !torch.int
%int512_23528 = torch.constant.int 512
%25810 = torch.prim.ListConstruct %25809, %int512_23528 : (!torch.int, !torch.int) -> !torch.list<int>
%25811 = torch.aten.view %25770, %25810 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25811, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25812 = torch.aten.mm %25811, %25786 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25812, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23529 = torch.constant.int 4
%int4096_23530 = torch.constant.int 4096
%25813 = torch.prim.ListConstruct %int4_23529, %25116, %int4096_23530 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25814 = torch.aten.view %25812, %25813 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23531 = torch.constant.int 4
%25815 = torch.aten.mul.int %int4_23531, %25131 : !torch.int, !torch.int -> !torch.int
%int512_23532 = torch.constant.int 512
%25816 = torch.prim.ListConstruct %25815, %int512_23532 : (!torch.int, !torch.int) -> !torch.list<int>
%25817 = torch.aten.view %25772, %25816 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25817, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25818 = torch.aten.mm %25817, %25788 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25818, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23533 = torch.constant.int 4
%int4096_23534 = torch.constant.int 4096
%25819 = torch.prim.ListConstruct %int4_23533, %25131, %int4096_23534 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25820 = torch.aten.view %25818, %25819 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23535 = torch.constant.int 4
%25821 = torch.aten.mul.int %int4_23535, %25146 : !torch.int, !torch.int -> !torch.int
%int512_23536 = torch.constant.int 512
%25822 = torch.prim.ListConstruct %25821, %int512_23536 : (!torch.int, !torch.int) -> !torch.list<int>
%25823 = torch.aten.view %25774, %25822 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25823, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25824 = torch.aten.mm %25823, %25790 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25824, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23537 = torch.constant.int 4
%int4096_23538 = torch.constant.int 4096
%25825 = torch.prim.ListConstruct %int4_23537, %25146, %int4096_23538 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25826 = torch.aten.view %25824, %25825 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23539 = torch.constant.int 4
%25827 = torch.aten.mul.int %int4_23539, %25161 : !torch.int, !torch.int -> !torch.int
%int512_23540 = torch.constant.int 512
%25828 = torch.prim.ListConstruct %25827, %int512_23540 : (!torch.int, !torch.int) -> !torch.list<int>
%25829 = torch.aten.view %25776, %25828 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25829, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25830 = torch.aten.mm %25829, %25792 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25830, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23541 = torch.constant.int 4
%int4096_23542 = torch.constant.int 4096
%25831 = torch.prim.ListConstruct %int4_23541, %25161, %int4096_23542 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25832 = torch.aten.view %25830, %25831 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23543 = torch.constant.int 4
%25833 = torch.aten.mul.int %int4_23543, %25176 : !torch.int, !torch.int -> !torch.int
%int512_23544 = torch.constant.int 512
%25834 = torch.prim.ListConstruct %25833, %int512_23544 : (!torch.int, !torch.int) -> !torch.list<int>
%25835 = torch.aten.view %25778, %25834 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25835, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25836 = torch.aten.mm %25835, %25794 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25836, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23545 = torch.constant.int 4
%int4096_23546 = torch.constant.int 4096
%25837 = torch.prim.ListConstruct %int4_23545, %25176, %int4096_23546 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25838 = torch.aten.view %25836, %25837 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_23547 = torch.constant.int 4
%25839 = torch.aten.mul.int %int4_23547, %25191 : !torch.int, !torch.int -> !torch.int
%int512_23548 = torch.constant.int 512
%25840 = torch.prim.ListConstruct %25839, %int512_23548 : (!torch.int, !torch.int) -> !torch.list<int>
%25841 = torch.aten.view %25780, %25840 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %25841, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%25842 = torch.aten.mm %25841, %25796 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %25842, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23549 = torch.constant.int 4
%int4096_23550 = torch.constant.int 4096
%25843 = torch.prim.ListConstruct %int4_23549, %25191, %int4096_23550 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%25844 = torch.aten.view %25842, %25843 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25845 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23551 = arith.constant 1 : index
%dim_23552 = tensor.dim %25845, %c1_23551 : tensor<4x?x4096xf16>
%25846 = flow.tensor.transfer %25845 : tensor<4x?x4096xf16>{%dim_23552} to #hal.device.promise<@__device_0>
%25847 = torch_c.from_builtin_tensor %25846 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25848 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23553 = arith.constant 1 : index
%dim_23554 = tensor.dim %25848, %c1_23553 : tensor<4x?x4096xf16>
%25849 = flow.tensor.transfer %25848 : tensor<4x?x4096xf16>{%dim_23554} to #hal.device.promise<@__device_0>
%25850 = torch_c.from_builtin_tensor %25849 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25851 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23555 = arith.constant 1 : index
%dim_23556 = tensor.dim %25851, %c1_23555 : tensor<4x?x4096xf16>
%25852 = flow.tensor.transfer %25851 : tensor<4x?x4096xf16>{%dim_23556} to #hal.device.promise<@__device_0>
%25853 = torch_c.from_builtin_tensor %25852 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25854 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23557 = arith.constant 1 : index
%dim_23558 = tensor.dim %25854, %c1_23557 : tensor<4x?x4096xf16>
%25855 = flow.tensor.transfer %25854 : tensor<4x?x4096xf16>{%dim_23558} to #hal.device.promise<@__device_0>
%25856 = torch_c.from_builtin_tensor %25855 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25857 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23559 = arith.constant 1 : index
%dim_23560 = tensor.dim %25857, %c1_23559 : tensor<4x?x4096xf16>
%25858 = flow.tensor.transfer %25857 : tensor<4x?x4096xf16>{%dim_23560} to #hal.device.promise<@__device_0>
%25859 = torch_c.from_builtin_tensor %25858 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25860 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23561 = arith.constant 1 : index
%dim_23562 = tensor.dim %25860, %c1_23561 : tensor<4x?x4096xf16>
%25861 = flow.tensor.transfer %25860 : tensor<4x?x4096xf16>{%dim_23562} to #hal.device.promise<@__device_0>
%25862 = torch_c.from_builtin_tensor %25861 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25863 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23563 = arith.constant 1 : index
%dim_23564 = tensor.dim %25863, %c1_23563 : tensor<4x?x4096xf16>
%25864 = flow.tensor.transfer %25863 : tensor<4x?x4096xf16>{%dim_23564} to #hal.device.promise<@__device_0>
%25865 = torch_c.from_builtin_tensor %25864 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23565 = torch.constant.int 1
%25866 = torch.aten.add.Tensor %25802, %25847, %int1_23565 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23566 = torch.constant.int 1
%25867 = torch.aten.add.Tensor %25866, %25850, %int1_23566 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23567 = torch.constant.int 1
%25868 = torch.aten.add.Tensor %25867, %25853, %int1_23567 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23568 = torch.constant.int 1
%25869 = torch.aten.add.Tensor %25868, %25856, %int1_23568 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23569 = torch.constant.int 1
%25870 = torch.aten.add.Tensor %25869, %25859, %int1_23569 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23570 = torch.constant.int 1
%25871 = torch.aten.add.Tensor %25870, %25862, %int1_23570 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23571 = torch.constant.int 1
%25872 = torch.aten.add.Tensor %25871, %25865, %int1_23571 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25873 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23572 = arith.constant 1 : index
%dim_23573 = tensor.dim %25873, %c1_23572 : tensor<4x?x4096xf16>
%25874 = flow.tensor.transfer %25873 : tensor<4x?x4096xf16>{%dim_23573} to #hal.device.promise<@__device_1>
%25875 = torch_c.from_builtin_tensor %25874 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25876 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23574 = arith.constant 1 : index
%dim_23575 = tensor.dim %25876, %c1_23574 : tensor<4x?x4096xf16>
%25877 = flow.tensor.transfer %25876 : tensor<4x?x4096xf16>{%dim_23575} to #hal.device.promise<@__device_1>
%25878 = torch_c.from_builtin_tensor %25877 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25879 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23576 = arith.constant 1 : index
%dim_23577 = tensor.dim %25879, %c1_23576 : tensor<4x?x4096xf16>
%25880 = flow.tensor.transfer %25879 : tensor<4x?x4096xf16>{%dim_23577} to #hal.device.promise<@__device_1>
%25881 = torch_c.from_builtin_tensor %25880 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25882 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23578 = arith.constant 1 : index
%dim_23579 = tensor.dim %25882, %c1_23578 : tensor<4x?x4096xf16>
%25883 = flow.tensor.transfer %25882 : tensor<4x?x4096xf16>{%dim_23579} to #hal.device.promise<@__device_1>
%25884 = torch_c.from_builtin_tensor %25883 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25885 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23580 = arith.constant 1 : index
%dim_23581 = tensor.dim %25885, %c1_23580 : tensor<4x?x4096xf16>
%25886 = flow.tensor.transfer %25885 : tensor<4x?x4096xf16>{%dim_23581} to #hal.device.promise<@__device_1>
%25887 = torch_c.from_builtin_tensor %25886 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25888 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23582 = arith.constant 1 : index
%dim_23583 = tensor.dim %25888, %c1_23582 : tensor<4x?x4096xf16>
%25889 = flow.tensor.transfer %25888 : tensor<4x?x4096xf16>{%dim_23583} to #hal.device.promise<@__device_1>
%25890 = torch_c.from_builtin_tensor %25889 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25891 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23584 = arith.constant 1 : index
%dim_23585 = tensor.dim %25891, %c1_23584 : tensor<4x?x4096xf16>
%25892 = flow.tensor.transfer %25891 : tensor<4x?x4096xf16>{%dim_23585} to #hal.device.promise<@__device_1>
%25893 = torch_c.from_builtin_tensor %25892 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23586 = torch.constant.int 1
%25894 = torch.aten.add.Tensor %25875, %25808, %int1_23586 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23587 = torch.constant.int 1
%25895 = torch.aten.add.Tensor %25894, %25878, %int1_23587 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23588 = torch.constant.int 1
%25896 = torch.aten.add.Tensor %25895, %25881, %int1_23588 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23589 = torch.constant.int 1
%25897 = torch.aten.add.Tensor %25896, %25884, %int1_23589 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23590 = torch.constant.int 1
%25898 = torch.aten.add.Tensor %25897, %25887, %int1_23590 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23591 = torch.constant.int 1
%25899 = torch.aten.add.Tensor %25898, %25890, %int1_23591 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23592 = torch.constant.int 1
%25900 = torch.aten.add.Tensor %25899, %25893, %int1_23592 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25901 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23593 = arith.constant 1 : index
%dim_23594 = tensor.dim %25901, %c1_23593 : tensor<4x?x4096xf16>
%25902 = flow.tensor.transfer %25901 : tensor<4x?x4096xf16>{%dim_23594} to #hal.device.promise<@__device_2>
%25903 = torch_c.from_builtin_tensor %25902 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25904 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23595 = arith.constant 1 : index
%dim_23596 = tensor.dim %25904, %c1_23595 : tensor<4x?x4096xf16>
%25905 = flow.tensor.transfer %25904 : tensor<4x?x4096xf16>{%dim_23596} to #hal.device.promise<@__device_2>
%25906 = torch_c.from_builtin_tensor %25905 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25907 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23597 = arith.constant 1 : index
%dim_23598 = tensor.dim %25907, %c1_23597 : tensor<4x?x4096xf16>
%25908 = flow.tensor.transfer %25907 : tensor<4x?x4096xf16>{%dim_23598} to #hal.device.promise<@__device_2>
%25909 = torch_c.from_builtin_tensor %25908 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25910 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23599 = arith.constant 1 : index
%dim_23600 = tensor.dim %25910, %c1_23599 : tensor<4x?x4096xf16>
%25911 = flow.tensor.transfer %25910 : tensor<4x?x4096xf16>{%dim_23600} to #hal.device.promise<@__device_2>
%25912 = torch_c.from_builtin_tensor %25911 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25913 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23601 = arith.constant 1 : index
%dim_23602 = tensor.dim %25913, %c1_23601 : tensor<4x?x4096xf16>
%25914 = flow.tensor.transfer %25913 : tensor<4x?x4096xf16>{%dim_23602} to #hal.device.promise<@__device_2>
%25915 = torch_c.from_builtin_tensor %25914 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25916 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23603 = arith.constant 1 : index
%dim_23604 = tensor.dim %25916, %c1_23603 : tensor<4x?x4096xf16>
%25917 = flow.tensor.transfer %25916 : tensor<4x?x4096xf16>{%dim_23604} to #hal.device.promise<@__device_2>
%25918 = torch_c.from_builtin_tensor %25917 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25919 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23605 = arith.constant 1 : index
%dim_23606 = tensor.dim %25919, %c1_23605 : tensor<4x?x4096xf16>
%25920 = flow.tensor.transfer %25919 : tensor<4x?x4096xf16>{%dim_23606} to #hal.device.promise<@__device_2>
%25921 = torch_c.from_builtin_tensor %25920 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23607 = torch.constant.int 1
%25922 = torch.aten.add.Tensor %25903, %25906, %int1_23607 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23608 = torch.constant.int 1
%25923 = torch.aten.add.Tensor %25922, %25814, %int1_23608 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23609 = torch.constant.int 1
%25924 = torch.aten.add.Tensor %25923, %25909, %int1_23609 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23610 = torch.constant.int 1
%25925 = torch.aten.add.Tensor %25924, %25912, %int1_23610 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23611 = torch.constant.int 1
%25926 = torch.aten.add.Tensor %25925, %25915, %int1_23611 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23612 = torch.constant.int 1
%25927 = torch.aten.add.Tensor %25926, %25918, %int1_23612 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23613 = torch.constant.int 1
%25928 = torch.aten.add.Tensor %25927, %25921, %int1_23613 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25929 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23614 = arith.constant 1 : index
%dim_23615 = tensor.dim %25929, %c1_23614 : tensor<4x?x4096xf16>
%25930 = flow.tensor.transfer %25929 : tensor<4x?x4096xf16>{%dim_23615} to #hal.device.promise<@__device_3>
%25931 = torch_c.from_builtin_tensor %25930 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25932 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23616 = arith.constant 1 : index
%dim_23617 = tensor.dim %25932, %c1_23616 : tensor<4x?x4096xf16>
%25933 = flow.tensor.transfer %25932 : tensor<4x?x4096xf16>{%dim_23617} to #hal.device.promise<@__device_3>
%25934 = torch_c.from_builtin_tensor %25933 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25935 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23618 = arith.constant 1 : index
%dim_23619 = tensor.dim %25935, %c1_23618 : tensor<4x?x4096xf16>
%25936 = flow.tensor.transfer %25935 : tensor<4x?x4096xf16>{%dim_23619} to #hal.device.promise<@__device_3>
%25937 = torch_c.from_builtin_tensor %25936 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25938 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23620 = arith.constant 1 : index
%dim_23621 = tensor.dim %25938, %c1_23620 : tensor<4x?x4096xf16>
%25939 = flow.tensor.transfer %25938 : tensor<4x?x4096xf16>{%dim_23621} to #hal.device.promise<@__device_3>
%25940 = torch_c.from_builtin_tensor %25939 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25941 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23622 = arith.constant 1 : index
%dim_23623 = tensor.dim %25941, %c1_23622 : tensor<4x?x4096xf16>
%25942 = flow.tensor.transfer %25941 : tensor<4x?x4096xf16>{%dim_23623} to #hal.device.promise<@__device_3>
%25943 = torch_c.from_builtin_tensor %25942 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25944 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23624 = arith.constant 1 : index
%dim_23625 = tensor.dim %25944, %c1_23624 : tensor<4x?x4096xf16>
%25945 = flow.tensor.transfer %25944 : tensor<4x?x4096xf16>{%dim_23625} to #hal.device.promise<@__device_3>
%25946 = torch_c.from_builtin_tensor %25945 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25947 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23626 = arith.constant 1 : index
%dim_23627 = tensor.dim %25947, %c1_23626 : tensor<4x?x4096xf16>
%25948 = flow.tensor.transfer %25947 : tensor<4x?x4096xf16>{%dim_23627} to #hal.device.promise<@__device_3>
%25949 = torch_c.from_builtin_tensor %25948 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23628 = torch.constant.int 1
%25950 = torch.aten.add.Tensor %25931, %25934, %int1_23628 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23629 = torch.constant.int 1
%25951 = torch.aten.add.Tensor %25950, %25937, %int1_23629 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23630 = torch.constant.int 1
%25952 = torch.aten.add.Tensor %25951, %25820, %int1_23630 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23631 = torch.constant.int 1
%25953 = torch.aten.add.Tensor %25952, %25940, %int1_23631 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23632 = torch.constant.int 1
%25954 = torch.aten.add.Tensor %25953, %25943, %int1_23632 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23633 = torch.constant.int 1
%25955 = torch.aten.add.Tensor %25954, %25946, %int1_23633 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23634 = torch.constant.int 1
%25956 = torch.aten.add.Tensor %25955, %25949, %int1_23634 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25957 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23635 = arith.constant 1 : index
%dim_23636 = tensor.dim %25957, %c1_23635 : tensor<4x?x4096xf16>
%25958 = flow.tensor.transfer %25957 : tensor<4x?x4096xf16>{%dim_23636} to #hal.device.promise<@__device_4>
%25959 = torch_c.from_builtin_tensor %25958 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25960 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23637 = arith.constant 1 : index
%dim_23638 = tensor.dim %25960, %c1_23637 : tensor<4x?x4096xf16>
%25961 = flow.tensor.transfer %25960 : tensor<4x?x4096xf16>{%dim_23638} to #hal.device.promise<@__device_4>
%25962 = torch_c.from_builtin_tensor %25961 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25963 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23639 = arith.constant 1 : index
%dim_23640 = tensor.dim %25963, %c1_23639 : tensor<4x?x4096xf16>
%25964 = flow.tensor.transfer %25963 : tensor<4x?x4096xf16>{%dim_23640} to #hal.device.promise<@__device_4>
%25965 = torch_c.from_builtin_tensor %25964 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25966 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23641 = arith.constant 1 : index
%dim_23642 = tensor.dim %25966, %c1_23641 : tensor<4x?x4096xf16>
%25967 = flow.tensor.transfer %25966 : tensor<4x?x4096xf16>{%dim_23642} to #hal.device.promise<@__device_4>
%25968 = torch_c.from_builtin_tensor %25967 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25969 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23643 = arith.constant 1 : index
%dim_23644 = tensor.dim %25969, %c1_23643 : tensor<4x?x4096xf16>
%25970 = flow.tensor.transfer %25969 : tensor<4x?x4096xf16>{%dim_23644} to #hal.device.promise<@__device_4>
%25971 = torch_c.from_builtin_tensor %25970 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25972 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23645 = arith.constant 1 : index
%dim_23646 = tensor.dim %25972, %c1_23645 : tensor<4x?x4096xf16>
%25973 = flow.tensor.transfer %25972 : tensor<4x?x4096xf16>{%dim_23646} to #hal.device.promise<@__device_4>
%25974 = torch_c.from_builtin_tensor %25973 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25975 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23647 = arith.constant 1 : index
%dim_23648 = tensor.dim %25975, %c1_23647 : tensor<4x?x4096xf16>
%25976 = flow.tensor.transfer %25975 : tensor<4x?x4096xf16>{%dim_23648} to #hal.device.promise<@__device_4>
%25977 = torch_c.from_builtin_tensor %25976 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23649 = torch.constant.int 1
%25978 = torch.aten.add.Tensor %25959, %25962, %int1_23649 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23650 = torch.constant.int 1
%25979 = torch.aten.add.Tensor %25978, %25965, %int1_23650 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23651 = torch.constant.int 1
%25980 = torch.aten.add.Tensor %25979, %25968, %int1_23651 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23652 = torch.constant.int 1
%25981 = torch.aten.add.Tensor %25980, %25826, %int1_23652 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23653 = torch.constant.int 1
%25982 = torch.aten.add.Tensor %25981, %25971, %int1_23653 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23654 = torch.constant.int 1
%25983 = torch.aten.add.Tensor %25982, %25974, %int1_23654 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23655 = torch.constant.int 1
%25984 = torch.aten.add.Tensor %25983, %25977, %int1_23655 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25985 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23656 = arith.constant 1 : index
%dim_23657 = tensor.dim %25985, %c1_23656 : tensor<4x?x4096xf16>
%25986 = flow.tensor.transfer %25985 : tensor<4x?x4096xf16>{%dim_23657} to #hal.device.promise<@__device_5>
%25987 = torch_c.from_builtin_tensor %25986 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25988 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23658 = arith.constant 1 : index
%dim_23659 = tensor.dim %25988, %c1_23658 : tensor<4x?x4096xf16>
%25989 = flow.tensor.transfer %25988 : tensor<4x?x4096xf16>{%dim_23659} to #hal.device.promise<@__device_5>
%25990 = torch_c.from_builtin_tensor %25989 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25991 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23660 = arith.constant 1 : index
%dim_23661 = tensor.dim %25991, %c1_23660 : tensor<4x?x4096xf16>
%25992 = flow.tensor.transfer %25991 : tensor<4x?x4096xf16>{%dim_23661} to #hal.device.promise<@__device_5>
%25993 = torch_c.from_builtin_tensor %25992 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25994 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23662 = arith.constant 1 : index
%dim_23663 = tensor.dim %25994, %c1_23662 : tensor<4x?x4096xf16>
%25995 = flow.tensor.transfer %25994 : tensor<4x?x4096xf16>{%dim_23663} to #hal.device.promise<@__device_5>
%25996 = torch_c.from_builtin_tensor %25995 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%25997 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23664 = arith.constant 1 : index
%dim_23665 = tensor.dim %25997, %c1_23664 : tensor<4x?x4096xf16>
%25998 = flow.tensor.transfer %25997 : tensor<4x?x4096xf16>{%dim_23665} to #hal.device.promise<@__device_5>
%25999 = torch_c.from_builtin_tensor %25998 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %25999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26000 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23666 = arith.constant 1 : index
%dim_23667 = tensor.dim %26000, %c1_23666 : tensor<4x?x4096xf16>
%26001 = flow.tensor.transfer %26000 : tensor<4x?x4096xf16>{%dim_23667} to #hal.device.promise<@__device_5>
%26002 = torch_c.from_builtin_tensor %26001 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26003 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23668 = arith.constant 1 : index
%dim_23669 = tensor.dim %26003, %c1_23668 : tensor<4x?x4096xf16>
%26004 = flow.tensor.transfer %26003 : tensor<4x?x4096xf16>{%dim_23669} to #hal.device.promise<@__device_5>
%26005 = torch_c.from_builtin_tensor %26004 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23670 = torch.constant.int 1
%26006 = torch.aten.add.Tensor %25987, %25990, %int1_23670 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23671 = torch.constant.int 1
%26007 = torch.aten.add.Tensor %26006, %25993, %int1_23671 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23672 = torch.constant.int 1
%26008 = torch.aten.add.Tensor %26007, %25996, %int1_23672 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23673 = torch.constant.int 1
%26009 = torch.aten.add.Tensor %26008, %25999, %int1_23673 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23674 = torch.constant.int 1
%26010 = torch.aten.add.Tensor %26009, %25832, %int1_23674 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23675 = torch.constant.int 1
%26011 = torch.aten.add.Tensor %26010, %26002, %int1_23675 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23676 = torch.constant.int 1
%26012 = torch.aten.add.Tensor %26011, %26005, %int1_23676 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26013 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23677 = arith.constant 1 : index
%dim_23678 = tensor.dim %26013, %c1_23677 : tensor<4x?x4096xf16>
%26014 = flow.tensor.transfer %26013 : tensor<4x?x4096xf16>{%dim_23678} to #hal.device.promise<@__device_6>
%26015 = torch_c.from_builtin_tensor %26014 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26016 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23679 = arith.constant 1 : index
%dim_23680 = tensor.dim %26016, %c1_23679 : tensor<4x?x4096xf16>
%26017 = flow.tensor.transfer %26016 : tensor<4x?x4096xf16>{%dim_23680} to #hal.device.promise<@__device_6>
%26018 = torch_c.from_builtin_tensor %26017 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26019 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23681 = arith.constant 1 : index
%dim_23682 = tensor.dim %26019, %c1_23681 : tensor<4x?x4096xf16>
%26020 = flow.tensor.transfer %26019 : tensor<4x?x4096xf16>{%dim_23682} to #hal.device.promise<@__device_6>
%26021 = torch_c.from_builtin_tensor %26020 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26022 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23683 = arith.constant 1 : index
%dim_23684 = tensor.dim %26022, %c1_23683 : tensor<4x?x4096xf16>
%26023 = flow.tensor.transfer %26022 : tensor<4x?x4096xf16>{%dim_23684} to #hal.device.promise<@__device_6>
%26024 = torch_c.from_builtin_tensor %26023 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26025 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23685 = arith.constant 1 : index
%dim_23686 = tensor.dim %26025, %c1_23685 : tensor<4x?x4096xf16>
%26026 = flow.tensor.transfer %26025 : tensor<4x?x4096xf16>{%dim_23686} to #hal.device.promise<@__device_6>
%26027 = torch_c.from_builtin_tensor %26026 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26028 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23687 = arith.constant 1 : index
%dim_23688 = tensor.dim %26028, %c1_23687 : tensor<4x?x4096xf16>
%26029 = flow.tensor.transfer %26028 : tensor<4x?x4096xf16>{%dim_23688} to #hal.device.promise<@__device_6>
%26030 = torch_c.from_builtin_tensor %26029 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26031 = torch_c.to_builtin_tensor %25844 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23689 = arith.constant 1 : index
%dim_23690 = tensor.dim %26031, %c1_23689 : tensor<4x?x4096xf16>
%26032 = flow.tensor.transfer %26031 : tensor<4x?x4096xf16>{%dim_23690} to #hal.device.promise<@__device_6>
%26033 = torch_c.from_builtin_tensor %26032 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23691 = torch.constant.int 1
%26034 = torch.aten.add.Tensor %26015, %26018, %int1_23691 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23692 = torch.constant.int 1
%26035 = torch.aten.add.Tensor %26034, %26021, %int1_23692 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23693 = torch.constant.int 1
%26036 = torch.aten.add.Tensor %26035, %26024, %int1_23693 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23694 = torch.constant.int 1
%26037 = torch.aten.add.Tensor %26036, %26027, %int1_23694 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23695 = torch.constant.int 1
%26038 = torch.aten.add.Tensor %26037, %26030, %int1_23695 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23696 = torch.constant.int 1
%26039 = torch.aten.add.Tensor %26038, %25838, %int1_23696 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23697 = torch.constant.int 1
%26040 = torch.aten.add.Tensor %26039, %26033, %int1_23697 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26041 = torch_c.to_builtin_tensor %25802 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23698 = arith.constant 1 : index
%dim_23699 = tensor.dim %26041, %c1_23698 : tensor<4x?x4096xf16>
%26042 = flow.tensor.transfer %26041 : tensor<4x?x4096xf16>{%dim_23699} to #hal.device.promise<@__device_7>
%26043 = torch_c.from_builtin_tensor %26042 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26044 = torch_c.to_builtin_tensor %25808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23700 = arith.constant 1 : index
%dim_23701 = tensor.dim %26044, %c1_23700 : tensor<4x?x4096xf16>
%26045 = flow.tensor.transfer %26044 : tensor<4x?x4096xf16>{%dim_23701} to #hal.device.promise<@__device_7>
%26046 = torch_c.from_builtin_tensor %26045 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26047 = torch_c.to_builtin_tensor %25814 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23702 = arith.constant 1 : index
%dim_23703 = tensor.dim %26047, %c1_23702 : tensor<4x?x4096xf16>
%26048 = flow.tensor.transfer %26047 : tensor<4x?x4096xf16>{%dim_23703} to #hal.device.promise<@__device_7>
%26049 = torch_c.from_builtin_tensor %26048 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26050 = torch_c.to_builtin_tensor %25820 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23704 = arith.constant 1 : index
%dim_23705 = tensor.dim %26050, %c1_23704 : tensor<4x?x4096xf16>
%26051 = flow.tensor.transfer %26050 : tensor<4x?x4096xf16>{%dim_23705} to #hal.device.promise<@__device_7>
%26052 = torch_c.from_builtin_tensor %26051 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26053 = torch_c.to_builtin_tensor %25826 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23706 = arith.constant 1 : index
%dim_23707 = tensor.dim %26053, %c1_23706 : tensor<4x?x4096xf16>
%26054 = flow.tensor.transfer %26053 : tensor<4x?x4096xf16>{%dim_23707} to #hal.device.promise<@__device_7>
%26055 = torch_c.from_builtin_tensor %26054 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26056 = torch_c.to_builtin_tensor %25832 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23708 = arith.constant 1 : index
%dim_23709 = tensor.dim %26056, %c1_23708 : tensor<4x?x4096xf16>
%26057 = flow.tensor.transfer %26056 : tensor<4x?x4096xf16>{%dim_23709} to #hal.device.promise<@__device_7>
%26058 = torch_c.from_builtin_tensor %26057 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26059 = torch_c.to_builtin_tensor %25838 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23710 = arith.constant 1 : index
%dim_23711 = tensor.dim %26059, %c1_23710 : tensor<4x?x4096xf16>
%26060 = flow.tensor.transfer %26059 : tensor<4x?x4096xf16>{%dim_23711} to #hal.device.promise<@__device_7>
%26061 = torch_c.from_builtin_tensor %26060 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23712 = torch.constant.int 1
%26062 = torch.aten.add.Tensor %26043, %26046, %int1_23712 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23713 = torch.constant.int 1
%26063 = torch.aten.add.Tensor %26062, %26049, %int1_23713 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23714 = torch.constant.int 1
%26064 = torch.aten.add.Tensor %26063, %26052, %int1_23714 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23715 = torch.constant.int 1
%26065 = torch.aten.add.Tensor %26064, %26055, %int1_23715 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23716 = torch.constant.int 1
%26066 = torch.aten.add.Tensor %26065, %26058, %int1_23716 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23717 = torch.constant.int 1
%26067 = torch.aten.add.Tensor %26066, %26061, %int1_23717 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23718 = torch.constant.int 1
%26068 = torch.aten.add.Tensor %26067, %25844, %int1_23718 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23719 = torch.constant.int 1
%26069 = torch.aten.add.Tensor %24728, %25872, %int1_23719 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23720 = torch.constant.int 1
%26070 = torch.aten.add.Tensor %24729, %25900, %int1_23720 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23721 = torch.constant.int 1
%26071 = torch.aten.add.Tensor %24730, %25928, %int1_23721 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23722 = torch.constant.int 1
%26072 = torch.aten.add.Tensor %24731, %25956, %int1_23722 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23723 = torch.constant.int 1
%26073 = torch.aten.add.Tensor %24732, %25984, %int1_23723 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23724 = torch.constant.int 1
%26074 = torch.aten.add.Tensor %24733, %26012, %int1_23724 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23725 = torch.constant.int 1
%26075 = torch.aten.add.Tensor %24734, %26040, %int1_23725 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23726 = torch.constant.int 1
%26076 = torch.aten.add.Tensor %24735, %26068, %int1_23726 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_23727 = torch.constant.int 6
%26077 = torch.prims.convert_element_type %26069, %int6_23727 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23728 = torch.constant.int 6
%26078 = torch.prims.convert_element_type %26070, %int6_23728 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23729 = torch.constant.int 6
%26079 = torch.prims.convert_element_type %26071, %int6_23729 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23730 = torch.constant.int 6
%26080 = torch.prims.convert_element_type %26072, %int6_23730 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23731 = torch.constant.int 6
%26081 = torch.prims.convert_element_type %26073, %int6_23731 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23732 = torch.constant.int 6
%26082 = torch.prims.convert_element_type %26074, %int6_23732 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23733 = torch.constant.int 6
%26083 = torch.prims.convert_element_type %26075, %int6_23733 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_23734 = torch.constant.int 6
%26084 = torch.prims.convert_element_type %26076, %int6_23734 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23735 = torch.constant.int 2
%26085 = torch.aten.pow.Tensor_Scalar %26077, %int2_23735 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23736 = torch.constant.int 2
%26086 = torch.aten.pow.Tensor_Scalar %26078, %int2_23736 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23737 = torch.constant.int 2
%26087 = torch.aten.pow.Tensor_Scalar %26079, %int2_23737 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23738 = torch.constant.int 2
%26088 = torch.aten.pow.Tensor_Scalar %26080, %int2_23738 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23739 = torch.constant.int 2
%26089 = torch.aten.pow.Tensor_Scalar %26081, %int2_23739 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23740 = torch.constant.int 2
%26090 = torch.aten.pow.Tensor_Scalar %26082, %int2_23740 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23741 = torch.constant.int 2
%26091 = torch.aten.pow.Tensor_Scalar %26083, %int2_23741 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_23742 = torch.constant.int 2
%26092 = torch.aten.pow.Tensor_Scalar %26084, %int2_23742 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_23743 = torch.constant.int -1
%26093 = torch.prim.ListConstruct %int-1_23743 : (!torch.int) -> !torch.list<int>
%true_23744 = torch.constant.bool true
%none_23745 = torch.constant.none
%26094 = torch.aten.mean.dim %26085, %26093, %true_23744, %none_23745 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23746 = torch.constant.int -1
%26095 = torch.prim.ListConstruct %int-1_23746 : (!torch.int) -> !torch.list<int>
%true_23747 = torch.constant.bool true
%none_23748 = torch.constant.none
%26096 = torch.aten.mean.dim %26086, %26095, %true_23747, %none_23748 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23749 = torch.constant.int -1
%26097 = torch.prim.ListConstruct %int-1_23749 : (!torch.int) -> !torch.list<int>
%true_23750 = torch.constant.bool true
%none_23751 = torch.constant.none
%26098 = torch.aten.mean.dim %26087, %26097, %true_23750, %none_23751 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23752 = torch.constant.int -1
%26099 = torch.prim.ListConstruct %int-1_23752 : (!torch.int) -> !torch.list<int>
%true_23753 = torch.constant.bool true
%none_23754 = torch.constant.none
%26100 = torch.aten.mean.dim %26088, %26099, %true_23753, %none_23754 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23755 = torch.constant.int -1
%26101 = torch.prim.ListConstruct %int-1_23755 : (!torch.int) -> !torch.list<int>
%true_23756 = torch.constant.bool true
%none_23757 = torch.constant.none
%26102 = torch.aten.mean.dim %26089, %26101, %true_23756, %none_23757 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23758 = torch.constant.int -1
%26103 = torch.prim.ListConstruct %int-1_23758 : (!torch.int) -> !torch.list<int>
%true_23759 = torch.constant.bool true
%none_23760 = torch.constant.none
%26104 = torch.aten.mean.dim %26090, %26103, %true_23759, %none_23760 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23761 = torch.constant.int -1
%26105 = torch.prim.ListConstruct %int-1_23761 : (!torch.int) -> !torch.list<int>
%true_23762 = torch.constant.bool true
%none_23763 = torch.constant.none
%26106 = torch.aten.mean.dim %26091, %26105, %true_23762, %none_23763 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_23764 = torch.constant.int -1
%26107 = torch.prim.ListConstruct %int-1_23764 : (!torch.int) -> !torch.list<int>
%true_23765 = torch.constant.bool true
%none_23766 = torch.constant.none
%26108 = torch.aten.mean.dim %26092, %26107, %true_23765, %none_23766 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23767 = torch.constant.float 9.9999997473787516E-6
%int1_23768 = torch.constant.int 1
%26109 = torch.aten.add.Scalar %26094, %float9.999990e-06_23767, %int1_23768 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23769 = torch.constant.float 9.9999997473787516E-6
%int1_23770 = torch.constant.int 1
%26110 = torch.aten.add.Scalar %26096, %float9.999990e-06_23769, %int1_23770 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23771 = torch.constant.float 9.9999997473787516E-6
%int1_23772 = torch.constant.int 1
%26111 = torch.aten.add.Scalar %26098, %float9.999990e-06_23771, %int1_23772 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23773 = torch.constant.float 9.9999997473787516E-6
%int1_23774 = torch.constant.int 1
%26112 = torch.aten.add.Scalar %26100, %float9.999990e-06_23773, %int1_23774 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23775 = torch.constant.float 9.9999997473787516E-6
%int1_23776 = torch.constant.int 1
%26113 = torch.aten.add.Scalar %26102, %float9.999990e-06_23775, %int1_23776 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23777 = torch.constant.float 9.9999997473787516E-6
%int1_23778 = torch.constant.int 1
%26114 = torch.aten.add.Scalar %26104, %float9.999990e-06_23777, %int1_23778 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23779 = torch.constant.float 9.9999997473787516E-6
%int1_23780 = torch.constant.int 1
%26115 = torch.aten.add.Scalar %26106, %float9.999990e-06_23779, %int1_23780 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_23781 = torch.constant.float 9.9999997473787516E-6
%int1_23782 = torch.constant.int 1
%26116 = torch.aten.add.Scalar %26108, %float9.999990e-06_23781, %int1_23782 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26117 = torch.aten.rsqrt %26109 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26118 = torch.aten.rsqrt %26110 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26119 = torch.aten.rsqrt %26111 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26120 = torch.aten.rsqrt %26112 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26121 = torch.aten.rsqrt %26113 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26122 = torch.aten.rsqrt %26114 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26123 = torch.aten.rsqrt %26115 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26124 = torch.aten.rsqrt %26116 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26124, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26125 = torch.aten.mul.Tensor %26077, %26117 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26126 = torch.aten.mul.Tensor %26078, %26118 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26127 = torch.aten.mul.Tensor %26079, %26119 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26128 = torch.aten.mul.Tensor %26080, %26120 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26129 = torch.aten.mul.Tensor %26081, %26121 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26130 = torch.aten.mul.Tensor %26082, %26122 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26131 = torch.aten.mul.Tensor %26083, %26123 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26132 = torch.aten.mul.Tensor %26084, %26124 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26133 = torch.aten.mul.Tensor %912, %26125 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26134 = torch.aten.mul.Tensor %913, %26126 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26135 = torch.aten.mul.Tensor %914, %26127 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26136 = torch.aten.mul.Tensor %915, %26128 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26137 = torch.aten.mul.Tensor %916, %26129 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26138 = torch.aten.mul.Tensor %917, %26130 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26139 = torch.aten.mul.Tensor %918, %26131 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26140 = torch.aten.mul.Tensor %919, %26132 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_23783 = torch.constant.int 5
%26141 = torch.prims.convert_element_type %26133, %int5_23783 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23784 = torch.constant.int 5
%26142 = torch.prims.convert_element_type %26134, %int5_23784 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23785 = torch.constant.int 5
%26143 = torch.prims.convert_element_type %26135, %int5_23785 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23786 = torch.constant.int 5
%26144 = torch.prims.convert_element_type %26136, %int5_23786 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23787 = torch.constant.int 5
%26145 = torch.prims.convert_element_type %26137, %int5_23787 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23788 = torch.constant.int 5
%26146 = torch.prims.convert_element_type %26138, %int5_23788 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23789 = torch.constant.int 5
%26147 = torch.prims.convert_element_type %26139, %int5_23789 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_23790 = torch.constant.int 5
%26148 = torch.prims.convert_element_type %26140, %int5_23790 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23791 = torch.constant.int 1
%int0_23792 = torch.constant.int 0
%26149 = torch.prim.ListConstruct %int1_23791, %int0_23792 : (!torch.int, !torch.int) -> !torch.list<int>
%26150 = torch.aten.permute %920, %26149 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23793 = torch.constant.int 1
%int0_23794 = torch.constant.int 0
%26151 = torch.prim.ListConstruct %int1_23793, %int0_23794 : (!torch.int, !torch.int) -> !torch.list<int>
%26152 = torch.aten.permute %921, %26151 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23795 = torch.constant.int 1
%int0_23796 = torch.constant.int 0
%26153 = torch.prim.ListConstruct %int1_23795, %int0_23796 : (!torch.int, !torch.int) -> !torch.list<int>
%26154 = torch.aten.permute %922, %26153 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23797 = torch.constant.int 1
%int0_23798 = torch.constant.int 0
%26155 = torch.prim.ListConstruct %int1_23797, %int0_23798 : (!torch.int, !torch.int) -> !torch.list<int>
%26156 = torch.aten.permute %923, %26155 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23799 = torch.constant.int 1
%int0_23800 = torch.constant.int 0
%26157 = torch.prim.ListConstruct %int1_23799, %int0_23800 : (!torch.int, !torch.int) -> !torch.list<int>
%26158 = torch.aten.permute %924, %26157 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23801 = torch.constant.int 1
%int0_23802 = torch.constant.int 0
%26159 = torch.prim.ListConstruct %int1_23801, %int0_23802 : (!torch.int, !torch.int) -> !torch.list<int>
%26160 = torch.aten.permute %925, %26159 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23803 = torch.constant.int 1
%int0_23804 = torch.constant.int 0
%26161 = torch.prim.ListConstruct %int1_23803, %int0_23804 : (!torch.int, !torch.int) -> !torch.list<int>
%26162 = torch.aten.permute %926, %26161 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23805 = torch.constant.int 1
%int0_23806 = torch.constant.int 0
%26163 = torch.prim.ListConstruct %int1_23805, %int0_23806 : (!torch.int, !torch.int) -> !torch.list<int>
%26164 = torch.aten.permute %927, %26163 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_23807 = torch.constant.int 4
%26165 = torch.aten.mul.int %int4_23807, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23808 = torch.constant.int 4096
%26166 = torch.prim.ListConstruct %26165, %int4096_23808 : (!torch.int, !torch.int) -> !torch.list<int>
%26167 = torch.aten.view %26141, %26166 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26167, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26168 = torch.aten.mm %26167, %26150 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26168, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23809 = torch.constant.int 4
%int1792_23810 = torch.constant.int 1792
%26169 = torch.prim.ListConstruct %int4_23809, %2482, %int1792_23810 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26170 = torch.aten.view %26168, %26169 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23811 = torch.constant.int 4
%26171 = torch.aten.mul.int %int4_23811, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23812 = torch.constant.int 4096
%26172 = torch.prim.ListConstruct %26171, %int4096_23812 : (!torch.int, !torch.int) -> !torch.list<int>
%26173 = torch.aten.view %26142, %26172 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26173, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26174 = torch.aten.mm %26173, %26152 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26174, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23813 = torch.constant.int 4
%int1792_23814 = torch.constant.int 1792
%26175 = torch.prim.ListConstruct %int4_23813, %2482, %int1792_23814 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26176 = torch.aten.view %26174, %26175 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23815 = torch.constant.int 4
%26177 = torch.aten.mul.int %int4_23815, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23816 = torch.constant.int 4096
%26178 = torch.prim.ListConstruct %26177, %int4096_23816 : (!torch.int, !torch.int) -> !torch.list<int>
%26179 = torch.aten.view %26143, %26178 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26179, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26180 = torch.aten.mm %26179, %26154 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26180, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23817 = torch.constant.int 4
%int1792_23818 = torch.constant.int 1792
%26181 = torch.prim.ListConstruct %int4_23817, %2482, %int1792_23818 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26182 = torch.aten.view %26180, %26181 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23819 = torch.constant.int 4
%26183 = torch.aten.mul.int %int4_23819, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23820 = torch.constant.int 4096
%26184 = torch.prim.ListConstruct %26183, %int4096_23820 : (!torch.int, !torch.int) -> !torch.list<int>
%26185 = torch.aten.view %26144, %26184 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26185, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26186 = torch.aten.mm %26185, %26156 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26186, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23821 = torch.constant.int 4
%int1792_23822 = torch.constant.int 1792
%26187 = torch.prim.ListConstruct %int4_23821, %2482, %int1792_23822 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26188 = torch.aten.view %26186, %26187 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23823 = torch.constant.int 4
%26189 = torch.aten.mul.int %int4_23823, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23824 = torch.constant.int 4096
%26190 = torch.prim.ListConstruct %26189, %int4096_23824 : (!torch.int, !torch.int) -> !torch.list<int>
%26191 = torch.aten.view %26145, %26190 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26191, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26192 = torch.aten.mm %26191, %26158 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26192, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23825 = torch.constant.int 4
%int1792_23826 = torch.constant.int 1792
%26193 = torch.prim.ListConstruct %int4_23825, %2482, %int1792_23826 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26194 = torch.aten.view %26192, %26193 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23827 = torch.constant.int 4
%26195 = torch.aten.mul.int %int4_23827, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23828 = torch.constant.int 4096
%26196 = torch.prim.ListConstruct %26195, %int4096_23828 : (!torch.int, !torch.int) -> !torch.list<int>
%26197 = torch.aten.view %26146, %26196 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26197, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26198 = torch.aten.mm %26197, %26160 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26198, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23829 = torch.constant.int 4
%int1792_23830 = torch.constant.int 1792
%26199 = torch.prim.ListConstruct %int4_23829, %2482, %int1792_23830 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26200 = torch.aten.view %26198, %26199 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23831 = torch.constant.int 4
%26201 = torch.aten.mul.int %int4_23831, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23832 = torch.constant.int 4096
%26202 = torch.prim.ListConstruct %26201, %int4096_23832 : (!torch.int, !torch.int) -> !torch.list<int>
%26203 = torch.aten.view %26147, %26202 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26203, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26204 = torch.aten.mm %26203, %26162 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26204, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23833 = torch.constant.int 4
%int1792_23834 = torch.constant.int 1792
%26205 = torch.prim.ListConstruct %int4_23833, %2482, %int1792_23834 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26206 = torch.aten.view %26204, %26205 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23835 = torch.constant.int 4
%26207 = torch.aten.mul.int %int4_23835, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23836 = torch.constant.int 4096
%26208 = torch.prim.ListConstruct %26207, %int4096_23836 : (!torch.int, !torch.int) -> !torch.list<int>
%26209 = torch.aten.view %26148, %26208 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26209, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26210 = torch.aten.mm %26209, %26164 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26210, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23837 = torch.constant.int 4
%int1792_23838 = torch.constant.int 1792
%26211 = torch.prim.ListConstruct %int4_23837, %2482, %int1792_23838 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26212 = torch.aten.view %26210, %26211 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26213 = torch.aten.silu %26170 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26214 = torch.aten.silu %26176 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26215 = torch.aten.silu %26182 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26216 = torch.aten.silu %26188 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26217 = torch.aten.silu %26194 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26218 = torch.aten.silu %26200 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26219 = torch.aten.silu %26206 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26220 = torch.aten.silu %26212 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_23839 = torch.constant.int 1
%int0_23840 = torch.constant.int 0
%26221 = torch.prim.ListConstruct %int1_23839, %int0_23840 : (!torch.int, !torch.int) -> !torch.list<int>
%26222 = torch.aten.permute %928, %26221 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23841 = torch.constant.int 1
%int0_23842 = torch.constant.int 0
%26223 = torch.prim.ListConstruct %int1_23841, %int0_23842 : (!torch.int, !torch.int) -> !torch.list<int>
%26224 = torch.aten.permute %929, %26223 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23843 = torch.constant.int 1
%int0_23844 = torch.constant.int 0
%26225 = torch.prim.ListConstruct %int1_23843, %int0_23844 : (!torch.int, !torch.int) -> !torch.list<int>
%26226 = torch.aten.permute %930, %26225 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23845 = torch.constant.int 1
%int0_23846 = torch.constant.int 0
%26227 = torch.prim.ListConstruct %int1_23845, %int0_23846 : (!torch.int, !torch.int) -> !torch.list<int>
%26228 = torch.aten.permute %931, %26227 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23847 = torch.constant.int 1
%int0_23848 = torch.constant.int 0
%26229 = torch.prim.ListConstruct %int1_23847, %int0_23848 : (!torch.int, !torch.int) -> !torch.list<int>
%26230 = torch.aten.permute %932, %26229 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23849 = torch.constant.int 1
%int0_23850 = torch.constant.int 0
%26231 = torch.prim.ListConstruct %int1_23849, %int0_23850 : (!torch.int, !torch.int) -> !torch.list<int>
%26232 = torch.aten.permute %933, %26231 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23851 = torch.constant.int 1
%int0_23852 = torch.constant.int 0
%26233 = torch.prim.ListConstruct %int1_23851, %int0_23852 : (!torch.int, !torch.int) -> !torch.list<int>
%26234 = torch.aten.permute %934, %26233 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_23853 = torch.constant.int 1
%int0_23854 = torch.constant.int 0
%26235 = torch.prim.ListConstruct %int1_23853, %int0_23854 : (!torch.int, !torch.int) -> !torch.list<int>
%26236 = torch.aten.permute %935, %26235 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_23855 = torch.constant.int 4
%26237 = torch.aten.mul.int %int4_23855, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23856 = torch.constant.int 4096
%26238 = torch.prim.ListConstruct %26237, %int4096_23856 : (!torch.int, !torch.int) -> !torch.list<int>
%26239 = torch.aten.view %26141, %26238 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26239, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26240 = torch.aten.mm %26239, %26222 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26240, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23857 = torch.constant.int 4
%int1792_23858 = torch.constant.int 1792
%26241 = torch.prim.ListConstruct %int4_23857, %2482, %int1792_23858 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26242 = torch.aten.view %26240, %26241 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23859 = torch.constant.int 4
%26243 = torch.aten.mul.int %int4_23859, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23860 = torch.constant.int 4096
%26244 = torch.prim.ListConstruct %26243, %int4096_23860 : (!torch.int, !torch.int) -> !torch.list<int>
%26245 = torch.aten.view %26142, %26244 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26245, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26246 = torch.aten.mm %26245, %26224 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26246, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23861 = torch.constant.int 4
%int1792_23862 = torch.constant.int 1792
%26247 = torch.prim.ListConstruct %int4_23861, %2482, %int1792_23862 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26248 = torch.aten.view %26246, %26247 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23863 = torch.constant.int 4
%26249 = torch.aten.mul.int %int4_23863, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23864 = torch.constant.int 4096
%26250 = torch.prim.ListConstruct %26249, %int4096_23864 : (!torch.int, !torch.int) -> !torch.list<int>
%26251 = torch.aten.view %26143, %26250 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26251, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26252 = torch.aten.mm %26251, %26226 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26252, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23865 = torch.constant.int 4
%int1792_23866 = torch.constant.int 1792
%26253 = torch.prim.ListConstruct %int4_23865, %2482, %int1792_23866 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26254 = torch.aten.view %26252, %26253 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23867 = torch.constant.int 4
%26255 = torch.aten.mul.int %int4_23867, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23868 = torch.constant.int 4096
%26256 = torch.prim.ListConstruct %26255, %int4096_23868 : (!torch.int, !torch.int) -> !torch.list<int>
%26257 = torch.aten.view %26144, %26256 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26257, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26258 = torch.aten.mm %26257, %26228 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26258, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23869 = torch.constant.int 4
%int1792_23870 = torch.constant.int 1792
%26259 = torch.prim.ListConstruct %int4_23869, %2482, %int1792_23870 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26260 = torch.aten.view %26258, %26259 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23871 = torch.constant.int 4
%26261 = torch.aten.mul.int %int4_23871, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23872 = torch.constant.int 4096
%26262 = torch.prim.ListConstruct %26261, %int4096_23872 : (!torch.int, !torch.int) -> !torch.list<int>
%26263 = torch.aten.view %26145, %26262 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26263, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26264 = torch.aten.mm %26263, %26230 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26264, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23873 = torch.constant.int 4
%int1792_23874 = torch.constant.int 1792
%26265 = torch.prim.ListConstruct %int4_23873, %2482, %int1792_23874 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26266 = torch.aten.view %26264, %26265 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23875 = torch.constant.int 4
%26267 = torch.aten.mul.int %int4_23875, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23876 = torch.constant.int 4096
%26268 = torch.prim.ListConstruct %26267, %int4096_23876 : (!torch.int, !torch.int) -> !torch.list<int>
%26269 = torch.aten.view %26146, %26268 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26269, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26270 = torch.aten.mm %26269, %26232 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26270, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23877 = torch.constant.int 4
%int1792_23878 = torch.constant.int 1792
%26271 = torch.prim.ListConstruct %int4_23877, %2482, %int1792_23878 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26272 = torch.aten.view %26270, %26271 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23879 = torch.constant.int 4
%26273 = torch.aten.mul.int %int4_23879, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23880 = torch.constant.int 4096
%26274 = torch.prim.ListConstruct %26273, %int4096_23880 : (!torch.int, !torch.int) -> !torch.list<int>
%26275 = torch.aten.view %26147, %26274 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26275, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26276 = torch.aten.mm %26275, %26234 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26276, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23881 = torch.constant.int 4
%int1792_23882 = torch.constant.int 1792
%26277 = torch.prim.ListConstruct %int4_23881, %2482, %int1792_23882 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26278 = torch.aten.view %26276, %26277 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_23883 = torch.constant.int 4
%26279 = torch.aten.mul.int %int4_23883, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_23884 = torch.constant.int 4096
%26280 = torch.prim.ListConstruct %26279, %int4096_23884 : (!torch.int, !torch.int) -> !torch.list<int>
%26281 = torch.aten.view %26148, %26280 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26281, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26282 = torch.aten.mm %26281, %26236 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26282, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_23885 = torch.constant.int 4
%int1792_23886 = torch.constant.int 1792
%26283 = torch.prim.ListConstruct %int4_23885, %2482, %int1792_23886 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26284 = torch.aten.view %26282, %26283 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26285 = torch.aten.mul.Tensor %26213, %26242 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26286 = torch.aten.mul.Tensor %26214, %26248 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26287 = torch.aten.mul.Tensor %26215, %26254 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26288 = torch.aten.mul.Tensor %26216, %26260 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26289 = torch.aten.mul.Tensor %26217, %26266 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26290 = torch.aten.mul.Tensor %26218, %26272 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26291 = torch.aten.mul.Tensor %26219, %26278 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%26292 = torch.aten.mul.Tensor %26220, %26284 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %26292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_23887 = torch.constant.int 1
%int0_23888 = torch.constant.int 0
%26293 = torch.prim.ListConstruct %int1_23887, %int0_23888 : (!torch.int, !torch.int) -> !torch.list<int>
%26294 = torch.aten.permute %936, %26293 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23889 = torch.constant.int 1
%int0_23890 = torch.constant.int 0
%26295 = torch.prim.ListConstruct %int1_23889, %int0_23890 : (!torch.int, !torch.int) -> !torch.list<int>
%26296 = torch.aten.permute %937, %26295 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23891 = torch.constant.int 1
%int0_23892 = torch.constant.int 0
%26297 = torch.prim.ListConstruct %int1_23891, %int0_23892 : (!torch.int, !torch.int) -> !torch.list<int>
%26298 = torch.aten.permute %938, %26297 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23893 = torch.constant.int 1
%int0_23894 = torch.constant.int 0
%26299 = torch.prim.ListConstruct %int1_23893, %int0_23894 : (!torch.int, !torch.int) -> !torch.list<int>
%26300 = torch.aten.permute %939, %26299 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23895 = torch.constant.int 1
%int0_23896 = torch.constant.int 0
%26301 = torch.prim.ListConstruct %int1_23895, %int0_23896 : (!torch.int, !torch.int) -> !torch.list<int>
%26302 = torch.aten.permute %940, %26301 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23897 = torch.constant.int 1
%int0_23898 = torch.constant.int 0
%26303 = torch.prim.ListConstruct %int1_23897, %int0_23898 : (!torch.int, !torch.int) -> !torch.list<int>
%26304 = torch.aten.permute %941, %26303 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23899 = torch.constant.int 1
%int0_23900 = torch.constant.int 0
%26305 = torch.prim.ListConstruct %int1_23899, %int0_23900 : (!torch.int, !torch.int) -> !torch.list<int>
%26306 = torch.aten.permute %942, %26305 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23901 = torch.constant.int 1
%int0_23902 = torch.constant.int 0
%26307 = torch.prim.ListConstruct %int1_23901, %int0_23902 : (!torch.int, !torch.int) -> !torch.list<int>
%26308 = torch.aten.permute %943, %26307 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_23903 = torch.constant.int 1
%26309 = torch.aten.size.int %26170, %int1_23903 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23904 = torch.constant.int 4
%26310 = torch.aten.mul.int %int4_23904, %26309 : !torch.int, !torch.int -> !torch.int
%int1792_23905 = torch.constant.int 1792
%26311 = torch.prim.ListConstruct %26310, %int1792_23905 : (!torch.int, !torch.int) -> !torch.list<int>
%26312 = torch.aten.view %26285, %26311 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26312, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26313 = torch.aten.mm %26312, %26294 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26313, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23906 = torch.constant.int 4
%int4096_23907 = torch.constant.int 4096
%26314 = torch.prim.ListConstruct %int4_23906, %26309, %int4096_23907 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26315 = torch.aten.view %26313, %26314 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23908 = torch.constant.int 1
%26316 = torch.aten.size.int %26176, %int1_23908 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23909 = torch.constant.int 4
%26317 = torch.aten.mul.int %int4_23909, %26316 : !torch.int, !torch.int -> !torch.int
%int1792_23910 = torch.constant.int 1792
%26318 = torch.prim.ListConstruct %26317, %int1792_23910 : (!torch.int, !torch.int) -> !torch.list<int>
%26319 = torch.aten.view %26286, %26318 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26319, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26320 = torch.aten.mm %26319, %26296 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26320, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23911 = torch.constant.int 4
%int4096_23912 = torch.constant.int 4096
%26321 = torch.prim.ListConstruct %int4_23911, %26316, %int4096_23912 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26322 = torch.aten.view %26320, %26321 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23913 = torch.constant.int 1
%26323 = torch.aten.size.int %26182, %int1_23913 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23914 = torch.constant.int 4
%26324 = torch.aten.mul.int %int4_23914, %26323 : !torch.int, !torch.int -> !torch.int
%int1792_23915 = torch.constant.int 1792
%26325 = torch.prim.ListConstruct %26324, %int1792_23915 : (!torch.int, !torch.int) -> !torch.list<int>
%26326 = torch.aten.view %26287, %26325 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26326, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26327 = torch.aten.mm %26326, %26298 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26327, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23916 = torch.constant.int 4
%int4096_23917 = torch.constant.int 4096
%26328 = torch.prim.ListConstruct %int4_23916, %26323, %int4096_23917 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26329 = torch.aten.view %26327, %26328 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23918 = torch.constant.int 1
%26330 = torch.aten.size.int %26188, %int1_23918 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23919 = torch.constant.int 4
%26331 = torch.aten.mul.int %int4_23919, %26330 : !torch.int, !torch.int -> !torch.int
%int1792_23920 = torch.constant.int 1792
%26332 = torch.prim.ListConstruct %26331, %int1792_23920 : (!torch.int, !torch.int) -> !torch.list<int>
%26333 = torch.aten.view %26288, %26332 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26333, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26334 = torch.aten.mm %26333, %26300 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26334, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23921 = torch.constant.int 4
%int4096_23922 = torch.constant.int 4096
%26335 = torch.prim.ListConstruct %int4_23921, %26330, %int4096_23922 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26336 = torch.aten.view %26334, %26335 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23923 = torch.constant.int 1
%26337 = torch.aten.size.int %26194, %int1_23923 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23924 = torch.constant.int 4
%26338 = torch.aten.mul.int %int4_23924, %26337 : !torch.int, !torch.int -> !torch.int
%int1792_23925 = torch.constant.int 1792
%26339 = torch.prim.ListConstruct %26338, %int1792_23925 : (!torch.int, !torch.int) -> !torch.list<int>
%26340 = torch.aten.view %26289, %26339 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26340, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26341 = torch.aten.mm %26340, %26302 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26341, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23926 = torch.constant.int 4
%int4096_23927 = torch.constant.int 4096
%26342 = torch.prim.ListConstruct %int4_23926, %26337, %int4096_23927 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26343 = torch.aten.view %26341, %26342 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23928 = torch.constant.int 1
%26344 = torch.aten.size.int %26200, %int1_23928 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23929 = torch.constant.int 4
%26345 = torch.aten.mul.int %int4_23929, %26344 : !torch.int, !torch.int -> !torch.int
%int1792_23930 = torch.constant.int 1792
%26346 = torch.prim.ListConstruct %26345, %int1792_23930 : (!torch.int, !torch.int) -> !torch.list<int>
%26347 = torch.aten.view %26290, %26346 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26347, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26348 = torch.aten.mm %26347, %26304 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26348, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23931 = torch.constant.int 4
%int4096_23932 = torch.constant.int 4096
%26349 = torch.prim.ListConstruct %int4_23931, %26344, %int4096_23932 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26350 = torch.aten.view %26348, %26349 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23933 = torch.constant.int 1
%26351 = torch.aten.size.int %26206, %int1_23933 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23934 = torch.constant.int 4
%26352 = torch.aten.mul.int %int4_23934, %26351 : !torch.int, !torch.int -> !torch.int
%int1792_23935 = torch.constant.int 1792
%26353 = torch.prim.ListConstruct %26352, %int1792_23935 : (!torch.int, !torch.int) -> !torch.list<int>
%26354 = torch.aten.view %26291, %26353 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26354, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26355 = torch.aten.mm %26354, %26306 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26355, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23936 = torch.constant.int 4
%int4096_23937 = torch.constant.int 4096
%26356 = torch.prim.ListConstruct %int4_23936, %26351, %int4096_23937 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26357 = torch.aten.view %26355, %26356 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23938 = torch.constant.int 1
%26358 = torch.aten.size.int %26212, %int1_23938 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_23939 = torch.constant.int 4
%26359 = torch.aten.mul.int %int4_23939, %26358 : !torch.int, !torch.int -> !torch.int
%int1792_23940 = torch.constant.int 1792
%26360 = torch.prim.ListConstruct %26359, %int1792_23940 : (!torch.int, !torch.int) -> !torch.list<int>
%26361 = torch.aten.view %26292, %26360 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %26361, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%26362 = torch.aten.mm %26361, %26308 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26362, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_23941 = torch.constant.int 4
%int4096_23942 = torch.constant.int 4096
%26363 = torch.prim.ListConstruct %int4_23941, %26358, %int4096_23942 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26364 = torch.aten.view %26362, %26363 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26365 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23943 = arith.constant 1 : index
%dim_23944 = tensor.dim %26365, %c1_23943 : tensor<4x?x4096xf16>
%26366 = flow.tensor.transfer %26365 : tensor<4x?x4096xf16>{%dim_23944} to #hal.device.promise<@__device_0>
%26367 = torch_c.from_builtin_tensor %26366 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26368 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23945 = arith.constant 1 : index
%dim_23946 = tensor.dim %26368, %c1_23945 : tensor<4x?x4096xf16>
%26369 = flow.tensor.transfer %26368 : tensor<4x?x4096xf16>{%dim_23946} to #hal.device.promise<@__device_0>
%26370 = torch_c.from_builtin_tensor %26369 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26371 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23947 = arith.constant 1 : index
%dim_23948 = tensor.dim %26371, %c1_23947 : tensor<4x?x4096xf16>
%26372 = flow.tensor.transfer %26371 : tensor<4x?x4096xf16>{%dim_23948} to #hal.device.promise<@__device_0>
%26373 = torch_c.from_builtin_tensor %26372 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26374 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23949 = arith.constant 1 : index
%dim_23950 = tensor.dim %26374, %c1_23949 : tensor<4x?x4096xf16>
%26375 = flow.tensor.transfer %26374 : tensor<4x?x4096xf16>{%dim_23950} to #hal.device.promise<@__device_0>
%26376 = torch_c.from_builtin_tensor %26375 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26377 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23951 = arith.constant 1 : index
%dim_23952 = tensor.dim %26377, %c1_23951 : tensor<4x?x4096xf16>
%26378 = flow.tensor.transfer %26377 : tensor<4x?x4096xf16>{%dim_23952} to #hal.device.promise<@__device_0>
%26379 = torch_c.from_builtin_tensor %26378 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26380 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23953 = arith.constant 1 : index
%dim_23954 = tensor.dim %26380, %c1_23953 : tensor<4x?x4096xf16>
%26381 = flow.tensor.transfer %26380 : tensor<4x?x4096xf16>{%dim_23954} to #hal.device.promise<@__device_0>
%26382 = torch_c.from_builtin_tensor %26381 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26383 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23955 = arith.constant 1 : index
%dim_23956 = tensor.dim %26383, %c1_23955 : tensor<4x?x4096xf16>
%26384 = flow.tensor.transfer %26383 : tensor<4x?x4096xf16>{%dim_23956} to #hal.device.promise<@__device_0>
%26385 = torch_c.from_builtin_tensor %26384 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23957 = torch.constant.int 1
%26386 = torch.aten.add.Tensor %26315, %26367, %int1_23957 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23958 = torch.constant.int 1
%26387 = torch.aten.add.Tensor %26386, %26370, %int1_23958 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23959 = torch.constant.int 1
%26388 = torch.aten.add.Tensor %26387, %26373, %int1_23959 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23960 = torch.constant.int 1
%26389 = torch.aten.add.Tensor %26388, %26376, %int1_23960 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23961 = torch.constant.int 1
%26390 = torch.aten.add.Tensor %26389, %26379, %int1_23961 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23962 = torch.constant.int 1
%26391 = torch.aten.add.Tensor %26390, %26382, %int1_23962 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23963 = torch.constant.int 1
%26392 = torch.aten.add.Tensor %26391, %26385, %int1_23963 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26393 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23964 = arith.constant 1 : index
%dim_23965 = tensor.dim %26393, %c1_23964 : tensor<4x?x4096xf16>
%26394 = flow.tensor.transfer %26393 : tensor<4x?x4096xf16>{%dim_23965} to #hal.device.promise<@__device_1>
%26395 = torch_c.from_builtin_tensor %26394 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26396 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23966 = arith.constant 1 : index
%dim_23967 = tensor.dim %26396, %c1_23966 : tensor<4x?x4096xf16>
%26397 = flow.tensor.transfer %26396 : tensor<4x?x4096xf16>{%dim_23967} to #hal.device.promise<@__device_1>
%26398 = torch_c.from_builtin_tensor %26397 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26399 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23968 = arith.constant 1 : index
%dim_23969 = tensor.dim %26399, %c1_23968 : tensor<4x?x4096xf16>
%26400 = flow.tensor.transfer %26399 : tensor<4x?x4096xf16>{%dim_23969} to #hal.device.promise<@__device_1>
%26401 = torch_c.from_builtin_tensor %26400 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26402 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23970 = arith.constant 1 : index
%dim_23971 = tensor.dim %26402, %c1_23970 : tensor<4x?x4096xf16>
%26403 = flow.tensor.transfer %26402 : tensor<4x?x4096xf16>{%dim_23971} to #hal.device.promise<@__device_1>
%26404 = torch_c.from_builtin_tensor %26403 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26405 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23972 = arith.constant 1 : index
%dim_23973 = tensor.dim %26405, %c1_23972 : tensor<4x?x4096xf16>
%26406 = flow.tensor.transfer %26405 : tensor<4x?x4096xf16>{%dim_23973} to #hal.device.promise<@__device_1>
%26407 = torch_c.from_builtin_tensor %26406 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26408 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23974 = arith.constant 1 : index
%dim_23975 = tensor.dim %26408, %c1_23974 : tensor<4x?x4096xf16>
%26409 = flow.tensor.transfer %26408 : tensor<4x?x4096xf16>{%dim_23975} to #hal.device.promise<@__device_1>
%26410 = torch_c.from_builtin_tensor %26409 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26411 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23976 = arith.constant 1 : index
%dim_23977 = tensor.dim %26411, %c1_23976 : tensor<4x?x4096xf16>
%26412 = flow.tensor.transfer %26411 : tensor<4x?x4096xf16>{%dim_23977} to #hal.device.promise<@__device_1>
%26413 = torch_c.from_builtin_tensor %26412 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23978 = torch.constant.int 1
%26414 = torch.aten.add.Tensor %26395, %26322, %int1_23978 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23979 = torch.constant.int 1
%26415 = torch.aten.add.Tensor %26414, %26398, %int1_23979 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23980 = torch.constant.int 1
%26416 = torch.aten.add.Tensor %26415, %26401, %int1_23980 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23981 = torch.constant.int 1
%26417 = torch.aten.add.Tensor %26416, %26404, %int1_23981 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23982 = torch.constant.int 1
%26418 = torch.aten.add.Tensor %26417, %26407, %int1_23982 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23983 = torch.constant.int 1
%26419 = torch.aten.add.Tensor %26418, %26410, %int1_23983 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23984 = torch.constant.int 1
%26420 = torch.aten.add.Tensor %26419, %26413, %int1_23984 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26421 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23985 = arith.constant 1 : index
%dim_23986 = tensor.dim %26421, %c1_23985 : tensor<4x?x4096xf16>
%26422 = flow.tensor.transfer %26421 : tensor<4x?x4096xf16>{%dim_23986} to #hal.device.promise<@__device_2>
%26423 = torch_c.from_builtin_tensor %26422 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26424 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23987 = arith.constant 1 : index
%dim_23988 = tensor.dim %26424, %c1_23987 : tensor<4x?x4096xf16>
%26425 = flow.tensor.transfer %26424 : tensor<4x?x4096xf16>{%dim_23988} to #hal.device.promise<@__device_2>
%26426 = torch_c.from_builtin_tensor %26425 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26427 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23989 = arith.constant 1 : index
%dim_23990 = tensor.dim %26427, %c1_23989 : tensor<4x?x4096xf16>
%26428 = flow.tensor.transfer %26427 : tensor<4x?x4096xf16>{%dim_23990} to #hal.device.promise<@__device_2>
%26429 = torch_c.from_builtin_tensor %26428 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26430 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23991 = arith.constant 1 : index
%dim_23992 = tensor.dim %26430, %c1_23991 : tensor<4x?x4096xf16>
%26431 = flow.tensor.transfer %26430 : tensor<4x?x4096xf16>{%dim_23992} to #hal.device.promise<@__device_2>
%26432 = torch_c.from_builtin_tensor %26431 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26433 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23993 = arith.constant 1 : index
%dim_23994 = tensor.dim %26433, %c1_23993 : tensor<4x?x4096xf16>
%26434 = flow.tensor.transfer %26433 : tensor<4x?x4096xf16>{%dim_23994} to #hal.device.promise<@__device_2>
%26435 = torch_c.from_builtin_tensor %26434 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26436 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23995 = arith.constant 1 : index
%dim_23996 = tensor.dim %26436, %c1_23995 : tensor<4x?x4096xf16>
%26437 = flow.tensor.transfer %26436 : tensor<4x?x4096xf16>{%dim_23996} to #hal.device.promise<@__device_2>
%26438 = torch_c.from_builtin_tensor %26437 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26439 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_23997 = arith.constant 1 : index
%dim_23998 = tensor.dim %26439, %c1_23997 : tensor<4x?x4096xf16>
%26440 = flow.tensor.transfer %26439 : tensor<4x?x4096xf16>{%dim_23998} to #hal.device.promise<@__device_2>
%26441 = torch_c.from_builtin_tensor %26440 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_23999 = torch.constant.int 1
%26442 = torch.aten.add.Tensor %26423, %26426, %int1_23999 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24000 = torch.constant.int 1
%26443 = torch.aten.add.Tensor %26442, %26329, %int1_24000 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24001 = torch.constant.int 1
%26444 = torch.aten.add.Tensor %26443, %26429, %int1_24001 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24002 = torch.constant.int 1
%26445 = torch.aten.add.Tensor %26444, %26432, %int1_24002 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24003 = torch.constant.int 1
%26446 = torch.aten.add.Tensor %26445, %26435, %int1_24003 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24004 = torch.constant.int 1
%26447 = torch.aten.add.Tensor %26446, %26438, %int1_24004 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24005 = torch.constant.int 1
%26448 = torch.aten.add.Tensor %26447, %26441, %int1_24005 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26449 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24006 = arith.constant 1 : index
%dim_24007 = tensor.dim %26449, %c1_24006 : tensor<4x?x4096xf16>
%26450 = flow.tensor.transfer %26449 : tensor<4x?x4096xf16>{%dim_24007} to #hal.device.promise<@__device_3>
%26451 = torch_c.from_builtin_tensor %26450 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26452 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24008 = arith.constant 1 : index
%dim_24009 = tensor.dim %26452, %c1_24008 : tensor<4x?x4096xf16>
%26453 = flow.tensor.transfer %26452 : tensor<4x?x4096xf16>{%dim_24009} to #hal.device.promise<@__device_3>
%26454 = torch_c.from_builtin_tensor %26453 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26455 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24010 = arith.constant 1 : index
%dim_24011 = tensor.dim %26455, %c1_24010 : tensor<4x?x4096xf16>
%26456 = flow.tensor.transfer %26455 : tensor<4x?x4096xf16>{%dim_24011} to #hal.device.promise<@__device_3>
%26457 = torch_c.from_builtin_tensor %26456 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26458 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24012 = arith.constant 1 : index
%dim_24013 = tensor.dim %26458, %c1_24012 : tensor<4x?x4096xf16>
%26459 = flow.tensor.transfer %26458 : tensor<4x?x4096xf16>{%dim_24013} to #hal.device.promise<@__device_3>
%26460 = torch_c.from_builtin_tensor %26459 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26461 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24014 = arith.constant 1 : index
%dim_24015 = tensor.dim %26461, %c1_24014 : tensor<4x?x4096xf16>
%26462 = flow.tensor.transfer %26461 : tensor<4x?x4096xf16>{%dim_24015} to #hal.device.promise<@__device_3>
%26463 = torch_c.from_builtin_tensor %26462 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26464 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24016 = arith.constant 1 : index
%dim_24017 = tensor.dim %26464, %c1_24016 : tensor<4x?x4096xf16>
%26465 = flow.tensor.transfer %26464 : tensor<4x?x4096xf16>{%dim_24017} to #hal.device.promise<@__device_3>
%26466 = torch_c.from_builtin_tensor %26465 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26467 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24018 = arith.constant 1 : index
%dim_24019 = tensor.dim %26467, %c1_24018 : tensor<4x?x4096xf16>
%26468 = flow.tensor.transfer %26467 : tensor<4x?x4096xf16>{%dim_24019} to #hal.device.promise<@__device_3>
%26469 = torch_c.from_builtin_tensor %26468 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24020 = torch.constant.int 1
%26470 = torch.aten.add.Tensor %26451, %26454, %int1_24020 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24021 = torch.constant.int 1
%26471 = torch.aten.add.Tensor %26470, %26457, %int1_24021 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24022 = torch.constant.int 1
%26472 = torch.aten.add.Tensor %26471, %26336, %int1_24022 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24023 = torch.constant.int 1
%26473 = torch.aten.add.Tensor %26472, %26460, %int1_24023 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24024 = torch.constant.int 1
%26474 = torch.aten.add.Tensor %26473, %26463, %int1_24024 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24025 = torch.constant.int 1
%26475 = torch.aten.add.Tensor %26474, %26466, %int1_24025 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24026 = torch.constant.int 1
%26476 = torch.aten.add.Tensor %26475, %26469, %int1_24026 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26477 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24027 = arith.constant 1 : index
%dim_24028 = tensor.dim %26477, %c1_24027 : tensor<4x?x4096xf16>
%26478 = flow.tensor.transfer %26477 : tensor<4x?x4096xf16>{%dim_24028} to #hal.device.promise<@__device_4>
%26479 = torch_c.from_builtin_tensor %26478 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26480 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24029 = arith.constant 1 : index
%dim_24030 = tensor.dim %26480, %c1_24029 : tensor<4x?x4096xf16>
%26481 = flow.tensor.transfer %26480 : tensor<4x?x4096xf16>{%dim_24030} to #hal.device.promise<@__device_4>
%26482 = torch_c.from_builtin_tensor %26481 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26483 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24031 = arith.constant 1 : index
%dim_24032 = tensor.dim %26483, %c1_24031 : tensor<4x?x4096xf16>
%26484 = flow.tensor.transfer %26483 : tensor<4x?x4096xf16>{%dim_24032} to #hal.device.promise<@__device_4>
%26485 = torch_c.from_builtin_tensor %26484 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26486 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24033 = arith.constant 1 : index
%dim_24034 = tensor.dim %26486, %c1_24033 : tensor<4x?x4096xf16>
%26487 = flow.tensor.transfer %26486 : tensor<4x?x4096xf16>{%dim_24034} to #hal.device.promise<@__device_4>
%26488 = torch_c.from_builtin_tensor %26487 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26489 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24035 = arith.constant 1 : index
%dim_24036 = tensor.dim %26489, %c1_24035 : tensor<4x?x4096xf16>
%26490 = flow.tensor.transfer %26489 : tensor<4x?x4096xf16>{%dim_24036} to #hal.device.promise<@__device_4>
%26491 = torch_c.from_builtin_tensor %26490 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26492 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24037 = arith.constant 1 : index
%dim_24038 = tensor.dim %26492, %c1_24037 : tensor<4x?x4096xf16>
%26493 = flow.tensor.transfer %26492 : tensor<4x?x4096xf16>{%dim_24038} to #hal.device.promise<@__device_4>
%26494 = torch_c.from_builtin_tensor %26493 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26495 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24039 = arith.constant 1 : index
%dim_24040 = tensor.dim %26495, %c1_24039 : tensor<4x?x4096xf16>
%26496 = flow.tensor.transfer %26495 : tensor<4x?x4096xf16>{%dim_24040} to #hal.device.promise<@__device_4>
%26497 = torch_c.from_builtin_tensor %26496 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24041 = torch.constant.int 1
%26498 = torch.aten.add.Tensor %26479, %26482, %int1_24041 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24042 = torch.constant.int 1
%26499 = torch.aten.add.Tensor %26498, %26485, %int1_24042 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24043 = torch.constant.int 1
%26500 = torch.aten.add.Tensor %26499, %26488, %int1_24043 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24044 = torch.constant.int 1
%26501 = torch.aten.add.Tensor %26500, %26343, %int1_24044 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24045 = torch.constant.int 1
%26502 = torch.aten.add.Tensor %26501, %26491, %int1_24045 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24046 = torch.constant.int 1
%26503 = torch.aten.add.Tensor %26502, %26494, %int1_24046 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24047 = torch.constant.int 1
%26504 = torch.aten.add.Tensor %26503, %26497, %int1_24047 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26505 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24048 = arith.constant 1 : index
%dim_24049 = tensor.dim %26505, %c1_24048 : tensor<4x?x4096xf16>
%26506 = flow.tensor.transfer %26505 : tensor<4x?x4096xf16>{%dim_24049} to #hal.device.promise<@__device_5>
%26507 = torch_c.from_builtin_tensor %26506 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26508 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24050 = arith.constant 1 : index
%dim_24051 = tensor.dim %26508, %c1_24050 : tensor<4x?x4096xf16>
%26509 = flow.tensor.transfer %26508 : tensor<4x?x4096xf16>{%dim_24051} to #hal.device.promise<@__device_5>
%26510 = torch_c.from_builtin_tensor %26509 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26511 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24052 = arith.constant 1 : index
%dim_24053 = tensor.dim %26511, %c1_24052 : tensor<4x?x4096xf16>
%26512 = flow.tensor.transfer %26511 : tensor<4x?x4096xf16>{%dim_24053} to #hal.device.promise<@__device_5>
%26513 = torch_c.from_builtin_tensor %26512 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26514 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24054 = arith.constant 1 : index
%dim_24055 = tensor.dim %26514, %c1_24054 : tensor<4x?x4096xf16>
%26515 = flow.tensor.transfer %26514 : tensor<4x?x4096xf16>{%dim_24055} to #hal.device.promise<@__device_5>
%26516 = torch_c.from_builtin_tensor %26515 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26517 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24056 = arith.constant 1 : index
%dim_24057 = tensor.dim %26517, %c1_24056 : tensor<4x?x4096xf16>
%26518 = flow.tensor.transfer %26517 : tensor<4x?x4096xf16>{%dim_24057} to #hal.device.promise<@__device_5>
%26519 = torch_c.from_builtin_tensor %26518 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26520 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24058 = arith.constant 1 : index
%dim_24059 = tensor.dim %26520, %c1_24058 : tensor<4x?x4096xf16>
%26521 = flow.tensor.transfer %26520 : tensor<4x?x4096xf16>{%dim_24059} to #hal.device.promise<@__device_5>
%26522 = torch_c.from_builtin_tensor %26521 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26523 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24060 = arith.constant 1 : index
%dim_24061 = tensor.dim %26523, %c1_24060 : tensor<4x?x4096xf16>
%26524 = flow.tensor.transfer %26523 : tensor<4x?x4096xf16>{%dim_24061} to #hal.device.promise<@__device_5>
%26525 = torch_c.from_builtin_tensor %26524 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24062 = torch.constant.int 1
%26526 = torch.aten.add.Tensor %26507, %26510, %int1_24062 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24063 = torch.constant.int 1
%26527 = torch.aten.add.Tensor %26526, %26513, %int1_24063 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24064 = torch.constant.int 1
%26528 = torch.aten.add.Tensor %26527, %26516, %int1_24064 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24065 = torch.constant.int 1
%26529 = torch.aten.add.Tensor %26528, %26519, %int1_24065 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24066 = torch.constant.int 1
%26530 = torch.aten.add.Tensor %26529, %26350, %int1_24066 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24067 = torch.constant.int 1
%26531 = torch.aten.add.Tensor %26530, %26522, %int1_24067 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24068 = torch.constant.int 1
%26532 = torch.aten.add.Tensor %26531, %26525, %int1_24068 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26533 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24069 = arith.constant 1 : index
%dim_24070 = tensor.dim %26533, %c1_24069 : tensor<4x?x4096xf16>
%26534 = flow.tensor.transfer %26533 : tensor<4x?x4096xf16>{%dim_24070} to #hal.device.promise<@__device_6>
%26535 = torch_c.from_builtin_tensor %26534 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26536 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24071 = arith.constant 1 : index
%dim_24072 = tensor.dim %26536, %c1_24071 : tensor<4x?x4096xf16>
%26537 = flow.tensor.transfer %26536 : tensor<4x?x4096xf16>{%dim_24072} to #hal.device.promise<@__device_6>
%26538 = torch_c.from_builtin_tensor %26537 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26539 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24073 = arith.constant 1 : index
%dim_24074 = tensor.dim %26539, %c1_24073 : tensor<4x?x4096xf16>
%26540 = flow.tensor.transfer %26539 : tensor<4x?x4096xf16>{%dim_24074} to #hal.device.promise<@__device_6>
%26541 = torch_c.from_builtin_tensor %26540 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26541, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26542 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24075 = arith.constant 1 : index
%dim_24076 = tensor.dim %26542, %c1_24075 : tensor<4x?x4096xf16>
%26543 = flow.tensor.transfer %26542 : tensor<4x?x4096xf16>{%dim_24076} to #hal.device.promise<@__device_6>
%26544 = torch_c.from_builtin_tensor %26543 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26545 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24077 = arith.constant 1 : index
%dim_24078 = tensor.dim %26545, %c1_24077 : tensor<4x?x4096xf16>
%26546 = flow.tensor.transfer %26545 : tensor<4x?x4096xf16>{%dim_24078} to #hal.device.promise<@__device_6>
%26547 = torch_c.from_builtin_tensor %26546 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26548 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24079 = arith.constant 1 : index
%dim_24080 = tensor.dim %26548, %c1_24079 : tensor<4x?x4096xf16>
%26549 = flow.tensor.transfer %26548 : tensor<4x?x4096xf16>{%dim_24080} to #hal.device.promise<@__device_6>
%26550 = torch_c.from_builtin_tensor %26549 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26551 = torch_c.to_builtin_tensor %26364 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24081 = arith.constant 1 : index
%dim_24082 = tensor.dim %26551, %c1_24081 : tensor<4x?x4096xf16>
%26552 = flow.tensor.transfer %26551 : tensor<4x?x4096xf16>{%dim_24082} to #hal.device.promise<@__device_6>
%26553 = torch_c.from_builtin_tensor %26552 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24083 = torch.constant.int 1
%26554 = torch.aten.add.Tensor %26535, %26538, %int1_24083 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24084 = torch.constant.int 1
%26555 = torch.aten.add.Tensor %26554, %26541, %int1_24084 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24085 = torch.constant.int 1
%26556 = torch.aten.add.Tensor %26555, %26544, %int1_24085 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24086 = torch.constant.int 1
%26557 = torch.aten.add.Tensor %26556, %26547, %int1_24086 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24087 = torch.constant.int 1
%26558 = torch.aten.add.Tensor %26557, %26550, %int1_24087 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24088 = torch.constant.int 1
%26559 = torch.aten.add.Tensor %26558, %26357, %int1_24088 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24089 = torch.constant.int 1
%26560 = torch.aten.add.Tensor %26559, %26553, %int1_24089 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26561 = torch_c.to_builtin_tensor %26315 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24090 = arith.constant 1 : index
%dim_24091 = tensor.dim %26561, %c1_24090 : tensor<4x?x4096xf16>
%26562 = flow.tensor.transfer %26561 : tensor<4x?x4096xf16>{%dim_24091} to #hal.device.promise<@__device_7>
%26563 = torch_c.from_builtin_tensor %26562 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26564 = torch_c.to_builtin_tensor %26322 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24092 = arith.constant 1 : index
%dim_24093 = tensor.dim %26564, %c1_24092 : tensor<4x?x4096xf16>
%26565 = flow.tensor.transfer %26564 : tensor<4x?x4096xf16>{%dim_24093} to #hal.device.promise<@__device_7>
%26566 = torch_c.from_builtin_tensor %26565 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26567 = torch_c.to_builtin_tensor %26329 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24094 = arith.constant 1 : index
%dim_24095 = tensor.dim %26567, %c1_24094 : tensor<4x?x4096xf16>
%26568 = flow.tensor.transfer %26567 : tensor<4x?x4096xf16>{%dim_24095} to #hal.device.promise<@__device_7>
%26569 = torch_c.from_builtin_tensor %26568 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26570 = torch_c.to_builtin_tensor %26336 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24096 = arith.constant 1 : index
%dim_24097 = tensor.dim %26570, %c1_24096 : tensor<4x?x4096xf16>
%26571 = flow.tensor.transfer %26570 : tensor<4x?x4096xf16>{%dim_24097} to #hal.device.promise<@__device_7>
%26572 = torch_c.from_builtin_tensor %26571 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26573 = torch_c.to_builtin_tensor %26343 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24098 = arith.constant 1 : index
%dim_24099 = tensor.dim %26573, %c1_24098 : tensor<4x?x4096xf16>
%26574 = flow.tensor.transfer %26573 : tensor<4x?x4096xf16>{%dim_24099} to #hal.device.promise<@__device_7>
%26575 = torch_c.from_builtin_tensor %26574 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26576 = torch_c.to_builtin_tensor %26350 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24100 = arith.constant 1 : index
%dim_24101 = tensor.dim %26576, %c1_24100 : tensor<4x?x4096xf16>
%26577 = flow.tensor.transfer %26576 : tensor<4x?x4096xf16>{%dim_24101} to #hal.device.promise<@__device_7>
%26578 = torch_c.from_builtin_tensor %26577 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%26579 = torch_c.to_builtin_tensor %26357 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_24102 = arith.constant 1 : index
%dim_24103 = tensor.dim %26579, %c1_24102 : tensor<4x?x4096xf16>
%26580 = flow.tensor.transfer %26579 : tensor<4x?x4096xf16>{%dim_24103} to #hal.device.promise<@__device_7>
%26581 = torch_c.from_builtin_tensor %26580 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24104 = torch.constant.int 1
%26582 = torch.aten.add.Tensor %26563, %26566, %int1_24104 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24105 = torch.constant.int 1
%26583 = torch.aten.add.Tensor %26582, %26569, %int1_24105 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24106 = torch.constant.int 1
%26584 = torch.aten.add.Tensor %26583, %26572, %int1_24106 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24107 = torch.constant.int 1
%26585 = torch.aten.add.Tensor %26584, %26575, %int1_24107 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24108 = torch.constant.int 1
%26586 = torch.aten.add.Tensor %26585, %26578, %int1_24108 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24109 = torch.constant.int 1
%26587 = torch.aten.add.Tensor %26586, %26581, %int1_24109 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24110 = torch.constant.int 1
%26588 = torch.aten.add.Tensor %26587, %26364, %int1_24110 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24111 = torch.constant.int 1
%26589 = torch.aten.add.Tensor %26069, %26392, %int1_24111 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24112 = torch.constant.int 1
%26590 = torch.aten.add.Tensor %26070, %26420, %int1_24112 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24113 = torch.constant.int 1
%26591 = torch.aten.add.Tensor %26071, %26448, %int1_24113 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24114 = torch.constant.int 1
%26592 = torch.aten.add.Tensor %26072, %26476, %int1_24114 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24115 = torch.constant.int 1
%26593 = torch.aten.add.Tensor %26073, %26504, %int1_24115 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24116 = torch.constant.int 1
%26594 = torch.aten.add.Tensor %26074, %26532, %int1_24116 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24117 = torch.constant.int 1
%26595 = torch.aten.add.Tensor %26075, %26560, %int1_24117 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24118 = torch.constant.int 1
%26596 = torch.aten.add.Tensor %26076, %26588, %int1_24118 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_24119 = torch.constant.int 6
%26597 = torch.prims.convert_element_type %26589, %int6_24119 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24120 = torch.constant.int 6
%26598 = torch.prims.convert_element_type %26590, %int6_24120 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24121 = torch.constant.int 6
%26599 = torch.prims.convert_element_type %26591, %int6_24121 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24122 = torch.constant.int 6
%26600 = torch.prims.convert_element_type %26592, %int6_24122 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24123 = torch.constant.int 6
%26601 = torch.prims.convert_element_type %26593, %int6_24123 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24124 = torch.constant.int 6
%26602 = torch.prims.convert_element_type %26594, %int6_24124 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24125 = torch.constant.int 6
%26603 = torch.prims.convert_element_type %26595, %int6_24125 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_24126 = torch.constant.int 6
%26604 = torch.prims.convert_element_type %26596, %int6_24126 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24127 = torch.constant.int 2
%26605 = torch.aten.pow.Tensor_Scalar %26597, %int2_24127 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24128 = torch.constant.int 2
%26606 = torch.aten.pow.Tensor_Scalar %26598, %int2_24128 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24129 = torch.constant.int 2
%26607 = torch.aten.pow.Tensor_Scalar %26599, %int2_24129 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24130 = torch.constant.int 2
%26608 = torch.aten.pow.Tensor_Scalar %26600, %int2_24130 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24131 = torch.constant.int 2
%26609 = torch.aten.pow.Tensor_Scalar %26601, %int2_24131 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24132 = torch.constant.int 2
%26610 = torch.aten.pow.Tensor_Scalar %26602, %int2_24132 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24133 = torch.constant.int 2
%26611 = torch.aten.pow.Tensor_Scalar %26603, %int2_24133 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_24134 = torch.constant.int 2
%26612 = torch.aten.pow.Tensor_Scalar %26604, %int2_24134 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_24135 = torch.constant.int -1
%26613 = torch.prim.ListConstruct %int-1_24135 : (!torch.int) -> !torch.list<int>
%true_24136 = torch.constant.bool true
%none_24137 = torch.constant.none
%26614 = torch.aten.mean.dim %26605, %26613, %true_24136, %none_24137 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24138 = torch.constant.int -1
%26615 = torch.prim.ListConstruct %int-1_24138 : (!torch.int) -> !torch.list<int>
%true_24139 = torch.constant.bool true
%none_24140 = torch.constant.none
%26616 = torch.aten.mean.dim %26606, %26615, %true_24139, %none_24140 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24141 = torch.constant.int -1
%26617 = torch.prim.ListConstruct %int-1_24141 : (!torch.int) -> !torch.list<int>
%true_24142 = torch.constant.bool true
%none_24143 = torch.constant.none
%26618 = torch.aten.mean.dim %26607, %26617, %true_24142, %none_24143 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24144 = torch.constant.int -1
%26619 = torch.prim.ListConstruct %int-1_24144 : (!torch.int) -> !torch.list<int>
%true_24145 = torch.constant.bool true
%none_24146 = torch.constant.none
%26620 = torch.aten.mean.dim %26608, %26619, %true_24145, %none_24146 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24147 = torch.constant.int -1
%26621 = torch.prim.ListConstruct %int-1_24147 : (!torch.int) -> !torch.list<int>
%true_24148 = torch.constant.bool true
%none_24149 = torch.constant.none
%26622 = torch.aten.mean.dim %26609, %26621, %true_24148, %none_24149 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24150 = torch.constant.int -1
%26623 = torch.prim.ListConstruct %int-1_24150 : (!torch.int) -> !torch.list<int>
%true_24151 = torch.constant.bool true
%none_24152 = torch.constant.none
%26624 = torch.aten.mean.dim %26610, %26623, %true_24151, %none_24152 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24153 = torch.constant.int -1
%26625 = torch.prim.ListConstruct %int-1_24153 : (!torch.int) -> !torch.list<int>
%true_24154 = torch.constant.bool true
%none_24155 = torch.constant.none
%26626 = torch.aten.mean.dim %26611, %26625, %true_24154, %none_24155 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_24156 = torch.constant.int -1
%26627 = torch.prim.ListConstruct %int-1_24156 : (!torch.int) -> !torch.list<int>
%true_24157 = torch.constant.bool true
%none_24158 = torch.constant.none
%26628 = torch.aten.mean.dim %26612, %26627, %true_24157, %none_24158 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24159 = torch.constant.float 9.9999997473787516E-6
%int1_24160 = torch.constant.int 1
%26629 = torch.aten.add.Scalar %26614, %float9.999990e-06_24159, %int1_24160 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24161 = torch.constant.float 9.9999997473787516E-6
%int1_24162 = torch.constant.int 1
%26630 = torch.aten.add.Scalar %26616, %float9.999990e-06_24161, %int1_24162 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24163 = torch.constant.float 9.9999997473787516E-6
%int1_24164 = torch.constant.int 1
%26631 = torch.aten.add.Scalar %26618, %float9.999990e-06_24163, %int1_24164 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24165 = torch.constant.float 9.9999997473787516E-6
%int1_24166 = torch.constant.int 1
%26632 = torch.aten.add.Scalar %26620, %float9.999990e-06_24165, %int1_24166 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24167 = torch.constant.float 9.9999997473787516E-6
%int1_24168 = torch.constant.int 1
%26633 = torch.aten.add.Scalar %26622, %float9.999990e-06_24167, %int1_24168 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24169 = torch.constant.float 9.9999997473787516E-6
%int1_24170 = torch.constant.int 1
%26634 = torch.aten.add.Scalar %26624, %float9.999990e-06_24169, %int1_24170 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24171 = torch.constant.float 9.9999997473787516E-6
%int1_24172 = torch.constant.int 1
%26635 = torch.aten.add.Scalar %26626, %float9.999990e-06_24171, %int1_24172 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_24173 = torch.constant.float 9.9999997473787516E-6
%int1_24174 = torch.constant.int 1
%26636 = torch.aten.add.Scalar %26628, %float9.999990e-06_24173, %int1_24174 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26637 = torch.aten.rsqrt %26629 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26638 = torch.aten.rsqrt %26630 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26639 = torch.aten.rsqrt %26631 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26640 = torch.aten.rsqrt %26632 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26641 = torch.aten.rsqrt %26633 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26642 = torch.aten.rsqrt %26634 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26643 = torch.aten.rsqrt %26635 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26644 = torch.aten.rsqrt %26636 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %26644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%26645 = torch.aten.mul.Tensor %26597, %26637 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26646 = torch.aten.mul.Tensor %26598, %26638 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26647 = torch.aten.mul.Tensor %26599, %26639 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26648 = torch.aten.mul.Tensor %26600, %26640 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26649 = torch.aten.mul.Tensor %26601, %26641 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26650 = torch.aten.mul.Tensor %26602, %26642 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26651 = torch.aten.mul.Tensor %26603, %26643 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26652 = torch.aten.mul.Tensor %26604, %26644 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26653 = torch.aten.mul.Tensor %944, %26645 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26654 = torch.aten.mul.Tensor %945, %26646 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26655 = torch.aten.mul.Tensor %946, %26647 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26656 = torch.aten.mul.Tensor %947, %26648 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26657 = torch.aten.mul.Tensor %948, %26649 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26658 = torch.aten.mul.Tensor %949, %26650 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26659 = torch.aten.mul.Tensor %950, %26651 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%26660 = torch.aten.mul.Tensor %951, %26652 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %26660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_24175 = torch.constant.int 5
%26661 = torch.prims.convert_element_type %26653, %int5_24175 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24176 = torch.constant.int 5
%26662 = torch.prims.convert_element_type %26654, %int5_24176 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24177 = torch.constant.int 5
%26663 = torch.prims.convert_element_type %26655, %int5_24177 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24178 = torch.constant.int 5
%26664 = torch.prims.convert_element_type %26656, %int5_24178 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24179 = torch.constant.int 5
%26665 = torch.prims.convert_element_type %26657, %int5_24179 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24180 = torch.constant.int 5
%26666 = torch.prims.convert_element_type %26658, %int5_24180 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24181 = torch.constant.int 5
%26667 = torch.prims.convert_element_type %26659, %int5_24181 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_24182 = torch.constant.int 5
%26668 = torch.prims.convert_element_type %26660, %int5_24182 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %26668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_24183 = torch.constant.int 1
%int0_24184 = torch.constant.int 0
%26669 = torch.prim.ListConstruct %int1_24183, %int0_24184 : (!torch.int, !torch.int) -> !torch.list<int>
%26670 = torch.aten.permute %952, %26669 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24185 = torch.constant.int 1
%int0_24186 = torch.constant.int 0
%26671 = torch.prim.ListConstruct %int1_24185, %int0_24186 : (!torch.int, !torch.int) -> !torch.list<int>
%26672 = torch.aten.permute %953, %26671 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24187 = torch.constant.int 1
%int0_24188 = torch.constant.int 0
%26673 = torch.prim.ListConstruct %int1_24187, %int0_24188 : (!torch.int, !torch.int) -> !torch.list<int>
%26674 = torch.aten.permute %954, %26673 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24189 = torch.constant.int 1
%int0_24190 = torch.constant.int 0
%26675 = torch.prim.ListConstruct %int1_24189, %int0_24190 : (!torch.int, !torch.int) -> !torch.list<int>
%26676 = torch.aten.permute %955, %26675 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24191 = torch.constant.int 1
%int0_24192 = torch.constant.int 0
%26677 = torch.prim.ListConstruct %int1_24191, %int0_24192 : (!torch.int, !torch.int) -> !torch.list<int>
%26678 = torch.aten.permute %956, %26677 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24193 = torch.constant.int 1
%int0_24194 = torch.constant.int 0
%26679 = torch.prim.ListConstruct %int1_24193, %int0_24194 : (!torch.int, !torch.int) -> !torch.list<int>
%26680 = torch.aten.permute %957, %26679 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24195 = torch.constant.int 1
%int0_24196 = torch.constant.int 0
%26681 = torch.prim.ListConstruct %int1_24195, %int0_24196 : (!torch.int, !torch.int) -> !torch.list<int>
%26682 = torch.aten.permute %958, %26681 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_24197 = torch.constant.int 1
%int0_24198 = torch.constant.int 0
%26683 = torch.prim.ListConstruct %int1_24197, %int0_24198 : (!torch.int, !torch.int) -> !torch.list<int>
%26684 = torch.aten.permute %959, %26683 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_24199 = torch.constant.int 4
%26685 = torch.aten.mul.int %int4_24199, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24200 = torch.constant.int 4096
%26686 = torch.prim.ListConstruct %26685, %int4096_24200 : (!torch.int, !torch.int) -> !torch.list<int>
%26687 = torch.aten.view %26661, %26686 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26687, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26688 = torch.aten.mm %26687, %26670 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26688, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24201 = torch.constant.int 4
%int512_24202 = torch.constant.int 512
%26689 = torch.prim.ListConstruct %int4_24201, %2482, %int512_24202 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26690 = torch.aten.view %26688, %26689 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24203 = torch.constant.int 4
%26691 = torch.aten.mul.int %int4_24203, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24204 = torch.constant.int 4096
%26692 = torch.prim.ListConstruct %26691, %int4096_24204 : (!torch.int, !torch.int) -> !torch.list<int>
%26693 = torch.aten.view %26662, %26692 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26693, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26694 = torch.aten.mm %26693, %26672 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26694, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24205 = torch.constant.int 4
%int512_24206 = torch.constant.int 512
%26695 = torch.prim.ListConstruct %int4_24205, %2482, %int512_24206 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26696 = torch.aten.view %26694, %26695 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24207 = torch.constant.int 4
%26697 = torch.aten.mul.int %int4_24207, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24208 = torch.constant.int 4096
%26698 = torch.prim.ListConstruct %26697, %int4096_24208 : (!torch.int, !torch.int) -> !torch.list<int>
%26699 = torch.aten.view %26663, %26698 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26699, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26700 = torch.aten.mm %26699, %26674 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26700, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24209 = torch.constant.int 4
%int512_24210 = torch.constant.int 512
%26701 = torch.prim.ListConstruct %int4_24209, %2482, %int512_24210 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26702 = torch.aten.view %26700, %26701 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24211 = torch.constant.int 4
%26703 = torch.aten.mul.int %int4_24211, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24212 = torch.constant.int 4096
%26704 = torch.prim.ListConstruct %26703, %int4096_24212 : (!torch.int, !torch.int) -> !torch.list<int>
%26705 = torch.aten.view %26664, %26704 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26705, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26706 = torch.aten.mm %26705, %26676 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26706, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24213 = torch.constant.int 4
%int512_24214 = torch.constant.int 512
%26707 = torch.prim.ListConstruct %int4_24213, %2482, %int512_24214 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26708 = torch.aten.view %26706, %26707 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24215 = torch.constant.int 4
%26709 = torch.aten.mul.int %int4_24215, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24216 = torch.constant.int 4096
%26710 = torch.prim.ListConstruct %26709, %int4096_24216 : (!torch.int, !torch.int) -> !torch.list<int>
%26711 = torch.aten.view %26665, %26710 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26711, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26712 = torch.aten.mm %26711, %26678 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26712, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24217 = torch.constant.int 4
%int512_24218 = torch.constant.int 512
%26713 = torch.prim.ListConstruct %int4_24217, %2482, %int512_24218 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26714 = torch.aten.view %26712, %26713 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24219 = torch.constant.int 4
%26715 = torch.aten.mul.int %int4_24219, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24220 = torch.constant.int 4096
%26716 = torch.prim.ListConstruct %26715, %int4096_24220 : (!torch.int, !torch.int) -> !torch.list<int>
%26717 = torch.aten.view %26666, %26716 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26717, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26718 = torch.aten.mm %26717, %26680 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26718, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24221 = torch.constant.int 4
%int512_24222 = torch.constant.int 512
%26719 = torch.prim.ListConstruct %int4_24221, %2482, %int512_24222 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26720 = torch.aten.view %26718, %26719 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24223 = torch.constant.int 4
%26721 = torch.aten.mul.int %int4_24223, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24224 = torch.constant.int 4096
%26722 = torch.prim.ListConstruct %26721, %int4096_24224 : (!torch.int, !torch.int) -> !torch.list<int>
%26723 = torch.aten.view %26667, %26722 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26723, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26724 = torch.aten.mm %26723, %26682 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26724, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24225 = torch.constant.int 4
%int512_24226 = torch.constant.int 512
%26725 = torch.prim.ListConstruct %int4_24225, %2482, %int512_24226 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26726 = torch.aten.view %26724, %26725 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_24227 = torch.constant.int 4
%26727 = torch.aten.mul.int %int4_24227, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24228 = torch.constant.int 4096
%26728 = torch.prim.ListConstruct %26727, %int4096_24228 : (!torch.int, !torch.int) -> !torch.list<int>
%26729 = torch.aten.view %26668, %26728 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26729, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26730 = torch.aten.mm %26729, %26684 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %26730, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_24229 = torch.constant.int 4
%int512_24230 = torch.constant.int 512
%26731 = torch.prim.ListConstruct %int4_24229, %2482, %int512_24230 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26732 = torch.aten.view %26730, %26731 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %26732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_24231 = torch.constant.int 1
%int0_24232 = torch.constant.int 0
%26733 = torch.prim.ListConstruct %int1_24231, %int0_24232 : (!torch.int, !torch.int) -> !torch.list<int>
%26734 = torch.aten.permute %960, %26733 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24233 = torch.constant.int 1
%int0_24234 = torch.constant.int 0
%26735 = torch.prim.ListConstruct %int1_24233, %int0_24234 : (!torch.int, !torch.int) -> !torch.list<int>
%26736 = torch.aten.permute %961, %26735 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24235 = torch.constant.int 1
%int0_24236 = torch.constant.int 0
%26737 = torch.prim.ListConstruct %int1_24235, %int0_24236 : (!torch.int, !torch.int) -> !torch.list<int>
%26738 = torch.aten.permute %962, %26737 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24237 = torch.constant.int 1
%int0_24238 = torch.constant.int 0
%26739 = torch.prim.ListConstruct %int1_24237, %int0_24238 : (!torch.int, !torch.int) -> !torch.list<int>
%26740 = torch.aten.permute %963, %26739 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24239 = torch.constant.int 1
%int0_24240 = torch.constant.int 0
%26741 = torch.prim.ListConstruct %int1_24239, %int0_24240 : (!torch.int, !torch.int) -> !torch.list<int>
%26742 = torch.aten.permute %964, %26741 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24241 = torch.constant.int 1
%int0_24242 = torch.constant.int 0
%26743 = torch.prim.ListConstruct %int1_24241, %int0_24242 : (!torch.int, !torch.int) -> !torch.list<int>
%26744 = torch.aten.permute %965, %26743 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24243 = torch.constant.int 1
%int0_24244 = torch.constant.int 0
%26745 = torch.prim.ListConstruct %int1_24243, %int0_24244 : (!torch.int, !torch.int) -> !torch.list<int>
%26746 = torch.aten.permute %966, %26745 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24245 = torch.constant.int 1
%int0_24246 = torch.constant.int 0
%26747 = torch.prim.ListConstruct %int1_24245, %int0_24246 : (!torch.int, !torch.int) -> !torch.list<int>
%26748 = torch.aten.permute %967, %26747 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_24247 = torch.constant.int 4
%26749 = torch.aten.mul.int %int4_24247, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24248 = torch.constant.int 4096
%26750 = torch.prim.ListConstruct %26749, %int4096_24248 : (!torch.int, !torch.int) -> !torch.list<int>
%26751 = torch.aten.view %26661, %26750 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26751, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26752 = torch.aten.mm %26751, %26734 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26752, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24249 = torch.constant.int 4
%int128_24250 = torch.constant.int 128
%26753 = torch.prim.ListConstruct %int4_24249, %2482, %int128_24250 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26754 = torch.aten.view %26752, %26753 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24251 = torch.constant.int 4
%26755 = torch.aten.mul.int %int4_24251, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24252 = torch.constant.int 4096
%26756 = torch.prim.ListConstruct %26755, %int4096_24252 : (!torch.int, !torch.int) -> !torch.list<int>
%26757 = torch.aten.view %26662, %26756 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26757, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26758 = torch.aten.mm %26757, %26736 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26758, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24253 = torch.constant.int 4
%int128_24254 = torch.constant.int 128
%26759 = torch.prim.ListConstruct %int4_24253, %2482, %int128_24254 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26760 = torch.aten.view %26758, %26759 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24255 = torch.constant.int 4
%26761 = torch.aten.mul.int %int4_24255, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24256 = torch.constant.int 4096
%26762 = torch.prim.ListConstruct %26761, %int4096_24256 : (!torch.int, !torch.int) -> !torch.list<int>
%26763 = torch.aten.view %26663, %26762 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26763, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26764 = torch.aten.mm %26763, %26738 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26764, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24257 = torch.constant.int 4
%int128_24258 = torch.constant.int 128
%26765 = torch.prim.ListConstruct %int4_24257, %2482, %int128_24258 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26766 = torch.aten.view %26764, %26765 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24259 = torch.constant.int 4
%26767 = torch.aten.mul.int %int4_24259, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24260 = torch.constant.int 4096
%26768 = torch.prim.ListConstruct %26767, %int4096_24260 : (!torch.int, !torch.int) -> !torch.list<int>
%26769 = torch.aten.view %26664, %26768 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26769, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26770 = torch.aten.mm %26769, %26740 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26770, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24261 = torch.constant.int 4
%int128_24262 = torch.constant.int 128
%26771 = torch.prim.ListConstruct %int4_24261, %2482, %int128_24262 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26772 = torch.aten.view %26770, %26771 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24263 = torch.constant.int 4
%26773 = torch.aten.mul.int %int4_24263, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24264 = torch.constant.int 4096
%26774 = torch.prim.ListConstruct %26773, %int4096_24264 : (!torch.int, !torch.int) -> !torch.list<int>
%26775 = torch.aten.view %26665, %26774 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26775, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26776 = torch.aten.mm %26775, %26742 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26776, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24265 = torch.constant.int 4
%int128_24266 = torch.constant.int 128
%26777 = torch.prim.ListConstruct %int4_24265, %2482, %int128_24266 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26778 = torch.aten.view %26776, %26777 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24267 = torch.constant.int 4
%26779 = torch.aten.mul.int %int4_24267, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24268 = torch.constant.int 4096
%26780 = torch.prim.ListConstruct %26779, %int4096_24268 : (!torch.int, !torch.int) -> !torch.list<int>
%26781 = torch.aten.view %26666, %26780 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26781, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26782 = torch.aten.mm %26781, %26744 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26782, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24269 = torch.constant.int 4
%int128_24270 = torch.constant.int 128
%26783 = torch.prim.ListConstruct %int4_24269, %2482, %int128_24270 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26784 = torch.aten.view %26782, %26783 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24271 = torch.constant.int 4
%26785 = torch.aten.mul.int %int4_24271, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24272 = torch.constant.int 4096
%26786 = torch.prim.ListConstruct %26785, %int4096_24272 : (!torch.int, !torch.int) -> !torch.list<int>
%26787 = torch.aten.view %26667, %26786 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26787, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26788 = torch.aten.mm %26787, %26746 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26788, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24273 = torch.constant.int 4
%int128_24274 = torch.constant.int 128
%26789 = torch.prim.ListConstruct %int4_24273, %2482, %int128_24274 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26790 = torch.aten.view %26788, %26789 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24275 = torch.constant.int 4
%26791 = torch.aten.mul.int %int4_24275, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24276 = torch.constant.int 4096
%26792 = torch.prim.ListConstruct %26791, %int4096_24276 : (!torch.int, !torch.int) -> !torch.list<int>
%26793 = torch.aten.view %26668, %26792 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26793, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26794 = torch.aten.mm %26793, %26748 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26794, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24277 = torch.constant.int 4
%int128_24278 = torch.constant.int 128
%26795 = torch.prim.ListConstruct %int4_24277, %2482, %int128_24278 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26796 = torch.aten.view %26794, %26795 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_24279 = torch.constant.int 1
%int0_24280 = torch.constant.int 0
%26797 = torch.prim.ListConstruct %int1_24279, %int0_24280 : (!torch.int, !torch.int) -> !torch.list<int>
%26798 = torch.aten.permute %968, %26797 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24281 = torch.constant.int 1
%int0_24282 = torch.constant.int 0
%26799 = torch.prim.ListConstruct %int1_24281, %int0_24282 : (!torch.int, !torch.int) -> !torch.list<int>
%26800 = torch.aten.permute %969, %26799 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24283 = torch.constant.int 1
%int0_24284 = torch.constant.int 0
%26801 = torch.prim.ListConstruct %int1_24283, %int0_24284 : (!torch.int, !torch.int) -> !torch.list<int>
%26802 = torch.aten.permute %970, %26801 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24285 = torch.constant.int 1
%int0_24286 = torch.constant.int 0
%26803 = torch.prim.ListConstruct %int1_24285, %int0_24286 : (!torch.int, !torch.int) -> !torch.list<int>
%26804 = torch.aten.permute %971, %26803 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24287 = torch.constant.int 1
%int0_24288 = torch.constant.int 0
%26805 = torch.prim.ListConstruct %int1_24287, %int0_24288 : (!torch.int, !torch.int) -> !torch.list<int>
%26806 = torch.aten.permute %972, %26805 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24289 = torch.constant.int 1
%int0_24290 = torch.constant.int 0
%26807 = torch.prim.ListConstruct %int1_24289, %int0_24290 : (!torch.int, !torch.int) -> !torch.list<int>
%26808 = torch.aten.permute %973, %26807 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24291 = torch.constant.int 1
%int0_24292 = torch.constant.int 0
%26809 = torch.prim.ListConstruct %int1_24291, %int0_24292 : (!torch.int, !torch.int) -> !torch.list<int>
%26810 = torch.aten.permute %974, %26809 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_24293 = torch.constant.int 1
%int0_24294 = torch.constant.int 0
%26811 = torch.prim.ListConstruct %int1_24293, %int0_24294 : (!torch.int, !torch.int) -> !torch.list<int>
%26812 = torch.aten.permute %975, %26811 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_24295 = torch.constant.int 4
%26813 = torch.aten.mul.int %int4_24295, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24296 = torch.constant.int 4096
%26814 = torch.prim.ListConstruct %26813, %int4096_24296 : (!torch.int, !torch.int) -> !torch.list<int>
%26815 = torch.aten.view %26661, %26814 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26815, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26816 = torch.aten.mm %26815, %26798 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26816, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24297 = torch.constant.int 4
%int128_24298 = torch.constant.int 128
%26817 = torch.prim.ListConstruct %int4_24297, %2482, %int128_24298 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26818 = torch.aten.view %26816, %26817 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24299 = torch.constant.int 4
%26819 = torch.aten.mul.int %int4_24299, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24300 = torch.constant.int 4096
%26820 = torch.prim.ListConstruct %26819, %int4096_24300 : (!torch.int, !torch.int) -> !torch.list<int>
%26821 = torch.aten.view %26662, %26820 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26821, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26822 = torch.aten.mm %26821, %26800 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26822, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24301 = torch.constant.int 4
%int128_24302 = torch.constant.int 128
%26823 = torch.prim.ListConstruct %int4_24301, %2482, %int128_24302 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26824 = torch.aten.view %26822, %26823 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24303 = torch.constant.int 4
%26825 = torch.aten.mul.int %int4_24303, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24304 = torch.constant.int 4096
%26826 = torch.prim.ListConstruct %26825, %int4096_24304 : (!torch.int, !torch.int) -> !torch.list<int>
%26827 = torch.aten.view %26663, %26826 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26827, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26828 = torch.aten.mm %26827, %26802 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26828, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24305 = torch.constant.int 4
%int128_24306 = torch.constant.int 128
%26829 = torch.prim.ListConstruct %int4_24305, %2482, %int128_24306 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26830 = torch.aten.view %26828, %26829 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24307 = torch.constant.int 4
%26831 = torch.aten.mul.int %int4_24307, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24308 = torch.constant.int 4096
%26832 = torch.prim.ListConstruct %26831, %int4096_24308 : (!torch.int, !torch.int) -> !torch.list<int>
%26833 = torch.aten.view %26664, %26832 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26833, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26834 = torch.aten.mm %26833, %26804 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26834, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24309 = torch.constant.int 4
%int128_24310 = torch.constant.int 128
%26835 = torch.prim.ListConstruct %int4_24309, %2482, %int128_24310 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26836 = torch.aten.view %26834, %26835 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24311 = torch.constant.int 4
%26837 = torch.aten.mul.int %int4_24311, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24312 = torch.constant.int 4096
%26838 = torch.prim.ListConstruct %26837, %int4096_24312 : (!torch.int, !torch.int) -> !torch.list<int>
%26839 = torch.aten.view %26665, %26838 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26839, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26840 = torch.aten.mm %26839, %26806 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26840, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24313 = torch.constant.int 4
%int128_24314 = torch.constant.int 128
%26841 = torch.prim.ListConstruct %int4_24313, %2482, %int128_24314 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26842 = torch.aten.view %26840, %26841 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24315 = torch.constant.int 4
%26843 = torch.aten.mul.int %int4_24315, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24316 = torch.constant.int 4096
%26844 = torch.prim.ListConstruct %26843, %int4096_24316 : (!torch.int, !torch.int) -> !torch.list<int>
%26845 = torch.aten.view %26666, %26844 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26845, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26846 = torch.aten.mm %26845, %26808 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26846, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24317 = torch.constant.int 4
%int128_24318 = torch.constant.int 128
%26847 = torch.prim.ListConstruct %int4_24317, %2482, %int128_24318 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26848 = torch.aten.view %26846, %26847 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24319 = torch.constant.int 4
%26849 = torch.aten.mul.int %int4_24319, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24320 = torch.constant.int 4096
%26850 = torch.prim.ListConstruct %26849, %int4096_24320 : (!torch.int, !torch.int) -> !torch.list<int>
%26851 = torch.aten.view %26667, %26850 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26851, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26852 = torch.aten.mm %26851, %26810 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26852, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24321 = torch.constant.int 4
%int128_24322 = torch.constant.int 128
%26853 = torch.prim.ListConstruct %int4_24321, %2482, %int128_24322 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26854 = torch.aten.view %26852, %26853 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24323 = torch.constant.int 4
%26855 = torch.aten.mul.int %int4_24323, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_24324 = torch.constant.int 4096
%26856 = torch.prim.ListConstruct %26855, %int4096_24324 : (!torch.int, !torch.int) -> !torch.list<int>
%26857 = torch.aten.view %26668, %26856 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %26857, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%26858 = torch.aten.mm %26857, %26812 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %26858, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_24325 = torch.constant.int 4
%int128_24326 = torch.constant.int 128
%26859 = torch.prim.ListConstruct %int4_24325, %2482, %int128_24326 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26860 = torch.aten.view %26858, %26859 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %26860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_24327 = torch.constant.int 4
%int4_24328 = torch.constant.int 4
%int128_24329 = torch.constant.int 128
%26861 = torch.prim.ListConstruct %int4_24327, %2482, %int4_24328, %int128_24329 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26862 = torch.aten.view %26690, %26861 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24330 = torch.constant.int 4
%int4_24331 = torch.constant.int 4
%int128_24332 = torch.constant.int 128
%26863 = torch.prim.ListConstruct %int4_24330, %2482, %int4_24331, %int128_24332 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26864 = torch.aten.view %26696, %26863 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24333 = torch.constant.int 4
%int4_24334 = torch.constant.int 4
%int128_24335 = torch.constant.int 128
%26865 = torch.prim.ListConstruct %int4_24333, %2482, %int4_24334, %int128_24335 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26866 = torch.aten.view %26702, %26865 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24336 = torch.constant.int 4
%int4_24337 = torch.constant.int 4
%int128_24338 = torch.constant.int 128
%26867 = torch.prim.ListConstruct %int4_24336, %2482, %int4_24337, %int128_24338 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26868 = torch.aten.view %26708, %26867 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24339 = torch.constant.int 4
%int4_24340 = torch.constant.int 4
%int128_24341 = torch.constant.int 128
%26869 = torch.prim.ListConstruct %int4_24339, %2482, %int4_24340, %int128_24341 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26870 = torch.aten.view %26714, %26869 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24342 = torch.constant.int 4
%int4_24343 = torch.constant.int 4
%int128_24344 = torch.constant.int 128
%26871 = torch.prim.ListConstruct %int4_24342, %2482, %int4_24343, %int128_24344 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26872 = torch.aten.view %26720, %26871 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24345 = torch.constant.int 4
%int4_24346 = torch.constant.int 4
%int128_24347 = torch.constant.int 128
%26873 = torch.prim.ListConstruct %int4_24345, %2482, %int4_24346, %int128_24347 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26874 = torch.aten.view %26726, %26873 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24348 = torch.constant.int 4
%int4_24349 = torch.constant.int 4
%int128_24350 = torch.constant.int 128
%26875 = torch.prim.ListConstruct %int4_24348, %2482, %int4_24349, %int128_24350 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26876 = torch.aten.view %26732, %26875 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_24351 = torch.constant.int 4
%int1_24352 = torch.constant.int 1
%int128_24353 = torch.constant.int 128
%26877 = torch.prim.ListConstruct %int4_24351, %2482, %int1_24352, %int128_24353 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26878 = torch.aten.view %26754, %26877 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24354 = torch.constant.int 4
%int1_24355 = torch.constant.int 1
%int128_24356 = torch.constant.int 128
%26879 = torch.prim.ListConstruct %int4_24354, %2482, %int1_24355, %int128_24356 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26880 = torch.aten.view %26760, %26879 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24357 = torch.constant.int 4
%int1_24358 = torch.constant.int 1
%int128_24359 = torch.constant.int 128
%26881 = torch.prim.ListConstruct %int4_24357, %2482, %int1_24358, %int128_24359 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26882 = torch.aten.view %26766, %26881 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24360 = torch.constant.int 4
%int1_24361 = torch.constant.int 1
%int128_24362 = torch.constant.int 128
%26883 = torch.prim.ListConstruct %int4_24360, %2482, %int1_24361, %int128_24362 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26884 = torch.aten.view %26772, %26883 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24363 = torch.constant.int 4
%int1_24364 = torch.constant.int 1
%int128_24365 = torch.constant.int 128
%26885 = torch.prim.ListConstruct %int4_24363, %2482, %int1_24364, %int128_24365 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26886 = torch.aten.view %26778, %26885 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24366 = torch.constant.int 4
%int1_24367 = torch.constant.int 1
%int128_24368 = torch.constant.int 128
%26887 = torch.prim.ListConstruct %int4_24366, %2482, %int1_24367, %int128_24368 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26888 = torch.aten.view %26784, %26887 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24369 = torch.constant.int 4
%int1_24370 = torch.constant.int 1
%int128_24371 = torch.constant.int 128
%26889 = torch.prim.ListConstruct %int4_24369, %2482, %int1_24370, %int128_24371 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26890 = torch.aten.view %26790, %26889 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24372 = torch.constant.int 4
%int1_24373 = torch.constant.int 1
%int128_24374 = torch.constant.int 128
%26891 = torch.prim.ListConstruct %int4_24372, %2482, %int1_24373, %int128_24374 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26892 = torch.aten.view %26796, %26891 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24375 = torch.constant.int 4
%int1_24376 = torch.constant.int 1
%int128_24377 = torch.constant.int 128
%26893 = torch.prim.ListConstruct %int4_24375, %2482, %int1_24376, %int128_24377 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26894 = torch.aten.view %26818, %26893 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24378 = torch.constant.int 4
%int1_24379 = torch.constant.int 1
%int128_24380 = torch.constant.int 128
%26895 = torch.prim.ListConstruct %int4_24378, %2482, %int1_24379, %int128_24380 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26896 = torch.aten.view %26824, %26895 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24381 = torch.constant.int 4
%int1_24382 = torch.constant.int 1
%int128_24383 = torch.constant.int 128
%26897 = torch.prim.ListConstruct %int4_24381, %2482, %int1_24382, %int128_24383 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26898 = torch.aten.view %26830, %26897 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24384 = torch.constant.int 4
%int1_24385 = torch.constant.int 1
%int128_24386 = torch.constant.int 128
%26899 = torch.prim.ListConstruct %int4_24384, %2482, %int1_24385, %int128_24386 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26900 = torch.aten.view %26836, %26899 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24387 = torch.constant.int 4
%int1_24388 = torch.constant.int 1
%int128_24389 = torch.constant.int 128
%26901 = torch.prim.ListConstruct %int4_24387, %2482, %int1_24388, %int128_24389 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26902 = torch.aten.view %26842, %26901 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24390 = torch.constant.int 4
%int1_24391 = torch.constant.int 1
%int128_24392 = torch.constant.int 128
%26903 = torch.prim.ListConstruct %int4_24390, %2482, %int1_24391, %int128_24392 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26904 = torch.aten.view %26848, %26903 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24393 = torch.constant.int 4
%int1_24394 = torch.constant.int 1
%int128_24395 = torch.constant.int 128
%26905 = torch.prim.ListConstruct %int4_24393, %2482, %int1_24394, %int128_24395 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26906 = torch.aten.view %26854, %26905 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_24396 = torch.constant.int 4
%int1_24397 = torch.constant.int 1
%int128_24398 = torch.constant.int 128
%26907 = torch.prim.ListConstruct %int4_24396, %2482, %int1_24397, %int128_24398 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%26908 = torch.aten.view %26860, %26907 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %26908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_24399 = torch.constant.int 131072
%none_24400 = torch.constant.none
%none_24401 = torch.constant.none
%cpu_24402 = torch.constant.device "cpu"
%false_24403 = torch.constant.bool false
%26909 = torch.aten.arange %int131072_24399, %none_24400, %none_24401, %cpu_24402, %false_24403 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_24404 = torch.constant.int 0
%int128_24405 = torch.constant.int 128
%int2_24406 = torch.constant.int 2
%none_24407 = torch.constant.none
%none_24408 = torch.constant.none
%cpu_24409 = torch.constant.device "cpu"
%false_24410 = torch.constant.bool false
%26910 = torch.aten.arange.start_step %int0_24404, %int128_24405, %int2_24406, %none_24407, %none_24408, %cpu_24409, %false_24410 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_24411 = torch.constant.int 0
%int0_24412 = torch.constant.int 0
%int64_24413 = torch.constant.int 64
%int1_24414 = torch.constant.int 1
%26911 = torch.aten.slice.Tensor %26910, %int0_24411, %int0_24412, %int64_24413, %int1_24414 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_24415 = torch.constant.int 6
%26912 = torch.prims.convert_element_type %26911, %int6_24415 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_24416 = torch.constant.int 128
%26913 = torch.aten.div.Scalar %26912, %int128_24416 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_24417 = torch.constant.float 5.000000e+05
%26914 = torch.aten.pow.Scalar %float5.000000e05_24417, %26913 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%26915 = torch.aten.reciprocal %26914 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_24418 = torch.constant.float 1.000000e+00
%26916 = torch.aten.mul.Scalar %26915, %float1.000000e00_24418 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_24419 = torch.constant.int 131072
%int1_24420 = torch.constant.int 1
%26917 = torch.prim.ListConstruct %int131072_24419, %int1_24420 : (!torch.int, !torch.int) -> !torch.list<int>
%26918 = torch.aten.view %26909, %26917 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%26919 = torch.aten.mul.Tensor %26918, %26916 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%26920 = torch.aten.cos %26919 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%26921 = torch.aten.sin %26919 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%26922 = torch.aten.complex %26920, %26921 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%26923 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26924 = flow.tensor.transfer %26923 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%26925 = torch_c.from_builtin_tensor %26924 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26926 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26927 = flow.tensor.transfer %26926 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%26928 = torch_c.from_builtin_tensor %26927 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26929 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26930 = flow.tensor.transfer %26929 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%26931 = torch_c.from_builtin_tensor %26930 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26932 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26933 = flow.tensor.transfer %26932 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%26934 = torch_c.from_builtin_tensor %26933 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26935 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26936 = flow.tensor.transfer %26935 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%26937 = torch_c.from_builtin_tensor %26936 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26938 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26939 = flow.tensor.transfer %26938 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%26940 = torch_c.from_builtin_tensor %26939 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26941 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26942 = flow.tensor.transfer %26941 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%26943 = torch_c.from_builtin_tensor %26942 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%26944 = torch_c.to_builtin_tensor %26922 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%26945 = flow.tensor.transfer %26944 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%26946 = torch_c.from_builtin_tensor %26945 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_24421 = torch.constant.int 1
%26947 = torch.aten.size.int %26690, %int1_24421 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24422 = torch.constant.int 0
%26948 = torch.aten.add.int %int0_24422, %26947 : !torch.int, !torch.int -> !torch.int
%int0_24423 = torch.constant.int 0
%int0_24424 = torch.constant.int 0
%int1_24425 = torch.constant.int 1
%26949 = torch.aten.slice.Tensor %26925, %int0_24423, %int0_24424, %26948, %int1_24425 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26949, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24426 = torch.constant.int 1
%int0_24427 = torch.constant.int 0
%int9223372036854775807_24428 = torch.constant.int 9223372036854775807
%int1_24429 = torch.constant.int 1
%26950 = torch.aten.slice.Tensor %26949, %int1_24426, %int0_24427, %int9223372036854775807_24428, %int1_24429 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26950, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24430 = torch.constant.int 0
%26951 = torch.aten.unsqueeze %26950, %int0_24430 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %26951, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24431 = torch.constant.int 2
%26952 = torch.aten.unsqueeze %26951, %int2_24431 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26952, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24432 = torch.constant.int 3
%int0_24433 = torch.constant.int 0
%int9223372036854775807_24434 = torch.constant.int 9223372036854775807
%int1_24435 = torch.constant.int 1
%26953 = torch.aten.slice.Tensor %26952, %int3_24432, %int0_24433, %int9223372036854775807_24434, %int1_24435 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26953, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%26954 = torch_c.to_builtin_tensor %26862 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24436 = arith.constant 1 : index
%dim_24437 = tensor.dim %26954, %c1_24436 : tensor<4x?x4x128xf16>
%26955 = flow.tensor.bitcast %26954 : tensor<4x?x4x128xf16>{%dim_24437} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24437}
%26956 = torch_c.from_builtin_tensor %26955 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %26956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%26957 = torch.aten.mul.Tensor %26956, %26953 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %26957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%26958 = torch_c.to_builtin_tensor %26957 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24438 = arith.constant 1 : index
%dim_24439 = tensor.dim %26958, %c1_24438 : tensor<4x?x4x64xcomplex<f32>>
%26959 = flow.tensor.bitcast %26958 : tensor<4x?x4x64xcomplex<f32>>{%dim_24439} -> tensor<4x?x4x128xf32>{%dim_24439}
%26960 = torch_c.from_builtin_tensor %26959 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %26960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24440 = torch.constant.int 5
%26961 = torch.prims.convert_element_type %26960, %int5_24440 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24441 = torch.constant.int 1
%26962 = torch.aten.size.int %26696, %int1_24441 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24442 = torch.constant.int 0
%26963 = torch.aten.add.int %int0_24442, %26962 : !torch.int, !torch.int -> !torch.int
%int0_24443 = torch.constant.int 0
%int0_24444 = torch.constant.int 0
%int1_24445 = torch.constant.int 1
%26964 = torch.aten.slice.Tensor %26928, %int0_24443, %int0_24444, %26963, %int1_24445 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26964, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24446 = torch.constant.int 1
%int0_24447 = torch.constant.int 0
%int9223372036854775807_24448 = torch.constant.int 9223372036854775807
%int1_24449 = torch.constant.int 1
%26965 = torch.aten.slice.Tensor %26964, %int1_24446, %int0_24447, %int9223372036854775807_24448, %int1_24449 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26965, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24450 = torch.constant.int 0
%26966 = torch.aten.unsqueeze %26965, %int0_24450 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %26966, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24451 = torch.constant.int 2
%26967 = torch.aten.unsqueeze %26966, %int2_24451 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26967, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24452 = torch.constant.int 3
%int0_24453 = torch.constant.int 0
%int9223372036854775807_24454 = torch.constant.int 9223372036854775807
%int1_24455 = torch.constant.int 1
%26968 = torch.aten.slice.Tensor %26967, %int3_24452, %int0_24453, %int9223372036854775807_24454, %int1_24455 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26968, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%26969 = torch_c.to_builtin_tensor %26864 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24456 = arith.constant 1 : index
%dim_24457 = tensor.dim %26969, %c1_24456 : tensor<4x?x4x128xf16>
%26970 = flow.tensor.bitcast %26969 : tensor<4x?x4x128xf16>{%dim_24457} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24457}
%26971 = torch_c.from_builtin_tensor %26970 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %26971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%26972 = torch.aten.mul.Tensor %26971, %26968 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %26972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%26973 = torch_c.to_builtin_tensor %26972 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24458 = arith.constant 1 : index
%dim_24459 = tensor.dim %26973, %c1_24458 : tensor<4x?x4x64xcomplex<f32>>
%26974 = flow.tensor.bitcast %26973 : tensor<4x?x4x64xcomplex<f32>>{%dim_24459} -> tensor<4x?x4x128xf32>{%dim_24459}
%26975 = torch_c.from_builtin_tensor %26974 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %26975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24460 = torch.constant.int 5
%26976 = torch.prims.convert_element_type %26975, %int5_24460 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24461 = torch.constant.int 1
%26977 = torch.aten.size.int %26702, %int1_24461 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24462 = torch.constant.int 0
%26978 = torch.aten.add.int %int0_24462, %26977 : !torch.int, !torch.int -> !torch.int
%int0_24463 = torch.constant.int 0
%int0_24464 = torch.constant.int 0
%int1_24465 = torch.constant.int 1
%26979 = torch.aten.slice.Tensor %26931, %int0_24463, %int0_24464, %26978, %int1_24465 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26979, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24466 = torch.constant.int 1
%int0_24467 = torch.constant.int 0
%int9223372036854775807_24468 = torch.constant.int 9223372036854775807
%int1_24469 = torch.constant.int 1
%26980 = torch.aten.slice.Tensor %26979, %int1_24466, %int0_24467, %int9223372036854775807_24468, %int1_24469 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26980, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24470 = torch.constant.int 0
%26981 = torch.aten.unsqueeze %26980, %int0_24470 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %26981, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24471 = torch.constant.int 2
%26982 = torch.aten.unsqueeze %26981, %int2_24471 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26982, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24472 = torch.constant.int 3
%int0_24473 = torch.constant.int 0
%int9223372036854775807_24474 = torch.constant.int 9223372036854775807
%int1_24475 = torch.constant.int 1
%26983 = torch.aten.slice.Tensor %26982, %int3_24472, %int0_24473, %int9223372036854775807_24474, %int1_24475 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26983, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%26984 = torch_c.to_builtin_tensor %26866 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24476 = arith.constant 1 : index
%dim_24477 = tensor.dim %26984, %c1_24476 : tensor<4x?x4x128xf16>
%26985 = flow.tensor.bitcast %26984 : tensor<4x?x4x128xf16>{%dim_24477} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24477}
%26986 = torch_c.from_builtin_tensor %26985 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %26986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%26987 = torch.aten.mul.Tensor %26986, %26983 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %26987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%26988 = torch_c.to_builtin_tensor %26987 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24478 = arith.constant 1 : index
%dim_24479 = tensor.dim %26988, %c1_24478 : tensor<4x?x4x64xcomplex<f32>>
%26989 = flow.tensor.bitcast %26988 : tensor<4x?x4x64xcomplex<f32>>{%dim_24479} -> tensor<4x?x4x128xf32>{%dim_24479}
%26990 = torch_c.from_builtin_tensor %26989 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %26990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24480 = torch.constant.int 5
%26991 = torch.prims.convert_element_type %26990, %int5_24480 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %26991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24481 = torch.constant.int 1
%26992 = torch.aten.size.int %26708, %int1_24481 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24482 = torch.constant.int 0
%26993 = torch.aten.add.int %int0_24482, %26992 : !torch.int, !torch.int -> !torch.int
%int0_24483 = torch.constant.int 0
%int0_24484 = torch.constant.int 0
%int1_24485 = torch.constant.int 1
%26994 = torch.aten.slice.Tensor %26934, %int0_24483, %int0_24484, %26993, %int1_24485 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26994, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24486 = torch.constant.int 1
%int0_24487 = torch.constant.int 0
%int9223372036854775807_24488 = torch.constant.int 9223372036854775807
%int1_24489 = torch.constant.int 1
%26995 = torch.aten.slice.Tensor %26994, %int1_24486, %int0_24487, %int9223372036854775807_24488, %int1_24489 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %26995, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24490 = torch.constant.int 0
%26996 = torch.aten.unsqueeze %26995, %int0_24490 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %26996, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24491 = torch.constant.int 2
%26997 = torch.aten.unsqueeze %26996, %int2_24491 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26997, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24492 = torch.constant.int 3
%int0_24493 = torch.constant.int 0
%int9223372036854775807_24494 = torch.constant.int 9223372036854775807
%int1_24495 = torch.constant.int 1
%26998 = torch.aten.slice.Tensor %26997, %int3_24492, %int0_24493, %int9223372036854775807_24494, %int1_24495 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %26998, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%26999 = torch_c.to_builtin_tensor %26868 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24496 = arith.constant 1 : index
%dim_24497 = tensor.dim %26999, %c1_24496 : tensor<4x?x4x128xf16>
%27000 = flow.tensor.bitcast %26999 : tensor<4x?x4x128xf16>{%dim_24497} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24497}
%27001 = torch_c.from_builtin_tensor %27000 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %27001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%27002 = torch.aten.mul.Tensor %27001, %26998 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %27002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%27003 = torch_c.to_builtin_tensor %27002 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24498 = arith.constant 1 : index
%dim_24499 = tensor.dim %27003, %c1_24498 : tensor<4x?x4x64xcomplex<f32>>
%27004 = flow.tensor.bitcast %27003 : tensor<4x?x4x64xcomplex<f32>>{%dim_24499} -> tensor<4x?x4x128xf32>{%dim_24499}
%27005 = torch_c.from_builtin_tensor %27004 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %27005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24500 = torch.constant.int 5
%27006 = torch.prims.convert_element_type %27005, %int5_24500 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24501 = torch.constant.int 1
%27007 = torch.aten.size.int %26714, %int1_24501 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24502 = torch.constant.int 0
%27008 = torch.aten.add.int %int0_24502, %27007 : !torch.int, !torch.int -> !torch.int
%int0_24503 = torch.constant.int 0
%int0_24504 = torch.constant.int 0
%int1_24505 = torch.constant.int 1
%27009 = torch.aten.slice.Tensor %26937, %int0_24503, %int0_24504, %27008, %int1_24505 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27009, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24506 = torch.constant.int 1
%int0_24507 = torch.constant.int 0
%int9223372036854775807_24508 = torch.constant.int 9223372036854775807
%int1_24509 = torch.constant.int 1
%27010 = torch.aten.slice.Tensor %27009, %int1_24506, %int0_24507, %int9223372036854775807_24508, %int1_24509 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27010, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24510 = torch.constant.int 0
%27011 = torch.aten.unsqueeze %27010, %int0_24510 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27011, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24511 = torch.constant.int 2
%27012 = torch.aten.unsqueeze %27011, %int2_24511 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27012, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24512 = torch.constant.int 3
%int0_24513 = torch.constant.int 0
%int9223372036854775807_24514 = torch.constant.int 9223372036854775807
%int1_24515 = torch.constant.int 1
%27013 = torch.aten.slice.Tensor %27012, %int3_24512, %int0_24513, %int9223372036854775807_24514, %int1_24515 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27013, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27014 = torch_c.to_builtin_tensor %26870 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24516 = arith.constant 1 : index
%dim_24517 = tensor.dim %27014, %c1_24516 : tensor<4x?x4x128xf16>
%27015 = flow.tensor.bitcast %27014 : tensor<4x?x4x128xf16>{%dim_24517} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24517}
%27016 = torch_c.from_builtin_tensor %27015 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %27016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%27017 = torch.aten.mul.Tensor %27016, %27013 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %27017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%27018 = torch_c.to_builtin_tensor %27017 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24518 = arith.constant 1 : index
%dim_24519 = tensor.dim %27018, %c1_24518 : tensor<4x?x4x64xcomplex<f32>>
%27019 = flow.tensor.bitcast %27018 : tensor<4x?x4x64xcomplex<f32>>{%dim_24519} -> tensor<4x?x4x128xf32>{%dim_24519}
%27020 = torch_c.from_builtin_tensor %27019 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %27020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24520 = torch.constant.int 5
%27021 = torch.prims.convert_element_type %27020, %int5_24520 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24521 = torch.constant.int 1
%27022 = torch.aten.size.int %26720, %int1_24521 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24522 = torch.constant.int 0
%27023 = torch.aten.add.int %int0_24522, %27022 : !torch.int, !torch.int -> !torch.int
%int0_24523 = torch.constant.int 0
%int0_24524 = torch.constant.int 0
%int1_24525 = torch.constant.int 1
%27024 = torch.aten.slice.Tensor %26940, %int0_24523, %int0_24524, %27023, %int1_24525 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27024, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24526 = torch.constant.int 1
%int0_24527 = torch.constant.int 0
%int9223372036854775807_24528 = torch.constant.int 9223372036854775807
%int1_24529 = torch.constant.int 1
%27025 = torch.aten.slice.Tensor %27024, %int1_24526, %int0_24527, %int9223372036854775807_24528, %int1_24529 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27025, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24530 = torch.constant.int 0
%27026 = torch.aten.unsqueeze %27025, %int0_24530 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27026, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24531 = torch.constant.int 2
%27027 = torch.aten.unsqueeze %27026, %int2_24531 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27027, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24532 = torch.constant.int 3
%int0_24533 = torch.constant.int 0
%int9223372036854775807_24534 = torch.constant.int 9223372036854775807
%int1_24535 = torch.constant.int 1
%27028 = torch.aten.slice.Tensor %27027, %int3_24532, %int0_24533, %int9223372036854775807_24534, %int1_24535 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27028, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27029 = torch_c.to_builtin_tensor %26872 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24536 = arith.constant 1 : index
%dim_24537 = tensor.dim %27029, %c1_24536 : tensor<4x?x4x128xf16>
%27030 = flow.tensor.bitcast %27029 : tensor<4x?x4x128xf16>{%dim_24537} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24537}
%27031 = torch_c.from_builtin_tensor %27030 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %27031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%27032 = torch.aten.mul.Tensor %27031, %27028 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %27032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%27033 = torch_c.to_builtin_tensor %27032 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24538 = arith.constant 1 : index
%dim_24539 = tensor.dim %27033, %c1_24538 : tensor<4x?x4x64xcomplex<f32>>
%27034 = flow.tensor.bitcast %27033 : tensor<4x?x4x64xcomplex<f32>>{%dim_24539} -> tensor<4x?x4x128xf32>{%dim_24539}
%27035 = torch_c.from_builtin_tensor %27034 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %27035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24540 = torch.constant.int 5
%27036 = torch.prims.convert_element_type %27035, %int5_24540 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24541 = torch.constant.int 1
%27037 = torch.aten.size.int %26726, %int1_24541 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24542 = torch.constant.int 0
%27038 = torch.aten.add.int %int0_24542, %27037 : !torch.int, !torch.int -> !torch.int
%int0_24543 = torch.constant.int 0
%int0_24544 = torch.constant.int 0
%int1_24545 = torch.constant.int 1
%27039 = torch.aten.slice.Tensor %26943, %int0_24543, %int0_24544, %27038, %int1_24545 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27039, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24546 = torch.constant.int 1
%int0_24547 = torch.constant.int 0
%int9223372036854775807_24548 = torch.constant.int 9223372036854775807
%int1_24549 = torch.constant.int 1
%27040 = torch.aten.slice.Tensor %27039, %int1_24546, %int0_24547, %int9223372036854775807_24548, %int1_24549 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27040, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24550 = torch.constant.int 0
%27041 = torch.aten.unsqueeze %27040, %int0_24550 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27041, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24551 = torch.constant.int 2
%27042 = torch.aten.unsqueeze %27041, %int2_24551 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27042, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24552 = torch.constant.int 3
%int0_24553 = torch.constant.int 0
%int9223372036854775807_24554 = torch.constant.int 9223372036854775807
%int1_24555 = torch.constant.int 1
%27043 = torch.aten.slice.Tensor %27042, %int3_24552, %int0_24553, %int9223372036854775807_24554, %int1_24555 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27043, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27044 = torch_c.to_builtin_tensor %26874 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24556 = arith.constant 1 : index
%dim_24557 = tensor.dim %27044, %c1_24556 : tensor<4x?x4x128xf16>
%27045 = flow.tensor.bitcast %27044 : tensor<4x?x4x128xf16>{%dim_24557} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24557}
%27046 = torch_c.from_builtin_tensor %27045 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %27046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%27047 = torch.aten.mul.Tensor %27046, %27043 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %27047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%27048 = torch_c.to_builtin_tensor %27047 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24558 = arith.constant 1 : index
%dim_24559 = tensor.dim %27048, %c1_24558 : tensor<4x?x4x64xcomplex<f32>>
%27049 = flow.tensor.bitcast %27048 : tensor<4x?x4x64xcomplex<f32>>{%dim_24559} -> tensor<4x?x4x128xf32>{%dim_24559}
%27050 = torch_c.from_builtin_tensor %27049 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %27050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24560 = torch.constant.int 5
%27051 = torch.prims.convert_element_type %27050, %int5_24560 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_24561 = torch.constant.int 1
%27052 = torch.aten.size.int %26732, %int1_24561 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_24562 = torch.constant.int 0
%27053 = torch.aten.add.int %int0_24562, %27052 : !torch.int, !torch.int -> !torch.int
%int0_24563 = torch.constant.int 0
%int0_24564 = torch.constant.int 0
%int1_24565 = torch.constant.int 1
%27054 = torch.aten.slice.Tensor %26946, %int0_24563, %int0_24564, %27053, %int1_24565 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27054, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24566 = torch.constant.int 1
%int0_24567 = torch.constant.int 0
%int9223372036854775807_24568 = torch.constant.int 9223372036854775807
%int1_24569 = torch.constant.int 1
%27055 = torch.aten.slice.Tensor %27054, %int1_24566, %int0_24567, %int9223372036854775807_24568, %int1_24569 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27055, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24570 = torch.constant.int 0
%27056 = torch.aten.unsqueeze %27055, %int0_24570 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27056, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24571 = torch.constant.int 2
%27057 = torch.aten.unsqueeze %27056, %int2_24571 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27057, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24572 = torch.constant.int 3
%int0_24573 = torch.constant.int 0
%int9223372036854775807_24574 = torch.constant.int 9223372036854775807
%int1_24575 = torch.constant.int 1
%27058 = torch.aten.slice.Tensor %27057, %int3_24572, %int0_24573, %int9223372036854775807_24574, %int1_24575 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27058, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27059 = torch_c.to_builtin_tensor %26876 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_24576 = arith.constant 1 : index
%dim_24577 = tensor.dim %27059, %c1_24576 : tensor<4x?x4x128xf16>
%27060 = flow.tensor.bitcast %27059 : tensor<4x?x4x128xf16>{%dim_24577} -> tensor<4x?x4x64xcomplex<f16>>{%dim_24577}
%27061 = torch_c.from_builtin_tensor %27060 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %27061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%27062 = torch.aten.mul.Tensor %27061, %27058 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %27062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%27063 = torch_c.to_builtin_tensor %27062 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_24578 = arith.constant 1 : index
%dim_24579 = tensor.dim %27063, %c1_24578 : tensor<4x?x4x64xcomplex<f32>>
%27064 = flow.tensor.bitcast %27063 : tensor<4x?x4x64xcomplex<f32>>{%dim_24579} -> tensor<4x?x4x128xf32>{%dim_24579}
%27065 = torch_c.from_builtin_tensor %27064 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %27065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_24580 = torch.constant.int 5
%27066 = torch.prims.convert_element_type %27065, %int5_24580 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_24581 = torch.constant.int 131072
%none_24582 = torch.constant.none
%none_24583 = torch.constant.none
%cpu_24584 = torch.constant.device "cpu"
%false_24585 = torch.constant.bool false
%27067 = torch.aten.arange %int131072_24581, %none_24582, %none_24583, %cpu_24584, %false_24585 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_24586 = torch.constant.int 0
%int128_24587 = torch.constant.int 128
%int2_24588 = torch.constant.int 2
%none_24589 = torch.constant.none
%none_24590 = torch.constant.none
%cpu_24591 = torch.constant.device "cpu"
%false_24592 = torch.constant.bool false
%27068 = torch.aten.arange.start_step %int0_24586, %int128_24587, %int2_24588, %none_24589, %none_24590, %cpu_24591, %false_24592 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_24593 = torch.constant.int 0
%int0_24594 = torch.constant.int 0
%int64_24595 = torch.constant.int 64
%int1_24596 = torch.constant.int 1
%27069 = torch.aten.slice.Tensor %27068, %int0_24593, %int0_24594, %int64_24595, %int1_24596 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_24597 = torch.constant.int 6
%27070 = torch.prims.convert_element_type %27069, %int6_24597 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_24598 = torch.constant.int 128
%27071 = torch.aten.div.Scalar %27070, %int128_24598 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_24599 = torch.constant.float 5.000000e+05
%27072 = torch.aten.pow.Scalar %float5.000000e05_24599, %27071 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%27073 = torch.aten.reciprocal %27072 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_24600 = torch.constant.float 1.000000e+00
%27074 = torch.aten.mul.Scalar %27073, %float1.000000e00_24600 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_24601 = torch.constant.int 131072
%int1_24602 = torch.constant.int 1
%27075 = torch.prim.ListConstruct %int131072_24601, %int1_24602 : (!torch.int, !torch.int) -> !torch.list<int>
%27076 = torch.aten.view %27067, %27075 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%27077 = torch.aten.mul.Tensor %27076, %27074 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%27078 = torch.aten.cos %27077 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%27079 = torch.aten.sin %27077 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%27080 = torch.aten.complex %27078, %27079 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%27081 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27082 = flow.tensor.transfer %27081 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%27083 = torch_c.from_builtin_tensor %27082 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27084 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27085 = flow.tensor.transfer %27084 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%27086 = torch_c.from_builtin_tensor %27085 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27087 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27088 = flow.tensor.transfer %27087 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%27089 = torch_c.from_builtin_tensor %27088 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27090 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27091 = flow.tensor.transfer %27090 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%27092 = torch_c.from_builtin_tensor %27091 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27093 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27094 = flow.tensor.transfer %27093 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%27095 = torch_c.from_builtin_tensor %27094 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27096 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27097 = flow.tensor.transfer %27096 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%27098 = torch_c.from_builtin_tensor %27097 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27099 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27100 = flow.tensor.transfer %27099 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%27101 = torch_c.from_builtin_tensor %27100 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%27102 = torch_c.to_builtin_tensor %27080 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%27103 = flow.tensor.transfer %27102 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%27104 = torch_c.from_builtin_tensor %27103 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_24603 = torch.constant.int 1
%27105 = torch.aten.size.int %26754, %int1_24603 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24604 = torch.constant.int 0
%27106 = torch.aten.add.int %int0_24604, %27105 : !torch.int, !torch.int -> !torch.int
%int0_24605 = torch.constant.int 0
%int0_24606 = torch.constant.int 0
%int1_24607 = torch.constant.int 1
%27107 = torch.aten.slice.Tensor %27083, %int0_24605, %int0_24606, %27106, %int1_24607 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27107, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24608 = torch.constant.int 1
%int0_24609 = torch.constant.int 0
%int9223372036854775807_24610 = torch.constant.int 9223372036854775807
%int1_24611 = torch.constant.int 1
%27108 = torch.aten.slice.Tensor %27107, %int1_24608, %int0_24609, %int9223372036854775807_24610, %int1_24611 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27108, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24612 = torch.constant.int 0
%27109 = torch.aten.unsqueeze %27108, %int0_24612 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27109, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24613 = torch.constant.int 2
%27110 = torch.aten.unsqueeze %27109, %int2_24613 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27110, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24614 = torch.constant.int 3
%int0_24615 = torch.constant.int 0
%int9223372036854775807_24616 = torch.constant.int 9223372036854775807
%int1_24617 = torch.constant.int 1
%27111 = torch.aten.slice.Tensor %27110, %int3_24614, %int0_24615, %int9223372036854775807_24616, %int1_24617 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27111, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27112 = torch_c.to_builtin_tensor %26878 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24618 = arith.constant 1 : index
%dim_24619 = tensor.dim %27112, %c1_24618 : tensor<4x?x1x128xf16>
%27113 = flow.tensor.bitcast %27112 : tensor<4x?x1x128xf16>{%dim_24619} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24619}
%27114 = torch_c.from_builtin_tensor %27113 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27115 = torch.aten.mul.Tensor %27114, %27111 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27116 = torch_c.to_builtin_tensor %27115 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24620 = arith.constant 1 : index
%dim_24621 = tensor.dim %27116, %c1_24620 : tensor<4x?x1x64xcomplex<f32>>
%27117 = flow.tensor.bitcast %27116 : tensor<4x?x1x64xcomplex<f32>>{%dim_24621} -> tensor<4x?x1x128xf32>{%dim_24621}
%27118 = torch_c.from_builtin_tensor %27117 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24622 = torch.constant.int 5
%27119 = torch.prims.convert_element_type %27118, %int5_24622 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24623 = torch.constant.int 1
%27120 = torch.aten.size.int %26760, %int1_24623 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24624 = torch.constant.int 0
%27121 = torch.aten.add.int %int0_24624, %27120 : !torch.int, !torch.int -> !torch.int
%int0_24625 = torch.constant.int 0
%int0_24626 = torch.constant.int 0
%int1_24627 = torch.constant.int 1
%27122 = torch.aten.slice.Tensor %27086, %int0_24625, %int0_24626, %27121, %int1_24627 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27122, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24628 = torch.constant.int 1
%int0_24629 = torch.constant.int 0
%int9223372036854775807_24630 = torch.constant.int 9223372036854775807
%int1_24631 = torch.constant.int 1
%27123 = torch.aten.slice.Tensor %27122, %int1_24628, %int0_24629, %int9223372036854775807_24630, %int1_24631 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27123, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24632 = torch.constant.int 0
%27124 = torch.aten.unsqueeze %27123, %int0_24632 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27124, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24633 = torch.constant.int 2
%27125 = torch.aten.unsqueeze %27124, %int2_24633 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27125, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24634 = torch.constant.int 3
%int0_24635 = torch.constant.int 0
%int9223372036854775807_24636 = torch.constant.int 9223372036854775807
%int1_24637 = torch.constant.int 1
%27126 = torch.aten.slice.Tensor %27125, %int3_24634, %int0_24635, %int9223372036854775807_24636, %int1_24637 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27126, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27127 = torch_c.to_builtin_tensor %26880 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24638 = arith.constant 1 : index
%dim_24639 = tensor.dim %27127, %c1_24638 : tensor<4x?x1x128xf16>
%27128 = flow.tensor.bitcast %27127 : tensor<4x?x1x128xf16>{%dim_24639} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24639}
%27129 = torch_c.from_builtin_tensor %27128 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27130 = torch.aten.mul.Tensor %27129, %27126 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27131 = torch_c.to_builtin_tensor %27130 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24640 = arith.constant 1 : index
%dim_24641 = tensor.dim %27131, %c1_24640 : tensor<4x?x1x64xcomplex<f32>>
%27132 = flow.tensor.bitcast %27131 : tensor<4x?x1x64xcomplex<f32>>{%dim_24641} -> tensor<4x?x1x128xf32>{%dim_24641}
%27133 = torch_c.from_builtin_tensor %27132 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24642 = torch.constant.int 5
%27134 = torch.prims.convert_element_type %27133, %int5_24642 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24643 = torch.constant.int 1
%27135 = torch.aten.size.int %26766, %int1_24643 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24644 = torch.constant.int 0
%27136 = torch.aten.add.int %int0_24644, %27135 : !torch.int, !torch.int -> !torch.int
%int0_24645 = torch.constant.int 0
%int0_24646 = torch.constant.int 0
%int1_24647 = torch.constant.int 1
%27137 = torch.aten.slice.Tensor %27089, %int0_24645, %int0_24646, %27136, %int1_24647 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27137, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24648 = torch.constant.int 1
%int0_24649 = torch.constant.int 0
%int9223372036854775807_24650 = torch.constant.int 9223372036854775807
%int1_24651 = torch.constant.int 1
%27138 = torch.aten.slice.Tensor %27137, %int1_24648, %int0_24649, %int9223372036854775807_24650, %int1_24651 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27138, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24652 = torch.constant.int 0
%27139 = torch.aten.unsqueeze %27138, %int0_24652 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27139, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24653 = torch.constant.int 2
%27140 = torch.aten.unsqueeze %27139, %int2_24653 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27140, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24654 = torch.constant.int 3
%int0_24655 = torch.constant.int 0
%int9223372036854775807_24656 = torch.constant.int 9223372036854775807
%int1_24657 = torch.constant.int 1
%27141 = torch.aten.slice.Tensor %27140, %int3_24654, %int0_24655, %int9223372036854775807_24656, %int1_24657 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27141, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27142 = torch_c.to_builtin_tensor %26882 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24658 = arith.constant 1 : index
%dim_24659 = tensor.dim %27142, %c1_24658 : tensor<4x?x1x128xf16>
%27143 = flow.tensor.bitcast %27142 : tensor<4x?x1x128xf16>{%dim_24659} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24659}
%27144 = torch_c.from_builtin_tensor %27143 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27145 = torch.aten.mul.Tensor %27144, %27141 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27146 = torch_c.to_builtin_tensor %27145 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24660 = arith.constant 1 : index
%dim_24661 = tensor.dim %27146, %c1_24660 : tensor<4x?x1x64xcomplex<f32>>
%27147 = flow.tensor.bitcast %27146 : tensor<4x?x1x64xcomplex<f32>>{%dim_24661} -> tensor<4x?x1x128xf32>{%dim_24661}
%27148 = torch_c.from_builtin_tensor %27147 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24662 = torch.constant.int 5
%27149 = torch.prims.convert_element_type %27148, %int5_24662 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24663 = torch.constant.int 1
%27150 = torch.aten.size.int %26772, %int1_24663 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24664 = torch.constant.int 0
%27151 = torch.aten.add.int %int0_24664, %27150 : !torch.int, !torch.int -> !torch.int
%int0_24665 = torch.constant.int 0
%int0_24666 = torch.constant.int 0
%int1_24667 = torch.constant.int 1
%27152 = torch.aten.slice.Tensor %27092, %int0_24665, %int0_24666, %27151, %int1_24667 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27152, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24668 = torch.constant.int 1
%int0_24669 = torch.constant.int 0
%int9223372036854775807_24670 = torch.constant.int 9223372036854775807
%int1_24671 = torch.constant.int 1
%27153 = torch.aten.slice.Tensor %27152, %int1_24668, %int0_24669, %int9223372036854775807_24670, %int1_24671 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27153, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24672 = torch.constant.int 0
%27154 = torch.aten.unsqueeze %27153, %int0_24672 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27154, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24673 = torch.constant.int 2
%27155 = torch.aten.unsqueeze %27154, %int2_24673 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27155, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24674 = torch.constant.int 3
%int0_24675 = torch.constant.int 0
%int9223372036854775807_24676 = torch.constant.int 9223372036854775807
%int1_24677 = torch.constant.int 1
%27156 = torch.aten.slice.Tensor %27155, %int3_24674, %int0_24675, %int9223372036854775807_24676, %int1_24677 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27156, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27157 = torch_c.to_builtin_tensor %26884 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24678 = arith.constant 1 : index
%dim_24679 = tensor.dim %27157, %c1_24678 : tensor<4x?x1x128xf16>
%27158 = flow.tensor.bitcast %27157 : tensor<4x?x1x128xf16>{%dim_24679} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24679}
%27159 = torch_c.from_builtin_tensor %27158 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27160 = torch.aten.mul.Tensor %27159, %27156 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27161 = torch_c.to_builtin_tensor %27160 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24680 = arith.constant 1 : index
%dim_24681 = tensor.dim %27161, %c1_24680 : tensor<4x?x1x64xcomplex<f32>>
%27162 = flow.tensor.bitcast %27161 : tensor<4x?x1x64xcomplex<f32>>{%dim_24681} -> tensor<4x?x1x128xf32>{%dim_24681}
%27163 = torch_c.from_builtin_tensor %27162 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24682 = torch.constant.int 5
%27164 = torch.prims.convert_element_type %27163, %int5_24682 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24683 = torch.constant.int 1
%27165 = torch.aten.size.int %26778, %int1_24683 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24684 = torch.constant.int 0
%27166 = torch.aten.add.int %int0_24684, %27165 : !torch.int, !torch.int -> !torch.int
%int0_24685 = torch.constant.int 0
%int0_24686 = torch.constant.int 0
%int1_24687 = torch.constant.int 1
%27167 = torch.aten.slice.Tensor %27095, %int0_24685, %int0_24686, %27166, %int1_24687 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27167, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24688 = torch.constant.int 1
%int0_24689 = torch.constant.int 0
%int9223372036854775807_24690 = torch.constant.int 9223372036854775807
%int1_24691 = torch.constant.int 1
%27168 = torch.aten.slice.Tensor %27167, %int1_24688, %int0_24689, %int9223372036854775807_24690, %int1_24691 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27168, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24692 = torch.constant.int 0
%27169 = torch.aten.unsqueeze %27168, %int0_24692 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27169, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24693 = torch.constant.int 2
%27170 = torch.aten.unsqueeze %27169, %int2_24693 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27170, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24694 = torch.constant.int 3
%int0_24695 = torch.constant.int 0
%int9223372036854775807_24696 = torch.constant.int 9223372036854775807
%int1_24697 = torch.constant.int 1
%27171 = torch.aten.slice.Tensor %27170, %int3_24694, %int0_24695, %int9223372036854775807_24696, %int1_24697 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27171, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27172 = torch_c.to_builtin_tensor %26886 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24698 = arith.constant 1 : index
%dim_24699 = tensor.dim %27172, %c1_24698 : tensor<4x?x1x128xf16>
%27173 = flow.tensor.bitcast %27172 : tensor<4x?x1x128xf16>{%dim_24699} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24699}
%27174 = torch_c.from_builtin_tensor %27173 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27175 = torch.aten.mul.Tensor %27174, %27171 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27176 = torch_c.to_builtin_tensor %27175 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24700 = arith.constant 1 : index
%dim_24701 = tensor.dim %27176, %c1_24700 : tensor<4x?x1x64xcomplex<f32>>
%27177 = flow.tensor.bitcast %27176 : tensor<4x?x1x64xcomplex<f32>>{%dim_24701} -> tensor<4x?x1x128xf32>{%dim_24701}
%27178 = torch_c.from_builtin_tensor %27177 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24702 = torch.constant.int 5
%27179 = torch.prims.convert_element_type %27178, %int5_24702 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24703 = torch.constant.int 1
%27180 = torch.aten.size.int %26784, %int1_24703 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24704 = torch.constant.int 0
%27181 = torch.aten.add.int %int0_24704, %27180 : !torch.int, !torch.int -> !torch.int
%int0_24705 = torch.constant.int 0
%int0_24706 = torch.constant.int 0
%int1_24707 = torch.constant.int 1
%27182 = torch.aten.slice.Tensor %27098, %int0_24705, %int0_24706, %27181, %int1_24707 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27182, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24708 = torch.constant.int 1
%int0_24709 = torch.constant.int 0
%int9223372036854775807_24710 = torch.constant.int 9223372036854775807
%int1_24711 = torch.constant.int 1
%27183 = torch.aten.slice.Tensor %27182, %int1_24708, %int0_24709, %int9223372036854775807_24710, %int1_24711 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27183, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24712 = torch.constant.int 0
%27184 = torch.aten.unsqueeze %27183, %int0_24712 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27184, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24713 = torch.constant.int 2
%27185 = torch.aten.unsqueeze %27184, %int2_24713 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27185, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24714 = torch.constant.int 3
%int0_24715 = torch.constant.int 0
%int9223372036854775807_24716 = torch.constant.int 9223372036854775807
%int1_24717 = torch.constant.int 1
%27186 = torch.aten.slice.Tensor %27185, %int3_24714, %int0_24715, %int9223372036854775807_24716, %int1_24717 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27186, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27187 = torch_c.to_builtin_tensor %26888 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24718 = arith.constant 1 : index
%dim_24719 = tensor.dim %27187, %c1_24718 : tensor<4x?x1x128xf16>
%27188 = flow.tensor.bitcast %27187 : tensor<4x?x1x128xf16>{%dim_24719} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24719}
%27189 = torch_c.from_builtin_tensor %27188 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27190 = torch.aten.mul.Tensor %27189, %27186 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27191 = torch_c.to_builtin_tensor %27190 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24720 = arith.constant 1 : index
%dim_24721 = tensor.dim %27191, %c1_24720 : tensor<4x?x1x64xcomplex<f32>>
%27192 = flow.tensor.bitcast %27191 : tensor<4x?x1x64xcomplex<f32>>{%dim_24721} -> tensor<4x?x1x128xf32>{%dim_24721}
%27193 = torch_c.from_builtin_tensor %27192 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24722 = torch.constant.int 5
%27194 = torch.prims.convert_element_type %27193, %int5_24722 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24723 = torch.constant.int 1
%27195 = torch.aten.size.int %26790, %int1_24723 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24724 = torch.constant.int 0
%27196 = torch.aten.add.int %int0_24724, %27195 : !torch.int, !torch.int -> !torch.int
%int0_24725 = torch.constant.int 0
%int0_24726 = torch.constant.int 0
%int1_24727 = torch.constant.int 1
%27197 = torch.aten.slice.Tensor %27101, %int0_24725, %int0_24726, %27196, %int1_24727 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27197, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24728 = torch.constant.int 1
%int0_24729 = torch.constant.int 0
%int9223372036854775807_24730 = torch.constant.int 9223372036854775807
%int1_24731 = torch.constant.int 1
%27198 = torch.aten.slice.Tensor %27197, %int1_24728, %int0_24729, %int9223372036854775807_24730, %int1_24731 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27198, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24732 = torch.constant.int 0
%27199 = torch.aten.unsqueeze %27198, %int0_24732 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27199, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24733 = torch.constant.int 2
%27200 = torch.aten.unsqueeze %27199, %int2_24733 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27200, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24734 = torch.constant.int 3
%int0_24735 = torch.constant.int 0
%int9223372036854775807_24736 = torch.constant.int 9223372036854775807
%int1_24737 = torch.constant.int 1
%27201 = torch.aten.slice.Tensor %27200, %int3_24734, %int0_24735, %int9223372036854775807_24736, %int1_24737 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27201, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27202 = torch_c.to_builtin_tensor %26890 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24738 = arith.constant 1 : index
%dim_24739 = tensor.dim %27202, %c1_24738 : tensor<4x?x1x128xf16>
%27203 = flow.tensor.bitcast %27202 : tensor<4x?x1x128xf16>{%dim_24739} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24739}
%27204 = torch_c.from_builtin_tensor %27203 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27205 = torch.aten.mul.Tensor %27204, %27201 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27206 = torch_c.to_builtin_tensor %27205 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24740 = arith.constant 1 : index
%dim_24741 = tensor.dim %27206, %c1_24740 : tensor<4x?x1x64xcomplex<f32>>
%27207 = flow.tensor.bitcast %27206 : tensor<4x?x1x64xcomplex<f32>>{%dim_24741} -> tensor<4x?x1x128xf32>{%dim_24741}
%27208 = torch_c.from_builtin_tensor %27207 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24742 = torch.constant.int 5
%27209 = torch.prims.convert_element_type %27208, %int5_24742 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_24743 = torch.constant.int 1
%27210 = torch.aten.size.int %26796, %int1_24743 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_24744 = torch.constant.int 0
%27211 = torch.aten.add.int %int0_24744, %27210 : !torch.int, !torch.int -> !torch.int
%int0_24745 = torch.constant.int 0
%int0_24746 = torch.constant.int 0
%int1_24747 = torch.constant.int 1
%27212 = torch.aten.slice.Tensor %27104, %int0_24745, %int0_24746, %27211, %int1_24747 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27212, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_24748 = torch.constant.int 1
%int0_24749 = torch.constant.int 0
%int9223372036854775807_24750 = torch.constant.int 9223372036854775807
%int1_24751 = torch.constant.int 1
%27213 = torch.aten.slice.Tensor %27212, %int1_24748, %int0_24749, %int9223372036854775807_24750, %int1_24751 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %27213, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_24752 = torch.constant.int 0
%27214 = torch.aten.unsqueeze %27213, %int0_24752 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %27214, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_24753 = torch.constant.int 2
%27215 = torch.aten.unsqueeze %27214, %int2_24753 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27215, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_24754 = torch.constant.int 3
%int0_24755 = torch.constant.int 0
%int9223372036854775807_24756 = torch.constant.int 9223372036854775807
%int1_24757 = torch.constant.int 1
%27216 = torch.aten.slice.Tensor %27215, %int3_24754, %int0_24755, %int9223372036854775807_24756, %int1_24757 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27216, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%27217 = torch_c.to_builtin_tensor %26892 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_24758 = arith.constant 1 : index
%dim_24759 = tensor.dim %27217, %c1_24758 : tensor<4x?x1x128xf16>
%27218 = flow.tensor.bitcast %27217 : tensor<4x?x1x128xf16>{%dim_24759} -> tensor<4x?x1x64xcomplex<f16>>{%dim_24759}
%27219 = torch_c.from_builtin_tensor %27218 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %27219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%27220 = torch.aten.mul.Tensor %27219, %27216 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %27220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%27221 = torch_c.to_builtin_tensor %27220 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_24760 = arith.constant 1 : index
%dim_24761 = tensor.dim %27221, %c1_24760 : tensor<4x?x1x64xcomplex<f32>>
%27222 = flow.tensor.bitcast %27221 : tensor<4x?x1x64xcomplex<f32>>{%dim_24761} -> tensor<4x?x1x128xf32>{%dim_24761}
%27223 = torch_c.from_builtin_tensor %27222 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %27223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_24762 = torch.constant.int 5
%27224 = torch.prims.convert_element_type %27223, %int5_24762 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %27224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_24763 = torch.constant.int 64
%27225 = torch.aten.mul.Scalar %2364, %int64_24763 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27225, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24764 = torch.constant.int 64
%27226 = torch.aten.mul.Scalar %2367, %int64_24764 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27226, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24765 = torch.constant.int 64
%27227 = torch.aten.mul.Scalar %2370, %int64_24765 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27227, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24766 = torch.constant.int 64
%27228 = torch.aten.mul.Scalar %2373, %int64_24766 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27228, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24767 = torch.constant.int 64
%27229 = torch.aten.mul.Scalar %2376, %int64_24767 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27229, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24768 = torch.constant.int 64
%27230 = torch.aten.mul.Scalar %2379, %int64_24768 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27230, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24769 = torch.constant.int 64
%27231 = torch.aten.mul.Scalar %2382, %int64_24769 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27231, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_24770 = torch.constant.int 64
%27232 = torch.aten.mul.Scalar %2385, %int64_24770 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27232, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26 = torch.constant.int 26
%int1_24771 = torch.constant.int 1
%27233 = torch.aten.add.Scalar %27225, %int26, %int1_24771 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27233, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24772 = torch.constant.int 26
%int1_24773 = torch.constant.int 1
%27234 = torch.aten.add.Scalar %27226, %int26_24772, %int1_24773 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27234, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24774 = torch.constant.int 26
%int1_24775 = torch.constant.int 1
%27235 = torch.aten.add.Scalar %27227, %int26_24774, %int1_24775 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27235, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24776 = torch.constant.int 26
%int1_24777 = torch.constant.int 1
%27236 = torch.aten.add.Scalar %27228, %int26_24776, %int1_24777 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27236, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24778 = torch.constant.int 26
%int1_24779 = torch.constant.int 1
%27237 = torch.aten.add.Scalar %27229, %int26_24778, %int1_24779 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27237, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24780 = torch.constant.int 26
%int1_24781 = torch.constant.int 1
%27238 = torch.aten.add.Scalar %27230, %int26_24780, %int1_24781 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27238, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24782 = torch.constant.int 26
%int1_24783 = torch.constant.int 1
%27239 = torch.aten.add.Scalar %27231, %int26_24782, %int1_24783 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27239, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int26_24784 = torch.constant.int 26
%int1_24785 = torch.constant.int 1
%27240 = torch.aten.add.Scalar %27232, %int26_24784, %int1_24785 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27240, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_24786 = torch.constant.int 4
%int16_24787 = torch.constant.int 16
%int1_24788 = torch.constant.int 1
%int128_24789 = torch.constant.int 128
%27241 = torch.prim.ListConstruct %int4_24786, %3095, %int16_24787, %int1_24788, %int128_24789 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27242 = torch.aten.view %27119, %27241 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27242, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24790 = torch.constant.int 4
%int16_24791 = torch.constant.int 16
%int1_24792 = torch.constant.int 1
%int128_24793 = torch.constant.int 128
%27243 = torch.prim.ListConstruct %int4_24790, %3095, %int16_24791, %int1_24792, %int128_24793 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27244 = torch.aten.view %27134, %27243 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27244, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24794 = torch.constant.int 4
%int16_24795 = torch.constant.int 16
%int1_24796 = torch.constant.int 1
%int128_24797 = torch.constant.int 128
%27245 = torch.prim.ListConstruct %int4_24794, %3095, %int16_24795, %int1_24796, %int128_24797 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27246 = torch.aten.view %27149, %27245 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27246, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24798 = torch.constant.int 4
%int16_24799 = torch.constant.int 16
%int1_24800 = torch.constant.int 1
%int128_24801 = torch.constant.int 128
%27247 = torch.prim.ListConstruct %int4_24798, %3095, %int16_24799, %int1_24800, %int128_24801 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27248 = torch.aten.view %27164, %27247 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27248, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24802 = torch.constant.int 4
%int16_24803 = torch.constant.int 16
%int1_24804 = torch.constant.int 1
%int128_24805 = torch.constant.int 128
%27249 = torch.prim.ListConstruct %int4_24802, %3095, %int16_24803, %int1_24804, %int128_24805 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27250 = torch.aten.view %27179, %27249 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27250, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24806 = torch.constant.int 4
%int16_24807 = torch.constant.int 16
%int1_24808 = torch.constant.int 1
%int128_24809 = torch.constant.int 128
%27251 = torch.prim.ListConstruct %int4_24806, %3095, %int16_24807, %int1_24808, %int128_24809 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27252 = torch.aten.view %27194, %27251 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27252, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24810 = torch.constant.int 4
%int16_24811 = torch.constant.int 16
%int1_24812 = torch.constant.int 1
%int128_24813 = torch.constant.int 128
%27253 = torch.prim.ListConstruct %int4_24810, %3095, %int16_24811, %int1_24812, %int128_24813 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27254 = torch.aten.view %27209, %27253 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27254, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24814 = torch.constant.int 4
%int16_24815 = torch.constant.int 16
%int1_24816 = torch.constant.int 1
%int128_24817 = torch.constant.int 128
%27255 = torch.prim.ListConstruct %int4_24814, %3095, %int16_24815, %int1_24816, %int128_24817 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27256 = torch.aten.view %27224, %27255 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27256, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24818 = torch.constant.int 4
%27257 = torch.aten.mul.int %int4_24818, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24819 = torch.constant.int 16
%int1_24820 = torch.constant.int 1
%int128_24821 = torch.constant.int 128
%27258 = torch.prim.ListConstruct %27257, %int16_24819, %int1_24820, %int128_24821 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27259 = torch.aten.view %27242, %27258 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27259, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24822 = torch.constant.int 4
%27260 = torch.aten.mul.int %int4_24822, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24823 = torch.constant.int 16
%int1_24824 = torch.constant.int 1
%int128_24825 = torch.constant.int 128
%27261 = torch.prim.ListConstruct %27260, %int16_24823, %int1_24824, %int128_24825 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27262 = torch.aten.view %27244, %27261 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27262, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24826 = torch.constant.int 4
%27263 = torch.aten.mul.int %int4_24826, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24827 = torch.constant.int 16
%int1_24828 = torch.constant.int 1
%int128_24829 = torch.constant.int 128
%27264 = torch.prim.ListConstruct %27263, %int16_24827, %int1_24828, %int128_24829 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27265 = torch.aten.view %27246, %27264 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27265, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24830 = torch.constant.int 4
%27266 = torch.aten.mul.int %int4_24830, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24831 = torch.constant.int 16
%int1_24832 = torch.constant.int 1
%int128_24833 = torch.constant.int 128
%27267 = torch.prim.ListConstruct %27266, %int16_24831, %int1_24832, %int128_24833 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27268 = torch.aten.view %27248, %27267 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27268, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24834 = torch.constant.int 4
%27269 = torch.aten.mul.int %int4_24834, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24835 = torch.constant.int 16
%int1_24836 = torch.constant.int 1
%int128_24837 = torch.constant.int 128
%27270 = torch.prim.ListConstruct %27269, %int16_24835, %int1_24836, %int128_24837 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27271 = torch.aten.view %27250, %27270 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27271, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24838 = torch.constant.int 4
%27272 = torch.aten.mul.int %int4_24838, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24839 = torch.constant.int 16
%int1_24840 = torch.constant.int 1
%int128_24841 = torch.constant.int 128
%27273 = torch.prim.ListConstruct %27272, %int16_24839, %int1_24840, %int128_24841 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27274 = torch.aten.view %27252, %27273 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27274, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24842 = torch.constant.int 4
%27275 = torch.aten.mul.int %int4_24842, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24843 = torch.constant.int 16
%int1_24844 = torch.constant.int 1
%int128_24845 = torch.constant.int 128
%27276 = torch.prim.ListConstruct %27275, %int16_24843, %int1_24844, %int128_24845 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27277 = torch.aten.view %27254, %27276 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27277, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24846 = torch.constant.int 4
%27278 = torch.aten.mul.int %int4_24846, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24847 = torch.constant.int 16
%int1_24848 = torch.constant.int 1
%int128_24849 = torch.constant.int 128
%27279 = torch.prim.ListConstruct %27278, %int16_24847, %int1_24848, %int128_24849 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27280 = torch.aten.view %27256, %27279 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27280, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24850 = torch.constant.int 4
%27281 = torch.aten.mul.int %int4_24850, %3095 : !torch.int, !torch.int -> !torch.int
%27282 = torch.prim.ListConstruct %27281 : (!torch.int) -> !torch.list<int>
%27283 = torch.aten.view %27233, %27282 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27283, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24851 = torch.constant.int 4
%27284 = torch.aten.mul.int %int4_24851, %3095 : !torch.int, !torch.int -> !torch.int
%27285 = torch.prim.ListConstruct %27284 : (!torch.int) -> !torch.list<int>
%27286 = torch.aten.view %27234, %27285 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27286, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24852 = torch.constant.int 4
%27287 = torch.aten.mul.int %int4_24852, %3095 : !torch.int, !torch.int -> !torch.int
%27288 = torch.prim.ListConstruct %27287 : (!torch.int) -> !torch.list<int>
%27289 = torch.aten.view %27235, %27288 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27289, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24853 = torch.constant.int 4
%27290 = torch.aten.mul.int %int4_24853, %3095 : !torch.int, !torch.int -> !torch.int
%27291 = torch.prim.ListConstruct %27290 : (!torch.int) -> !torch.list<int>
%27292 = torch.aten.view %27236, %27291 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27292, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24854 = torch.constant.int 4
%27293 = torch.aten.mul.int %int4_24854, %3095 : !torch.int, !torch.int -> !torch.int
%27294 = torch.prim.ListConstruct %27293 : (!torch.int) -> !torch.list<int>
%27295 = torch.aten.view %27237, %27294 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27295, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24855 = torch.constant.int 4
%27296 = torch.aten.mul.int %int4_24855, %3095 : !torch.int, !torch.int -> !torch.int
%27297 = torch.prim.ListConstruct %27296 : (!torch.int) -> !torch.list<int>
%27298 = torch.aten.view %27238, %27297 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27298, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24856 = torch.constant.int 4
%27299 = torch.aten.mul.int %int4_24856, %3095 : !torch.int, !torch.int -> !torch.int
%27300 = torch.prim.ListConstruct %27299 : (!torch.int) -> !torch.list<int>
%27301 = torch.aten.view %27239, %27300 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27301, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24857 = torch.constant.int 4
%27302 = torch.aten.mul.int %int4_24857, %3095 : !torch.int, !torch.int -> !torch.int
%27303 = torch.prim.ListConstruct %27302 : (!torch.int) -> !torch.list<int>
%27304 = torch.aten.view %27240, %27303 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27304, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24858 = torch.constant.int 4
%int16_24859 = torch.constant.int 16
%int1_24860 = torch.constant.int 1
%int128_24861 = torch.constant.int 128
%27305 = torch.prim.ListConstruct %int4_24858, %3095, %int16_24859, %int1_24860, %int128_24861 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27306 = torch.aten.view %26894, %27305 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27306, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24862 = torch.constant.int 4
%int16_24863 = torch.constant.int 16
%int1_24864 = torch.constant.int 1
%int128_24865 = torch.constant.int 128
%27307 = torch.prim.ListConstruct %int4_24862, %3095, %int16_24863, %int1_24864, %int128_24865 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27308 = torch.aten.view %26896, %27307 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27308, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24866 = torch.constant.int 4
%int16_24867 = torch.constant.int 16
%int1_24868 = torch.constant.int 1
%int128_24869 = torch.constant.int 128
%27309 = torch.prim.ListConstruct %int4_24866, %3095, %int16_24867, %int1_24868, %int128_24869 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27310 = torch.aten.view %26898, %27309 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27310, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24870 = torch.constant.int 4
%int16_24871 = torch.constant.int 16
%int1_24872 = torch.constant.int 1
%int128_24873 = torch.constant.int 128
%27311 = torch.prim.ListConstruct %int4_24870, %3095, %int16_24871, %int1_24872, %int128_24873 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27312 = torch.aten.view %26900, %27311 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27312, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24874 = torch.constant.int 4
%int16_24875 = torch.constant.int 16
%int1_24876 = torch.constant.int 1
%int128_24877 = torch.constant.int 128
%27313 = torch.prim.ListConstruct %int4_24874, %3095, %int16_24875, %int1_24876, %int128_24877 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27314 = torch.aten.view %26902, %27313 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27314, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24878 = torch.constant.int 4
%int16_24879 = torch.constant.int 16
%int1_24880 = torch.constant.int 1
%int128_24881 = torch.constant.int 128
%27315 = torch.prim.ListConstruct %int4_24878, %3095, %int16_24879, %int1_24880, %int128_24881 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27316 = torch.aten.view %26904, %27315 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27316, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24882 = torch.constant.int 4
%int16_24883 = torch.constant.int 16
%int1_24884 = torch.constant.int 1
%int128_24885 = torch.constant.int 128
%27317 = torch.prim.ListConstruct %int4_24882, %3095, %int16_24883, %int1_24884, %int128_24885 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27318 = torch.aten.view %26906, %27317 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27318, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24886 = torch.constant.int 4
%int16_24887 = torch.constant.int 16
%int1_24888 = torch.constant.int 1
%int128_24889 = torch.constant.int 128
%27319 = torch.prim.ListConstruct %int4_24886, %3095, %int16_24887, %int1_24888, %int128_24889 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27320 = torch.aten.view %26908, %27319 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %27320, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_24890 = torch.constant.int 4
%27321 = torch.aten.mul.int %int4_24890, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24891 = torch.constant.int 16
%int1_24892 = torch.constant.int 1
%int128_24893 = torch.constant.int 128
%27322 = torch.prim.ListConstruct %27321, %int16_24891, %int1_24892, %int128_24893 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27323 = torch.aten.view %27306, %27322 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27323, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24894 = torch.constant.int 4
%27324 = torch.aten.mul.int %int4_24894, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24895 = torch.constant.int 16
%int1_24896 = torch.constant.int 1
%int128_24897 = torch.constant.int 128
%27325 = torch.prim.ListConstruct %27324, %int16_24895, %int1_24896, %int128_24897 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27326 = torch.aten.view %27308, %27325 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27326, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24898 = torch.constant.int 4
%27327 = torch.aten.mul.int %int4_24898, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24899 = torch.constant.int 16
%int1_24900 = torch.constant.int 1
%int128_24901 = torch.constant.int 128
%27328 = torch.prim.ListConstruct %27327, %int16_24899, %int1_24900, %int128_24901 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27329 = torch.aten.view %27310, %27328 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27329, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24902 = torch.constant.int 4
%27330 = torch.aten.mul.int %int4_24902, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24903 = torch.constant.int 16
%int1_24904 = torch.constant.int 1
%int128_24905 = torch.constant.int 128
%27331 = torch.prim.ListConstruct %27330, %int16_24903, %int1_24904, %int128_24905 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27332 = torch.aten.view %27312, %27331 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27332, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24906 = torch.constant.int 4
%27333 = torch.aten.mul.int %int4_24906, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24907 = torch.constant.int 16
%int1_24908 = torch.constant.int 1
%int128_24909 = torch.constant.int 128
%27334 = torch.prim.ListConstruct %27333, %int16_24907, %int1_24908, %int128_24909 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27335 = torch.aten.view %27314, %27334 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27335, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24910 = torch.constant.int 4
%27336 = torch.aten.mul.int %int4_24910, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24911 = torch.constant.int 16
%int1_24912 = torch.constant.int 1
%int128_24913 = torch.constant.int 128
%27337 = torch.prim.ListConstruct %27336, %int16_24911, %int1_24912, %int128_24913 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27338 = torch.aten.view %27316, %27337 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27338, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24914 = torch.constant.int 4
%27339 = torch.aten.mul.int %int4_24914, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24915 = torch.constant.int 16
%int1_24916 = torch.constant.int 1
%int128_24917 = torch.constant.int 128
%27340 = torch.prim.ListConstruct %27339, %int16_24915, %int1_24916, %int128_24917 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27341 = torch.aten.view %27318, %27340 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27341, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_24918 = torch.constant.int 4
%27342 = torch.aten.mul.int %int4_24918, %3095 : !torch.int, !torch.int -> !torch.int
%int16_24919 = torch.constant.int 16
%int1_24920 = torch.constant.int 1
%int128_24921 = torch.constant.int 128
%27343 = torch.prim.ListConstruct %27342, %int16_24919, %int1_24920, %int128_24921 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27344 = torch.aten.view %27320, %27343 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27344, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_24922 = torch.constant.int 1
%int1_24923 = torch.constant.int 1
%27345 = torch.aten.add.Scalar %27233, %int1_24922, %int1_24923 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27345, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24924 = torch.constant.int 1
%int1_24925 = torch.constant.int 1
%27346 = torch.aten.add.Scalar %27234, %int1_24924, %int1_24925 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27346, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24926 = torch.constant.int 1
%int1_24927 = torch.constant.int 1
%27347 = torch.aten.add.Scalar %27235, %int1_24926, %int1_24927 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27347, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24928 = torch.constant.int 1
%int1_24929 = torch.constant.int 1
%27348 = torch.aten.add.Scalar %27236, %int1_24928, %int1_24929 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27348, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24930 = torch.constant.int 1
%int1_24931 = torch.constant.int 1
%27349 = torch.aten.add.Scalar %27237, %int1_24930, %int1_24931 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27349, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24932 = torch.constant.int 1
%int1_24933 = torch.constant.int 1
%27350 = torch.aten.add.Scalar %27238, %int1_24932, %int1_24933 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27350, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24934 = torch.constant.int 1
%int1_24935 = torch.constant.int 1
%27351 = torch.aten.add.Scalar %27239, %int1_24934, %int1_24935 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27351, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_24936 = torch.constant.int 1
%int1_24937 = torch.constant.int 1
%27352 = torch.aten.add.Scalar %27240, %int1_24936, %int1_24937 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %27352, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_24938 = torch.constant.int 4
%27353 = torch.aten.mul.int %int4_24938, %3095 : !torch.int, !torch.int -> !torch.int
%27354 = torch.prim.ListConstruct %27353 : (!torch.int) -> !torch.list<int>
%27355 = torch.aten.view %27345, %27354 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27355, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24939 = torch.constant.int 4
%27356 = torch.aten.mul.int %int4_24939, %3095 : !torch.int, !torch.int -> !torch.int
%27357 = torch.prim.ListConstruct %27356 : (!torch.int) -> !torch.list<int>
%27358 = torch.aten.view %27346, %27357 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27358, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24940 = torch.constant.int 4
%27359 = torch.aten.mul.int %int4_24940, %3095 : !torch.int, !torch.int -> !torch.int
%27360 = torch.prim.ListConstruct %27359 : (!torch.int) -> !torch.list<int>
%27361 = torch.aten.view %27347, %27360 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27361, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24941 = torch.constant.int 4
%27362 = torch.aten.mul.int %int4_24941, %3095 : !torch.int, !torch.int -> !torch.int
%27363 = torch.prim.ListConstruct %27362 : (!torch.int) -> !torch.list<int>
%27364 = torch.aten.view %27348, %27363 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27364, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24942 = torch.constant.int 4
%27365 = torch.aten.mul.int %int4_24942, %3095 : !torch.int, !torch.int -> !torch.int
%27366 = torch.prim.ListConstruct %27365 : (!torch.int) -> !torch.list<int>
%27367 = torch.aten.view %27349, %27366 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27367, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24943 = torch.constant.int 4
%27368 = torch.aten.mul.int %int4_24943, %3095 : !torch.int, !torch.int -> !torch.int
%27369 = torch.prim.ListConstruct %27368 : (!torch.int) -> !torch.list<int>
%27370 = torch.aten.view %27350, %27369 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27370, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24944 = torch.constant.int 4
%27371 = torch.aten.mul.int %int4_24944, %3095 : !torch.int, !torch.int -> !torch.int
%27372 = torch.prim.ListConstruct %27371 : (!torch.int) -> !torch.list<int>
%27373 = torch.aten.view %27351, %27372 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27373, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_24945 = torch.constant.int 4
%27374 = torch.aten.mul.int %int4_24945, %3095 : !torch.int, !torch.int -> !torch.int
%27375 = torch.prim.ListConstruct %27374 : (!torch.int) -> !torch.list<int>
%27376 = torch.aten.view %27352, %27375 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27376, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%27377 = torch.prim.ListConstruct %27283, %27355 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24946 = torch.constant.int 0
%27378 = torch.aten.cat %27377, %int0_24946 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27378, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27379 = torch.prim.ListConstruct %27286, %27358 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24947 = torch.constant.int 0
%27380 = torch.aten.cat %27379, %int0_24947 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27380, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27381 = torch.prim.ListConstruct %27289, %27361 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24948 = torch.constant.int 0
%27382 = torch.aten.cat %27381, %int0_24948 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27382, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27383 = torch.prim.ListConstruct %27292, %27364 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24949 = torch.constant.int 0
%27384 = torch.aten.cat %27383, %int0_24949 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27384, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27385 = torch.prim.ListConstruct %27295, %27367 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24950 = torch.constant.int 0
%27386 = torch.aten.cat %27385, %int0_24950 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27386, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27387 = torch.prim.ListConstruct %27298, %27370 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24951 = torch.constant.int 0
%27388 = torch.aten.cat %27387, %int0_24951 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27388, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27389 = torch.prim.ListConstruct %27301, %27373 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24952 = torch.constant.int 0
%27390 = torch.aten.cat %27389, %int0_24952 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27390, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27391 = torch.prim.ListConstruct %27304, %27376 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_24953 = torch.constant.int 0
%27392 = torch.aten.cat %27391, %int0_24953 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %27392, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%27393 = torch.prim.ListConstruct %27259, %27323 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24954 = torch.constant.int 0
%27394 = torch.aten.cat %27393, %int0_24954 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27394, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27395 = torch.prim.ListConstruct %27262, %27326 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24955 = torch.constant.int 0
%27396 = torch.aten.cat %27395, %int0_24955 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27396, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27397 = torch.prim.ListConstruct %27265, %27329 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24956 = torch.constant.int 0
%27398 = torch.aten.cat %27397, %int0_24956 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27398, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27399 = torch.prim.ListConstruct %27268, %27332 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24957 = torch.constant.int 0
%27400 = torch.aten.cat %27399, %int0_24957 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27400, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27401 = torch.prim.ListConstruct %27271, %27335 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24958 = torch.constant.int 0
%27402 = torch.aten.cat %27401, %int0_24958 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27402, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27403 = torch.prim.ListConstruct %27274, %27338 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24959 = torch.constant.int 0
%27404 = torch.aten.cat %27403, %int0_24959 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27404, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27405 = torch.prim.ListConstruct %27277, %27341 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24960 = torch.constant.int 0
%27406 = torch.aten.cat %27405, %int0_24960 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27406, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27407 = torch.prim.ListConstruct %27280, %27344 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_24961 = torch.constant.int 0
%27408 = torch.aten.cat %27407, %int0_24961 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27408, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_24962 = torch.constant.int 32
%int2_24963 = torch.constant.int 2
%int16_24964 = torch.constant.int 16
%int1_24965 = torch.constant.int 1
%int128_24966 = torch.constant.int 128
%27409 = torch.prim.ListConstruct %3023, %int32_24962, %int2_24963, %int16_24964, %int1_24965, %int128_24966 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27410 = torch.aten.view %25559, %27409 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27410, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_24967 = torch.constant.int 32
%27411 = torch.aten.mul.int %3023, %int32_24967 : !torch.int, !torch.int -> !torch.int
%int2_24968 = torch.constant.int 2
%27412 = torch.aten.mul.int %27411, %int2_24968 : !torch.int, !torch.int -> !torch.int
%int16_24969 = torch.constant.int 16
%int1_24970 = torch.constant.int 1
%int128_24971 = torch.constant.int 128
%27413 = torch.prim.ListConstruct %27412, %int16_24969, %int1_24970, %int128_24971 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27414 = torch.aten.view %27410, %27413 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27414, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27415 = torch.prim.ListConstruct %27378 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_24972 = torch.constant.bool false
%27416 = torch.aten.index_put %27414, %27415, %27394, %false_24972 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27416, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_24973 = torch.constant.int 32
%int2_24974 = torch.constant.int 2
%int16_24975 = torch.constant.int 16
%int1_24976 = torch.constant.int 1
%int128_24977 = torch.constant.int 128
%27417 = torch.prim.ListConstruct %3023, %int32_24973, %int2_24974, %int16_24975, %int1_24976, %int128_24977 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27418 = torch.aten.view %27416, %27417 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27418, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_24978 = torch.constant.int 131072
%27419 = torch.prim.ListConstruct %3023, %int131072_24978 : (!torch.int, !torch.int) -> !torch.list<int>
%27420 = torch.aten.view %27418, %27419 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27420, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_24979 = torch.constant.int 32
%int2_24980 = torch.constant.int 2
%int16_24981 = torch.constant.int 16
%int1_24982 = torch.constant.int 1
%int128_24983 = torch.constant.int 128
%27421 = torch.prim.ListConstruct %3026, %int32_24979, %int2_24980, %int16_24981, %int1_24982, %int128_24983 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27422 = torch.aten.view %25571, %27421 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27422, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_24984 = torch.constant.int 32
%27423 = torch.aten.mul.int %3026, %int32_24984 : !torch.int, !torch.int -> !torch.int
%int2_24985 = torch.constant.int 2
%27424 = torch.aten.mul.int %27423, %int2_24985 : !torch.int, !torch.int -> !torch.int
%int16_24986 = torch.constant.int 16
%int1_24987 = torch.constant.int 1
%int128_24988 = torch.constant.int 128
%27425 = torch.prim.ListConstruct %27424, %int16_24986, %int1_24987, %int128_24988 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27426 = torch.aten.view %27422, %27425 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27426, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27427 = torch.prim.ListConstruct %27380 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_24989 = torch.constant.bool false
%27428 = torch.aten.index_put %27426, %27427, %27396, %false_24989 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27428, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_24990 = torch.constant.int 32
%int2_24991 = torch.constant.int 2
%int16_24992 = torch.constant.int 16
%int1_24993 = torch.constant.int 1
%int128_24994 = torch.constant.int 128
%27429 = torch.prim.ListConstruct %3026, %int32_24990, %int2_24991, %int16_24992, %int1_24993, %int128_24994 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27430 = torch.aten.view %27428, %27429 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27430, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_24995 = torch.constant.int 131072
%27431 = torch.prim.ListConstruct %3026, %int131072_24995 : (!torch.int, !torch.int) -> !torch.list<int>
%27432 = torch.aten.view %27430, %27431 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27432, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_24996 = torch.constant.int 32
%int2_24997 = torch.constant.int 2
%int16_24998 = torch.constant.int 16
%int1_24999 = torch.constant.int 1
%int128_25000 = torch.constant.int 128
%27433 = torch.prim.ListConstruct %3029, %int32_24996, %int2_24997, %int16_24998, %int1_24999, %int128_25000 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27434 = torch.aten.view %25583, %27433 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27434, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_25001 = torch.constant.int 32
%27435 = torch.aten.mul.int %3029, %int32_25001 : !torch.int, !torch.int -> !torch.int
%int2_25002 = torch.constant.int 2
%27436 = torch.aten.mul.int %27435, %int2_25002 : !torch.int, !torch.int -> !torch.int
%int16_25003 = torch.constant.int 16
%int1_25004 = torch.constant.int 1
%int128_25005 = torch.constant.int 128
%27437 = torch.prim.ListConstruct %27436, %int16_25003, %int1_25004, %int128_25005 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27438 = torch.aten.view %27434, %27437 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27438, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27439 = torch.prim.ListConstruct %27382 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_25006 = torch.constant.bool false
%27440 = torch.aten.index_put %27438, %27439, %27398, %false_25006 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27440, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_25007 = torch.constant.int 32
%int2_25008 = torch.constant.int 2
%int16_25009 = torch.constant.int 16
%int1_25010 = torch.constant.int 1
%int128_25011 = torch.constant.int 128
%27441 = torch.prim.ListConstruct %3029, %int32_25007, %int2_25008, %int16_25009, %int1_25010, %int128_25011 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27442 = torch.aten.view %27440, %27441 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27442, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_25012 = torch.constant.int 131072
%27443 = torch.prim.ListConstruct %3029, %int131072_25012 : (!torch.int, !torch.int) -> !torch.list<int>
%27444 = torch.aten.view %27442, %27443 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27444, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_25013 = torch.constant.int 32
%int2_25014 = torch.constant.int 2
%int16_25015 = torch.constant.int 16
%int1_25016 = torch.constant.int 1
%int128_25017 = torch.constant.int 128
%27445 = torch.prim.ListConstruct %3032, %int32_25013, %int2_25014, %int16_25015, %int1_25016, %int128_25017 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27446 = torch.aten.view %25595, %27445 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27446, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_25018 = torch.constant.int 32
%27447 = torch.aten.mul.int %3032, %int32_25018 : !torch.int, !torch.int -> !torch.int
%int2_25019 = torch.constant.int 2
%27448 = torch.aten.mul.int %27447, %int2_25019 : !torch.int, !torch.int -> !torch.int
%int16_25020 = torch.constant.int 16
%int1_25021 = torch.constant.int 1
%int128_25022 = torch.constant.int 128
%27449 = torch.prim.ListConstruct %27448, %int16_25020, %int1_25021, %int128_25022 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27450 = torch.aten.view %27446, %27449 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27450, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27451 = torch.prim.ListConstruct %27384 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_25023 = torch.constant.bool false
%27452 = torch.aten.index_put %27450, %27451, %27400, %false_25023 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27452, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_25024 = torch.constant.int 32
%int2_25025 = torch.constant.int 2
%int16_25026 = torch.constant.int 16
%int1_25027 = torch.constant.int 1
%int128_25028 = torch.constant.int 128
%27453 = torch.prim.ListConstruct %3032, %int32_25024, %int2_25025, %int16_25026, %int1_25027, %int128_25028 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27454 = torch.aten.view %27452, %27453 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27454, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_25029 = torch.constant.int 131072
%27455 = torch.prim.ListConstruct %3032, %int131072_25029 : (!torch.int, !torch.int) -> !torch.list<int>
%27456 = torch.aten.view %27454, %27455 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27456, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_25030 = torch.constant.int 32
%int2_25031 = torch.constant.int 2
%int16_25032 = torch.constant.int 16
%int1_25033 = torch.constant.int 1
%int128_25034 = torch.constant.int 128
%27457 = torch.prim.ListConstruct %3035, %int32_25030, %int2_25031, %int16_25032, %int1_25033, %int128_25034 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27458 = torch.aten.view %25607, %27457 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27458, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_25035 = torch.constant.int 32
%27459 = torch.aten.mul.int %3035, %int32_25035 : !torch.int, !torch.int -> !torch.int
%int2_25036 = torch.constant.int 2
%27460 = torch.aten.mul.int %27459, %int2_25036 : !torch.int, !torch.int -> !torch.int
%int16_25037 = torch.constant.int 16
%int1_25038 = torch.constant.int 1
%int128_25039 = torch.constant.int 128
%27461 = torch.prim.ListConstruct %27460, %int16_25037, %int1_25038, %int128_25039 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27462 = torch.aten.view %27458, %27461 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27462, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27463 = torch.prim.ListConstruct %27386 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_25040 = torch.constant.bool false
%27464 = torch.aten.index_put %27462, %27463, %27402, %false_25040 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27464, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_25041 = torch.constant.int 32
%int2_25042 = torch.constant.int 2
%int16_25043 = torch.constant.int 16
%int1_25044 = torch.constant.int 1
%int128_25045 = torch.constant.int 128
%27465 = torch.prim.ListConstruct %3035, %int32_25041, %int2_25042, %int16_25043, %int1_25044, %int128_25045 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27466 = torch.aten.view %27464, %27465 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27466, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_25046 = torch.constant.int 131072
%27467 = torch.prim.ListConstruct %3035, %int131072_25046 : (!torch.int, !torch.int) -> !torch.list<int>
%27468 = torch.aten.view %27466, %27467 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27468, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_25047 = torch.constant.int 32
%int2_25048 = torch.constant.int 2
%int16_25049 = torch.constant.int 16
%int1_25050 = torch.constant.int 1
%int128_25051 = torch.constant.int 128
%27469 = torch.prim.ListConstruct %3038, %int32_25047, %int2_25048, %int16_25049, %int1_25050, %int128_25051 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27470 = torch.aten.view %25619, %27469 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27470, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_25052 = torch.constant.int 32
%27471 = torch.aten.mul.int %3038, %int32_25052 : !torch.int, !torch.int -> !torch.int
%int2_25053 = torch.constant.int 2
%27472 = torch.aten.mul.int %27471, %int2_25053 : !torch.int, !torch.int -> !torch.int
%int16_25054 = torch.constant.int 16
%int1_25055 = torch.constant.int 1
%int128_25056 = torch.constant.int 128
%27473 = torch.prim.ListConstruct %27472, %int16_25054, %int1_25055, %int128_25056 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27474 = torch.aten.view %27470, %27473 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27474, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27475 = torch.prim.ListConstruct %27388 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_25057 = torch.constant.bool false
%27476 = torch.aten.index_put %27474, %27475, %27404, %false_25057 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27476, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_25058 = torch.constant.int 32
%int2_25059 = torch.constant.int 2
%int16_25060 = torch.constant.int 16
%int1_25061 = torch.constant.int 1
%int128_25062 = torch.constant.int 128
%27477 = torch.prim.ListConstruct %3038, %int32_25058, %int2_25059, %int16_25060, %int1_25061, %int128_25062 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27478 = torch.aten.view %27476, %27477 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27478, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_25063 = torch.constant.int 131072
%27479 = torch.prim.ListConstruct %3038, %int131072_25063 : (!torch.int, !torch.int) -> !torch.list<int>
%27480 = torch.aten.view %27478, %27479 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27480, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_25064 = torch.constant.int 32
%int2_25065 = torch.constant.int 2
%int16_25066 = torch.constant.int 16
%int1_25067 = torch.constant.int 1
%int128_25068 = torch.constant.int 128
%27481 = torch.prim.ListConstruct %3041, %int32_25064, %int2_25065, %int16_25066, %int1_25067, %int128_25068 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27482 = torch.aten.view %25631, %27481 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27482, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_25069 = torch.constant.int 32
%27483 = torch.aten.mul.int %3041, %int32_25069 : !torch.int, !torch.int -> !torch.int
%int2_25070 = torch.constant.int 2
%27484 = torch.aten.mul.int %27483, %int2_25070 : !torch.int, !torch.int -> !torch.int
%int16_25071 = torch.constant.int 16
%int1_25072 = torch.constant.int 1
%int128_25073 = torch.constant.int 128
%27485 = torch.prim.ListConstruct %27484, %int16_25071, %int1_25072, %int128_25073 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27486 = torch.aten.view %27482, %27485 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27486, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27487 = torch.prim.ListConstruct %27390 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_25074 = torch.constant.bool false
%27488 = torch.aten.index_put %27486, %27487, %27406, %false_25074 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27488, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_25075 = torch.constant.int 32
%int2_25076 = torch.constant.int 2
%int16_25077 = torch.constant.int 16
%int1_25078 = torch.constant.int 1
%int128_25079 = torch.constant.int 128
%27489 = torch.prim.ListConstruct %3041, %int32_25075, %int2_25076, %int16_25077, %int1_25078, %int128_25079 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27490 = torch.aten.view %27488, %27489 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27490, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_25080 = torch.constant.int 131072
%27491 = torch.prim.ListConstruct %3041, %int131072_25080 : (!torch.int, !torch.int) -> !torch.list<int>
%27492 = torch.aten.view %27490, %27491 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27492, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_25081 = torch.constant.int 32
%int2_25082 = torch.constant.int 2
%int16_25083 = torch.constant.int 16
%int1_25084 = torch.constant.int 1
%int128_25085 = torch.constant.int 128
%27493 = torch.prim.ListConstruct %3044, %int32_25081, %int2_25082, %int16_25083, %int1_25084, %int128_25085 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27494 = torch.aten.view %25643, %27493 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27494, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_25086 = torch.constant.int 32
%27495 = torch.aten.mul.int %3044, %int32_25086 : !torch.int, !torch.int -> !torch.int
%int2_25087 = torch.constant.int 2
%27496 = torch.aten.mul.int %27495, %int2_25087 : !torch.int, !torch.int -> !torch.int
%int16_25088 = torch.constant.int 16
%int1_25089 = torch.constant.int 1
%int128_25090 = torch.constant.int 128
%27497 = torch.prim.ListConstruct %27496, %int16_25088, %int1_25089, %int128_25090 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27498 = torch.aten.view %27494, %27497 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27498, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%27499 = torch.prim.ListConstruct %27392 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_25091 = torch.constant.bool false
%27500 = torch.aten.index_put %27498, %27499, %27408, %false_25091 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %27500, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_25092 = torch.constant.int 32
%int2_25093 = torch.constant.int 2
%int16_25094 = torch.constant.int 16
%int1_25095 = torch.constant.int 1
%int128_25096 = torch.constant.int 128
%27501 = torch.prim.ListConstruct %3044, %int32_25092, %int2_25093, %int16_25094, %int1_25095, %int128_25096 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27502 = torch.aten.view %27500, %27501 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %27502, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_25097 = torch.constant.int 131072
%27503 = torch.prim.ListConstruct %3044, %int131072_25097 : (!torch.int, !torch.int) -> !torch.list<int>
%27504 = torch.aten.view %27502, %27503 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %27504, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_25098 = torch.constant.int -2
%27505 = torch.aten.unsqueeze %27119, %int-2_25098 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25099 = torch.constant.int -2
%27506 = torch.aten.unsqueeze %27134, %int-2_25099 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25100 = torch.constant.int -2
%27507 = torch.aten.unsqueeze %27149, %int-2_25100 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25101 = torch.constant.int -2
%27508 = torch.aten.unsqueeze %27164, %int-2_25101 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25102 = torch.constant.int -2
%27509 = torch.aten.unsqueeze %27179, %int-2_25102 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25103 = torch.constant.int -2
%27510 = torch.aten.unsqueeze %27194, %int-2_25103 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25104 = torch.constant.int -2
%27511 = torch.aten.unsqueeze %27209, %int-2_25104 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25105 = torch.constant.int -2
%27512 = torch.aten.unsqueeze %27224, %int-2_25105 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_25106 = torch.constant.int 4
%int1_25107 = torch.constant.int 1
%int4_25108 = torch.constant.int 4
%int128_25109 = torch.constant.int 128
%27513 = torch.prim.ListConstruct %int4_25106, %27105, %int1_25107, %int4_25108, %int128_25109 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25110 = torch.constant.bool false
%27514 = torch.aten.expand %27505, %27513, %false_25110 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25111 = torch.constant.int 4
%int1_25112 = torch.constant.int 1
%int4_25113 = torch.constant.int 4
%int128_25114 = torch.constant.int 128
%27515 = torch.prim.ListConstruct %int4_25111, %27105, %int1_25112, %int4_25113, %int128_25114 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25115 = torch.constant.bool false
%27516 = torch.aten.expand %27506, %27515, %false_25115 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25116 = torch.constant.int 4
%int1_25117 = torch.constant.int 1
%int4_25118 = torch.constant.int 4
%int128_25119 = torch.constant.int 128
%27517 = torch.prim.ListConstruct %int4_25116, %27105, %int1_25117, %int4_25118, %int128_25119 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25120 = torch.constant.bool false
%27518 = torch.aten.expand %27507, %27517, %false_25120 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25121 = torch.constant.int 4
%int1_25122 = torch.constant.int 1
%int4_25123 = torch.constant.int 4
%int128_25124 = torch.constant.int 128
%27519 = torch.prim.ListConstruct %int4_25121, %27105, %int1_25122, %int4_25123, %int128_25124 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25125 = torch.constant.bool false
%27520 = torch.aten.expand %27508, %27519, %false_25125 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25126 = torch.constant.int 4
%int1_25127 = torch.constant.int 1
%int4_25128 = torch.constant.int 4
%int128_25129 = torch.constant.int 128
%27521 = torch.prim.ListConstruct %int4_25126, %27105, %int1_25127, %int4_25128, %int128_25129 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25130 = torch.constant.bool false
%27522 = torch.aten.expand %27509, %27521, %false_25130 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25131 = torch.constant.int 4
%int1_25132 = torch.constant.int 1
%int4_25133 = torch.constant.int 4
%int128_25134 = torch.constant.int 128
%27523 = torch.prim.ListConstruct %int4_25131, %27105, %int1_25132, %int4_25133, %int128_25134 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25135 = torch.constant.bool false
%27524 = torch.aten.expand %27510, %27523, %false_25135 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25136 = torch.constant.int 4
%int1_25137 = torch.constant.int 1
%int4_25138 = torch.constant.int 4
%int128_25139 = torch.constant.int 128
%27525 = torch.prim.ListConstruct %int4_25136, %27105, %int1_25137, %int4_25138, %int128_25139 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25140 = torch.constant.bool false
%27526 = torch.aten.expand %27511, %27525, %false_25140 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25141 = torch.constant.int 4
%int1_25142 = torch.constant.int 1
%int4_25143 = torch.constant.int 4
%int128_25144 = torch.constant.int 128
%27527 = torch.prim.ListConstruct %int4_25141, %27105, %int1_25142, %int4_25143, %int128_25144 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25145 = torch.constant.bool false
%27528 = torch.aten.expand %27512, %27527, %false_25145 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25146 = torch.constant.int 4
%int4_25147 = torch.constant.int 4
%int128_25148 = torch.constant.int 128
%27529 = torch.prim.ListConstruct %int4_25146, %27105, %int4_25147, %int128_25148 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27530 = torch.aten.view %27514, %27529 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25149 = torch.constant.int 4
%int4_25150 = torch.constant.int 4
%int128_25151 = torch.constant.int 128
%27531 = torch.prim.ListConstruct %int4_25149, %27105, %int4_25150, %int128_25151 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27532 = torch.aten.view %27516, %27531 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25152 = torch.constant.int 4
%int4_25153 = torch.constant.int 4
%int128_25154 = torch.constant.int 128
%27533 = torch.prim.ListConstruct %int4_25152, %27105, %int4_25153, %int128_25154 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27534 = torch.aten.view %27518, %27533 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25155 = torch.constant.int 4
%int4_25156 = torch.constant.int 4
%int128_25157 = torch.constant.int 128
%27535 = torch.prim.ListConstruct %int4_25155, %27105, %int4_25156, %int128_25157 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27536 = torch.aten.view %27520, %27535 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25158 = torch.constant.int 4
%int4_25159 = torch.constant.int 4
%int128_25160 = torch.constant.int 128
%27537 = torch.prim.ListConstruct %int4_25158, %27105, %int4_25159, %int128_25160 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27538 = torch.aten.view %27522, %27537 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25161 = torch.constant.int 4
%int4_25162 = torch.constant.int 4
%int128_25163 = torch.constant.int 128
%27539 = torch.prim.ListConstruct %int4_25161, %27105, %int4_25162, %int128_25163 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27540 = torch.aten.view %27524, %27539 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25164 = torch.constant.int 4
%int4_25165 = torch.constant.int 4
%int128_25166 = torch.constant.int 128
%27541 = torch.prim.ListConstruct %int4_25164, %27105, %int4_25165, %int128_25166 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27542 = torch.aten.view %27526, %27541 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25167 = torch.constant.int 4
%int4_25168 = torch.constant.int 4
%int128_25169 = torch.constant.int 128
%27543 = torch.prim.ListConstruct %int4_25167, %27105, %int4_25168, %int128_25169 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27544 = torch.aten.view %27528, %27543 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_25170 = torch.constant.int -2
%27545 = torch.aten.unsqueeze %26894, %int-2_25170 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25171 = torch.constant.int -2
%27546 = torch.aten.unsqueeze %26896, %int-2_25171 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25172 = torch.constant.int -2
%27547 = torch.aten.unsqueeze %26898, %int-2_25172 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25173 = torch.constant.int -2
%27548 = torch.aten.unsqueeze %26900, %int-2_25173 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25174 = torch.constant.int -2
%27549 = torch.aten.unsqueeze %26902, %int-2_25174 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25175 = torch.constant.int -2
%27550 = torch.aten.unsqueeze %26904, %int-2_25175 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25176 = torch.constant.int -2
%27551 = torch.aten.unsqueeze %26906, %int-2_25176 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_25177 = torch.constant.int -2
%27552 = torch.aten.unsqueeze %26908, %int-2_25177 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %27552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_25178 = torch.constant.int 1
%27553 = torch.aten.size.int %26818, %int1_25178 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_25179 = torch.constant.int 4
%int1_25180 = torch.constant.int 1
%int4_25181 = torch.constant.int 4
%int128_25182 = torch.constant.int 128
%27554 = torch.prim.ListConstruct %int4_25179, %27553, %int1_25180, %int4_25181, %int128_25182 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25183 = torch.constant.bool false
%27555 = torch.aten.expand %27545, %27554, %false_25183 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25184 = torch.constant.int 4
%int1_25185 = torch.constant.int 1
%int4_25186 = torch.constant.int 4
%int128_25187 = torch.constant.int 128
%27556 = torch.prim.ListConstruct %int4_25184, %27553, %int1_25185, %int4_25186, %int128_25187 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25188 = torch.constant.bool false
%27557 = torch.aten.expand %27546, %27556, %false_25188 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25189 = torch.constant.int 4
%int1_25190 = torch.constant.int 1
%int4_25191 = torch.constant.int 4
%int128_25192 = torch.constant.int 128
%27558 = torch.prim.ListConstruct %int4_25189, %27553, %int1_25190, %int4_25191, %int128_25192 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25193 = torch.constant.bool false
%27559 = torch.aten.expand %27547, %27558, %false_25193 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25194 = torch.constant.int 4
%int1_25195 = torch.constant.int 1
%int4_25196 = torch.constant.int 4
%int128_25197 = torch.constant.int 128
%27560 = torch.prim.ListConstruct %int4_25194, %27553, %int1_25195, %int4_25196, %int128_25197 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25198 = torch.constant.bool false
%27561 = torch.aten.expand %27548, %27560, %false_25198 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25199 = torch.constant.int 4
%int1_25200 = torch.constant.int 1
%int4_25201 = torch.constant.int 4
%int128_25202 = torch.constant.int 128
%27562 = torch.prim.ListConstruct %int4_25199, %27553, %int1_25200, %int4_25201, %int128_25202 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25203 = torch.constant.bool false
%27563 = torch.aten.expand %27549, %27562, %false_25203 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25204 = torch.constant.int 4
%int1_25205 = torch.constant.int 1
%int4_25206 = torch.constant.int 4
%int128_25207 = torch.constant.int 128
%27564 = torch.prim.ListConstruct %int4_25204, %27553, %int1_25205, %int4_25206, %int128_25207 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25208 = torch.constant.bool false
%27565 = torch.aten.expand %27550, %27564, %false_25208 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25209 = torch.constant.int 4
%int1_25210 = torch.constant.int 1
%int4_25211 = torch.constant.int 4
%int128_25212 = torch.constant.int 128
%27566 = torch.prim.ListConstruct %int4_25209, %27553, %int1_25210, %int4_25211, %int128_25212 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25213 = torch.constant.bool false
%27567 = torch.aten.expand %27551, %27566, %false_25213 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25214 = torch.constant.int 4
%int1_25215 = torch.constant.int 1
%int4_25216 = torch.constant.int 4
%int128_25217 = torch.constant.int 128
%27568 = torch.prim.ListConstruct %int4_25214, %27553, %int1_25215, %int4_25216, %int128_25217 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_25218 = torch.constant.bool false
%27569 = torch.aten.expand %27552, %27568, %false_25218 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %27569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_25219 = torch.constant.int 4
%int4_25220 = torch.constant.int 4
%int128_25221 = torch.constant.int 128
%27570 = torch.prim.ListConstruct %int4_25219, %27553, %int4_25220, %int128_25221 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27571 = torch.aten.view %27555, %27570 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25222 = torch.constant.int 4
%int4_25223 = torch.constant.int 4
%int128_25224 = torch.constant.int 128
%27572 = torch.prim.ListConstruct %int4_25222, %27553, %int4_25223, %int128_25224 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27573 = torch.aten.view %27557, %27572 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25225 = torch.constant.int 4
%int4_25226 = torch.constant.int 4
%int128_25227 = torch.constant.int 128
%27574 = torch.prim.ListConstruct %int4_25225, %27553, %int4_25226, %int128_25227 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27575 = torch.aten.view %27559, %27574 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25228 = torch.constant.int 4
%int4_25229 = torch.constant.int 4
%int128_25230 = torch.constant.int 128
%27576 = torch.prim.ListConstruct %int4_25228, %27553, %int4_25229, %int128_25230 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27577 = torch.aten.view %27561, %27576 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25231 = torch.constant.int 4
%int4_25232 = torch.constant.int 4
%int128_25233 = torch.constant.int 128
%27578 = torch.prim.ListConstruct %int4_25231, %27553, %int4_25232, %int128_25233 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27579 = torch.aten.view %27563, %27578 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25234 = torch.constant.int 4
%int4_25235 = torch.constant.int 4
%int128_25236 = torch.constant.int 128
%27580 = torch.prim.ListConstruct %int4_25234, %27553, %int4_25235, %int128_25236 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27581 = torch.aten.view %27565, %27580 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25237 = torch.constant.int 4
%int4_25238 = torch.constant.int 4
%int128_25239 = torch.constant.int 128
%27582 = torch.prim.ListConstruct %int4_25237, %27553, %int4_25238, %int128_25239 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27583 = torch.aten.view %27567, %27582 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25240 = torch.constant.int 4
%int4_25241 = torch.constant.int 4
%int128_25242 = torch.constant.int 128
%27584 = torch.prim.ListConstruct %int4_25240, %27553, %int4_25241, %int128_25242 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27585 = torch.aten.view %27569, %27584 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25243 = torch.constant.int 1
%int2_25244 = torch.constant.int 2
%27586 = torch.aten.transpose.int %26961, %int1_25243, %int2_25244 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27586, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25245 = torch.constant.int 1
%int2_25246 = torch.constant.int 2
%27587 = torch.aten.transpose.int %26976, %int1_25245, %int2_25246 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27587, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25247 = torch.constant.int 1
%int2_25248 = torch.constant.int 2
%27588 = torch.aten.transpose.int %26991, %int1_25247, %int2_25248 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27588, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25249 = torch.constant.int 1
%int2_25250 = torch.constant.int 2
%27589 = torch.aten.transpose.int %27006, %int1_25249, %int2_25250 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27589, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25251 = torch.constant.int 1
%int2_25252 = torch.constant.int 2
%27590 = torch.aten.transpose.int %27021, %int1_25251, %int2_25252 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27590, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25253 = torch.constant.int 1
%int2_25254 = torch.constant.int 2
%27591 = torch.aten.transpose.int %27036, %int1_25253, %int2_25254 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27591, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25255 = torch.constant.int 1
%int2_25256 = torch.constant.int 2
%27592 = torch.aten.transpose.int %27051, %int1_25255, %int2_25256 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27592, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25257 = torch.constant.int 1
%int2_25258 = torch.constant.int 2
%27593 = torch.aten.transpose.int %27066, %int1_25257, %int2_25258 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27593, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25259 = torch.constant.int 1
%int2_25260 = torch.constant.int 2
%27594 = torch.aten.transpose.int %27530, %int1_25259, %int2_25260 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27594, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25261 = torch.constant.int 1
%int2_25262 = torch.constant.int 2
%27595 = torch.aten.transpose.int %27532, %int1_25261, %int2_25262 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27595, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25263 = torch.constant.int 1
%int2_25264 = torch.constant.int 2
%27596 = torch.aten.transpose.int %27534, %int1_25263, %int2_25264 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27596, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25265 = torch.constant.int 1
%int2_25266 = torch.constant.int 2
%27597 = torch.aten.transpose.int %27536, %int1_25265, %int2_25266 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27597, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25267 = torch.constant.int 1
%int2_25268 = torch.constant.int 2
%27598 = torch.aten.transpose.int %27538, %int1_25267, %int2_25268 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27598, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25269 = torch.constant.int 1
%int2_25270 = torch.constant.int 2
%27599 = torch.aten.transpose.int %27540, %int1_25269, %int2_25270 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27599, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25271 = torch.constant.int 1
%int2_25272 = torch.constant.int 2
%27600 = torch.aten.transpose.int %27542, %int1_25271, %int2_25272 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27600, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25273 = torch.constant.int 1
%int2_25274 = torch.constant.int 2
%27601 = torch.aten.transpose.int %27544, %int1_25273, %int2_25274 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27601, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25275 = torch.constant.int 1
%int2_25276 = torch.constant.int 2
%27602 = torch.aten.transpose.int %27571, %int1_25275, %int2_25276 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27602, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25277 = torch.constant.int 1
%int2_25278 = torch.constant.int 2
%27603 = torch.aten.transpose.int %27573, %int1_25277, %int2_25278 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27603, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25279 = torch.constant.int 1
%int2_25280 = torch.constant.int 2
%27604 = torch.aten.transpose.int %27575, %int1_25279, %int2_25280 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27604, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25281 = torch.constant.int 1
%int2_25282 = torch.constant.int 2
%27605 = torch.aten.transpose.int %27577, %int1_25281, %int2_25282 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27605, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25283 = torch.constant.int 1
%int2_25284 = torch.constant.int 2
%27606 = torch.aten.transpose.int %27579, %int1_25283, %int2_25284 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27606, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25285 = torch.constant.int 1
%int2_25286 = torch.constant.int 2
%27607 = torch.aten.transpose.int %27581, %int1_25285, %int2_25286 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27607, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25287 = torch.constant.int 1
%int2_25288 = torch.constant.int 2
%27608 = torch.aten.transpose.int %27583, %int1_25287, %int2_25288 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27608, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25289 = torch.constant.int 1
%int2_25290 = torch.constant.int 2
%27609 = torch.aten.transpose.int %27585, %int1_25289, %int2_25290 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %27609, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25291 = torch.constant.float 0.000000e+00
%true_25292 = torch.constant.bool true
%none_25293 = torch.constant.none
%none_25294 = torch.constant.none
%27610:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27586, %27594, %27602, %float0.000000e00_25291, %true_25292, %none_25293, %none_25294) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27610#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25295 = torch.constant.float 0.000000e+00
%true_25296 = torch.constant.bool true
%none_25297 = torch.constant.none
%none_25298 = torch.constant.none
%27611:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27587, %27595, %27603, %float0.000000e00_25295, %true_25296, %none_25297, %none_25298) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27611#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25299 = torch.constant.float 0.000000e+00
%true_25300 = torch.constant.bool true
%none_25301 = torch.constant.none
%none_25302 = torch.constant.none
%27612:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27588, %27596, %27604, %float0.000000e00_25299, %true_25300, %none_25301, %none_25302) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27612#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25303 = torch.constant.float 0.000000e+00
%true_25304 = torch.constant.bool true
%none_25305 = torch.constant.none
%none_25306 = torch.constant.none
%27613:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27589, %27597, %27605, %float0.000000e00_25303, %true_25304, %none_25305, %none_25306) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27613#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25307 = torch.constant.float 0.000000e+00
%true_25308 = torch.constant.bool true
%none_25309 = torch.constant.none
%none_25310 = torch.constant.none
%27614:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27590, %27598, %27606, %float0.000000e00_25307, %true_25308, %none_25309, %none_25310) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27614#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25311 = torch.constant.float 0.000000e+00
%true_25312 = torch.constant.bool true
%none_25313 = torch.constant.none
%none_25314 = torch.constant.none
%27615:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27591, %27599, %27607, %float0.000000e00_25311, %true_25312, %none_25313, %none_25314) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27615#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25315 = torch.constant.float 0.000000e+00
%true_25316 = torch.constant.bool true
%none_25317 = torch.constant.none
%none_25318 = torch.constant.none
%27616:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27592, %27600, %27608, %float0.000000e00_25315, %true_25316, %none_25317, %none_25318) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27616#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_25319 = torch.constant.float 0.000000e+00
%true_25320 = torch.constant.bool true
%none_25321 = torch.constant.none
%none_25322 = torch.constant.none
%27617:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%27593, %27601, %27609, %float0.000000e00_25319, %true_25320, %none_25321, %none_25322) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %27617#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_25323 = torch.constant.int 1
%int2_25324 = torch.constant.int 2
%27618 = torch.aten.transpose.int %27610#0, %int1_25323, %int2_25324 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25325 = torch.constant.int 1
%int2_25326 = torch.constant.int 2
%27619 = torch.aten.transpose.int %27611#0, %int1_25325, %int2_25326 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25327 = torch.constant.int 1
%int2_25328 = torch.constant.int 2
%27620 = torch.aten.transpose.int %27612#0, %int1_25327, %int2_25328 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25329 = torch.constant.int 1
%int2_25330 = torch.constant.int 2
%27621 = torch.aten.transpose.int %27613#0, %int1_25329, %int2_25330 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25331 = torch.constant.int 1
%int2_25332 = torch.constant.int 2
%27622 = torch.aten.transpose.int %27614#0, %int1_25331, %int2_25332 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25333 = torch.constant.int 1
%int2_25334 = torch.constant.int 2
%27623 = torch.aten.transpose.int %27615#0, %int1_25333, %int2_25334 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25335 = torch.constant.int 1
%int2_25336 = torch.constant.int 2
%27624 = torch.aten.transpose.int %27616#0, %int1_25335, %int2_25336 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_25337 = torch.constant.int 1
%int2_25338 = torch.constant.int 2
%27625 = torch.aten.transpose.int %27617#0, %int1_25337, %int2_25338 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %27625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_25339 = torch.constant.int 4
%int512_25340 = torch.constant.int 512
%27626 = torch.prim.ListConstruct %int4_25339, %26947, %int512_25340 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27627 = torch.aten.view %27618, %27626 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25341 = torch.constant.int 4
%int512_25342 = torch.constant.int 512
%27628 = torch.prim.ListConstruct %int4_25341, %26962, %int512_25342 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27629 = torch.aten.view %27619, %27628 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25343 = torch.constant.int 4
%int512_25344 = torch.constant.int 512
%27630 = torch.prim.ListConstruct %int4_25343, %26977, %int512_25344 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27631 = torch.aten.view %27620, %27630 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25345 = torch.constant.int 4
%int512_25346 = torch.constant.int 512
%27632 = torch.prim.ListConstruct %int4_25345, %26992, %int512_25346 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27633 = torch.aten.view %27621, %27632 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25347 = torch.constant.int 4
%int512_25348 = torch.constant.int 512
%27634 = torch.prim.ListConstruct %int4_25347, %27007, %int512_25348 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27635 = torch.aten.view %27622, %27634 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25349 = torch.constant.int 4
%int512_25350 = torch.constant.int 512
%27636 = torch.prim.ListConstruct %int4_25349, %27022, %int512_25350 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27637 = torch.aten.view %27623, %27636 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25351 = torch.constant.int 4
%int512_25352 = torch.constant.int 512
%27638 = torch.prim.ListConstruct %int4_25351, %27037, %int512_25352 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27639 = torch.aten.view %27624, %27638 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_25353 = torch.constant.int 4
%int512_25354 = torch.constant.int 512
%27640 = torch.prim.ListConstruct %int4_25353, %27052, %int512_25354 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27641 = torch.aten.view %27625, %27640 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %27641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_25355 = torch.constant.int 1
%int0_25356 = torch.constant.int 0
%27642 = torch.prim.ListConstruct %int1_25355, %int0_25356 : (!torch.int, !torch.int) -> !torch.list<int>
%27643 = torch.aten.permute %976, %27642 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25357 = torch.constant.int 1
%int0_25358 = torch.constant.int 0
%27644 = torch.prim.ListConstruct %int1_25357, %int0_25358 : (!torch.int, !torch.int) -> !torch.list<int>
%27645 = torch.aten.permute %977, %27644 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25359 = torch.constant.int 1
%int0_25360 = torch.constant.int 0
%27646 = torch.prim.ListConstruct %int1_25359, %int0_25360 : (!torch.int, !torch.int) -> !torch.list<int>
%27647 = torch.aten.permute %978, %27646 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25361 = torch.constant.int 1
%int0_25362 = torch.constant.int 0
%27648 = torch.prim.ListConstruct %int1_25361, %int0_25362 : (!torch.int, !torch.int) -> !torch.list<int>
%27649 = torch.aten.permute %979, %27648 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25363 = torch.constant.int 1
%int0_25364 = torch.constant.int 0
%27650 = torch.prim.ListConstruct %int1_25363, %int0_25364 : (!torch.int, !torch.int) -> !torch.list<int>
%27651 = torch.aten.permute %980, %27650 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25365 = torch.constant.int 1
%int0_25366 = torch.constant.int 0
%27652 = torch.prim.ListConstruct %int1_25365, %int0_25366 : (!torch.int, !torch.int) -> !torch.list<int>
%27653 = torch.aten.permute %981, %27652 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25367 = torch.constant.int 1
%int0_25368 = torch.constant.int 0
%27654 = torch.prim.ListConstruct %int1_25367, %int0_25368 : (!torch.int, !torch.int) -> !torch.list<int>
%27655 = torch.aten.permute %982, %27654 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_25369 = torch.constant.int 1
%int0_25370 = torch.constant.int 0
%27656 = torch.prim.ListConstruct %int1_25369, %int0_25370 : (!torch.int, !torch.int) -> !torch.list<int>
%27657 = torch.aten.permute %983, %27656 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_25371 = torch.constant.int 4
%27658 = torch.aten.mul.int %int4_25371, %26947 : !torch.int, !torch.int -> !torch.int
%int512_25372 = torch.constant.int 512
%27659 = torch.prim.ListConstruct %27658, %int512_25372 : (!torch.int, !torch.int) -> !torch.list<int>
%27660 = torch.aten.view %27627, %27659 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27660, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27661 = torch.aten.mm %27660, %27643 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27661, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25373 = torch.constant.int 4
%int4096_25374 = torch.constant.int 4096
%27662 = torch.prim.ListConstruct %int4_25373, %26947, %int4096_25374 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27663 = torch.aten.view %27661, %27662 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25375 = torch.constant.int 4
%27664 = torch.aten.mul.int %int4_25375, %26962 : !torch.int, !torch.int -> !torch.int
%int512_25376 = torch.constant.int 512
%27665 = torch.prim.ListConstruct %27664, %int512_25376 : (!torch.int, !torch.int) -> !torch.list<int>
%27666 = torch.aten.view %27629, %27665 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27666, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27667 = torch.aten.mm %27666, %27645 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27667, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25377 = torch.constant.int 4
%int4096_25378 = torch.constant.int 4096
%27668 = torch.prim.ListConstruct %int4_25377, %26962, %int4096_25378 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27669 = torch.aten.view %27667, %27668 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25379 = torch.constant.int 4
%27670 = torch.aten.mul.int %int4_25379, %26977 : !torch.int, !torch.int -> !torch.int
%int512_25380 = torch.constant.int 512
%27671 = torch.prim.ListConstruct %27670, %int512_25380 : (!torch.int, !torch.int) -> !torch.list<int>
%27672 = torch.aten.view %27631, %27671 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27672, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27673 = torch.aten.mm %27672, %27647 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27673, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25381 = torch.constant.int 4
%int4096_25382 = torch.constant.int 4096
%27674 = torch.prim.ListConstruct %int4_25381, %26977, %int4096_25382 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27675 = torch.aten.view %27673, %27674 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25383 = torch.constant.int 4
%27676 = torch.aten.mul.int %int4_25383, %26992 : !torch.int, !torch.int -> !torch.int
%int512_25384 = torch.constant.int 512
%27677 = torch.prim.ListConstruct %27676, %int512_25384 : (!torch.int, !torch.int) -> !torch.list<int>
%27678 = torch.aten.view %27633, %27677 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27678, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27679 = torch.aten.mm %27678, %27649 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27679, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25385 = torch.constant.int 4
%int4096_25386 = torch.constant.int 4096
%27680 = torch.prim.ListConstruct %int4_25385, %26992, %int4096_25386 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27681 = torch.aten.view %27679, %27680 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25387 = torch.constant.int 4
%27682 = torch.aten.mul.int %int4_25387, %27007 : !torch.int, !torch.int -> !torch.int
%int512_25388 = torch.constant.int 512
%27683 = torch.prim.ListConstruct %27682, %int512_25388 : (!torch.int, !torch.int) -> !torch.list<int>
%27684 = torch.aten.view %27635, %27683 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27684, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27685 = torch.aten.mm %27684, %27651 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27685, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25389 = torch.constant.int 4
%int4096_25390 = torch.constant.int 4096
%27686 = torch.prim.ListConstruct %int4_25389, %27007, %int4096_25390 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27687 = torch.aten.view %27685, %27686 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25391 = torch.constant.int 4
%27688 = torch.aten.mul.int %int4_25391, %27022 : !torch.int, !torch.int -> !torch.int
%int512_25392 = torch.constant.int 512
%27689 = torch.prim.ListConstruct %27688, %int512_25392 : (!torch.int, !torch.int) -> !torch.list<int>
%27690 = torch.aten.view %27637, %27689 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27690, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27691 = torch.aten.mm %27690, %27653 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27691, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25393 = torch.constant.int 4
%int4096_25394 = torch.constant.int 4096
%27692 = torch.prim.ListConstruct %int4_25393, %27022, %int4096_25394 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27693 = torch.aten.view %27691, %27692 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25395 = torch.constant.int 4
%27694 = torch.aten.mul.int %int4_25395, %27037 : !torch.int, !torch.int -> !torch.int
%int512_25396 = torch.constant.int 512
%27695 = torch.prim.ListConstruct %27694, %int512_25396 : (!torch.int, !torch.int) -> !torch.list<int>
%27696 = torch.aten.view %27639, %27695 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27696, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27697 = torch.aten.mm %27696, %27655 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27697, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25397 = torch.constant.int 4
%int4096_25398 = torch.constant.int 4096
%27698 = torch.prim.ListConstruct %int4_25397, %27037, %int4096_25398 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27699 = torch.aten.view %27697, %27698 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_25399 = torch.constant.int 4
%27700 = torch.aten.mul.int %int4_25399, %27052 : !torch.int, !torch.int -> !torch.int
%int512_25400 = torch.constant.int 512
%27701 = torch.prim.ListConstruct %27700, %int512_25400 : (!torch.int, !torch.int) -> !torch.list<int>
%27702 = torch.aten.view %27641, %27701 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %27702, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%27703 = torch.aten.mm %27702, %27657 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %27703, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25401 = torch.constant.int 4
%int4096_25402 = torch.constant.int 4096
%27704 = torch.prim.ListConstruct %int4_25401, %27052, %int4096_25402 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%27705 = torch.aten.view %27703, %27704 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27706 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25403 = arith.constant 1 : index
%dim_25404 = tensor.dim %27706, %c1_25403 : tensor<4x?x4096xf16>
%27707 = flow.tensor.transfer %27706 : tensor<4x?x4096xf16>{%dim_25404} to #hal.device.promise<@__device_0>
%27708 = torch_c.from_builtin_tensor %27707 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27709 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25405 = arith.constant 1 : index
%dim_25406 = tensor.dim %27709, %c1_25405 : tensor<4x?x4096xf16>
%27710 = flow.tensor.transfer %27709 : tensor<4x?x4096xf16>{%dim_25406} to #hal.device.promise<@__device_0>
%27711 = torch_c.from_builtin_tensor %27710 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27712 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25407 = arith.constant 1 : index
%dim_25408 = tensor.dim %27712, %c1_25407 : tensor<4x?x4096xf16>
%27713 = flow.tensor.transfer %27712 : tensor<4x?x4096xf16>{%dim_25408} to #hal.device.promise<@__device_0>
%27714 = torch_c.from_builtin_tensor %27713 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27715 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25409 = arith.constant 1 : index
%dim_25410 = tensor.dim %27715, %c1_25409 : tensor<4x?x4096xf16>
%27716 = flow.tensor.transfer %27715 : tensor<4x?x4096xf16>{%dim_25410} to #hal.device.promise<@__device_0>
%27717 = torch_c.from_builtin_tensor %27716 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27718 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25411 = arith.constant 1 : index
%dim_25412 = tensor.dim %27718, %c1_25411 : tensor<4x?x4096xf16>
%27719 = flow.tensor.transfer %27718 : tensor<4x?x4096xf16>{%dim_25412} to #hal.device.promise<@__device_0>
%27720 = torch_c.from_builtin_tensor %27719 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27721 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25413 = arith.constant 1 : index
%dim_25414 = tensor.dim %27721, %c1_25413 : tensor<4x?x4096xf16>
%27722 = flow.tensor.transfer %27721 : tensor<4x?x4096xf16>{%dim_25414} to #hal.device.promise<@__device_0>
%27723 = torch_c.from_builtin_tensor %27722 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27724 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25415 = arith.constant 1 : index
%dim_25416 = tensor.dim %27724, %c1_25415 : tensor<4x?x4096xf16>
%27725 = flow.tensor.transfer %27724 : tensor<4x?x4096xf16>{%dim_25416} to #hal.device.promise<@__device_0>
%27726 = torch_c.from_builtin_tensor %27725 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25417 = torch.constant.int 1
%27727 = torch.aten.add.Tensor %27663, %27708, %int1_25417 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25418 = torch.constant.int 1
%27728 = torch.aten.add.Tensor %27727, %27711, %int1_25418 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25419 = torch.constant.int 1
%27729 = torch.aten.add.Tensor %27728, %27714, %int1_25419 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25420 = torch.constant.int 1
%27730 = torch.aten.add.Tensor %27729, %27717, %int1_25420 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25421 = torch.constant.int 1
%27731 = torch.aten.add.Tensor %27730, %27720, %int1_25421 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25422 = torch.constant.int 1
%27732 = torch.aten.add.Tensor %27731, %27723, %int1_25422 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25423 = torch.constant.int 1
%27733 = torch.aten.add.Tensor %27732, %27726, %int1_25423 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27734 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25424 = arith.constant 1 : index
%dim_25425 = tensor.dim %27734, %c1_25424 : tensor<4x?x4096xf16>
%27735 = flow.tensor.transfer %27734 : tensor<4x?x4096xf16>{%dim_25425} to #hal.device.promise<@__device_1>
%27736 = torch_c.from_builtin_tensor %27735 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27737 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25426 = arith.constant 1 : index
%dim_25427 = tensor.dim %27737, %c1_25426 : tensor<4x?x4096xf16>
%27738 = flow.tensor.transfer %27737 : tensor<4x?x4096xf16>{%dim_25427} to #hal.device.promise<@__device_1>
%27739 = torch_c.from_builtin_tensor %27738 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27740 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25428 = arith.constant 1 : index
%dim_25429 = tensor.dim %27740, %c1_25428 : tensor<4x?x4096xf16>
%27741 = flow.tensor.transfer %27740 : tensor<4x?x4096xf16>{%dim_25429} to #hal.device.promise<@__device_1>
%27742 = torch_c.from_builtin_tensor %27741 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27743 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25430 = arith.constant 1 : index
%dim_25431 = tensor.dim %27743, %c1_25430 : tensor<4x?x4096xf16>
%27744 = flow.tensor.transfer %27743 : tensor<4x?x4096xf16>{%dim_25431} to #hal.device.promise<@__device_1>
%27745 = torch_c.from_builtin_tensor %27744 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27746 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25432 = arith.constant 1 : index
%dim_25433 = tensor.dim %27746, %c1_25432 : tensor<4x?x4096xf16>
%27747 = flow.tensor.transfer %27746 : tensor<4x?x4096xf16>{%dim_25433} to #hal.device.promise<@__device_1>
%27748 = torch_c.from_builtin_tensor %27747 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27749 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25434 = arith.constant 1 : index
%dim_25435 = tensor.dim %27749, %c1_25434 : tensor<4x?x4096xf16>
%27750 = flow.tensor.transfer %27749 : tensor<4x?x4096xf16>{%dim_25435} to #hal.device.promise<@__device_1>
%27751 = torch_c.from_builtin_tensor %27750 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27752 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25436 = arith.constant 1 : index
%dim_25437 = tensor.dim %27752, %c1_25436 : tensor<4x?x4096xf16>
%27753 = flow.tensor.transfer %27752 : tensor<4x?x4096xf16>{%dim_25437} to #hal.device.promise<@__device_1>
%27754 = torch_c.from_builtin_tensor %27753 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25438 = torch.constant.int 1
%27755 = torch.aten.add.Tensor %27736, %27669, %int1_25438 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25439 = torch.constant.int 1
%27756 = torch.aten.add.Tensor %27755, %27739, %int1_25439 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25440 = torch.constant.int 1
%27757 = torch.aten.add.Tensor %27756, %27742, %int1_25440 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25441 = torch.constant.int 1
%27758 = torch.aten.add.Tensor %27757, %27745, %int1_25441 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25442 = torch.constant.int 1
%27759 = torch.aten.add.Tensor %27758, %27748, %int1_25442 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25443 = torch.constant.int 1
%27760 = torch.aten.add.Tensor %27759, %27751, %int1_25443 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25444 = torch.constant.int 1
%27761 = torch.aten.add.Tensor %27760, %27754, %int1_25444 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27762 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25445 = arith.constant 1 : index
%dim_25446 = tensor.dim %27762, %c1_25445 : tensor<4x?x4096xf16>
%27763 = flow.tensor.transfer %27762 : tensor<4x?x4096xf16>{%dim_25446} to #hal.device.promise<@__device_2>
%27764 = torch_c.from_builtin_tensor %27763 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27765 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25447 = arith.constant 1 : index
%dim_25448 = tensor.dim %27765, %c1_25447 : tensor<4x?x4096xf16>
%27766 = flow.tensor.transfer %27765 : tensor<4x?x4096xf16>{%dim_25448} to #hal.device.promise<@__device_2>
%27767 = torch_c.from_builtin_tensor %27766 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27768 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25449 = arith.constant 1 : index
%dim_25450 = tensor.dim %27768, %c1_25449 : tensor<4x?x4096xf16>
%27769 = flow.tensor.transfer %27768 : tensor<4x?x4096xf16>{%dim_25450} to #hal.device.promise<@__device_2>
%27770 = torch_c.from_builtin_tensor %27769 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27771 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25451 = arith.constant 1 : index
%dim_25452 = tensor.dim %27771, %c1_25451 : tensor<4x?x4096xf16>
%27772 = flow.tensor.transfer %27771 : tensor<4x?x4096xf16>{%dim_25452} to #hal.device.promise<@__device_2>
%27773 = torch_c.from_builtin_tensor %27772 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27774 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25453 = arith.constant 1 : index
%dim_25454 = tensor.dim %27774, %c1_25453 : tensor<4x?x4096xf16>
%27775 = flow.tensor.transfer %27774 : tensor<4x?x4096xf16>{%dim_25454} to #hal.device.promise<@__device_2>
%27776 = torch_c.from_builtin_tensor %27775 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27777 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25455 = arith.constant 1 : index
%dim_25456 = tensor.dim %27777, %c1_25455 : tensor<4x?x4096xf16>
%27778 = flow.tensor.transfer %27777 : tensor<4x?x4096xf16>{%dim_25456} to #hal.device.promise<@__device_2>
%27779 = torch_c.from_builtin_tensor %27778 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27780 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25457 = arith.constant 1 : index
%dim_25458 = tensor.dim %27780, %c1_25457 : tensor<4x?x4096xf16>
%27781 = flow.tensor.transfer %27780 : tensor<4x?x4096xf16>{%dim_25458} to #hal.device.promise<@__device_2>
%27782 = torch_c.from_builtin_tensor %27781 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25459 = torch.constant.int 1
%27783 = torch.aten.add.Tensor %27764, %27767, %int1_25459 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25460 = torch.constant.int 1
%27784 = torch.aten.add.Tensor %27783, %27675, %int1_25460 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25461 = torch.constant.int 1
%27785 = torch.aten.add.Tensor %27784, %27770, %int1_25461 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25462 = torch.constant.int 1
%27786 = torch.aten.add.Tensor %27785, %27773, %int1_25462 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25463 = torch.constant.int 1
%27787 = torch.aten.add.Tensor %27786, %27776, %int1_25463 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25464 = torch.constant.int 1
%27788 = torch.aten.add.Tensor %27787, %27779, %int1_25464 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25465 = torch.constant.int 1
%27789 = torch.aten.add.Tensor %27788, %27782, %int1_25465 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27790 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25466 = arith.constant 1 : index
%dim_25467 = tensor.dim %27790, %c1_25466 : tensor<4x?x4096xf16>
%27791 = flow.tensor.transfer %27790 : tensor<4x?x4096xf16>{%dim_25467} to #hal.device.promise<@__device_3>
%27792 = torch_c.from_builtin_tensor %27791 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27793 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25468 = arith.constant 1 : index
%dim_25469 = tensor.dim %27793, %c1_25468 : tensor<4x?x4096xf16>
%27794 = flow.tensor.transfer %27793 : tensor<4x?x4096xf16>{%dim_25469} to #hal.device.promise<@__device_3>
%27795 = torch_c.from_builtin_tensor %27794 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27796 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25470 = arith.constant 1 : index
%dim_25471 = tensor.dim %27796, %c1_25470 : tensor<4x?x4096xf16>
%27797 = flow.tensor.transfer %27796 : tensor<4x?x4096xf16>{%dim_25471} to #hal.device.promise<@__device_3>
%27798 = torch_c.from_builtin_tensor %27797 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27799 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25472 = arith.constant 1 : index
%dim_25473 = tensor.dim %27799, %c1_25472 : tensor<4x?x4096xf16>
%27800 = flow.tensor.transfer %27799 : tensor<4x?x4096xf16>{%dim_25473} to #hal.device.promise<@__device_3>
%27801 = torch_c.from_builtin_tensor %27800 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27802 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25474 = arith.constant 1 : index
%dim_25475 = tensor.dim %27802, %c1_25474 : tensor<4x?x4096xf16>
%27803 = flow.tensor.transfer %27802 : tensor<4x?x4096xf16>{%dim_25475} to #hal.device.promise<@__device_3>
%27804 = torch_c.from_builtin_tensor %27803 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27805 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25476 = arith.constant 1 : index
%dim_25477 = tensor.dim %27805, %c1_25476 : tensor<4x?x4096xf16>
%27806 = flow.tensor.transfer %27805 : tensor<4x?x4096xf16>{%dim_25477} to #hal.device.promise<@__device_3>
%27807 = torch_c.from_builtin_tensor %27806 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27808 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25478 = arith.constant 1 : index
%dim_25479 = tensor.dim %27808, %c1_25478 : tensor<4x?x4096xf16>
%27809 = flow.tensor.transfer %27808 : tensor<4x?x4096xf16>{%dim_25479} to #hal.device.promise<@__device_3>
%27810 = torch_c.from_builtin_tensor %27809 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25480 = torch.constant.int 1
%27811 = torch.aten.add.Tensor %27792, %27795, %int1_25480 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25481 = torch.constant.int 1
%27812 = torch.aten.add.Tensor %27811, %27798, %int1_25481 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25482 = torch.constant.int 1
%27813 = torch.aten.add.Tensor %27812, %27681, %int1_25482 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25483 = torch.constant.int 1
%27814 = torch.aten.add.Tensor %27813, %27801, %int1_25483 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25484 = torch.constant.int 1
%27815 = torch.aten.add.Tensor %27814, %27804, %int1_25484 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25485 = torch.constant.int 1
%27816 = torch.aten.add.Tensor %27815, %27807, %int1_25485 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25486 = torch.constant.int 1
%27817 = torch.aten.add.Tensor %27816, %27810, %int1_25486 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27818 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25487 = arith.constant 1 : index
%dim_25488 = tensor.dim %27818, %c1_25487 : tensor<4x?x4096xf16>
%27819 = flow.tensor.transfer %27818 : tensor<4x?x4096xf16>{%dim_25488} to #hal.device.promise<@__device_4>
%27820 = torch_c.from_builtin_tensor %27819 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27821 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25489 = arith.constant 1 : index
%dim_25490 = tensor.dim %27821, %c1_25489 : tensor<4x?x4096xf16>
%27822 = flow.tensor.transfer %27821 : tensor<4x?x4096xf16>{%dim_25490} to #hal.device.promise<@__device_4>
%27823 = torch_c.from_builtin_tensor %27822 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27824 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25491 = arith.constant 1 : index
%dim_25492 = tensor.dim %27824, %c1_25491 : tensor<4x?x4096xf16>
%27825 = flow.tensor.transfer %27824 : tensor<4x?x4096xf16>{%dim_25492} to #hal.device.promise<@__device_4>
%27826 = torch_c.from_builtin_tensor %27825 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27827 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25493 = arith.constant 1 : index
%dim_25494 = tensor.dim %27827, %c1_25493 : tensor<4x?x4096xf16>
%27828 = flow.tensor.transfer %27827 : tensor<4x?x4096xf16>{%dim_25494} to #hal.device.promise<@__device_4>
%27829 = torch_c.from_builtin_tensor %27828 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27830 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25495 = arith.constant 1 : index
%dim_25496 = tensor.dim %27830, %c1_25495 : tensor<4x?x4096xf16>
%27831 = flow.tensor.transfer %27830 : tensor<4x?x4096xf16>{%dim_25496} to #hal.device.promise<@__device_4>
%27832 = torch_c.from_builtin_tensor %27831 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27833 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25497 = arith.constant 1 : index
%dim_25498 = tensor.dim %27833, %c1_25497 : tensor<4x?x4096xf16>
%27834 = flow.tensor.transfer %27833 : tensor<4x?x4096xf16>{%dim_25498} to #hal.device.promise<@__device_4>
%27835 = torch_c.from_builtin_tensor %27834 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27836 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25499 = arith.constant 1 : index
%dim_25500 = tensor.dim %27836, %c1_25499 : tensor<4x?x4096xf16>
%27837 = flow.tensor.transfer %27836 : tensor<4x?x4096xf16>{%dim_25500} to #hal.device.promise<@__device_4>
%27838 = torch_c.from_builtin_tensor %27837 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25501 = torch.constant.int 1
%27839 = torch.aten.add.Tensor %27820, %27823, %int1_25501 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25502 = torch.constant.int 1
%27840 = torch.aten.add.Tensor %27839, %27826, %int1_25502 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25503 = torch.constant.int 1
%27841 = torch.aten.add.Tensor %27840, %27829, %int1_25503 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25504 = torch.constant.int 1
%27842 = torch.aten.add.Tensor %27841, %27687, %int1_25504 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25505 = torch.constant.int 1
%27843 = torch.aten.add.Tensor %27842, %27832, %int1_25505 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25506 = torch.constant.int 1
%27844 = torch.aten.add.Tensor %27843, %27835, %int1_25506 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25507 = torch.constant.int 1
%27845 = torch.aten.add.Tensor %27844, %27838, %int1_25507 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27846 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25508 = arith.constant 1 : index
%dim_25509 = tensor.dim %27846, %c1_25508 : tensor<4x?x4096xf16>
%27847 = flow.tensor.transfer %27846 : tensor<4x?x4096xf16>{%dim_25509} to #hal.device.promise<@__device_5>
%27848 = torch_c.from_builtin_tensor %27847 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27849 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25510 = arith.constant 1 : index
%dim_25511 = tensor.dim %27849, %c1_25510 : tensor<4x?x4096xf16>
%27850 = flow.tensor.transfer %27849 : tensor<4x?x4096xf16>{%dim_25511} to #hal.device.promise<@__device_5>
%27851 = torch_c.from_builtin_tensor %27850 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27852 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25512 = arith.constant 1 : index
%dim_25513 = tensor.dim %27852, %c1_25512 : tensor<4x?x4096xf16>
%27853 = flow.tensor.transfer %27852 : tensor<4x?x4096xf16>{%dim_25513} to #hal.device.promise<@__device_5>
%27854 = torch_c.from_builtin_tensor %27853 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27855 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25514 = arith.constant 1 : index
%dim_25515 = tensor.dim %27855, %c1_25514 : tensor<4x?x4096xf16>
%27856 = flow.tensor.transfer %27855 : tensor<4x?x4096xf16>{%dim_25515} to #hal.device.promise<@__device_5>
%27857 = torch_c.from_builtin_tensor %27856 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27858 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25516 = arith.constant 1 : index
%dim_25517 = tensor.dim %27858, %c1_25516 : tensor<4x?x4096xf16>
%27859 = flow.tensor.transfer %27858 : tensor<4x?x4096xf16>{%dim_25517} to #hal.device.promise<@__device_5>
%27860 = torch_c.from_builtin_tensor %27859 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27861 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25518 = arith.constant 1 : index
%dim_25519 = tensor.dim %27861, %c1_25518 : tensor<4x?x4096xf16>
%27862 = flow.tensor.transfer %27861 : tensor<4x?x4096xf16>{%dim_25519} to #hal.device.promise<@__device_5>
%27863 = torch_c.from_builtin_tensor %27862 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27864 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25520 = arith.constant 1 : index
%dim_25521 = tensor.dim %27864, %c1_25520 : tensor<4x?x4096xf16>
%27865 = flow.tensor.transfer %27864 : tensor<4x?x4096xf16>{%dim_25521} to #hal.device.promise<@__device_5>
%27866 = torch_c.from_builtin_tensor %27865 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25522 = torch.constant.int 1
%27867 = torch.aten.add.Tensor %27848, %27851, %int1_25522 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25523 = torch.constant.int 1
%27868 = torch.aten.add.Tensor %27867, %27854, %int1_25523 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25524 = torch.constant.int 1
%27869 = torch.aten.add.Tensor %27868, %27857, %int1_25524 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25525 = torch.constant.int 1
%27870 = torch.aten.add.Tensor %27869, %27860, %int1_25525 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25526 = torch.constant.int 1
%27871 = torch.aten.add.Tensor %27870, %27693, %int1_25526 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25527 = torch.constant.int 1
%27872 = torch.aten.add.Tensor %27871, %27863, %int1_25527 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25528 = torch.constant.int 1
%27873 = torch.aten.add.Tensor %27872, %27866, %int1_25528 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27874 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25529 = arith.constant 1 : index
%dim_25530 = tensor.dim %27874, %c1_25529 : tensor<4x?x4096xf16>
%27875 = flow.tensor.transfer %27874 : tensor<4x?x4096xf16>{%dim_25530} to #hal.device.promise<@__device_6>
%27876 = torch_c.from_builtin_tensor %27875 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27877 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25531 = arith.constant 1 : index
%dim_25532 = tensor.dim %27877, %c1_25531 : tensor<4x?x4096xf16>
%27878 = flow.tensor.transfer %27877 : tensor<4x?x4096xf16>{%dim_25532} to #hal.device.promise<@__device_6>
%27879 = torch_c.from_builtin_tensor %27878 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27880 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25533 = arith.constant 1 : index
%dim_25534 = tensor.dim %27880, %c1_25533 : tensor<4x?x4096xf16>
%27881 = flow.tensor.transfer %27880 : tensor<4x?x4096xf16>{%dim_25534} to #hal.device.promise<@__device_6>
%27882 = torch_c.from_builtin_tensor %27881 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27883 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25535 = arith.constant 1 : index
%dim_25536 = tensor.dim %27883, %c1_25535 : tensor<4x?x4096xf16>
%27884 = flow.tensor.transfer %27883 : tensor<4x?x4096xf16>{%dim_25536} to #hal.device.promise<@__device_6>
%27885 = torch_c.from_builtin_tensor %27884 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27886 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25537 = arith.constant 1 : index
%dim_25538 = tensor.dim %27886, %c1_25537 : tensor<4x?x4096xf16>
%27887 = flow.tensor.transfer %27886 : tensor<4x?x4096xf16>{%dim_25538} to #hal.device.promise<@__device_6>
%27888 = torch_c.from_builtin_tensor %27887 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27889 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25539 = arith.constant 1 : index
%dim_25540 = tensor.dim %27889, %c1_25539 : tensor<4x?x4096xf16>
%27890 = flow.tensor.transfer %27889 : tensor<4x?x4096xf16>{%dim_25540} to #hal.device.promise<@__device_6>
%27891 = torch_c.from_builtin_tensor %27890 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27892 = torch_c.to_builtin_tensor %27705 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25541 = arith.constant 1 : index
%dim_25542 = tensor.dim %27892, %c1_25541 : tensor<4x?x4096xf16>
%27893 = flow.tensor.transfer %27892 : tensor<4x?x4096xf16>{%dim_25542} to #hal.device.promise<@__device_6>
%27894 = torch_c.from_builtin_tensor %27893 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25543 = torch.constant.int 1
%27895 = torch.aten.add.Tensor %27876, %27879, %int1_25543 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25544 = torch.constant.int 1
%27896 = torch.aten.add.Tensor %27895, %27882, %int1_25544 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25545 = torch.constant.int 1
%27897 = torch.aten.add.Tensor %27896, %27885, %int1_25545 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25546 = torch.constant.int 1
%27898 = torch.aten.add.Tensor %27897, %27888, %int1_25546 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25547 = torch.constant.int 1
%27899 = torch.aten.add.Tensor %27898, %27891, %int1_25547 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25548 = torch.constant.int 1
%27900 = torch.aten.add.Tensor %27899, %27699, %int1_25548 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25549 = torch.constant.int 1
%27901 = torch.aten.add.Tensor %27900, %27894, %int1_25549 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27902 = torch_c.to_builtin_tensor %27663 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25550 = arith.constant 1 : index
%dim_25551 = tensor.dim %27902, %c1_25550 : tensor<4x?x4096xf16>
%27903 = flow.tensor.transfer %27902 : tensor<4x?x4096xf16>{%dim_25551} to #hal.device.promise<@__device_7>
%27904 = torch_c.from_builtin_tensor %27903 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27905 = torch_c.to_builtin_tensor %27669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25552 = arith.constant 1 : index
%dim_25553 = tensor.dim %27905, %c1_25552 : tensor<4x?x4096xf16>
%27906 = flow.tensor.transfer %27905 : tensor<4x?x4096xf16>{%dim_25553} to #hal.device.promise<@__device_7>
%27907 = torch_c.from_builtin_tensor %27906 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27908 = torch_c.to_builtin_tensor %27675 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25554 = arith.constant 1 : index
%dim_25555 = tensor.dim %27908, %c1_25554 : tensor<4x?x4096xf16>
%27909 = flow.tensor.transfer %27908 : tensor<4x?x4096xf16>{%dim_25555} to #hal.device.promise<@__device_7>
%27910 = torch_c.from_builtin_tensor %27909 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27911 = torch_c.to_builtin_tensor %27681 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25556 = arith.constant 1 : index
%dim_25557 = tensor.dim %27911, %c1_25556 : tensor<4x?x4096xf16>
%27912 = flow.tensor.transfer %27911 : tensor<4x?x4096xf16>{%dim_25557} to #hal.device.promise<@__device_7>
%27913 = torch_c.from_builtin_tensor %27912 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27914 = torch_c.to_builtin_tensor %27687 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25558 = arith.constant 1 : index
%dim_25559 = tensor.dim %27914, %c1_25558 : tensor<4x?x4096xf16>
%27915 = flow.tensor.transfer %27914 : tensor<4x?x4096xf16>{%dim_25559} to #hal.device.promise<@__device_7>
%27916 = torch_c.from_builtin_tensor %27915 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27917 = torch_c.to_builtin_tensor %27693 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25560 = arith.constant 1 : index
%dim_25561 = tensor.dim %27917, %c1_25560 : tensor<4x?x4096xf16>
%27918 = flow.tensor.transfer %27917 : tensor<4x?x4096xf16>{%dim_25561} to #hal.device.promise<@__device_7>
%27919 = torch_c.from_builtin_tensor %27918 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%27920 = torch_c.to_builtin_tensor %27699 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25562 = arith.constant 1 : index
%dim_25563 = tensor.dim %27920, %c1_25562 : tensor<4x?x4096xf16>
%27921 = flow.tensor.transfer %27920 : tensor<4x?x4096xf16>{%dim_25563} to #hal.device.promise<@__device_7>
%27922 = torch_c.from_builtin_tensor %27921 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25564 = torch.constant.int 1
%27923 = torch.aten.add.Tensor %27904, %27907, %int1_25564 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25565 = torch.constant.int 1
%27924 = torch.aten.add.Tensor %27923, %27910, %int1_25565 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25566 = torch.constant.int 1
%27925 = torch.aten.add.Tensor %27924, %27913, %int1_25566 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25567 = torch.constant.int 1
%27926 = torch.aten.add.Tensor %27925, %27916, %int1_25567 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25568 = torch.constant.int 1
%27927 = torch.aten.add.Tensor %27926, %27919, %int1_25568 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25569 = torch.constant.int 1
%27928 = torch.aten.add.Tensor %27927, %27922, %int1_25569 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25570 = torch.constant.int 1
%27929 = torch.aten.add.Tensor %27928, %27705, %int1_25570 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25571 = torch.constant.int 1
%27930 = torch.aten.add.Tensor %26589, %27733, %int1_25571 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25572 = torch.constant.int 1
%27931 = torch.aten.add.Tensor %26590, %27761, %int1_25572 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25573 = torch.constant.int 1
%27932 = torch.aten.add.Tensor %26591, %27789, %int1_25573 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25574 = torch.constant.int 1
%27933 = torch.aten.add.Tensor %26592, %27817, %int1_25574 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25575 = torch.constant.int 1
%27934 = torch.aten.add.Tensor %26593, %27845, %int1_25575 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25576 = torch.constant.int 1
%27935 = torch.aten.add.Tensor %26594, %27873, %int1_25576 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25577 = torch.constant.int 1
%27936 = torch.aten.add.Tensor %26595, %27901, %int1_25577 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25578 = torch.constant.int 1
%27937 = torch.aten.add.Tensor %26596, %27929, %int1_25578 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %27937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_25579 = torch.constant.int 6
%27938 = torch.prims.convert_element_type %27930, %int6_25579 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25580 = torch.constant.int 6
%27939 = torch.prims.convert_element_type %27931, %int6_25580 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25581 = torch.constant.int 6
%27940 = torch.prims.convert_element_type %27932, %int6_25581 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25582 = torch.constant.int 6
%27941 = torch.prims.convert_element_type %27933, %int6_25582 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25583 = torch.constant.int 6
%27942 = torch.prims.convert_element_type %27934, %int6_25583 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25584 = torch.constant.int 6
%27943 = torch.prims.convert_element_type %27935, %int6_25584 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25585 = torch.constant.int 6
%27944 = torch.prims.convert_element_type %27936, %int6_25585 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25586 = torch.constant.int 6
%27945 = torch.prims.convert_element_type %27937, %int6_25586 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25587 = torch.constant.int 2
%27946 = torch.aten.pow.Tensor_Scalar %27938, %int2_25587 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25588 = torch.constant.int 2
%27947 = torch.aten.pow.Tensor_Scalar %27939, %int2_25588 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25589 = torch.constant.int 2
%27948 = torch.aten.pow.Tensor_Scalar %27940, %int2_25589 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25590 = torch.constant.int 2
%27949 = torch.aten.pow.Tensor_Scalar %27941, %int2_25590 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25591 = torch.constant.int 2
%27950 = torch.aten.pow.Tensor_Scalar %27942, %int2_25591 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25592 = torch.constant.int 2
%27951 = torch.aten.pow.Tensor_Scalar %27943, %int2_25592 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25593 = torch.constant.int 2
%27952 = torch.aten.pow.Tensor_Scalar %27944, %int2_25593 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25594 = torch.constant.int 2
%27953 = torch.aten.pow.Tensor_Scalar %27945, %int2_25594 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_25595 = torch.constant.int -1
%27954 = torch.prim.ListConstruct %int-1_25595 : (!torch.int) -> !torch.list<int>
%true_25596 = torch.constant.bool true
%none_25597 = torch.constant.none
%27955 = torch.aten.mean.dim %27946, %27954, %true_25596, %none_25597 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25598 = torch.constant.int -1
%27956 = torch.prim.ListConstruct %int-1_25598 : (!torch.int) -> !torch.list<int>
%true_25599 = torch.constant.bool true
%none_25600 = torch.constant.none
%27957 = torch.aten.mean.dim %27947, %27956, %true_25599, %none_25600 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25601 = torch.constant.int -1
%27958 = torch.prim.ListConstruct %int-1_25601 : (!torch.int) -> !torch.list<int>
%true_25602 = torch.constant.bool true
%none_25603 = torch.constant.none
%27959 = torch.aten.mean.dim %27948, %27958, %true_25602, %none_25603 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25604 = torch.constant.int -1
%27960 = torch.prim.ListConstruct %int-1_25604 : (!torch.int) -> !torch.list<int>
%true_25605 = torch.constant.bool true
%none_25606 = torch.constant.none
%27961 = torch.aten.mean.dim %27949, %27960, %true_25605, %none_25606 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25607 = torch.constant.int -1
%27962 = torch.prim.ListConstruct %int-1_25607 : (!torch.int) -> !torch.list<int>
%true_25608 = torch.constant.bool true
%none_25609 = torch.constant.none
%27963 = torch.aten.mean.dim %27950, %27962, %true_25608, %none_25609 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25610 = torch.constant.int -1
%27964 = torch.prim.ListConstruct %int-1_25610 : (!torch.int) -> !torch.list<int>
%true_25611 = torch.constant.bool true
%none_25612 = torch.constant.none
%27965 = torch.aten.mean.dim %27951, %27964, %true_25611, %none_25612 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25613 = torch.constant.int -1
%27966 = torch.prim.ListConstruct %int-1_25613 : (!torch.int) -> !torch.list<int>
%true_25614 = torch.constant.bool true
%none_25615 = torch.constant.none
%27967 = torch.aten.mean.dim %27952, %27966, %true_25614, %none_25615 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25616 = torch.constant.int -1
%27968 = torch.prim.ListConstruct %int-1_25616 : (!torch.int) -> !torch.list<int>
%true_25617 = torch.constant.bool true
%none_25618 = torch.constant.none
%27969 = torch.aten.mean.dim %27953, %27968, %true_25617, %none_25618 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25619 = torch.constant.float 9.9999997473787516E-6
%int1_25620 = torch.constant.int 1
%27970 = torch.aten.add.Scalar %27955, %float9.999990e-06_25619, %int1_25620 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25621 = torch.constant.float 9.9999997473787516E-6
%int1_25622 = torch.constant.int 1
%27971 = torch.aten.add.Scalar %27957, %float9.999990e-06_25621, %int1_25622 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25623 = torch.constant.float 9.9999997473787516E-6
%int1_25624 = torch.constant.int 1
%27972 = torch.aten.add.Scalar %27959, %float9.999990e-06_25623, %int1_25624 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25625 = torch.constant.float 9.9999997473787516E-6
%int1_25626 = torch.constant.int 1
%27973 = torch.aten.add.Scalar %27961, %float9.999990e-06_25625, %int1_25626 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25627 = torch.constant.float 9.9999997473787516E-6
%int1_25628 = torch.constant.int 1
%27974 = torch.aten.add.Scalar %27963, %float9.999990e-06_25627, %int1_25628 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25629 = torch.constant.float 9.9999997473787516E-6
%int1_25630 = torch.constant.int 1
%27975 = torch.aten.add.Scalar %27965, %float9.999990e-06_25629, %int1_25630 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25631 = torch.constant.float 9.9999997473787516E-6
%int1_25632 = torch.constant.int 1
%27976 = torch.aten.add.Scalar %27967, %float9.999990e-06_25631, %int1_25632 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_25633 = torch.constant.float 9.9999997473787516E-6
%int1_25634 = torch.constant.int 1
%27977 = torch.aten.add.Scalar %27969, %float9.999990e-06_25633, %int1_25634 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27978 = torch.aten.rsqrt %27970 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27979 = torch.aten.rsqrt %27971 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27980 = torch.aten.rsqrt %27972 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27981 = torch.aten.rsqrt %27973 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27982 = torch.aten.rsqrt %27974 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27983 = torch.aten.rsqrt %27975 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27984 = torch.aten.rsqrt %27976 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27985 = torch.aten.rsqrt %27977 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %27985, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%27986 = torch.aten.mul.Tensor %27938, %27978 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27987 = torch.aten.mul.Tensor %27939, %27979 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27988 = torch.aten.mul.Tensor %27940, %27980 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27989 = torch.aten.mul.Tensor %27941, %27981 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27990 = torch.aten.mul.Tensor %27942, %27982 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27991 = torch.aten.mul.Tensor %27943, %27983 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27992 = torch.aten.mul.Tensor %27944, %27984 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27993 = torch.aten.mul.Tensor %27945, %27985 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27994 = torch.aten.mul.Tensor %984, %27986 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27995 = torch.aten.mul.Tensor %985, %27987 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27996 = torch.aten.mul.Tensor %986, %27988 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27997 = torch.aten.mul.Tensor %987, %27989 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27998 = torch.aten.mul.Tensor %988, %27990 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%27999 = torch.aten.mul.Tensor %989, %27991 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %27999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28000 = torch.aten.mul.Tensor %990, %27992 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28001 = torch.aten.mul.Tensor %991, %27993 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_25635 = torch.constant.int 5
%28002 = torch.prims.convert_element_type %27994, %int5_25635 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25636 = torch.constant.int 5
%28003 = torch.prims.convert_element_type %27995, %int5_25636 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25637 = torch.constant.int 5
%28004 = torch.prims.convert_element_type %27996, %int5_25637 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25638 = torch.constant.int 5
%28005 = torch.prims.convert_element_type %27997, %int5_25638 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25639 = torch.constant.int 5
%28006 = torch.prims.convert_element_type %27998, %int5_25639 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25640 = torch.constant.int 5
%28007 = torch.prims.convert_element_type %27999, %int5_25640 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25641 = torch.constant.int 5
%28008 = torch.prims.convert_element_type %28000, %int5_25641 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_25642 = torch.constant.int 5
%28009 = torch.prims.convert_element_type %28001, %int5_25642 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25643 = torch.constant.int 1
%int0_25644 = torch.constant.int 0
%28010 = torch.prim.ListConstruct %int1_25643, %int0_25644 : (!torch.int, !torch.int) -> !torch.list<int>
%28011 = torch.aten.permute %992, %28010 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25645 = torch.constant.int 1
%int0_25646 = torch.constant.int 0
%28012 = torch.prim.ListConstruct %int1_25645, %int0_25646 : (!torch.int, !torch.int) -> !torch.list<int>
%28013 = torch.aten.permute %993, %28012 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25647 = torch.constant.int 1
%int0_25648 = torch.constant.int 0
%28014 = torch.prim.ListConstruct %int1_25647, %int0_25648 : (!torch.int, !torch.int) -> !torch.list<int>
%28015 = torch.aten.permute %994, %28014 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25649 = torch.constant.int 1
%int0_25650 = torch.constant.int 0
%28016 = torch.prim.ListConstruct %int1_25649, %int0_25650 : (!torch.int, !torch.int) -> !torch.list<int>
%28017 = torch.aten.permute %995, %28016 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25651 = torch.constant.int 1
%int0_25652 = torch.constant.int 0
%28018 = torch.prim.ListConstruct %int1_25651, %int0_25652 : (!torch.int, !torch.int) -> !torch.list<int>
%28019 = torch.aten.permute %996, %28018 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25653 = torch.constant.int 1
%int0_25654 = torch.constant.int 0
%28020 = torch.prim.ListConstruct %int1_25653, %int0_25654 : (!torch.int, !torch.int) -> !torch.list<int>
%28021 = torch.aten.permute %997, %28020 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25655 = torch.constant.int 1
%int0_25656 = torch.constant.int 0
%28022 = torch.prim.ListConstruct %int1_25655, %int0_25656 : (!torch.int, !torch.int) -> !torch.list<int>
%28023 = torch.aten.permute %998, %28022 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25657 = torch.constant.int 1
%int0_25658 = torch.constant.int 0
%28024 = torch.prim.ListConstruct %int1_25657, %int0_25658 : (!torch.int, !torch.int) -> !torch.list<int>
%28025 = torch.aten.permute %999, %28024 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_25659 = torch.constant.int 4
%28026 = torch.aten.mul.int %int4_25659, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25660 = torch.constant.int 4096
%28027 = torch.prim.ListConstruct %28026, %int4096_25660 : (!torch.int, !torch.int) -> !torch.list<int>
%28028 = torch.aten.view %28002, %28027 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28028, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28029 = torch.aten.mm %28028, %28011 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28029, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25661 = torch.constant.int 4
%int1792_25662 = torch.constant.int 1792
%28030 = torch.prim.ListConstruct %int4_25661, %2482, %int1792_25662 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28031 = torch.aten.view %28029, %28030 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25663 = torch.constant.int 4
%28032 = torch.aten.mul.int %int4_25663, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25664 = torch.constant.int 4096
%28033 = torch.prim.ListConstruct %28032, %int4096_25664 : (!torch.int, !torch.int) -> !torch.list<int>
%28034 = torch.aten.view %28003, %28033 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28034, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28035 = torch.aten.mm %28034, %28013 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28035, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25665 = torch.constant.int 4
%int1792_25666 = torch.constant.int 1792
%28036 = torch.prim.ListConstruct %int4_25665, %2482, %int1792_25666 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28037 = torch.aten.view %28035, %28036 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25667 = torch.constant.int 4
%28038 = torch.aten.mul.int %int4_25667, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25668 = torch.constant.int 4096
%28039 = torch.prim.ListConstruct %28038, %int4096_25668 : (!torch.int, !torch.int) -> !torch.list<int>
%28040 = torch.aten.view %28004, %28039 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28040, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28041 = torch.aten.mm %28040, %28015 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28041, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25669 = torch.constant.int 4
%int1792_25670 = torch.constant.int 1792
%28042 = torch.prim.ListConstruct %int4_25669, %2482, %int1792_25670 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28043 = torch.aten.view %28041, %28042 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25671 = torch.constant.int 4
%28044 = torch.aten.mul.int %int4_25671, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25672 = torch.constant.int 4096
%28045 = torch.prim.ListConstruct %28044, %int4096_25672 : (!torch.int, !torch.int) -> !torch.list<int>
%28046 = torch.aten.view %28005, %28045 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28046, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28047 = torch.aten.mm %28046, %28017 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28047, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25673 = torch.constant.int 4
%int1792_25674 = torch.constant.int 1792
%28048 = torch.prim.ListConstruct %int4_25673, %2482, %int1792_25674 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28049 = torch.aten.view %28047, %28048 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25675 = torch.constant.int 4
%28050 = torch.aten.mul.int %int4_25675, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25676 = torch.constant.int 4096
%28051 = torch.prim.ListConstruct %28050, %int4096_25676 : (!torch.int, !torch.int) -> !torch.list<int>
%28052 = torch.aten.view %28006, %28051 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28052, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28053 = torch.aten.mm %28052, %28019 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28053, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25677 = torch.constant.int 4
%int1792_25678 = torch.constant.int 1792
%28054 = torch.prim.ListConstruct %int4_25677, %2482, %int1792_25678 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28055 = torch.aten.view %28053, %28054 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25679 = torch.constant.int 4
%28056 = torch.aten.mul.int %int4_25679, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25680 = torch.constant.int 4096
%28057 = torch.prim.ListConstruct %28056, %int4096_25680 : (!torch.int, !torch.int) -> !torch.list<int>
%28058 = torch.aten.view %28007, %28057 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28058, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28059 = torch.aten.mm %28058, %28021 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28059, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25681 = torch.constant.int 4
%int1792_25682 = torch.constant.int 1792
%28060 = torch.prim.ListConstruct %int4_25681, %2482, %int1792_25682 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28061 = torch.aten.view %28059, %28060 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25683 = torch.constant.int 4
%28062 = torch.aten.mul.int %int4_25683, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25684 = torch.constant.int 4096
%28063 = torch.prim.ListConstruct %28062, %int4096_25684 : (!torch.int, !torch.int) -> !torch.list<int>
%28064 = torch.aten.view %28008, %28063 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28064, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28065 = torch.aten.mm %28064, %28023 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28065, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25685 = torch.constant.int 4
%int1792_25686 = torch.constant.int 1792
%28066 = torch.prim.ListConstruct %int4_25685, %2482, %int1792_25686 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28067 = torch.aten.view %28065, %28066 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25687 = torch.constant.int 4
%28068 = torch.aten.mul.int %int4_25687, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25688 = torch.constant.int 4096
%28069 = torch.prim.ListConstruct %28068, %int4096_25688 : (!torch.int, !torch.int) -> !torch.list<int>
%28070 = torch.aten.view %28009, %28069 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28070, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28071 = torch.aten.mm %28070, %28025 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28071, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25689 = torch.constant.int 4
%int1792_25690 = torch.constant.int 1792
%28072 = torch.prim.ListConstruct %int4_25689, %2482, %int1792_25690 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28073 = torch.aten.view %28071, %28072 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28074 = torch.aten.silu %28031 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28075 = torch.aten.silu %28037 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28076 = torch.aten.silu %28043 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28077 = torch.aten.silu %28049 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28078 = torch.aten.silu %28055 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28079 = torch.aten.silu %28061 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28080 = torch.aten.silu %28067 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28081 = torch.aten.silu %28073 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_25691 = torch.constant.int 1
%int0_25692 = torch.constant.int 0
%28082 = torch.prim.ListConstruct %int1_25691, %int0_25692 : (!torch.int, !torch.int) -> !torch.list<int>
%28083 = torch.aten.permute %1000, %28082 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25693 = torch.constant.int 1
%int0_25694 = torch.constant.int 0
%28084 = torch.prim.ListConstruct %int1_25693, %int0_25694 : (!torch.int, !torch.int) -> !torch.list<int>
%28085 = torch.aten.permute %1001, %28084 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25695 = torch.constant.int 1
%int0_25696 = torch.constant.int 0
%28086 = torch.prim.ListConstruct %int1_25695, %int0_25696 : (!torch.int, !torch.int) -> !torch.list<int>
%28087 = torch.aten.permute %1002, %28086 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25697 = torch.constant.int 1
%int0_25698 = torch.constant.int 0
%28088 = torch.prim.ListConstruct %int1_25697, %int0_25698 : (!torch.int, !torch.int) -> !torch.list<int>
%28089 = torch.aten.permute %1003, %28088 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25699 = torch.constant.int 1
%int0_25700 = torch.constant.int 0
%28090 = torch.prim.ListConstruct %int1_25699, %int0_25700 : (!torch.int, !torch.int) -> !torch.list<int>
%28091 = torch.aten.permute %1004, %28090 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25701 = torch.constant.int 1
%int0_25702 = torch.constant.int 0
%28092 = torch.prim.ListConstruct %int1_25701, %int0_25702 : (!torch.int, !torch.int) -> !torch.list<int>
%28093 = torch.aten.permute %1005, %28092 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25703 = torch.constant.int 1
%int0_25704 = torch.constant.int 0
%28094 = torch.prim.ListConstruct %int1_25703, %int0_25704 : (!torch.int, !torch.int) -> !torch.list<int>
%28095 = torch.aten.permute %1006, %28094 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_25705 = torch.constant.int 1
%int0_25706 = torch.constant.int 0
%28096 = torch.prim.ListConstruct %int1_25705, %int0_25706 : (!torch.int, !torch.int) -> !torch.list<int>
%28097 = torch.aten.permute %1007, %28096 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_25707 = torch.constant.int 4
%28098 = torch.aten.mul.int %int4_25707, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25708 = torch.constant.int 4096
%28099 = torch.prim.ListConstruct %28098, %int4096_25708 : (!torch.int, !torch.int) -> !torch.list<int>
%28100 = torch.aten.view %28002, %28099 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28100, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28101 = torch.aten.mm %28100, %28083 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28101, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25709 = torch.constant.int 4
%int1792_25710 = torch.constant.int 1792
%28102 = torch.prim.ListConstruct %int4_25709, %2482, %int1792_25710 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28103 = torch.aten.view %28101, %28102 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25711 = torch.constant.int 4
%28104 = torch.aten.mul.int %int4_25711, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25712 = torch.constant.int 4096
%28105 = torch.prim.ListConstruct %28104, %int4096_25712 : (!torch.int, !torch.int) -> !torch.list<int>
%28106 = torch.aten.view %28003, %28105 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28106, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28107 = torch.aten.mm %28106, %28085 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28107, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25713 = torch.constant.int 4
%int1792_25714 = torch.constant.int 1792
%28108 = torch.prim.ListConstruct %int4_25713, %2482, %int1792_25714 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28109 = torch.aten.view %28107, %28108 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25715 = torch.constant.int 4
%28110 = torch.aten.mul.int %int4_25715, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25716 = torch.constant.int 4096
%28111 = torch.prim.ListConstruct %28110, %int4096_25716 : (!torch.int, !torch.int) -> !torch.list<int>
%28112 = torch.aten.view %28004, %28111 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28112, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28113 = torch.aten.mm %28112, %28087 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28113, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25717 = torch.constant.int 4
%int1792_25718 = torch.constant.int 1792
%28114 = torch.prim.ListConstruct %int4_25717, %2482, %int1792_25718 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28115 = torch.aten.view %28113, %28114 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25719 = torch.constant.int 4
%28116 = torch.aten.mul.int %int4_25719, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25720 = torch.constant.int 4096
%28117 = torch.prim.ListConstruct %28116, %int4096_25720 : (!torch.int, !torch.int) -> !torch.list<int>
%28118 = torch.aten.view %28005, %28117 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28118, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28119 = torch.aten.mm %28118, %28089 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28119, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25721 = torch.constant.int 4
%int1792_25722 = torch.constant.int 1792
%28120 = torch.prim.ListConstruct %int4_25721, %2482, %int1792_25722 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28121 = torch.aten.view %28119, %28120 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25723 = torch.constant.int 4
%28122 = torch.aten.mul.int %int4_25723, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25724 = torch.constant.int 4096
%28123 = torch.prim.ListConstruct %28122, %int4096_25724 : (!torch.int, !torch.int) -> !torch.list<int>
%28124 = torch.aten.view %28006, %28123 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28124, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28125 = torch.aten.mm %28124, %28091 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28125, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25725 = torch.constant.int 4
%int1792_25726 = torch.constant.int 1792
%28126 = torch.prim.ListConstruct %int4_25725, %2482, %int1792_25726 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28127 = torch.aten.view %28125, %28126 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25727 = torch.constant.int 4
%28128 = torch.aten.mul.int %int4_25727, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25728 = torch.constant.int 4096
%28129 = torch.prim.ListConstruct %28128, %int4096_25728 : (!torch.int, !torch.int) -> !torch.list<int>
%28130 = torch.aten.view %28007, %28129 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28130, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28131 = torch.aten.mm %28130, %28093 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28131, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25729 = torch.constant.int 4
%int1792_25730 = torch.constant.int 1792
%28132 = torch.prim.ListConstruct %int4_25729, %2482, %int1792_25730 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28133 = torch.aten.view %28131, %28132 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25731 = torch.constant.int 4
%28134 = torch.aten.mul.int %int4_25731, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25732 = torch.constant.int 4096
%28135 = torch.prim.ListConstruct %28134, %int4096_25732 : (!torch.int, !torch.int) -> !torch.list<int>
%28136 = torch.aten.view %28008, %28135 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28136, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28137 = torch.aten.mm %28136, %28095 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28137, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25733 = torch.constant.int 4
%int1792_25734 = torch.constant.int 1792
%28138 = torch.prim.ListConstruct %int4_25733, %2482, %int1792_25734 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28139 = torch.aten.view %28137, %28138 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_25735 = torch.constant.int 4
%28140 = torch.aten.mul.int %int4_25735, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_25736 = torch.constant.int 4096
%28141 = torch.prim.ListConstruct %28140, %int4096_25736 : (!torch.int, !torch.int) -> !torch.list<int>
%28142 = torch.aten.view %28009, %28141 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28142, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28143 = torch.aten.mm %28142, %28097 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28143, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_25737 = torch.constant.int 4
%int1792_25738 = torch.constant.int 1792
%28144 = torch.prim.ListConstruct %int4_25737, %2482, %int1792_25738 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28145 = torch.aten.view %28143, %28144 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28146 = torch.aten.mul.Tensor %28074, %28103 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28147 = torch.aten.mul.Tensor %28075, %28109 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28148 = torch.aten.mul.Tensor %28076, %28115 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28149 = torch.aten.mul.Tensor %28077, %28121 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28150 = torch.aten.mul.Tensor %28078, %28127 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28151 = torch.aten.mul.Tensor %28079, %28133 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28152 = torch.aten.mul.Tensor %28080, %28139 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%28153 = torch.aten.mul.Tensor %28081, %28145 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %28153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_25739 = torch.constant.int 1
%int0_25740 = torch.constant.int 0
%28154 = torch.prim.ListConstruct %int1_25739, %int0_25740 : (!torch.int, !torch.int) -> !torch.list<int>
%28155 = torch.aten.permute %1008, %28154 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25741 = torch.constant.int 1
%int0_25742 = torch.constant.int 0
%28156 = torch.prim.ListConstruct %int1_25741, %int0_25742 : (!torch.int, !torch.int) -> !torch.list<int>
%28157 = torch.aten.permute %1009, %28156 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25743 = torch.constant.int 1
%int0_25744 = torch.constant.int 0
%28158 = torch.prim.ListConstruct %int1_25743, %int0_25744 : (!torch.int, !torch.int) -> !torch.list<int>
%28159 = torch.aten.permute %1010, %28158 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25745 = torch.constant.int 1
%int0_25746 = torch.constant.int 0
%28160 = torch.prim.ListConstruct %int1_25745, %int0_25746 : (!torch.int, !torch.int) -> !torch.list<int>
%28161 = torch.aten.permute %1011, %28160 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25747 = torch.constant.int 1
%int0_25748 = torch.constant.int 0
%28162 = torch.prim.ListConstruct %int1_25747, %int0_25748 : (!torch.int, !torch.int) -> !torch.list<int>
%28163 = torch.aten.permute %1012, %28162 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25749 = torch.constant.int 1
%int0_25750 = torch.constant.int 0
%28164 = torch.prim.ListConstruct %int1_25749, %int0_25750 : (!torch.int, !torch.int) -> !torch.list<int>
%28165 = torch.aten.permute %1013, %28164 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25751 = torch.constant.int 1
%int0_25752 = torch.constant.int 0
%28166 = torch.prim.ListConstruct %int1_25751, %int0_25752 : (!torch.int, !torch.int) -> !torch.list<int>
%28167 = torch.aten.permute %1014, %28166 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25753 = torch.constant.int 1
%int0_25754 = torch.constant.int 0
%28168 = torch.prim.ListConstruct %int1_25753, %int0_25754 : (!torch.int, !torch.int) -> !torch.list<int>
%28169 = torch.aten.permute %1015, %28168 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_25755 = torch.constant.int 1
%28170 = torch.aten.size.int %28031, %int1_25755 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25756 = torch.constant.int 4
%28171 = torch.aten.mul.int %int4_25756, %28170 : !torch.int, !torch.int -> !torch.int
%int1792_25757 = torch.constant.int 1792
%28172 = torch.prim.ListConstruct %28171, %int1792_25757 : (!torch.int, !torch.int) -> !torch.list<int>
%28173 = torch.aten.view %28146, %28172 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28173, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28174 = torch.aten.mm %28173, %28155 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28174, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25758 = torch.constant.int 4
%int4096_25759 = torch.constant.int 4096
%28175 = torch.prim.ListConstruct %int4_25758, %28170, %int4096_25759 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28176 = torch.aten.view %28174, %28175 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25760 = torch.constant.int 1
%28177 = torch.aten.size.int %28037, %int1_25760 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25761 = torch.constant.int 4
%28178 = torch.aten.mul.int %int4_25761, %28177 : !torch.int, !torch.int -> !torch.int
%int1792_25762 = torch.constant.int 1792
%28179 = torch.prim.ListConstruct %28178, %int1792_25762 : (!torch.int, !torch.int) -> !torch.list<int>
%28180 = torch.aten.view %28147, %28179 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28180, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28181 = torch.aten.mm %28180, %28157 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28181, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25763 = torch.constant.int 4
%int4096_25764 = torch.constant.int 4096
%28182 = torch.prim.ListConstruct %int4_25763, %28177, %int4096_25764 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28183 = torch.aten.view %28181, %28182 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25765 = torch.constant.int 1
%28184 = torch.aten.size.int %28043, %int1_25765 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25766 = torch.constant.int 4
%28185 = torch.aten.mul.int %int4_25766, %28184 : !torch.int, !torch.int -> !torch.int
%int1792_25767 = torch.constant.int 1792
%28186 = torch.prim.ListConstruct %28185, %int1792_25767 : (!torch.int, !torch.int) -> !torch.list<int>
%28187 = torch.aten.view %28148, %28186 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28187, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28188 = torch.aten.mm %28187, %28159 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28188, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25768 = torch.constant.int 4
%int4096_25769 = torch.constant.int 4096
%28189 = torch.prim.ListConstruct %int4_25768, %28184, %int4096_25769 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28190 = torch.aten.view %28188, %28189 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25770 = torch.constant.int 1
%28191 = torch.aten.size.int %28049, %int1_25770 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25771 = torch.constant.int 4
%28192 = torch.aten.mul.int %int4_25771, %28191 : !torch.int, !torch.int -> !torch.int
%int1792_25772 = torch.constant.int 1792
%28193 = torch.prim.ListConstruct %28192, %int1792_25772 : (!torch.int, !torch.int) -> !torch.list<int>
%28194 = torch.aten.view %28149, %28193 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28194, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28195 = torch.aten.mm %28194, %28161 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28195, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25773 = torch.constant.int 4
%int4096_25774 = torch.constant.int 4096
%28196 = torch.prim.ListConstruct %int4_25773, %28191, %int4096_25774 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28197 = torch.aten.view %28195, %28196 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25775 = torch.constant.int 1
%28198 = torch.aten.size.int %28055, %int1_25775 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25776 = torch.constant.int 4
%28199 = torch.aten.mul.int %int4_25776, %28198 : !torch.int, !torch.int -> !torch.int
%int1792_25777 = torch.constant.int 1792
%28200 = torch.prim.ListConstruct %28199, %int1792_25777 : (!torch.int, !torch.int) -> !torch.list<int>
%28201 = torch.aten.view %28150, %28200 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28201, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28202 = torch.aten.mm %28201, %28163 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28202, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25778 = torch.constant.int 4
%int4096_25779 = torch.constant.int 4096
%28203 = torch.prim.ListConstruct %int4_25778, %28198, %int4096_25779 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28204 = torch.aten.view %28202, %28203 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25780 = torch.constant.int 1
%28205 = torch.aten.size.int %28061, %int1_25780 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25781 = torch.constant.int 4
%28206 = torch.aten.mul.int %int4_25781, %28205 : !torch.int, !torch.int -> !torch.int
%int1792_25782 = torch.constant.int 1792
%28207 = torch.prim.ListConstruct %28206, %int1792_25782 : (!torch.int, !torch.int) -> !torch.list<int>
%28208 = torch.aten.view %28151, %28207 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28208, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28209 = torch.aten.mm %28208, %28165 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28209, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25783 = torch.constant.int 4
%int4096_25784 = torch.constant.int 4096
%28210 = torch.prim.ListConstruct %int4_25783, %28205, %int4096_25784 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28211 = torch.aten.view %28209, %28210 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25785 = torch.constant.int 1
%28212 = torch.aten.size.int %28067, %int1_25785 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25786 = torch.constant.int 4
%28213 = torch.aten.mul.int %int4_25786, %28212 : !torch.int, !torch.int -> !torch.int
%int1792_25787 = torch.constant.int 1792
%28214 = torch.prim.ListConstruct %28213, %int1792_25787 : (!torch.int, !torch.int) -> !torch.list<int>
%28215 = torch.aten.view %28152, %28214 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28215, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28216 = torch.aten.mm %28215, %28167 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28216, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25788 = torch.constant.int 4
%int4096_25789 = torch.constant.int 4096
%28217 = torch.prim.ListConstruct %int4_25788, %28212, %int4096_25789 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28218 = torch.aten.view %28216, %28217 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25790 = torch.constant.int 1
%28219 = torch.aten.size.int %28073, %int1_25790 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_25791 = torch.constant.int 4
%28220 = torch.aten.mul.int %int4_25791, %28219 : !torch.int, !torch.int -> !torch.int
%int1792_25792 = torch.constant.int 1792
%28221 = torch.prim.ListConstruct %28220, %int1792_25792 : (!torch.int, !torch.int) -> !torch.list<int>
%28222 = torch.aten.view %28153, %28221 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %28222, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%28223 = torch.aten.mm %28222, %28169 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28223, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_25793 = torch.constant.int 4
%int4096_25794 = torch.constant.int 4096
%28224 = torch.prim.ListConstruct %int4_25793, %28219, %int4096_25794 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28225 = torch.aten.view %28223, %28224 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28226 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25795 = arith.constant 1 : index
%dim_25796 = tensor.dim %28226, %c1_25795 : tensor<4x?x4096xf16>
%28227 = flow.tensor.transfer %28226 : tensor<4x?x4096xf16>{%dim_25796} to #hal.device.promise<@__device_0>
%28228 = torch_c.from_builtin_tensor %28227 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28229 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25797 = arith.constant 1 : index
%dim_25798 = tensor.dim %28229, %c1_25797 : tensor<4x?x4096xf16>
%28230 = flow.tensor.transfer %28229 : tensor<4x?x4096xf16>{%dim_25798} to #hal.device.promise<@__device_0>
%28231 = torch_c.from_builtin_tensor %28230 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28232 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25799 = arith.constant 1 : index
%dim_25800 = tensor.dim %28232, %c1_25799 : tensor<4x?x4096xf16>
%28233 = flow.tensor.transfer %28232 : tensor<4x?x4096xf16>{%dim_25800} to #hal.device.promise<@__device_0>
%28234 = torch_c.from_builtin_tensor %28233 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28235 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25801 = arith.constant 1 : index
%dim_25802 = tensor.dim %28235, %c1_25801 : tensor<4x?x4096xf16>
%28236 = flow.tensor.transfer %28235 : tensor<4x?x4096xf16>{%dim_25802} to #hal.device.promise<@__device_0>
%28237 = torch_c.from_builtin_tensor %28236 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28238 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25803 = arith.constant 1 : index
%dim_25804 = tensor.dim %28238, %c1_25803 : tensor<4x?x4096xf16>
%28239 = flow.tensor.transfer %28238 : tensor<4x?x4096xf16>{%dim_25804} to #hal.device.promise<@__device_0>
%28240 = torch_c.from_builtin_tensor %28239 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28241 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25805 = arith.constant 1 : index
%dim_25806 = tensor.dim %28241, %c1_25805 : tensor<4x?x4096xf16>
%28242 = flow.tensor.transfer %28241 : tensor<4x?x4096xf16>{%dim_25806} to #hal.device.promise<@__device_0>
%28243 = torch_c.from_builtin_tensor %28242 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28244 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25807 = arith.constant 1 : index
%dim_25808 = tensor.dim %28244, %c1_25807 : tensor<4x?x4096xf16>
%28245 = flow.tensor.transfer %28244 : tensor<4x?x4096xf16>{%dim_25808} to #hal.device.promise<@__device_0>
%28246 = torch_c.from_builtin_tensor %28245 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25809 = torch.constant.int 1
%28247 = torch.aten.add.Tensor %28176, %28228, %int1_25809 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25810 = torch.constant.int 1
%28248 = torch.aten.add.Tensor %28247, %28231, %int1_25810 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25811 = torch.constant.int 1
%28249 = torch.aten.add.Tensor %28248, %28234, %int1_25811 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25812 = torch.constant.int 1
%28250 = torch.aten.add.Tensor %28249, %28237, %int1_25812 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25813 = torch.constant.int 1
%28251 = torch.aten.add.Tensor %28250, %28240, %int1_25813 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25814 = torch.constant.int 1
%28252 = torch.aten.add.Tensor %28251, %28243, %int1_25814 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25815 = torch.constant.int 1
%28253 = torch.aten.add.Tensor %28252, %28246, %int1_25815 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28254 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25816 = arith.constant 1 : index
%dim_25817 = tensor.dim %28254, %c1_25816 : tensor<4x?x4096xf16>
%28255 = flow.tensor.transfer %28254 : tensor<4x?x4096xf16>{%dim_25817} to #hal.device.promise<@__device_1>
%28256 = torch_c.from_builtin_tensor %28255 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28257 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25818 = arith.constant 1 : index
%dim_25819 = tensor.dim %28257, %c1_25818 : tensor<4x?x4096xf16>
%28258 = flow.tensor.transfer %28257 : tensor<4x?x4096xf16>{%dim_25819} to #hal.device.promise<@__device_1>
%28259 = torch_c.from_builtin_tensor %28258 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28260 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25820 = arith.constant 1 : index
%dim_25821 = tensor.dim %28260, %c1_25820 : tensor<4x?x4096xf16>
%28261 = flow.tensor.transfer %28260 : tensor<4x?x4096xf16>{%dim_25821} to #hal.device.promise<@__device_1>
%28262 = torch_c.from_builtin_tensor %28261 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28263 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25822 = arith.constant 1 : index
%dim_25823 = tensor.dim %28263, %c1_25822 : tensor<4x?x4096xf16>
%28264 = flow.tensor.transfer %28263 : tensor<4x?x4096xf16>{%dim_25823} to #hal.device.promise<@__device_1>
%28265 = torch_c.from_builtin_tensor %28264 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28266 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25824 = arith.constant 1 : index
%dim_25825 = tensor.dim %28266, %c1_25824 : tensor<4x?x4096xf16>
%28267 = flow.tensor.transfer %28266 : tensor<4x?x4096xf16>{%dim_25825} to #hal.device.promise<@__device_1>
%28268 = torch_c.from_builtin_tensor %28267 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28269 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25826 = arith.constant 1 : index
%dim_25827 = tensor.dim %28269, %c1_25826 : tensor<4x?x4096xf16>
%28270 = flow.tensor.transfer %28269 : tensor<4x?x4096xf16>{%dim_25827} to #hal.device.promise<@__device_1>
%28271 = torch_c.from_builtin_tensor %28270 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28272 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25828 = arith.constant 1 : index
%dim_25829 = tensor.dim %28272, %c1_25828 : tensor<4x?x4096xf16>
%28273 = flow.tensor.transfer %28272 : tensor<4x?x4096xf16>{%dim_25829} to #hal.device.promise<@__device_1>
%28274 = torch_c.from_builtin_tensor %28273 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25830 = torch.constant.int 1
%28275 = torch.aten.add.Tensor %28256, %28183, %int1_25830 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25831 = torch.constant.int 1
%28276 = torch.aten.add.Tensor %28275, %28259, %int1_25831 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25832 = torch.constant.int 1
%28277 = torch.aten.add.Tensor %28276, %28262, %int1_25832 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25833 = torch.constant.int 1
%28278 = torch.aten.add.Tensor %28277, %28265, %int1_25833 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25834 = torch.constant.int 1
%28279 = torch.aten.add.Tensor %28278, %28268, %int1_25834 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25835 = torch.constant.int 1
%28280 = torch.aten.add.Tensor %28279, %28271, %int1_25835 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25836 = torch.constant.int 1
%28281 = torch.aten.add.Tensor %28280, %28274, %int1_25836 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28282 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25837 = arith.constant 1 : index
%dim_25838 = tensor.dim %28282, %c1_25837 : tensor<4x?x4096xf16>
%28283 = flow.tensor.transfer %28282 : tensor<4x?x4096xf16>{%dim_25838} to #hal.device.promise<@__device_2>
%28284 = torch_c.from_builtin_tensor %28283 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28285 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25839 = arith.constant 1 : index
%dim_25840 = tensor.dim %28285, %c1_25839 : tensor<4x?x4096xf16>
%28286 = flow.tensor.transfer %28285 : tensor<4x?x4096xf16>{%dim_25840} to #hal.device.promise<@__device_2>
%28287 = torch_c.from_builtin_tensor %28286 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28288 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25841 = arith.constant 1 : index
%dim_25842 = tensor.dim %28288, %c1_25841 : tensor<4x?x4096xf16>
%28289 = flow.tensor.transfer %28288 : tensor<4x?x4096xf16>{%dim_25842} to #hal.device.promise<@__device_2>
%28290 = torch_c.from_builtin_tensor %28289 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28291 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25843 = arith.constant 1 : index
%dim_25844 = tensor.dim %28291, %c1_25843 : tensor<4x?x4096xf16>
%28292 = flow.tensor.transfer %28291 : tensor<4x?x4096xf16>{%dim_25844} to #hal.device.promise<@__device_2>
%28293 = torch_c.from_builtin_tensor %28292 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28294 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25845 = arith.constant 1 : index
%dim_25846 = tensor.dim %28294, %c1_25845 : tensor<4x?x4096xf16>
%28295 = flow.tensor.transfer %28294 : tensor<4x?x4096xf16>{%dim_25846} to #hal.device.promise<@__device_2>
%28296 = torch_c.from_builtin_tensor %28295 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28297 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25847 = arith.constant 1 : index
%dim_25848 = tensor.dim %28297, %c1_25847 : tensor<4x?x4096xf16>
%28298 = flow.tensor.transfer %28297 : tensor<4x?x4096xf16>{%dim_25848} to #hal.device.promise<@__device_2>
%28299 = torch_c.from_builtin_tensor %28298 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28300 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25849 = arith.constant 1 : index
%dim_25850 = tensor.dim %28300, %c1_25849 : tensor<4x?x4096xf16>
%28301 = flow.tensor.transfer %28300 : tensor<4x?x4096xf16>{%dim_25850} to #hal.device.promise<@__device_2>
%28302 = torch_c.from_builtin_tensor %28301 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25851 = torch.constant.int 1
%28303 = torch.aten.add.Tensor %28284, %28287, %int1_25851 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25852 = torch.constant.int 1
%28304 = torch.aten.add.Tensor %28303, %28190, %int1_25852 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25853 = torch.constant.int 1
%28305 = torch.aten.add.Tensor %28304, %28290, %int1_25853 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25854 = torch.constant.int 1
%28306 = torch.aten.add.Tensor %28305, %28293, %int1_25854 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25855 = torch.constant.int 1
%28307 = torch.aten.add.Tensor %28306, %28296, %int1_25855 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25856 = torch.constant.int 1
%28308 = torch.aten.add.Tensor %28307, %28299, %int1_25856 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25857 = torch.constant.int 1
%28309 = torch.aten.add.Tensor %28308, %28302, %int1_25857 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28310 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25858 = arith.constant 1 : index
%dim_25859 = tensor.dim %28310, %c1_25858 : tensor<4x?x4096xf16>
%28311 = flow.tensor.transfer %28310 : tensor<4x?x4096xf16>{%dim_25859} to #hal.device.promise<@__device_3>
%28312 = torch_c.from_builtin_tensor %28311 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28313 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25860 = arith.constant 1 : index
%dim_25861 = tensor.dim %28313, %c1_25860 : tensor<4x?x4096xf16>
%28314 = flow.tensor.transfer %28313 : tensor<4x?x4096xf16>{%dim_25861} to #hal.device.promise<@__device_3>
%28315 = torch_c.from_builtin_tensor %28314 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28316 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25862 = arith.constant 1 : index
%dim_25863 = tensor.dim %28316, %c1_25862 : tensor<4x?x4096xf16>
%28317 = flow.tensor.transfer %28316 : tensor<4x?x4096xf16>{%dim_25863} to #hal.device.promise<@__device_3>
%28318 = torch_c.from_builtin_tensor %28317 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28319 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25864 = arith.constant 1 : index
%dim_25865 = tensor.dim %28319, %c1_25864 : tensor<4x?x4096xf16>
%28320 = flow.tensor.transfer %28319 : tensor<4x?x4096xf16>{%dim_25865} to #hal.device.promise<@__device_3>
%28321 = torch_c.from_builtin_tensor %28320 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28322 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25866 = arith.constant 1 : index
%dim_25867 = tensor.dim %28322, %c1_25866 : tensor<4x?x4096xf16>
%28323 = flow.tensor.transfer %28322 : tensor<4x?x4096xf16>{%dim_25867} to #hal.device.promise<@__device_3>
%28324 = torch_c.from_builtin_tensor %28323 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28325 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25868 = arith.constant 1 : index
%dim_25869 = tensor.dim %28325, %c1_25868 : tensor<4x?x4096xf16>
%28326 = flow.tensor.transfer %28325 : tensor<4x?x4096xf16>{%dim_25869} to #hal.device.promise<@__device_3>
%28327 = torch_c.from_builtin_tensor %28326 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28328 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25870 = arith.constant 1 : index
%dim_25871 = tensor.dim %28328, %c1_25870 : tensor<4x?x4096xf16>
%28329 = flow.tensor.transfer %28328 : tensor<4x?x4096xf16>{%dim_25871} to #hal.device.promise<@__device_3>
%28330 = torch_c.from_builtin_tensor %28329 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25872 = torch.constant.int 1
%28331 = torch.aten.add.Tensor %28312, %28315, %int1_25872 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25873 = torch.constant.int 1
%28332 = torch.aten.add.Tensor %28331, %28318, %int1_25873 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25874 = torch.constant.int 1
%28333 = torch.aten.add.Tensor %28332, %28197, %int1_25874 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25875 = torch.constant.int 1
%28334 = torch.aten.add.Tensor %28333, %28321, %int1_25875 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25876 = torch.constant.int 1
%28335 = torch.aten.add.Tensor %28334, %28324, %int1_25876 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25877 = torch.constant.int 1
%28336 = torch.aten.add.Tensor %28335, %28327, %int1_25877 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25878 = torch.constant.int 1
%28337 = torch.aten.add.Tensor %28336, %28330, %int1_25878 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28338 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25879 = arith.constant 1 : index
%dim_25880 = tensor.dim %28338, %c1_25879 : tensor<4x?x4096xf16>
%28339 = flow.tensor.transfer %28338 : tensor<4x?x4096xf16>{%dim_25880} to #hal.device.promise<@__device_4>
%28340 = torch_c.from_builtin_tensor %28339 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28341 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25881 = arith.constant 1 : index
%dim_25882 = tensor.dim %28341, %c1_25881 : tensor<4x?x4096xf16>
%28342 = flow.tensor.transfer %28341 : tensor<4x?x4096xf16>{%dim_25882} to #hal.device.promise<@__device_4>
%28343 = torch_c.from_builtin_tensor %28342 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28344 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25883 = arith.constant 1 : index
%dim_25884 = tensor.dim %28344, %c1_25883 : tensor<4x?x4096xf16>
%28345 = flow.tensor.transfer %28344 : tensor<4x?x4096xf16>{%dim_25884} to #hal.device.promise<@__device_4>
%28346 = torch_c.from_builtin_tensor %28345 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28347 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25885 = arith.constant 1 : index
%dim_25886 = tensor.dim %28347, %c1_25885 : tensor<4x?x4096xf16>
%28348 = flow.tensor.transfer %28347 : tensor<4x?x4096xf16>{%dim_25886} to #hal.device.promise<@__device_4>
%28349 = torch_c.from_builtin_tensor %28348 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28350 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25887 = arith.constant 1 : index
%dim_25888 = tensor.dim %28350, %c1_25887 : tensor<4x?x4096xf16>
%28351 = flow.tensor.transfer %28350 : tensor<4x?x4096xf16>{%dim_25888} to #hal.device.promise<@__device_4>
%28352 = torch_c.from_builtin_tensor %28351 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28353 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25889 = arith.constant 1 : index
%dim_25890 = tensor.dim %28353, %c1_25889 : tensor<4x?x4096xf16>
%28354 = flow.tensor.transfer %28353 : tensor<4x?x4096xf16>{%dim_25890} to #hal.device.promise<@__device_4>
%28355 = torch_c.from_builtin_tensor %28354 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28356 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25891 = arith.constant 1 : index
%dim_25892 = tensor.dim %28356, %c1_25891 : tensor<4x?x4096xf16>
%28357 = flow.tensor.transfer %28356 : tensor<4x?x4096xf16>{%dim_25892} to #hal.device.promise<@__device_4>
%28358 = torch_c.from_builtin_tensor %28357 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25893 = torch.constant.int 1
%28359 = torch.aten.add.Tensor %28340, %28343, %int1_25893 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25894 = torch.constant.int 1
%28360 = torch.aten.add.Tensor %28359, %28346, %int1_25894 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25895 = torch.constant.int 1
%28361 = torch.aten.add.Tensor %28360, %28349, %int1_25895 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25896 = torch.constant.int 1
%28362 = torch.aten.add.Tensor %28361, %28204, %int1_25896 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25897 = torch.constant.int 1
%28363 = torch.aten.add.Tensor %28362, %28352, %int1_25897 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25898 = torch.constant.int 1
%28364 = torch.aten.add.Tensor %28363, %28355, %int1_25898 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25899 = torch.constant.int 1
%28365 = torch.aten.add.Tensor %28364, %28358, %int1_25899 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28366 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25900 = arith.constant 1 : index
%dim_25901 = tensor.dim %28366, %c1_25900 : tensor<4x?x4096xf16>
%28367 = flow.tensor.transfer %28366 : tensor<4x?x4096xf16>{%dim_25901} to #hal.device.promise<@__device_5>
%28368 = torch_c.from_builtin_tensor %28367 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28369 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25902 = arith.constant 1 : index
%dim_25903 = tensor.dim %28369, %c1_25902 : tensor<4x?x4096xf16>
%28370 = flow.tensor.transfer %28369 : tensor<4x?x4096xf16>{%dim_25903} to #hal.device.promise<@__device_5>
%28371 = torch_c.from_builtin_tensor %28370 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28372 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25904 = arith.constant 1 : index
%dim_25905 = tensor.dim %28372, %c1_25904 : tensor<4x?x4096xf16>
%28373 = flow.tensor.transfer %28372 : tensor<4x?x4096xf16>{%dim_25905} to #hal.device.promise<@__device_5>
%28374 = torch_c.from_builtin_tensor %28373 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28375 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25906 = arith.constant 1 : index
%dim_25907 = tensor.dim %28375, %c1_25906 : tensor<4x?x4096xf16>
%28376 = flow.tensor.transfer %28375 : tensor<4x?x4096xf16>{%dim_25907} to #hal.device.promise<@__device_5>
%28377 = torch_c.from_builtin_tensor %28376 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28378 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25908 = arith.constant 1 : index
%dim_25909 = tensor.dim %28378, %c1_25908 : tensor<4x?x4096xf16>
%28379 = flow.tensor.transfer %28378 : tensor<4x?x4096xf16>{%dim_25909} to #hal.device.promise<@__device_5>
%28380 = torch_c.from_builtin_tensor %28379 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28381 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25910 = arith.constant 1 : index
%dim_25911 = tensor.dim %28381, %c1_25910 : tensor<4x?x4096xf16>
%28382 = flow.tensor.transfer %28381 : tensor<4x?x4096xf16>{%dim_25911} to #hal.device.promise<@__device_5>
%28383 = torch_c.from_builtin_tensor %28382 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28384 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25912 = arith.constant 1 : index
%dim_25913 = tensor.dim %28384, %c1_25912 : tensor<4x?x4096xf16>
%28385 = flow.tensor.transfer %28384 : tensor<4x?x4096xf16>{%dim_25913} to #hal.device.promise<@__device_5>
%28386 = torch_c.from_builtin_tensor %28385 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25914 = torch.constant.int 1
%28387 = torch.aten.add.Tensor %28368, %28371, %int1_25914 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25915 = torch.constant.int 1
%28388 = torch.aten.add.Tensor %28387, %28374, %int1_25915 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25916 = torch.constant.int 1
%28389 = torch.aten.add.Tensor %28388, %28377, %int1_25916 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25917 = torch.constant.int 1
%28390 = torch.aten.add.Tensor %28389, %28380, %int1_25917 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25918 = torch.constant.int 1
%28391 = torch.aten.add.Tensor %28390, %28211, %int1_25918 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25919 = torch.constant.int 1
%28392 = torch.aten.add.Tensor %28391, %28383, %int1_25919 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25920 = torch.constant.int 1
%28393 = torch.aten.add.Tensor %28392, %28386, %int1_25920 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28394 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25921 = arith.constant 1 : index
%dim_25922 = tensor.dim %28394, %c1_25921 : tensor<4x?x4096xf16>
%28395 = flow.tensor.transfer %28394 : tensor<4x?x4096xf16>{%dim_25922} to #hal.device.promise<@__device_6>
%28396 = torch_c.from_builtin_tensor %28395 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28397 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25923 = arith.constant 1 : index
%dim_25924 = tensor.dim %28397, %c1_25923 : tensor<4x?x4096xf16>
%28398 = flow.tensor.transfer %28397 : tensor<4x?x4096xf16>{%dim_25924} to #hal.device.promise<@__device_6>
%28399 = torch_c.from_builtin_tensor %28398 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28400 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25925 = arith.constant 1 : index
%dim_25926 = tensor.dim %28400, %c1_25925 : tensor<4x?x4096xf16>
%28401 = flow.tensor.transfer %28400 : tensor<4x?x4096xf16>{%dim_25926} to #hal.device.promise<@__device_6>
%28402 = torch_c.from_builtin_tensor %28401 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28402, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28403 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25927 = arith.constant 1 : index
%dim_25928 = tensor.dim %28403, %c1_25927 : tensor<4x?x4096xf16>
%28404 = flow.tensor.transfer %28403 : tensor<4x?x4096xf16>{%dim_25928} to #hal.device.promise<@__device_6>
%28405 = torch_c.from_builtin_tensor %28404 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28406 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25929 = arith.constant 1 : index
%dim_25930 = tensor.dim %28406, %c1_25929 : tensor<4x?x4096xf16>
%28407 = flow.tensor.transfer %28406 : tensor<4x?x4096xf16>{%dim_25930} to #hal.device.promise<@__device_6>
%28408 = torch_c.from_builtin_tensor %28407 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28409 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25931 = arith.constant 1 : index
%dim_25932 = tensor.dim %28409, %c1_25931 : tensor<4x?x4096xf16>
%28410 = flow.tensor.transfer %28409 : tensor<4x?x4096xf16>{%dim_25932} to #hal.device.promise<@__device_6>
%28411 = torch_c.from_builtin_tensor %28410 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28412 = torch_c.to_builtin_tensor %28225 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25933 = arith.constant 1 : index
%dim_25934 = tensor.dim %28412, %c1_25933 : tensor<4x?x4096xf16>
%28413 = flow.tensor.transfer %28412 : tensor<4x?x4096xf16>{%dim_25934} to #hal.device.promise<@__device_6>
%28414 = torch_c.from_builtin_tensor %28413 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25935 = torch.constant.int 1
%28415 = torch.aten.add.Tensor %28396, %28399, %int1_25935 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25936 = torch.constant.int 1
%28416 = torch.aten.add.Tensor %28415, %28402, %int1_25936 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25937 = torch.constant.int 1
%28417 = torch.aten.add.Tensor %28416, %28405, %int1_25937 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25938 = torch.constant.int 1
%28418 = torch.aten.add.Tensor %28417, %28408, %int1_25938 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25939 = torch.constant.int 1
%28419 = torch.aten.add.Tensor %28418, %28411, %int1_25939 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25940 = torch.constant.int 1
%28420 = torch.aten.add.Tensor %28419, %28218, %int1_25940 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25941 = torch.constant.int 1
%28421 = torch.aten.add.Tensor %28420, %28414, %int1_25941 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28422 = torch_c.to_builtin_tensor %28176 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25942 = arith.constant 1 : index
%dim_25943 = tensor.dim %28422, %c1_25942 : tensor<4x?x4096xf16>
%28423 = flow.tensor.transfer %28422 : tensor<4x?x4096xf16>{%dim_25943} to #hal.device.promise<@__device_7>
%28424 = torch_c.from_builtin_tensor %28423 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28425 = torch_c.to_builtin_tensor %28183 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25944 = arith.constant 1 : index
%dim_25945 = tensor.dim %28425, %c1_25944 : tensor<4x?x4096xf16>
%28426 = flow.tensor.transfer %28425 : tensor<4x?x4096xf16>{%dim_25945} to #hal.device.promise<@__device_7>
%28427 = torch_c.from_builtin_tensor %28426 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28428 = torch_c.to_builtin_tensor %28190 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25946 = arith.constant 1 : index
%dim_25947 = tensor.dim %28428, %c1_25946 : tensor<4x?x4096xf16>
%28429 = flow.tensor.transfer %28428 : tensor<4x?x4096xf16>{%dim_25947} to #hal.device.promise<@__device_7>
%28430 = torch_c.from_builtin_tensor %28429 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28431 = torch_c.to_builtin_tensor %28197 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25948 = arith.constant 1 : index
%dim_25949 = tensor.dim %28431, %c1_25948 : tensor<4x?x4096xf16>
%28432 = flow.tensor.transfer %28431 : tensor<4x?x4096xf16>{%dim_25949} to #hal.device.promise<@__device_7>
%28433 = torch_c.from_builtin_tensor %28432 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28434 = torch_c.to_builtin_tensor %28204 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25950 = arith.constant 1 : index
%dim_25951 = tensor.dim %28434, %c1_25950 : tensor<4x?x4096xf16>
%28435 = flow.tensor.transfer %28434 : tensor<4x?x4096xf16>{%dim_25951} to #hal.device.promise<@__device_7>
%28436 = torch_c.from_builtin_tensor %28435 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28437 = torch_c.to_builtin_tensor %28211 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25952 = arith.constant 1 : index
%dim_25953 = tensor.dim %28437, %c1_25952 : tensor<4x?x4096xf16>
%28438 = flow.tensor.transfer %28437 : tensor<4x?x4096xf16>{%dim_25953} to #hal.device.promise<@__device_7>
%28439 = torch_c.from_builtin_tensor %28438 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%28440 = torch_c.to_builtin_tensor %28218 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_25954 = arith.constant 1 : index
%dim_25955 = tensor.dim %28440, %c1_25954 : tensor<4x?x4096xf16>
%28441 = flow.tensor.transfer %28440 : tensor<4x?x4096xf16>{%dim_25955} to #hal.device.promise<@__device_7>
%28442 = torch_c.from_builtin_tensor %28441 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25956 = torch.constant.int 1
%28443 = torch.aten.add.Tensor %28424, %28427, %int1_25956 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25957 = torch.constant.int 1
%28444 = torch.aten.add.Tensor %28443, %28430, %int1_25957 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25958 = torch.constant.int 1
%28445 = torch.aten.add.Tensor %28444, %28433, %int1_25958 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25959 = torch.constant.int 1
%28446 = torch.aten.add.Tensor %28445, %28436, %int1_25959 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25960 = torch.constant.int 1
%28447 = torch.aten.add.Tensor %28446, %28439, %int1_25960 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25961 = torch.constant.int 1
%28448 = torch.aten.add.Tensor %28447, %28442, %int1_25961 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25962 = torch.constant.int 1
%28449 = torch.aten.add.Tensor %28448, %28225, %int1_25962 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25963 = torch.constant.int 1
%28450 = torch.aten.add.Tensor %27930, %28253, %int1_25963 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25964 = torch.constant.int 1
%28451 = torch.aten.add.Tensor %27931, %28281, %int1_25964 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25965 = torch.constant.int 1
%28452 = torch.aten.add.Tensor %27932, %28309, %int1_25965 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25966 = torch.constant.int 1
%28453 = torch.aten.add.Tensor %27933, %28337, %int1_25966 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25967 = torch.constant.int 1
%28454 = torch.aten.add.Tensor %27934, %28365, %int1_25967 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25968 = torch.constant.int 1
%28455 = torch.aten.add.Tensor %27935, %28393, %int1_25968 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25969 = torch.constant.int 1
%28456 = torch.aten.add.Tensor %27936, %28421, %int1_25969 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_25970 = torch.constant.int 1
%28457 = torch.aten.add.Tensor %27937, %28449, %int1_25970 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_25971 = torch.constant.int 6
%28458 = torch.prims.convert_element_type %28450, %int6_25971 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25972 = torch.constant.int 6
%28459 = torch.prims.convert_element_type %28451, %int6_25972 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25973 = torch.constant.int 6
%28460 = torch.prims.convert_element_type %28452, %int6_25973 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25974 = torch.constant.int 6
%28461 = torch.prims.convert_element_type %28453, %int6_25974 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25975 = torch.constant.int 6
%28462 = torch.prims.convert_element_type %28454, %int6_25975 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25976 = torch.constant.int 6
%28463 = torch.prims.convert_element_type %28455, %int6_25976 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25977 = torch.constant.int 6
%28464 = torch.prims.convert_element_type %28456, %int6_25977 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_25978 = torch.constant.int 6
%28465 = torch.prims.convert_element_type %28457, %int6_25978 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25979 = torch.constant.int 2
%28466 = torch.aten.pow.Tensor_Scalar %28458, %int2_25979 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28466, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25980 = torch.constant.int 2
%28467 = torch.aten.pow.Tensor_Scalar %28459, %int2_25980 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25981 = torch.constant.int 2
%28468 = torch.aten.pow.Tensor_Scalar %28460, %int2_25981 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25982 = torch.constant.int 2
%28469 = torch.aten.pow.Tensor_Scalar %28461, %int2_25982 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25983 = torch.constant.int 2
%28470 = torch.aten.pow.Tensor_Scalar %28462, %int2_25983 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25984 = torch.constant.int 2
%28471 = torch.aten.pow.Tensor_Scalar %28463, %int2_25984 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25985 = torch.constant.int 2
%28472 = torch.aten.pow.Tensor_Scalar %28464, %int2_25985 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_25986 = torch.constant.int 2
%28473 = torch.aten.pow.Tensor_Scalar %28465, %int2_25986 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_25987 = torch.constant.int -1
%28474 = torch.prim.ListConstruct %int-1_25987 : (!torch.int) -> !torch.list<int>
%true_25988 = torch.constant.bool true
%none_25989 = torch.constant.none
%28475 = torch.aten.mean.dim %28466, %28474, %true_25988, %none_25989 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25990 = torch.constant.int -1
%28476 = torch.prim.ListConstruct %int-1_25990 : (!torch.int) -> !torch.list<int>
%true_25991 = torch.constant.bool true
%none_25992 = torch.constant.none
%28477 = torch.aten.mean.dim %28467, %28476, %true_25991, %none_25992 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25993 = torch.constant.int -1
%28478 = torch.prim.ListConstruct %int-1_25993 : (!torch.int) -> !torch.list<int>
%true_25994 = torch.constant.bool true
%none_25995 = torch.constant.none
%28479 = torch.aten.mean.dim %28468, %28478, %true_25994, %none_25995 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25996 = torch.constant.int -1
%28480 = torch.prim.ListConstruct %int-1_25996 : (!torch.int) -> !torch.list<int>
%true_25997 = torch.constant.bool true
%none_25998 = torch.constant.none
%28481 = torch.aten.mean.dim %28469, %28480, %true_25997, %none_25998 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_25999 = torch.constant.int -1
%28482 = torch.prim.ListConstruct %int-1_25999 : (!torch.int) -> !torch.list<int>
%true_26000 = torch.constant.bool true
%none_26001 = torch.constant.none
%28483 = torch.aten.mean.dim %28470, %28482, %true_26000, %none_26001 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_26002 = torch.constant.int -1
%28484 = torch.prim.ListConstruct %int-1_26002 : (!torch.int) -> !torch.list<int>
%true_26003 = torch.constant.bool true
%none_26004 = torch.constant.none
%28485 = torch.aten.mean.dim %28471, %28484, %true_26003, %none_26004 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_26005 = torch.constant.int -1
%28486 = torch.prim.ListConstruct %int-1_26005 : (!torch.int) -> !torch.list<int>
%true_26006 = torch.constant.bool true
%none_26007 = torch.constant.none
%28487 = torch.aten.mean.dim %28472, %28486, %true_26006, %none_26007 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_26008 = torch.constant.int -1
%28488 = torch.prim.ListConstruct %int-1_26008 : (!torch.int) -> !torch.list<int>
%true_26009 = torch.constant.bool true
%none_26010 = torch.constant.none
%28489 = torch.aten.mean.dim %28473, %28488, %true_26009, %none_26010 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26011 = torch.constant.float 9.9999997473787516E-6
%int1_26012 = torch.constant.int 1
%28490 = torch.aten.add.Scalar %28475, %float9.999990e-06_26011, %int1_26012 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26013 = torch.constant.float 9.9999997473787516E-6
%int1_26014 = torch.constant.int 1
%28491 = torch.aten.add.Scalar %28477, %float9.999990e-06_26013, %int1_26014 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26015 = torch.constant.float 9.9999997473787516E-6
%int1_26016 = torch.constant.int 1
%28492 = torch.aten.add.Scalar %28479, %float9.999990e-06_26015, %int1_26016 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26017 = torch.constant.float 9.9999997473787516E-6
%int1_26018 = torch.constant.int 1
%28493 = torch.aten.add.Scalar %28481, %float9.999990e-06_26017, %int1_26018 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26019 = torch.constant.float 9.9999997473787516E-6
%int1_26020 = torch.constant.int 1
%28494 = torch.aten.add.Scalar %28483, %float9.999990e-06_26019, %int1_26020 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26021 = torch.constant.float 9.9999997473787516E-6
%int1_26022 = torch.constant.int 1
%28495 = torch.aten.add.Scalar %28485, %float9.999990e-06_26021, %int1_26022 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26023 = torch.constant.float 9.9999997473787516E-6
%int1_26024 = torch.constant.int 1
%28496 = torch.aten.add.Scalar %28487, %float9.999990e-06_26023, %int1_26024 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_26025 = torch.constant.float 9.9999997473787516E-6
%int1_26026 = torch.constant.int 1
%28497 = torch.aten.add.Scalar %28489, %float9.999990e-06_26025, %int1_26026 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28498 = torch.aten.rsqrt %28490 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28499 = torch.aten.rsqrt %28491 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28500 = torch.aten.rsqrt %28492 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28501 = torch.aten.rsqrt %28493 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28502 = torch.aten.rsqrt %28494 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28503 = torch.aten.rsqrt %28495 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28504 = torch.aten.rsqrt %28496 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28505 = torch.aten.rsqrt %28497 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %28505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%28506 = torch.aten.mul.Tensor %28458, %28498 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28507 = torch.aten.mul.Tensor %28459, %28499 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28508 = torch.aten.mul.Tensor %28460, %28500 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28509 = torch.aten.mul.Tensor %28461, %28501 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28510 = torch.aten.mul.Tensor %28462, %28502 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28511 = torch.aten.mul.Tensor %28463, %28503 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28512 = torch.aten.mul.Tensor %28464, %28504 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28513 = torch.aten.mul.Tensor %28465, %28505 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28514 = torch.aten.mul.Tensor %1016, %28506 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28515 = torch.aten.mul.Tensor %1017, %28507 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28516 = torch.aten.mul.Tensor %1018, %28508 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28517 = torch.aten.mul.Tensor %1019, %28509 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28518 = torch.aten.mul.Tensor %1020, %28510 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28519 = torch.aten.mul.Tensor %1021, %28511 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28520 = torch.aten.mul.Tensor %1022, %28512 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%28521 = torch.aten.mul.Tensor %1023, %28513 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %28521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_26027 = torch.constant.int 5
%28522 = torch.prims.convert_element_type %28514, %int5_26027 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26028 = torch.constant.int 5
%28523 = torch.prims.convert_element_type %28515, %int5_26028 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26029 = torch.constant.int 5
%28524 = torch.prims.convert_element_type %28516, %int5_26029 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26030 = torch.constant.int 5
%28525 = torch.prims.convert_element_type %28517, %int5_26030 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26031 = torch.constant.int 5
%28526 = torch.prims.convert_element_type %28518, %int5_26031 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26032 = torch.constant.int 5
%28527 = torch.prims.convert_element_type %28519, %int5_26032 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26033 = torch.constant.int 5
%28528 = torch.prims.convert_element_type %28520, %int5_26033 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_26034 = torch.constant.int 5
%28529 = torch.prims.convert_element_type %28521, %int5_26034 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %28529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_26035 = torch.constant.int 1
%int0_26036 = torch.constant.int 0
%28530 = torch.prim.ListConstruct %int1_26035, %int0_26036 : (!torch.int, !torch.int) -> !torch.list<int>
%28531 = torch.aten.permute %1024, %28530 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26037 = torch.constant.int 1
%int0_26038 = torch.constant.int 0
%28532 = torch.prim.ListConstruct %int1_26037, %int0_26038 : (!torch.int, !torch.int) -> !torch.list<int>
%28533 = torch.aten.permute %1025, %28532 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26039 = torch.constant.int 1
%int0_26040 = torch.constant.int 0
%28534 = torch.prim.ListConstruct %int1_26039, %int0_26040 : (!torch.int, !torch.int) -> !torch.list<int>
%28535 = torch.aten.permute %1026, %28534 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26041 = torch.constant.int 1
%int0_26042 = torch.constant.int 0
%28536 = torch.prim.ListConstruct %int1_26041, %int0_26042 : (!torch.int, !torch.int) -> !torch.list<int>
%28537 = torch.aten.permute %1027, %28536 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26043 = torch.constant.int 1
%int0_26044 = torch.constant.int 0
%28538 = torch.prim.ListConstruct %int1_26043, %int0_26044 : (!torch.int, !torch.int) -> !torch.list<int>
%28539 = torch.aten.permute %1028, %28538 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26045 = torch.constant.int 1
%int0_26046 = torch.constant.int 0
%28540 = torch.prim.ListConstruct %int1_26045, %int0_26046 : (!torch.int, !torch.int) -> !torch.list<int>
%28541 = torch.aten.permute %1029, %28540 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26047 = torch.constant.int 1
%int0_26048 = torch.constant.int 0
%28542 = torch.prim.ListConstruct %int1_26047, %int0_26048 : (!torch.int, !torch.int) -> !torch.list<int>
%28543 = torch.aten.permute %1030, %28542 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_26049 = torch.constant.int 1
%int0_26050 = torch.constant.int 0
%28544 = torch.prim.ListConstruct %int1_26049, %int0_26050 : (!torch.int, !torch.int) -> !torch.list<int>
%28545 = torch.aten.permute %1031, %28544 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_26051 = torch.constant.int 4
%28546 = torch.aten.mul.int %int4_26051, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26052 = torch.constant.int 4096
%28547 = torch.prim.ListConstruct %28546, %int4096_26052 : (!torch.int, !torch.int) -> !torch.list<int>
%28548 = torch.aten.view %28522, %28547 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28548, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28549 = torch.aten.mm %28548, %28531 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28549, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26053 = torch.constant.int 4
%int512_26054 = torch.constant.int 512
%28550 = torch.prim.ListConstruct %int4_26053, %2482, %int512_26054 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28551 = torch.aten.view %28549, %28550 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26055 = torch.constant.int 4
%28552 = torch.aten.mul.int %int4_26055, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26056 = torch.constant.int 4096
%28553 = torch.prim.ListConstruct %28552, %int4096_26056 : (!torch.int, !torch.int) -> !torch.list<int>
%28554 = torch.aten.view %28523, %28553 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28554, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28555 = torch.aten.mm %28554, %28533 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28555, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26057 = torch.constant.int 4
%int512_26058 = torch.constant.int 512
%28556 = torch.prim.ListConstruct %int4_26057, %2482, %int512_26058 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28557 = torch.aten.view %28555, %28556 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26059 = torch.constant.int 4
%28558 = torch.aten.mul.int %int4_26059, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26060 = torch.constant.int 4096
%28559 = torch.prim.ListConstruct %28558, %int4096_26060 : (!torch.int, !torch.int) -> !torch.list<int>
%28560 = torch.aten.view %28524, %28559 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28560, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28561 = torch.aten.mm %28560, %28535 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28561, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26061 = torch.constant.int 4
%int512_26062 = torch.constant.int 512
%28562 = torch.prim.ListConstruct %int4_26061, %2482, %int512_26062 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28563 = torch.aten.view %28561, %28562 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26063 = torch.constant.int 4
%28564 = torch.aten.mul.int %int4_26063, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26064 = torch.constant.int 4096
%28565 = torch.prim.ListConstruct %28564, %int4096_26064 : (!torch.int, !torch.int) -> !torch.list<int>
%28566 = torch.aten.view %28525, %28565 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28566, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28567 = torch.aten.mm %28566, %28537 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28567, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26065 = torch.constant.int 4
%int512_26066 = torch.constant.int 512
%28568 = torch.prim.ListConstruct %int4_26065, %2482, %int512_26066 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28569 = torch.aten.view %28567, %28568 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26067 = torch.constant.int 4
%28570 = torch.aten.mul.int %int4_26067, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26068 = torch.constant.int 4096
%28571 = torch.prim.ListConstruct %28570, %int4096_26068 : (!torch.int, !torch.int) -> !torch.list<int>
%28572 = torch.aten.view %28526, %28571 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28572, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28573 = torch.aten.mm %28572, %28539 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28573, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26069 = torch.constant.int 4
%int512_26070 = torch.constant.int 512
%28574 = torch.prim.ListConstruct %int4_26069, %2482, %int512_26070 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28575 = torch.aten.view %28573, %28574 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26071 = torch.constant.int 4
%28576 = torch.aten.mul.int %int4_26071, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26072 = torch.constant.int 4096
%28577 = torch.prim.ListConstruct %28576, %int4096_26072 : (!torch.int, !torch.int) -> !torch.list<int>
%28578 = torch.aten.view %28527, %28577 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28578, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28579 = torch.aten.mm %28578, %28541 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28579, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26073 = torch.constant.int 4
%int512_26074 = torch.constant.int 512
%28580 = torch.prim.ListConstruct %int4_26073, %2482, %int512_26074 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28581 = torch.aten.view %28579, %28580 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26075 = torch.constant.int 4
%28582 = torch.aten.mul.int %int4_26075, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26076 = torch.constant.int 4096
%28583 = torch.prim.ListConstruct %28582, %int4096_26076 : (!torch.int, !torch.int) -> !torch.list<int>
%28584 = torch.aten.view %28528, %28583 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28584, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28585 = torch.aten.mm %28584, %28543 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28585, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26077 = torch.constant.int 4
%int512_26078 = torch.constant.int 512
%28586 = torch.prim.ListConstruct %int4_26077, %2482, %int512_26078 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28587 = torch.aten.view %28585, %28586 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_26079 = torch.constant.int 4
%28588 = torch.aten.mul.int %int4_26079, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26080 = torch.constant.int 4096
%28589 = torch.prim.ListConstruct %28588, %int4096_26080 : (!torch.int, !torch.int) -> !torch.list<int>
%28590 = torch.aten.view %28529, %28589 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28590, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28591 = torch.aten.mm %28590, %28545 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %28591, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_26081 = torch.constant.int 4
%int512_26082 = torch.constant.int 512
%28592 = torch.prim.ListConstruct %int4_26081, %2482, %int512_26082 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28593 = torch.aten.view %28591, %28592 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %28593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_26083 = torch.constant.int 1
%int0_26084 = torch.constant.int 0
%28594 = torch.prim.ListConstruct %int1_26083, %int0_26084 : (!torch.int, !torch.int) -> !torch.list<int>
%28595 = torch.aten.permute %1032, %28594 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26085 = torch.constant.int 1
%int0_26086 = torch.constant.int 0
%28596 = torch.prim.ListConstruct %int1_26085, %int0_26086 : (!torch.int, !torch.int) -> !torch.list<int>
%28597 = torch.aten.permute %1033, %28596 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26087 = torch.constant.int 1
%int0_26088 = torch.constant.int 0
%28598 = torch.prim.ListConstruct %int1_26087, %int0_26088 : (!torch.int, !torch.int) -> !torch.list<int>
%28599 = torch.aten.permute %1034, %28598 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26089 = torch.constant.int 1
%int0_26090 = torch.constant.int 0
%28600 = torch.prim.ListConstruct %int1_26089, %int0_26090 : (!torch.int, !torch.int) -> !torch.list<int>
%28601 = torch.aten.permute %1035, %28600 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26091 = torch.constant.int 1
%int0_26092 = torch.constant.int 0
%28602 = torch.prim.ListConstruct %int1_26091, %int0_26092 : (!torch.int, !torch.int) -> !torch.list<int>
%28603 = torch.aten.permute %1036, %28602 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26093 = torch.constant.int 1
%int0_26094 = torch.constant.int 0
%28604 = torch.prim.ListConstruct %int1_26093, %int0_26094 : (!torch.int, !torch.int) -> !torch.list<int>
%28605 = torch.aten.permute %1037, %28604 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26095 = torch.constant.int 1
%int0_26096 = torch.constant.int 0
%28606 = torch.prim.ListConstruct %int1_26095, %int0_26096 : (!torch.int, !torch.int) -> !torch.list<int>
%28607 = torch.aten.permute %1038, %28606 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26097 = torch.constant.int 1
%int0_26098 = torch.constant.int 0
%28608 = torch.prim.ListConstruct %int1_26097, %int0_26098 : (!torch.int, !torch.int) -> !torch.list<int>
%28609 = torch.aten.permute %1039, %28608 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_26099 = torch.constant.int 4
%28610 = torch.aten.mul.int %int4_26099, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26100 = torch.constant.int 4096
%28611 = torch.prim.ListConstruct %28610, %int4096_26100 : (!torch.int, !torch.int) -> !torch.list<int>
%28612 = torch.aten.view %28522, %28611 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28612, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28613 = torch.aten.mm %28612, %28595 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28613, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26101 = torch.constant.int 4
%int128_26102 = torch.constant.int 128
%28614 = torch.prim.ListConstruct %int4_26101, %2482, %int128_26102 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28615 = torch.aten.view %28613, %28614 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26103 = torch.constant.int 4
%28616 = torch.aten.mul.int %int4_26103, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26104 = torch.constant.int 4096
%28617 = torch.prim.ListConstruct %28616, %int4096_26104 : (!torch.int, !torch.int) -> !torch.list<int>
%28618 = torch.aten.view %28523, %28617 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28618, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28619 = torch.aten.mm %28618, %28597 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28619, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26105 = torch.constant.int 4
%int128_26106 = torch.constant.int 128
%28620 = torch.prim.ListConstruct %int4_26105, %2482, %int128_26106 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28621 = torch.aten.view %28619, %28620 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26107 = torch.constant.int 4
%28622 = torch.aten.mul.int %int4_26107, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26108 = torch.constant.int 4096
%28623 = torch.prim.ListConstruct %28622, %int4096_26108 : (!torch.int, !torch.int) -> !torch.list<int>
%28624 = torch.aten.view %28524, %28623 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28624, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28625 = torch.aten.mm %28624, %28599 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28625, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26109 = torch.constant.int 4
%int128_26110 = torch.constant.int 128
%28626 = torch.prim.ListConstruct %int4_26109, %2482, %int128_26110 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28627 = torch.aten.view %28625, %28626 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26111 = torch.constant.int 4
%28628 = torch.aten.mul.int %int4_26111, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26112 = torch.constant.int 4096
%28629 = torch.prim.ListConstruct %28628, %int4096_26112 : (!torch.int, !torch.int) -> !torch.list<int>
%28630 = torch.aten.view %28525, %28629 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28630, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28631 = torch.aten.mm %28630, %28601 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28631, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26113 = torch.constant.int 4
%int128_26114 = torch.constant.int 128
%28632 = torch.prim.ListConstruct %int4_26113, %2482, %int128_26114 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28633 = torch.aten.view %28631, %28632 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26115 = torch.constant.int 4
%28634 = torch.aten.mul.int %int4_26115, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26116 = torch.constant.int 4096
%28635 = torch.prim.ListConstruct %28634, %int4096_26116 : (!torch.int, !torch.int) -> !torch.list<int>
%28636 = torch.aten.view %28526, %28635 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28636, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28637 = torch.aten.mm %28636, %28603 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28637, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26117 = torch.constant.int 4
%int128_26118 = torch.constant.int 128
%28638 = torch.prim.ListConstruct %int4_26117, %2482, %int128_26118 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28639 = torch.aten.view %28637, %28638 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26119 = torch.constant.int 4
%28640 = torch.aten.mul.int %int4_26119, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26120 = torch.constant.int 4096
%28641 = torch.prim.ListConstruct %28640, %int4096_26120 : (!torch.int, !torch.int) -> !torch.list<int>
%28642 = torch.aten.view %28527, %28641 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28642, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28643 = torch.aten.mm %28642, %28605 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28643, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26121 = torch.constant.int 4
%int128_26122 = torch.constant.int 128
%28644 = torch.prim.ListConstruct %int4_26121, %2482, %int128_26122 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28645 = torch.aten.view %28643, %28644 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26123 = torch.constant.int 4
%28646 = torch.aten.mul.int %int4_26123, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26124 = torch.constant.int 4096
%28647 = torch.prim.ListConstruct %28646, %int4096_26124 : (!torch.int, !torch.int) -> !torch.list<int>
%28648 = torch.aten.view %28528, %28647 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28648, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28649 = torch.aten.mm %28648, %28607 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28649, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26125 = torch.constant.int 4
%int128_26126 = torch.constant.int 128
%28650 = torch.prim.ListConstruct %int4_26125, %2482, %int128_26126 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28651 = torch.aten.view %28649, %28650 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26127 = torch.constant.int 4
%28652 = torch.aten.mul.int %int4_26127, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26128 = torch.constant.int 4096
%28653 = torch.prim.ListConstruct %28652, %int4096_26128 : (!torch.int, !torch.int) -> !torch.list<int>
%28654 = torch.aten.view %28529, %28653 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28654, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28655 = torch.aten.mm %28654, %28609 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28655, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26129 = torch.constant.int 4
%int128_26130 = torch.constant.int 128
%28656 = torch.prim.ListConstruct %int4_26129, %2482, %int128_26130 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28657 = torch.aten.view %28655, %28656 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_26131 = torch.constant.int 1
%int0_26132 = torch.constant.int 0
%28658 = torch.prim.ListConstruct %int1_26131, %int0_26132 : (!torch.int, !torch.int) -> !torch.list<int>
%28659 = torch.aten.permute %1040, %28658 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26133 = torch.constant.int 1
%int0_26134 = torch.constant.int 0
%28660 = torch.prim.ListConstruct %int1_26133, %int0_26134 : (!torch.int, !torch.int) -> !torch.list<int>
%28661 = torch.aten.permute %1041, %28660 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26135 = torch.constant.int 1
%int0_26136 = torch.constant.int 0
%28662 = torch.prim.ListConstruct %int1_26135, %int0_26136 : (!torch.int, !torch.int) -> !torch.list<int>
%28663 = torch.aten.permute %1042, %28662 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26137 = torch.constant.int 1
%int0_26138 = torch.constant.int 0
%28664 = torch.prim.ListConstruct %int1_26137, %int0_26138 : (!torch.int, !torch.int) -> !torch.list<int>
%28665 = torch.aten.permute %1043, %28664 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26139 = torch.constant.int 1
%int0_26140 = torch.constant.int 0
%28666 = torch.prim.ListConstruct %int1_26139, %int0_26140 : (!torch.int, !torch.int) -> !torch.list<int>
%28667 = torch.aten.permute %1044, %28666 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26141 = torch.constant.int 1
%int0_26142 = torch.constant.int 0
%28668 = torch.prim.ListConstruct %int1_26141, %int0_26142 : (!torch.int, !torch.int) -> !torch.list<int>
%28669 = torch.aten.permute %1045, %28668 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26143 = torch.constant.int 1
%int0_26144 = torch.constant.int 0
%28670 = torch.prim.ListConstruct %int1_26143, %int0_26144 : (!torch.int, !torch.int) -> !torch.list<int>
%28671 = torch.aten.permute %1046, %28670 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_26145 = torch.constant.int 1
%int0_26146 = torch.constant.int 0
%28672 = torch.prim.ListConstruct %int1_26145, %int0_26146 : (!torch.int, !torch.int) -> !torch.list<int>
%28673 = torch.aten.permute %1047, %28672 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_26147 = torch.constant.int 4
%28674 = torch.aten.mul.int %int4_26147, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26148 = torch.constant.int 4096
%28675 = torch.prim.ListConstruct %28674, %int4096_26148 : (!torch.int, !torch.int) -> !torch.list<int>
%28676 = torch.aten.view %28522, %28675 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28676, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28677 = torch.aten.mm %28676, %28659 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28677, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26149 = torch.constant.int 4
%int128_26150 = torch.constant.int 128
%28678 = torch.prim.ListConstruct %int4_26149, %2482, %int128_26150 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28679 = torch.aten.view %28677, %28678 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26151 = torch.constant.int 4
%28680 = torch.aten.mul.int %int4_26151, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26152 = torch.constant.int 4096
%28681 = torch.prim.ListConstruct %28680, %int4096_26152 : (!torch.int, !torch.int) -> !torch.list<int>
%28682 = torch.aten.view %28523, %28681 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28682, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28683 = torch.aten.mm %28682, %28661 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28683, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26153 = torch.constant.int 4
%int128_26154 = torch.constant.int 128
%28684 = torch.prim.ListConstruct %int4_26153, %2482, %int128_26154 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28685 = torch.aten.view %28683, %28684 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26155 = torch.constant.int 4
%28686 = torch.aten.mul.int %int4_26155, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26156 = torch.constant.int 4096
%28687 = torch.prim.ListConstruct %28686, %int4096_26156 : (!torch.int, !torch.int) -> !torch.list<int>
%28688 = torch.aten.view %28524, %28687 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28688, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28689 = torch.aten.mm %28688, %28663 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28689, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26157 = torch.constant.int 4
%int128_26158 = torch.constant.int 128
%28690 = torch.prim.ListConstruct %int4_26157, %2482, %int128_26158 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28691 = torch.aten.view %28689, %28690 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26159 = torch.constant.int 4
%28692 = torch.aten.mul.int %int4_26159, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26160 = torch.constant.int 4096
%28693 = torch.prim.ListConstruct %28692, %int4096_26160 : (!torch.int, !torch.int) -> !torch.list<int>
%28694 = torch.aten.view %28525, %28693 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28694, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28695 = torch.aten.mm %28694, %28665 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28695, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26161 = torch.constant.int 4
%int128_26162 = torch.constant.int 128
%28696 = torch.prim.ListConstruct %int4_26161, %2482, %int128_26162 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28697 = torch.aten.view %28695, %28696 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26163 = torch.constant.int 4
%28698 = torch.aten.mul.int %int4_26163, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26164 = torch.constant.int 4096
%28699 = torch.prim.ListConstruct %28698, %int4096_26164 : (!torch.int, !torch.int) -> !torch.list<int>
%28700 = torch.aten.view %28526, %28699 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28700, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28701 = torch.aten.mm %28700, %28667 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28701, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26165 = torch.constant.int 4
%int128_26166 = torch.constant.int 128
%28702 = torch.prim.ListConstruct %int4_26165, %2482, %int128_26166 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28703 = torch.aten.view %28701, %28702 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26167 = torch.constant.int 4
%28704 = torch.aten.mul.int %int4_26167, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26168 = torch.constant.int 4096
%28705 = torch.prim.ListConstruct %28704, %int4096_26168 : (!torch.int, !torch.int) -> !torch.list<int>
%28706 = torch.aten.view %28527, %28705 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28706, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28707 = torch.aten.mm %28706, %28669 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28707, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26169 = torch.constant.int 4
%int128_26170 = torch.constant.int 128
%28708 = torch.prim.ListConstruct %int4_26169, %2482, %int128_26170 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28709 = torch.aten.view %28707, %28708 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26171 = torch.constant.int 4
%28710 = torch.aten.mul.int %int4_26171, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26172 = torch.constant.int 4096
%28711 = torch.prim.ListConstruct %28710, %int4096_26172 : (!torch.int, !torch.int) -> !torch.list<int>
%28712 = torch.aten.view %28528, %28711 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28712, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28713 = torch.aten.mm %28712, %28671 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28713, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26173 = torch.constant.int 4
%int128_26174 = torch.constant.int 128
%28714 = torch.prim.ListConstruct %int4_26173, %2482, %int128_26174 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28715 = torch.aten.view %28713, %28714 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26175 = torch.constant.int 4
%28716 = torch.aten.mul.int %int4_26175, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_26176 = torch.constant.int 4096
%28717 = torch.prim.ListConstruct %28716, %int4096_26176 : (!torch.int, !torch.int) -> !torch.list<int>
%28718 = torch.aten.view %28529, %28717 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %28718, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%28719 = torch.aten.mm %28718, %28673 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %28719, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_26177 = torch.constant.int 4
%int128_26178 = torch.constant.int 128
%28720 = torch.prim.ListConstruct %int4_26177, %2482, %int128_26178 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28721 = torch.aten.view %28719, %28720 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %28721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_26179 = torch.constant.int 4
%int4_26180 = torch.constant.int 4
%int128_26181 = torch.constant.int 128
%28722 = torch.prim.ListConstruct %int4_26179, %2482, %int4_26180, %int128_26181 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28723 = torch.aten.view %28551, %28722 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26182 = torch.constant.int 4
%int4_26183 = torch.constant.int 4
%int128_26184 = torch.constant.int 128
%28724 = torch.prim.ListConstruct %int4_26182, %2482, %int4_26183, %int128_26184 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28725 = torch.aten.view %28557, %28724 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26185 = torch.constant.int 4
%int4_26186 = torch.constant.int 4
%int128_26187 = torch.constant.int 128
%28726 = torch.prim.ListConstruct %int4_26185, %2482, %int4_26186, %int128_26187 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28727 = torch.aten.view %28563, %28726 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26188 = torch.constant.int 4
%int4_26189 = torch.constant.int 4
%int128_26190 = torch.constant.int 128
%28728 = torch.prim.ListConstruct %int4_26188, %2482, %int4_26189, %int128_26190 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28729 = torch.aten.view %28569, %28728 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26191 = torch.constant.int 4
%int4_26192 = torch.constant.int 4
%int128_26193 = torch.constant.int 128
%28730 = torch.prim.ListConstruct %int4_26191, %2482, %int4_26192, %int128_26193 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28731 = torch.aten.view %28575, %28730 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26194 = torch.constant.int 4
%int4_26195 = torch.constant.int 4
%int128_26196 = torch.constant.int 128
%28732 = torch.prim.ListConstruct %int4_26194, %2482, %int4_26195, %int128_26196 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28733 = torch.aten.view %28581, %28732 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26197 = torch.constant.int 4
%int4_26198 = torch.constant.int 4
%int128_26199 = torch.constant.int 128
%28734 = torch.prim.ListConstruct %int4_26197, %2482, %int4_26198, %int128_26199 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28735 = torch.aten.view %28587, %28734 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26200 = torch.constant.int 4
%int4_26201 = torch.constant.int 4
%int128_26202 = torch.constant.int 128
%28736 = torch.prim.ListConstruct %int4_26200, %2482, %int4_26201, %int128_26202 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28737 = torch.aten.view %28593, %28736 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_26203 = torch.constant.int 4
%int1_26204 = torch.constant.int 1
%int128_26205 = torch.constant.int 128
%28738 = torch.prim.ListConstruct %int4_26203, %2482, %int1_26204, %int128_26205 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28739 = torch.aten.view %28615, %28738 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26206 = torch.constant.int 4
%int1_26207 = torch.constant.int 1
%int128_26208 = torch.constant.int 128
%28740 = torch.prim.ListConstruct %int4_26206, %2482, %int1_26207, %int128_26208 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28741 = torch.aten.view %28621, %28740 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26209 = torch.constant.int 4
%int1_26210 = torch.constant.int 1
%int128_26211 = torch.constant.int 128
%28742 = torch.prim.ListConstruct %int4_26209, %2482, %int1_26210, %int128_26211 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28743 = torch.aten.view %28627, %28742 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26212 = torch.constant.int 4
%int1_26213 = torch.constant.int 1
%int128_26214 = torch.constant.int 128
%28744 = torch.prim.ListConstruct %int4_26212, %2482, %int1_26213, %int128_26214 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28745 = torch.aten.view %28633, %28744 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26215 = torch.constant.int 4
%int1_26216 = torch.constant.int 1
%int128_26217 = torch.constant.int 128
%28746 = torch.prim.ListConstruct %int4_26215, %2482, %int1_26216, %int128_26217 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28747 = torch.aten.view %28639, %28746 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26218 = torch.constant.int 4
%int1_26219 = torch.constant.int 1
%int128_26220 = torch.constant.int 128
%28748 = torch.prim.ListConstruct %int4_26218, %2482, %int1_26219, %int128_26220 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28749 = torch.aten.view %28645, %28748 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26221 = torch.constant.int 4
%int1_26222 = torch.constant.int 1
%int128_26223 = torch.constant.int 128
%28750 = torch.prim.ListConstruct %int4_26221, %2482, %int1_26222, %int128_26223 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28751 = torch.aten.view %28651, %28750 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26224 = torch.constant.int 4
%int1_26225 = torch.constant.int 1
%int128_26226 = torch.constant.int 128
%28752 = torch.prim.ListConstruct %int4_26224, %2482, %int1_26225, %int128_26226 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28753 = torch.aten.view %28657, %28752 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26227 = torch.constant.int 4
%int1_26228 = torch.constant.int 1
%int128_26229 = torch.constant.int 128
%28754 = torch.prim.ListConstruct %int4_26227, %2482, %int1_26228, %int128_26229 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28755 = torch.aten.view %28679, %28754 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26230 = torch.constant.int 4
%int1_26231 = torch.constant.int 1
%int128_26232 = torch.constant.int 128
%28756 = torch.prim.ListConstruct %int4_26230, %2482, %int1_26231, %int128_26232 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28757 = torch.aten.view %28685, %28756 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26233 = torch.constant.int 4
%int1_26234 = torch.constant.int 1
%int128_26235 = torch.constant.int 128
%28758 = torch.prim.ListConstruct %int4_26233, %2482, %int1_26234, %int128_26235 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28759 = torch.aten.view %28691, %28758 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26236 = torch.constant.int 4
%int1_26237 = torch.constant.int 1
%int128_26238 = torch.constant.int 128
%28760 = torch.prim.ListConstruct %int4_26236, %2482, %int1_26237, %int128_26238 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28761 = torch.aten.view %28697, %28760 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26239 = torch.constant.int 4
%int1_26240 = torch.constant.int 1
%int128_26241 = torch.constant.int 128
%28762 = torch.prim.ListConstruct %int4_26239, %2482, %int1_26240, %int128_26241 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28763 = torch.aten.view %28703, %28762 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26242 = torch.constant.int 4
%int1_26243 = torch.constant.int 1
%int128_26244 = torch.constant.int 128
%28764 = torch.prim.ListConstruct %int4_26242, %2482, %int1_26243, %int128_26244 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28765 = torch.aten.view %28709, %28764 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26245 = torch.constant.int 4
%int1_26246 = torch.constant.int 1
%int128_26247 = torch.constant.int 128
%28766 = torch.prim.ListConstruct %int4_26245, %2482, %int1_26246, %int128_26247 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28767 = torch.aten.view %28715, %28766 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_26248 = torch.constant.int 4
%int1_26249 = torch.constant.int 1
%int128_26250 = torch.constant.int 128
%28768 = torch.prim.ListConstruct %int4_26248, %2482, %int1_26249, %int128_26250 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%28769 = torch.aten.view %28721, %28768 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_26251 = torch.constant.int 131072
%none_26252 = torch.constant.none
%none_26253 = torch.constant.none
%cpu_26254 = torch.constant.device "cpu"
%false_26255 = torch.constant.bool false
%28770 = torch.aten.arange %int131072_26251, %none_26252, %none_26253, %cpu_26254, %false_26255 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_26256 = torch.constant.int 0
%int128_26257 = torch.constant.int 128
%int2_26258 = torch.constant.int 2
%none_26259 = torch.constant.none
%none_26260 = torch.constant.none
%cpu_26261 = torch.constant.device "cpu"
%false_26262 = torch.constant.bool false
%28771 = torch.aten.arange.start_step %int0_26256, %int128_26257, %int2_26258, %none_26259, %none_26260, %cpu_26261, %false_26262 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_26263 = torch.constant.int 0
%int0_26264 = torch.constant.int 0
%int64_26265 = torch.constant.int 64
%int1_26266 = torch.constant.int 1
%28772 = torch.aten.slice.Tensor %28771, %int0_26263, %int0_26264, %int64_26265, %int1_26266 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_26267 = torch.constant.int 6
%28773 = torch.prims.convert_element_type %28772, %int6_26267 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_26268 = torch.constant.int 128
%28774 = torch.aten.div.Scalar %28773, %int128_26268 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_26269 = torch.constant.float 5.000000e+05
%28775 = torch.aten.pow.Scalar %float5.000000e05_26269, %28774 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%28776 = torch.aten.reciprocal %28775 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_26270 = torch.constant.float 1.000000e+00
%28777 = torch.aten.mul.Scalar %28776, %float1.000000e00_26270 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_26271 = torch.constant.int 131072
%int1_26272 = torch.constant.int 1
%28778 = torch.prim.ListConstruct %int131072_26271, %int1_26272 : (!torch.int, !torch.int) -> !torch.list<int>
%28779 = torch.aten.view %28770, %28778 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%28780 = torch.aten.mul.Tensor %28779, %28777 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%28781 = torch.aten.cos %28780 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%28782 = torch.aten.sin %28780 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%28783 = torch.aten.complex %28781, %28782 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%28784 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28785 = flow.tensor.transfer %28784 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%28786 = torch_c.from_builtin_tensor %28785 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28787 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28788 = flow.tensor.transfer %28787 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%28789 = torch_c.from_builtin_tensor %28788 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28790 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28791 = flow.tensor.transfer %28790 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%28792 = torch_c.from_builtin_tensor %28791 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28793 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28794 = flow.tensor.transfer %28793 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%28795 = torch_c.from_builtin_tensor %28794 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28796 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28797 = flow.tensor.transfer %28796 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%28798 = torch_c.from_builtin_tensor %28797 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28799 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28800 = flow.tensor.transfer %28799 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%28801 = torch_c.from_builtin_tensor %28800 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28802 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28803 = flow.tensor.transfer %28802 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%28804 = torch_c.from_builtin_tensor %28803 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28805 = torch_c.to_builtin_tensor %28783 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28806 = flow.tensor.transfer %28805 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%28807 = torch_c.from_builtin_tensor %28806 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_26273 = torch.constant.int 1
%28808 = torch.aten.size.int %28551, %int1_26273 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26274 = torch.constant.int 0
%28809 = torch.aten.add.int %int0_26274, %28808 : !torch.int, !torch.int -> !torch.int
%int0_26275 = torch.constant.int 0
%int0_26276 = torch.constant.int 0
%int1_26277 = torch.constant.int 1
%28810 = torch.aten.slice.Tensor %28786, %int0_26275, %int0_26276, %28809, %int1_26277 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28810, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26278 = torch.constant.int 1
%int0_26279 = torch.constant.int 0
%int9223372036854775807_26280 = torch.constant.int 9223372036854775807
%int1_26281 = torch.constant.int 1
%28811 = torch.aten.slice.Tensor %28810, %int1_26278, %int0_26279, %int9223372036854775807_26280, %int1_26281 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28811, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26282 = torch.constant.int 0
%28812 = torch.aten.unsqueeze %28811, %int0_26282 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28812, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26283 = torch.constant.int 2
%28813 = torch.aten.unsqueeze %28812, %int2_26283 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28813, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26284 = torch.constant.int 3
%int0_26285 = torch.constant.int 0
%int9223372036854775807_26286 = torch.constant.int 9223372036854775807
%int1_26287 = torch.constant.int 1
%28814 = torch.aten.slice.Tensor %28813, %int3_26284, %int0_26285, %int9223372036854775807_26286, %int1_26287 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28814, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28815 = torch_c.to_builtin_tensor %28723 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26288 = arith.constant 1 : index
%dim_26289 = tensor.dim %28815, %c1_26288 : tensor<4x?x4x128xf16>
%28816 = flow.tensor.bitcast %28815 : tensor<4x?x4x128xf16>{%dim_26289} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26289}
%28817 = torch_c.from_builtin_tensor %28816 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28818 = torch.aten.mul.Tensor %28817, %28814 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28819 = torch_c.to_builtin_tensor %28818 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26290 = arith.constant 1 : index
%dim_26291 = tensor.dim %28819, %c1_26290 : tensor<4x?x4x64xcomplex<f32>>
%28820 = flow.tensor.bitcast %28819 : tensor<4x?x4x64xcomplex<f32>>{%dim_26291} -> tensor<4x?x4x128xf32>{%dim_26291}
%28821 = torch_c.from_builtin_tensor %28820 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26292 = torch.constant.int 5
%28822 = torch.prims.convert_element_type %28821, %int5_26292 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26293 = torch.constant.int 1
%28823 = torch.aten.size.int %28557, %int1_26293 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26294 = torch.constant.int 0
%28824 = torch.aten.add.int %int0_26294, %28823 : !torch.int, !torch.int -> !torch.int
%int0_26295 = torch.constant.int 0
%int0_26296 = torch.constant.int 0
%int1_26297 = torch.constant.int 1
%28825 = torch.aten.slice.Tensor %28789, %int0_26295, %int0_26296, %28824, %int1_26297 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28825, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26298 = torch.constant.int 1
%int0_26299 = torch.constant.int 0
%int9223372036854775807_26300 = torch.constant.int 9223372036854775807
%int1_26301 = torch.constant.int 1
%28826 = torch.aten.slice.Tensor %28825, %int1_26298, %int0_26299, %int9223372036854775807_26300, %int1_26301 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28826, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26302 = torch.constant.int 0
%28827 = torch.aten.unsqueeze %28826, %int0_26302 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28827, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26303 = torch.constant.int 2
%28828 = torch.aten.unsqueeze %28827, %int2_26303 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28828, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26304 = torch.constant.int 3
%int0_26305 = torch.constant.int 0
%int9223372036854775807_26306 = torch.constant.int 9223372036854775807
%int1_26307 = torch.constant.int 1
%28829 = torch.aten.slice.Tensor %28828, %int3_26304, %int0_26305, %int9223372036854775807_26306, %int1_26307 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28829, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28830 = torch_c.to_builtin_tensor %28725 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26308 = arith.constant 1 : index
%dim_26309 = tensor.dim %28830, %c1_26308 : tensor<4x?x4x128xf16>
%28831 = flow.tensor.bitcast %28830 : tensor<4x?x4x128xf16>{%dim_26309} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26309}
%28832 = torch_c.from_builtin_tensor %28831 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28833 = torch.aten.mul.Tensor %28832, %28829 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28834 = torch_c.to_builtin_tensor %28833 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26310 = arith.constant 1 : index
%dim_26311 = tensor.dim %28834, %c1_26310 : tensor<4x?x4x64xcomplex<f32>>
%28835 = flow.tensor.bitcast %28834 : tensor<4x?x4x64xcomplex<f32>>{%dim_26311} -> tensor<4x?x4x128xf32>{%dim_26311}
%28836 = torch_c.from_builtin_tensor %28835 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26312 = torch.constant.int 5
%28837 = torch.prims.convert_element_type %28836, %int5_26312 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26313 = torch.constant.int 1
%28838 = torch.aten.size.int %28563, %int1_26313 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26314 = torch.constant.int 0
%28839 = torch.aten.add.int %int0_26314, %28838 : !torch.int, !torch.int -> !torch.int
%int0_26315 = torch.constant.int 0
%int0_26316 = torch.constant.int 0
%int1_26317 = torch.constant.int 1
%28840 = torch.aten.slice.Tensor %28792, %int0_26315, %int0_26316, %28839, %int1_26317 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28840, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26318 = torch.constant.int 1
%int0_26319 = torch.constant.int 0
%int9223372036854775807_26320 = torch.constant.int 9223372036854775807
%int1_26321 = torch.constant.int 1
%28841 = torch.aten.slice.Tensor %28840, %int1_26318, %int0_26319, %int9223372036854775807_26320, %int1_26321 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28841, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26322 = torch.constant.int 0
%28842 = torch.aten.unsqueeze %28841, %int0_26322 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28842, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26323 = torch.constant.int 2
%28843 = torch.aten.unsqueeze %28842, %int2_26323 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28843, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26324 = torch.constant.int 3
%int0_26325 = torch.constant.int 0
%int9223372036854775807_26326 = torch.constant.int 9223372036854775807
%int1_26327 = torch.constant.int 1
%28844 = torch.aten.slice.Tensor %28843, %int3_26324, %int0_26325, %int9223372036854775807_26326, %int1_26327 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28844, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28845 = torch_c.to_builtin_tensor %28727 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26328 = arith.constant 1 : index
%dim_26329 = tensor.dim %28845, %c1_26328 : tensor<4x?x4x128xf16>
%28846 = flow.tensor.bitcast %28845 : tensor<4x?x4x128xf16>{%dim_26329} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26329}
%28847 = torch_c.from_builtin_tensor %28846 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28848 = torch.aten.mul.Tensor %28847, %28844 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28849 = torch_c.to_builtin_tensor %28848 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26330 = arith.constant 1 : index
%dim_26331 = tensor.dim %28849, %c1_26330 : tensor<4x?x4x64xcomplex<f32>>
%28850 = flow.tensor.bitcast %28849 : tensor<4x?x4x64xcomplex<f32>>{%dim_26331} -> tensor<4x?x4x128xf32>{%dim_26331}
%28851 = torch_c.from_builtin_tensor %28850 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26332 = torch.constant.int 5
%28852 = torch.prims.convert_element_type %28851, %int5_26332 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26333 = torch.constant.int 1
%28853 = torch.aten.size.int %28569, %int1_26333 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26334 = torch.constant.int 0
%28854 = torch.aten.add.int %int0_26334, %28853 : !torch.int, !torch.int -> !torch.int
%int0_26335 = torch.constant.int 0
%int0_26336 = torch.constant.int 0
%int1_26337 = torch.constant.int 1
%28855 = torch.aten.slice.Tensor %28795, %int0_26335, %int0_26336, %28854, %int1_26337 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28855, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26338 = torch.constant.int 1
%int0_26339 = torch.constant.int 0
%int9223372036854775807_26340 = torch.constant.int 9223372036854775807
%int1_26341 = torch.constant.int 1
%28856 = torch.aten.slice.Tensor %28855, %int1_26338, %int0_26339, %int9223372036854775807_26340, %int1_26341 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28856, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26342 = torch.constant.int 0
%28857 = torch.aten.unsqueeze %28856, %int0_26342 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28857, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26343 = torch.constant.int 2
%28858 = torch.aten.unsqueeze %28857, %int2_26343 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28858, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26344 = torch.constant.int 3
%int0_26345 = torch.constant.int 0
%int9223372036854775807_26346 = torch.constant.int 9223372036854775807
%int1_26347 = torch.constant.int 1
%28859 = torch.aten.slice.Tensor %28858, %int3_26344, %int0_26345, %int9223372036854775807_26346, %int1_26347 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28859, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28860 = torch_c.to_builtin_tensor %28729 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26348 = arith.constant 1 : index
%dim_26349 = tensor.dim %28860, %c1_26348 : tensor<4x?x4x128xf16>
%28861 = flow.tensor.bitcast %28860 : tensor<4x?x4x128xf16>{%dim_26349} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26349}
%28862 = torch_c.from_builtin_tensor %28861 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28863 = torch.aten.mul.Tensor %28862, %28859 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28864 = torch_c.to_builtin_tensor %28863 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26350 = arith.constant 1 : index
%dim_26351 = tensor.dim %28864, %c1_26350 : tensor<4x?x4x64xcomplex<f32>>
%28865 = flow.tensor.bitcast %28864 : tensor<4x?x4x64xcomplex<f32>>{%dim_26351} -> tensor<4x?x4x128xf32>{%dim_26351}
%28866 = torch_c.from_builtin_tensor %28865 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26352 = torch.constant.int 5
%28867 = torch.prims.convert_element_type %28866, %int5_26352 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26353 = torch.constant.int 1
%28868 = torch.aten.size.int %28575, %int1_26353 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26354 = torch.constant.int 0
%28869 = torch.aten.add.int %int0_26354, %28868 : !torch.int, !torch.int -> !torch.int
%int0_26355 = torch.constant.int 0
%int0_26356 = torch.constant.int 0
%int1_26357 = torch.constant.int 1
%28870 = torch.aten.slice.Tensor %28798, %int0_26355, %int0_26356, %28869, %int1_26357 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28870, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26358 = torch.constant.int 1
%int0_26359 = torch.constant.int 0
%int9223372036854775807_26360 = torch.constant.int 9223372036854775807
%int1_26361 = torch.constant.int 1
%28871 = torch.aten.slice.Tensor %28870, %int1_26358, %int0_26359, %int9223372036854775807_26360, %int1_26361 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28871, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26362 = torch.constant.int 0
%28872 = torch.aten.unsqueeze %28871, %int0_26362 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28872, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26363 = torch.constant.int 2
%28873 = torch.aten.unsqueeze %28872, %int2_26363 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28873, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26364 = torch.constant.int 3
%int0_26365 = torch.constant.int 0
%int9223372036854775807_26366 = torch.constant.int 9223372036854775807
%int1_26367 = torch.constant.int 1
%28874 = torch.aten.slice.Tensor %28873, %int3_26364, %int0_26365, %int9223372036854775807_26366, %int1_26367 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28874, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28875 = torch_c.to_builtin_tensor %28731 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26368 = arith.constant 1 : index
%dim_26369 = tensor.dim %28875, %c1_26368 : tensor<4x?x4x128xf16>
%28876 = flow.tensor.bitcast %28875 : tensor<4x?x4x128xf16>{%dim_26369} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26369}
%28877 = torch_c.from_builtin_tensor %28876 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28878 = torch.aten.mul.Tensor %28877, %28874 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28879 = torch_c.to_builtin_tensor %28878 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26370 = arith.constant 1 : index
%dim_26371 = tensor.dim %28879, %c1_26370 : tensor<4x?x4x64xcomplex<f32>>
%28880 = flow.tensor.bitcast %28879 : tensor<4x?x4x64xcomplex<f32>>{%dim_26371} -> tensor<4x?x4x128xf32>{%dim_26371}
%28881 = torch_c.from_builtin_tensor %28880 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26372 = torch.constant.int 5
%28882 = torch.prims.convert_element_type %28881, %int5_26372 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26373 = torch.constant.int 1
%28883 = torch.aten.size.int %28581, %int1_26373 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26374 = torch.constant.int 0
%28884 = torch.aten.add.int %int0_26374, %28883 : !torch.int, !torch.int -> !torch.int
%int0_26375 = torch.constant.int 0
%int0_26376 = torch.constant.int 0
%int1_26377 = torch.constant.int 1
%28885 = torch.aten.slice.Tensor %28801, %int0_26375, %int0_26376, %28884, %int1_26377 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28885, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26378 = torch.constant.int 1
%int0_26379 = torch.constant.int 0
%int9223372036854775807_26380 = torch.constant.int 9223372036854775807
%int1_26381 = torch.constant.int 1
%28886 = torch.aten.slice.Tensor %28885, %int1_26378, %int0_26379, %int9223372036854775807_26380, %int1_26381 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28886, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26382 = torch.constant.int 0
%28887 = torch.aten.unsqueeze %28886, %int0_26382 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28887, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26383 = torch.constant.int 2
%28888 = torch.aten.unsqueeze %28887, %int2_26383 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28888, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26384 = torch.constant.int 3
%int0_26385 = torch.constant.int 0
%int9223372036854775807_26386 = torch.constant.int 9223372036854775807
%int1_26387 = torch.constant.int 1
%28889 = torch.aten.slice.Tensor %28888, %int3_26384, %int0_26385, %int9223372036854775807_26386, %int1_26387 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28889, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28890 = torch_c.to_builtin_tensor %28733 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26388 = arith.constant 1 : index
%dim_26389 = tensor.dim %28890, %c1_26388 : tensor<4x?x4x128xf16>
%28891 = flow.tensor.bitcast %28890 : tensor<4x?x4x128xf16>{%dim_26389} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26389}
%28892 = torch_c.from_builtin_tensor %28891 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28893 = torch.aten.mul.Tensor %28892, %28889 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28894 = torch_c.to_builtin_tensor %28893 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26390 = arith.constant 1 : index
%dim_26391 = tensor.dim %28894, %c1_26390 : tensor<4x?x4x64xcomplex<f32>>
%28895 = flow.tensor.bitcast %28894 : tensor<4x?x4x64xcomplex<f32>>{%dim_26391} -> tensor<4x?x4x128xf32>{%dim_26391}
%28896 = torch_c.from_builtin_tensor %28895 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26392 = torch.constant.int 5
%28897 = torch.prims.convert_element_type %28896, %int5_26392 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26393 = torch.constant.int 1
%28898 = torch.aten.size.int %28587, %int1_26393 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26394 = torch.constant.int 0
%28899 = torch.aten.add.int %int0_26394, %28898 : !torch.int, !torch.int -> !torch.int
%int0_26395 = torch.constant.int 0
%int0_26396 = torch.constant.int 0
%int1_26397 = torch.constant.int 1
%28900 = torch.aten.slice.Tensor %28804, %int0_26395, %int0_26396, %28899, %int1_26397 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28900, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26398 = torch.constant.int 1
%int0_26399 = torch.constant.int 0
%int9223372036854775807_26400 = torch.constant.int 9223372036854775807
%int1_26401 = torch.constant.int 1
%28901 = torch.aten.slice.Tensor %28900, %int1_26398, %int0_26399, %int9223372036854775807_26400, %int1_26401 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28901, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26402 = torch.constant.int 0
%28902 = torch.aten.unsqueeze %28901, %int0_26402 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28902, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26403 = torch.constant.int 2
%28903 = torch.aten.unsqueeze %28902, %int2_26403 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28903, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26404 = torch.constant.int 3
%int0_26405 = torch.constant.int 0
%int9223372036854775807_26406 = torch.constant.int 9223372036854775807
%int1_26407 = torch.constant.int 1
%28904 = torch.aten.slice.Tensor %28903, %int3_26404, %int0_26405, %int9223372036854775807_26406, %int1_26407 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28904, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28905 = torch_c.to_builtin_tensor %28735 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26408 = arith.constant 1 : index
%dim_26409 = tensor.dim %28905, %c1_26408 : tensor<4x?x4x128xf16>
%28906 = flow.tensor.bitcast %28905 : tensor<4x?x4x128xf16>{%dim_26409} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26409}
%28907 = torch_c.from_builtin_tensor %28906 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28908 = torch.aten.mul.Tensor %28907, %28904 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28909 = torch_c.to_builtin_tensor %28908 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26410 = arith.constant 1 : index
%dim_26411 = tensor.dim %28909, %c1_26410 : tensor<4x?x4x64xcomplex<f32>>
%28910 = flow.tensor.bitcast %28909 : tensor<4x?x4x64xcomplex<f32>>{%dim_26411} -> tensor<4x?x4x128xf32>{%dim_26411}
%28911 = torch_c.from_builtin_tensor %28910 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26412 = torch.constant.int 5
%28912 = torch.prims.convert_element_type %28911, %int5_26412 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_26413 = torch.constant.int 1
%28913 = torch.aten.size.int %28593, %int1_26413 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_26414 = torch.constant.int 0
%28914 = torch.aten.add.int %int0_26414, %28913 : !torch.int, !torch.int -> !torch.int
%int0_26415 = torch.constant.int 0
%int0_26416 = torch.constant.int 0
%int1_26417 = torch.constant.int 1
%28915 = torch.aten.slice.Tensor %28807, %int0_26415, %int0_26416, %28914, %int1_26417 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28915, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26418 = torch.constant.int 1
%int0_26419 = torch.constant.int 0
%int9223372036854775807_26420 = torch.constant.int 9223372036854775807
%int1_26421 = torch.constant.int 1
%28916 = torch.aten.slice.Tensor %28915, %int1_26418, %int0_26419, %int9223372036854775807_26420, %int1_26421 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28916, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26422 = torch.constant.int 0
%28917 = torch.aten.unsqueeze %28916, %int0_26422 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28917, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26423 = torch.constant.int 2
%28918 = torch.aten.unsqueeze %28917, %int2_26423 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28918, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26424 = torch.constant.int 3
%int0_26425 = torch.constant.int 0
%int9223372036854775807_26426 = torch.constant.int 9223372036854775807
%int1_26427 = torch.constant.int 1
%28919 = torch.aten.slice.Tensor %28918, %int3_26424, %int0_26425, %int9223372036854775807_26426, %int1_26427 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28919, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28920 = torch_c.to_builtin_tensor %28737 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_26428 = arith.constant 1 : index
%dim_26429 = tensor.dim %28920, %c1_26428 : tensor<4x?x4x128xf16>
%28921 = flow.tensor.bitcast %28920 : tensor<4x?x4x128xf16>{%dim_26429} -> tensor<4x?x4x64xcomplex<f16>>{%dim_26429}
%28922 = torch_c.from_builtin_tensor %28921 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %28922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%28923 = torch.aten.mul.Tensor %28922, %28919 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %28923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%28924 = torch_c.to_builtin_tensor %28923 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_26430 = arith.constant 1 : index
%dim_26431 = tensor.dim %28924, %c1_26430 : tensor<4x?x4x64xcomplex<f32>>
%28925 = flow.tensor.bitcast %28924 : tensor<4x?x4x64xcomplex<f32>>{%dim_26431} -> tensor<4x?x4x128xf32>{%dim_26431}
%28926 = torch_c.from_builtin_tensor %28925 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %28926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_26432 = torch.constant.int 5
%28927 = torch.prims.convert_element_type %28926, %int5_26432 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %28927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_26433 = torch.constant.int 131072
%none_26434 = torch.constant.none
%none_26435 = torch.constant.none
%cpu_26436 = torch.constant.device "cpu"
%false_26437 = torch.constant.bool false
%28928 = torch.aten.arange %int131072_26433, %none_26434, %none_26435, %cpu_26436, %false_26437 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_26438 = torch.constant.int 0
%int128_26439 = torch.constant.int 128
%int2_26440 = torch.constant.int 2
%none_26441 = torch.constant.none
%none_26442 = torch.constant.none
%cpu_26443 = torch.constant.device "cpu"
%false_26444 = torch.constant.bool false
%28929 = torch.aten.arange.start_step %int0_26438, %int128_26439, %int2_26440, %none_26441, %none_26442, %cpu_26443, %false_26444 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_26445 = torch.constant.int 0
%int0_26446 = torch.constant.int 0
%int64_26447 = torch.constant.int 64
%int1_26448 = torch.constant.int 1
%28930 = torch.aten.slice.Tensor %28929, %int0_26445, %int0_26446, %int64_26447, %int1_26448 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_26449 = torch.constant.int 6
%28931 = torch.prims.convert_element_type %28930, %int6_26449 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_26450 = torch.constant.int 128
%28932 = torch.aten.div.Scalar %28931, %int128_26450 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_26451 = torch.constant.float 5.000000e+05
%28933 = torch.aten.pow.Scalar %float5.000000e05_26451, %28932 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%28934 = torch.aten.reciprocal %28933 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_26452 = torch.constant.float 1.000000e+00
%28935 = torch.aten.mul.Scalar %28934, %float1.000000e00_26452 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_26453 = torch.constant.int 131072
%int1_26454 = torch.constant.int 1
%28936 = torch.prim.ListConstruct %int131072_26453, %int1_26454 : (!torch.int, !torch.int) -> !torch.list<int>
%28937 = torch.aten.view %28928, %28936 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%28938 = torch.aten.mul.Tensor %28937, %28935 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%28939 = torch.aten.cos %28938 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%28940 = torch.aten.sin %28938 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%28941 = torch.aten.complex %28939, %28940 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%28942 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28943 = flow.tensor.transfer %28942 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%28944 = torch_c.from_builtin_tensor %28943 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28945 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28946 = flow.tensor.transfer %28945 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%28947 = torch_c.from_builtin_tensor %28946 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28948 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28949 = flow.tensor.transfer %28948 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%28950 = torch_c.from_builtin_tensor %28949 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28951 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28952 = flow.tensor.transfer %28951 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%28953 = torch_c.from_builtin_tensor %28952 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28954 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28955 = flow.tensor.transfer %28954 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%28956 = torch_c.from_builtin_tensor %28955 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28957 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28958 = flow.tensor.transfer %28957 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%28959 = torch_c.from_builtin_tensor %28958 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28960 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28961 = flow.tensor.transfer %28960 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%28962 = torch_c.from_builtin_tensor %28961 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%28963 = torch_c.to_builtin_tensor %28941 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%28964 = flow.tensor.transfer %28963 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%28965 = torch_c.from_builtin_tensor %28964 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_26455 = torch.constant.int 1
%28966 = torch.aten.size.int %28615, %int1_26455 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26456 = torch.constant.int 0
%28967 = torch.aten.add.int %int0_26456, %28966 : !torch.int, !torch.int -> !torch.int
%int0_26457 = torch.constant.int 0
%int0_26458 = torch.constant.int 0
%int1_26459 = torch.constant.int 1
%28968 = torch.aten.slice.Tensor %28944, %int0_26457, %int0_26458, %28967, %int1_26459 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28968, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26460 = torch.constant.int 1
%int0_26461 = torch.constant.int 0
%int9223372036854775807_26462 = torch.constant.int 9223372036854775807
%int1_26463 = torch.constant.int 1
%28969 = torch.aten.slice.Tensor %28968, %int1_26460, %int0_26461, %int9223372036854775807_26462, %int1_26463 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28969, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26464 = torch.constant.int 0
%28970 = torch.aten.unsqueeze %28969, %int0_26464 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28970, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26465 = torch.constant.int 2
%28971 = torch.aten.unsqueeze %28970, %int2_26465 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28971, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26466 = torch.constant.int 3
%int0_26467 = torch.constant.int 0
%int9223372036854775807_26468 = torch.constant.int 9223372036854775807
%int1_26469 = torch.constant.int 1
%28972 = torch.aten.slice.Tensor %28971, %int3_26466, %int0_26467, %int9223372036854775807_26468, %int1_26469 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28972, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28973 = torch_c.to_builtin_tensor %28739 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26470 = arith.constant 1 : index
%dim_26471 = tensor.dim %28973, %c1_26470 : tensor<4x?x1x128xf16>
%28974 = flow.tensor.bitcast %28973 : tensor<4x?x1x128xf16>{%dim_26471} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26471}
%28975 = torch_c.from_builtin_tensor %28974 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %28975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%28976 = torch.aten.mul.Tensor %28975, %28972 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%28977 = torch_c.to_builtin_tensor %28976 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26472 = arith.constant 1 : index
%dim_26473 = tensor.dim %28977, %c1_26472 : tensor<4x?x1x64xcomplex<f32>>
%28978 = flow.tensor.bitcast %28977 : tensor<4x?x1x64xcomplex<f32>>{%dim_26473} -> tensor<4x?x1x128xf32>{%dim_26473}
%28979 = torch_c.from_builtin_tensor %28978 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %28979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26474 = torch.constant.int 5
%28980 = torch.prims.convert_element_type %28979, %int5_26474 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26475 = torch.constant.int 1
%28981 = torch.aten.size.int %28621, %int1_26475 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26476 = torch.constant.int 0
%28982 = torch.aten.add.int %int0_26476, %28981 : !torch.int, !torch.int -> !torch.int
%int0_26477 = torch.constant.int 0
%int0_26478 = torch.constant.int 0
%int1_26479 = torch.constant.int 1
%28983 = torch.aten.slice.Tensor %28947, %int0_26477, %int0_26478, %28982, %int1_26479 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28983, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26480 = torch.constant.int 1
%int0_26481 = torch.constant.int 0
%int9223372036854775807_26482 = torch.constant.int 9223372036854775807
%int1_26483 = torch.constant.int 1
%28984 = torch.aten.slice.Tensor %28983, %int1_26480, %int0_26481, %int9223372036854775807_26482, %int1_26483 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28984, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26484 = torch.constant.int 0
%28985 = torch.aten.unsqueeze %28984, %int0_26484 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %28985, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26485 = torch.constant.int 2
%28986 = torch.aten.unsqueeze %28985, %int2_26485 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28986, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26486 = torch.constant.int 3
%int0_26487 = torch.constant.int 0
%int9223372036854775807_26488 = torch.constant.int 9223372036854775807
%int1_26489 = torch.constant.int 1
%28987 = torch.aten.slice.Tensor %28986, %int3_26486, %int0_26487, %int9223372036854775807_26488, %int1_26489 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28987, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%28988 = torch_c.to_builtin_tensor %28741 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26490 = arith.constant 1 : index
%dim_26491 = tensor.dim %28988, %c1_26490 : tensor<4x?x1x128xf16>
%28989 = flow.tensor.bitcast %28988 : tensor<4x?x1x128xf16>{%dim_26491} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26491}
%28990 = torch_c.from_builtin_tensor %28989 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %28990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%28991 = torch.aten.mul.Tensor %28990, %28987 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %28991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%28992 = torch_c.to_builtin_tensor %28991 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26492 = arith.constant 1 : index
%dim_26493 = tensor.dim %28992, %c1_26492 : tensor<4x?x1x64xcomplex<f32>>
%28993 = flow.tensor.bitcast %28992 : tensor<4x?x1x64xcomplex<f32>>{%dim_26493} -> tensor<4x?x1x128xf32>{%dim_26493}
%28994 = torch_c.from_builtin_tensor %28993 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %28994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26494 = torch.constant.int 5
%28995 = torch.prims.convert_element_type %28994, %int5_26494 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %28995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26495 = torch.constant.int 1
%28996 = torch.aten.size.int %28627, %int1_26495 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26496 = torch.constant.int 0
%28997 = torch.aten.add.int %int0_26496, %28996 : !torch.int, !torch.int -> !torch.int
%int0_26497 = torch.constant.int 0
%int0_26498 = torch.constant.int 0
%int1_26499 = torch.constant.int 1
%28998 = torch.aten.slice.Tensor %28950, %int0_26497, %int0_26498, %28997, %int1_26499 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28998, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26500 = torch.constant.int 1
%int0_26501 = torch.constant.int 0
%int9223372036854775807_26502 = torch.constant.int 9223372036854775807
%int1_26503 = torch.constant.int 1
%28999 = torch.aten.slice.Tensor %28998, %int1_26500, %int0_26501, %int9223372036854775807_26502, %int1_26503 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %28999, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26504 = torch.constant.int 0
%29000 = torch.aten.unsqueeze %28999, %int0_26504 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %29000, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26505 = torch.constant.int 2
%29001 = torch.aten.unsqueeze %29000, %int2_26505 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29001, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26506 = torch.constant.int 3
%int0_26507 = torch.constant.int 0
%int9223372036854775807_26508 = torch.constant.int 9223372036854775807
%int1_26509 = torch.constant.int 1
%29002 = torch.aten.slice.Tensor %29001, %int3_26506, %int0_26507, %int9223372036854775807_26508, %int1_26509 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29002, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%29003 = torch_c.to_builtin_tensor %28743 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26510 = arith.constant 1 : index
%dim_26511 = tensor.dim %29003, %c1_26510 : tensor<4x?x1x128xf16>
%29004 = flow.tensor.bitcast %29003 : tensor<4x?x1x128xf16>{%dim_26511} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26511}
%29005 = torch_c.from_builtin_tensor %29004 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %29005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%29006 = torch.aten.mul.Tensor %29005, %29002 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%29007 = torch_c.to_builtin_tensor %29006 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26512 = arith.constant 1 : index
%dim_26513 = tensor.dim %29007, %c1_26512 : tensor<4x?x1x64xcomplex<f32>>
%29008 = flow.tensor.bitcast %29007 : tensor<4x?x1x64xcomplex<f32>>{%dim_26513} -> tensor<4x?x1x128xf32>{%dim_26513}
%29009 = torch_c.from_builtin_tensor %29008 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %29009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26514 = torch.constant.int 5
%29010 = torch.prims.convert_element_type %29009, %int5_26514 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %29010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26515 = torch.constant.int 1
%29011 = torch.aten.size.int %28633, %int1_26515 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26516 = torch.constant.int 0
%29012 = torch.aten.add.int %int0_26516, %29011 : !torch.int, !torch.int -> !torch.int
%int0_26517 = torch.constant.int 0
%int0_26518 = torch.constant.int 0
%int1_26519 = torch.constant.int 1
%29013 = torch.aten.slice.Tensor %28953, %int0_26517, %int0_26518, %29012, %int1_26519 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29013, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26520 = torch.constant.int 1
%int0_26521 = torch.constant.int 0
%int9223372036854775807_26522 = torch.constant.int 9223372036854775807
%int1_26523 = torch.constant.int 1
%29014 = torch.aten.slice.Tensor %29013, %int1_26520, %int0_26521, %int9223372036854775807_26522, %int1_26523 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29014, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26524 = torch.constant.int 0
%29015 = torch.aten.unsqueeze %29014, %int0_26524 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %29015, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26525 = torch.constant.int 2
%29016 = torch.aten.unsqueeze %29015, %int2_26525 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29016, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26526 = torch.constant.int 3
%int0_26527 = torch.constant.int 0
%int9223372036854775807_26528 = torch.constant.int 9223372036854775807
%int1_26529 = torch.constant.int 1
%29017 = torch.aten.slice.Tensor %29016, %int3_26526, %int0_26527, %int9223372036854775807_26528, %int1_26529 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29017, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%29018 = torch_c.to_builtin_tensor %28745 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26530 = arith.constant 1 : index
%dim_26531 = tensor.dim %29018, %c1_26530 : tensor<4x?x1x128xf16>
%29019 = flow.tensor.bitcast %29018 : tensor<4x?x1x128xf16>{%dim_26531} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26531}
%29020 = torch_c.from_builtin_tensor %29019 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %29020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%29021 = torch.aten.mul.Tensor %29020, %29017 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%29022 = torch_c.to_builtin_tensor %29021 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26532 = arith.constant 1 : index
%dim_26533 = tensor.dim %29022, %c1_26532 : tensor<4x?x1x64xcomplex<f32>>
%29023 = flow.tensor.bitcast %29022 : tensor<4x?x1x64xcomplex<f32>>{%dim_26533} -> tensor<4x?x1x128xf32>{%dim_26533}
%29024 = torch_c.from_builtin_tensor %29023 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %29024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26534 = torch.constant.int 5
%29025 = torch.prims.convert_element_type %29024, %int5_26534 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %29025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26535 = torch.constant.int 1
%29026 = torch.aten.size.int %28639, %int1_26535 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26536 = torch.constant.int 0
%29027 = torch.aten.add.int %int0_26536, %29026 : !torch.int, !torch.int -> !torch.int
%int0_26537 = torch.constant.int 0
%int0_26538 = torch.constant.int 0
%int1_26539 = torch.constant.int 1
%29028 = torch.aten.slice.Tensor %28956, %int0_26537, %int0_26538, %29027, %int1_26539 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29028, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26540 = torch.constant.int 1
%int0_26541 = torch.constant.int 0
%int9223372036854775807_26542 = torch.constant.int 9223372036854775807
%int1_26543 = torch.constant.int 1
%29029 = torch.aten.slice.Tensor %29028, %int1_26540, %int0_26541, %int9223372036854775807_26542, %int1_26543 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29029, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26544 = torch.constant.int 0
%29030 = torch.aten.unsqueeze %29029, %int0_26544 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %29030, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26545 = torch.constant.int 2
%29031 = torch.aten.unsqueeze %29030, %int2_26545 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29031, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26546 = torch.constant.int 3
%int0_26547 = torch.constant.int 0
%int9223372036854775807_26548 = torch.constant.int 9223372036854775807
%int1_26549 = torch.constant.int 1
%29032 = torch.aten.slice.Tensor %29031, %int3_26546, %int0_26547, %int9223372036854775807_26548, %int1_26549 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29032, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%29033 = torch_c.to_builtin_tensor %28747 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26550 = arith.constant 1 : index
%dim_26551 = tensor.dim %29033, %c1_26550 : tensor<4x?x1x128xf16>
%29034 = flow.tensor.bitcast %29033 : tensor<4x?x1x128xf16>{%dim_26551} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26551}
%29035 = torch_c.from_builtin_tensor %29034 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %29035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%29036 = torch.aten.mul.Tensor %29035, %29032 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%29037 = torch_c.to_builtin_tensor %29036 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26552 = arith.constant 1 : index
%dim_26553 = tensor.dim %29037, %c1_26552 : tensor<4x?x1x64xcomplex<f32>>
%29038 = flow.tensor.bitcast %29037 : tensor<4x?x1x64xcomplex<f32>>{%dim_26553} -> tensor<4x?x1x128xf32>{%dim_26553}
%29039 = torch_c.from_builtin_tensor %29038 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %29039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26554 = torch.constant.int 5
%29040 = torch.prims.convert_element_type %29039, %int5_26554 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %29040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26555 = torch.constant.int 1
%29041 = torch.aten.size.int %28645, %int1_26555 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26556 = torch.constant.int 0
%29042 = torch.aten.add.int %int0_26556, %29041 : !torch.int, !torch.int -> !torch.int
%int0_26557 = torch.constant.int 0
%int0_26558 = torch.constant.int 0
%int1_26559 = torch.constant.int 1
%29043 = torch.aten.slice.Tensor %28959, %int0_26557, %int0_26558, %29042, %int1_26559 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29043, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26560 = torch.constant.int 1
%int0_26561 = torch.constant.int 0
%int9223372036854775807_26562 = torch.constant.int 9223372036854775807
%int1_26563 = torch.constant.int 1
%29044 = torch.aten.slice.Tensor %29043, %int1_26560, %int0_26561, %int9223372036854775807_26562, %int1_26563 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29044, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26564 = torch.constant.int 0
%29045 = torch.aten.unsqueeze %29044, %int0_26564 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %29045, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26565 = torch.constant.int 2
%29046 = torch.aten.unsqueeze %29045, %int2_26565 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29046, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26566 = torch.constant.int 3
%int0_26567 = torch.constant.int 0
%int9223372036854775807_26568 = torch.constant.int 9223372036854775807
%int1_26569 = torch.constant.int 1
%29047 = torch.aten.slice.Tensor %29046, %int3_26566, %int0_26567, %int9223372036854775807_26568, %int1_26569 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29047, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%29048 = torch_c.to_builtin_tensor %28749 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26570 = arith.constant 1 : index
%dim_26571 = tensor.dim %29048, %c1_26570 : tensor<4x?x1x128xf16>
%29049 = flow.tensor.bitcast %29048 : tensor<4x?x1x128xf16>{%dim_26571} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26571}
%29050 = torch_c.from_builtin_tensor %29049 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %29050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%29051 = torch.aten.mul.Tensor %29050, %29047 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%29052 = torch_c.to_builtin_tensor %29051 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26572 = arith.constant 1 : index
%dim_26573 = tensor.dim %29052, %c1_26572 : tensor<4x?x1x64xcomplex<f32>>
%29053 = flow.tensor.bitcast %29052 : tensor<4x?x1x64xcomplex<f32>>{%dim_26573} -> tensor<4x?x1x128xf32>{%dim_26573}
%29054 = torch_c.from_builtin_tensor %29053 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %29054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26574 = torch.constant.int 5
%29055 = torch.prims.convert_element_type %29054, %int5_26574 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %29055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26575 = torch.constant.int 1
%29056 = torch.aten.size.int %28651, %int1_26575 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26576 = torch.constant.int 0
%29057 = torch.aten.add.int %int0_26576, %29056 : !torch.int, !torch.int -> !torch.int
%int0_26577 = torch.constant.int 0
%int0_26578 = torch.constant.int 0
%int1_26579 = torch.constant.int 1
%29058 = torch.aten.slice.Tensor %28962, %int0_26577, %int0_26578, %29057, %int1_26579 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29058, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26580 = torch.constant.int 1
%int0_26581 = torch.constant.int 0
%int9223372036854775807_26582 = torch.constant.int 9223372036854775807
%int1_26583 = torch.constant.int 1
%29059 = torch.aten.slice.Tensor %29058, %int1_26580, %int0_26581, %int9223372036854775807_26582, %int1_26583 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29059, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26584 = torch.constant.int 0
%29060 = torch.aten.unsqueeze %29059, %int0_26584 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %29060, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26585 = torch.constant.int 2
%29061 = torch.aten.unsqueeze %29060, %int2_26585 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29061, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26586 = torch.constant.int 3
%int0_26587 = torch.constant.int 0
%int9223372036854775807_26588 = torch.constant.int 9223372036854775807
%int1_26589 = torch.constant.int 1
%29062 = torch.aten.slice.Tensor %29061, %int3_26586, %int0_26587, %int9223372036854775807_26588, %int1_26589 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29062, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%29063 = torch_c.to_builtin_tensor %28751 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26590 = arith.constant 1 : index
%dim_26591 = tensor.dim %29063, %c1_26590 : tensor<4x?x1x128xf16>
%29064 = flow.tensor.bitcast %29063 : tensor<4x?x1x128xf16>{%dim_26591} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26591}
%29065 = torch_c.from_builtin_tensor %29064 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %29065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%29066 = torch.aten.mul.Tensor %29065, %29062 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%29067 = torch_c.to_builtin_tensor %29066 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26592 = arith.constant 1 : index
%dim_26593 = tensor.dim %29067, %c1_26592 : tensor<4x?x1x64xcomplex<f32>>
%29068 = flow.tensor.bitcast %29067 : tensor<4x?x1x64xcomplex<f32>>{%dim_26593} -> tensor<4x?x1x128xf32>{%dim_26593}
%29069 = torch_c.from_builtin_tensor %29068 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %29069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26594 = torch.constant.int 5
%29070 = torch.prims.convert_element_type %29069, %int5_26594 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %29070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_26595 = torch.constant.int 1
%29071 = torch.aten.size.int %28657, %int1_26595 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_26596 = torch.constant.int 0
%29072 = torch.aten.add.int %int0_26596, %29071 : !torch.int, !torch.int -> !torch.int
%int0_26597 = torch.constant.int 0
%int0_26598 = torch.constant.int 0
%int1_26599 = torch.constant.int 1
%29073 = torch.aten.slice.Tensor %28965, %int0_26597, %int0_26598, %29072, %int1_26599 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29073, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_26600 = torch.constant.int 1
%int0_26601 = torch.constant.int 0
%int9223372036854775807_26602 = torch.constant.int 9223372036854775807
%int1_26603 = torch.constant.int 1
%29074 = torch.aten.slice.Tensor %29073, %int1_26600, %int0_26601, %int9223372036854775807_26602, %int1_26603 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %29074, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_26604 = torch.constant.int 0
%29075 = torch.aten.unsqueeze %29074, %int0_26604 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %29075, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_26605 = torch.constant.int 2
%29076 = torch.aten.unsqueeze %29075, %int2_26605 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29076, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_26606 = torch.constant.int 3
%int0_26607 = torch.constant.int 0
%int9223372036854775807_26608 = torch.constant.int 9223372036854775807
%int1_26609 = torch.constant.int 1
%29077 = torch.aten.slice.Tensor %29076, %int3_26606, %int0_26607, %int9223372036854775807_26608, %int1_26609 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29077, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%29078 = torch_c.to_builtin_tensor %28753 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_26610 = arith.constant 1 : index
%dim_26611 = tensor.dim %29078, %c1_26610 : tensor<4x?x1x128xf16>
%29079 = flow.tensor.bitcast %29078 : tensor<4x?x1x128xf16>{%dim_26611} -> tensor<4x?x1x64xcomplex<f16>>{%dim_26611}
%29080 = torch_c.from_builtin_tensor %29079 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %29080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%29081 = torch.aten.mul.Tensor %29080, %29077 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %29081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%29082 = torch_c.to_builtin_tensor %29081 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_26612 = arith.constant 1 : index
%dim_26613 = tensor.dim %29082, %c1_26612 : tensor<4x?x1x64xcomplex<f32>>
%29083 = flow.tensor.bitcast %29082 : tensor<4x?x1x64xcomplex<f32>>{%dim_26613} -> tensor<4x?x1x128xf32>{%dim_26613}
%29084 = torch_c.from_builtin_tensor %29083 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %29084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_26614 = torch.constant.int 5
%29085 = torch.prims.convert_element_type %29084, %int5_26614 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %29085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_26615 = torch.constant.int 64
%29086 = torch.aten.mul.Scalar %2364, %int64_26615 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29086, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26616 = torch.constant.int 64
%29087 = torch.aten.mul.Scalar %2367, %int64_26616 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29087, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26617 = torch.constant.int 64
%29088 = torch.aten.mul.Scalar %2370, %int64_26617 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29088, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26618 = torch.constant.int 64
%29089 = torch.aten.mul.Scalar %2373, %int64_26618 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29089, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26619 = torch.constant.int 64
%29090 = torch.aten.mul.Scalar %2376, %int64_26619 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29090, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26620 = torch.constant.int 64
%29091 = torch.aten.mul.Scalar %2379, %int64_26620 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29091, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26621 = torch.constant.int 64
%29092 = torch.aten.mul.Scalar %2382, %int64_26621 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29092, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_26622 = torch.constant.int 64
%29093 = torch.aten.mul.Scalar %2385, %int64_26622 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29093, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28 = torch.constant.int 28
%int1_26623 = torch.constant.int 1
%29094 = torch.aten.add.Scalar %29086, %int28, %int1_26623 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29094, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26624 = torch.constant.int 28
%int1_26625 = torch.constant.int 1
%29095 = torch.aten.add.Scalar %29087, %int28_26624, %int1_26625 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29095, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26626 = torch.constant.int 28
%int1_26627 = torch.constant.int 1
%29096 = torch.aten.add.Scalar %29088, %int28_26626, %int1_26627 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29096, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26628 = torch.constant.int 28
%int1_26629 = torch.constant.int 1
%29097 = torch.aten.add.Scalar %29089, %int28_26628, %int1_26629 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29097, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26630 = torch.constant.int 28
%int1_26631 = torch.constant.int 1
%29098 = torch.aten.add.Scalar %29090, %int28_26630, %int1_26631 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29098, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26632 = torch.constant.int 28
%int1_26633 = torch.constant.int 1
%29099 = torch.aten.add.Scalar %29091, %int28_26632, %int1_26633 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29099, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26634 = torch.constant.int 28
%int1_26635 = torch.constant.int 1
%29100 = torch.aten.add.Scalar %29092, %int28_26634, %int1_26635 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29100, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int28_26636 = torch.constant.int 28
%int1_26637 = torch.constant.int 1
%29101 = torch.aten.add.Scalar %29093, %int28_26636, %int1_26637 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29101, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_26638 = torch.constant.int 4
%int16_26639 = torch.constant.int 16
%int1_26640 = torch.constant.int 1
%int128_26641 = torch.constant.int 128
%29102 = torch.prim.ListConstruct %int4_26638, %3095, %int16_26639, %int1_26640, %int128_26641 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29103 = torch.aten.view %28980, %29102 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29103, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26642 = torch.constant.int 4
%int16_26643 = torch.constant.int 16
%int1_26644 = torch.constant.int 1
%int128_26645 = torch.constant.int 128
%29104 = torch.prim.ListConstruct %int4_26642, %3095, %int16_26643, %int1_26644, %int128_26645 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29105 = torch.aten.view %28995, %29104 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29105, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26646 = torch.constant.int 4
%int16_26647 = torch.constant.int 16
%int1_26648 = torch.constant.int 1
%int128_26649 = torch.constant.int 128
%29106 = torch.prim.ListConstruct %int4_26646, %3095, %int16_26647, %int1_26648, %int128_26649 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29107 = torch.aten.view %29010, %29106 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29107, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26650 = torch.constant.int 4
%int16_26651 = torch.constant.int 16
%int1_26652 = torch.constant.int 1
%int128_26653 = torch.constant.int 128
%29108 = torch.prim.ListConstruct %int4_26650, %3095, %int16_26651, %int1_26652, %int128_26653 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29109 = torch.aten.view %29025, %29108 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29109, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26654 = torch.constant.int 4
%int16_26655 = torch.constant.int 16
%int1_26656 = torch.constant.int 1
%int128_26657 = torch.constant.int 128
%29110 = torch.prim.ListConstruct %int4_26654, %3095, %int16_26655, %int1_26656, %int128_26657 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29111 = torch.aten.view %29040, %29110 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29111, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26658 = torch.constant.int 4
%int16_26659 = torch.constant.int 16
%int1_26660 = torch.constant.int 1
%int128_26661 = torch.constant.int 128
%29112 = torch.prim.ListConstruct %int4_26658, %3095, %int16_26659, %int1_26660, %int128_26661 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29113 = torch.aten.view %29055, %29112 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29113, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26662 = torch.constant.int 4
%int16_26663 = torch.constant.int 16
%int1_26664 = torch.constant.int 1
%int128_26665 = torch.constant.int 128
%29114 = torch.prim.ListConstruct %int4_26662, %3095, %int16_26663, %int1_26664, %int128_26665 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29115 = torch.aten.view %29070, %29114 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29115, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26666 = torch.constant.int 4
%int16_26667 = torch.constant.int 16
%int1_26668 = torch.constant.int 1
%int128_26669 = torch.constant.int 128
%29116 = torch.prim.ListConstruct %int4_26666, %3095, %int16_26667, %int1_26668, %int128_26669 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29117 = torch.aten.view %29085, %29116 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29117, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26670 = torch.constant.int 4
%29118 = torch.aten.mul.int %int4_26670, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26671 = torch.constant.int 16
%int1_26672 = torch.constant.int 1
%int128_26673 = torch.constant.int 128
%29119 = torch.prim.ListConstruct %29118, %int16_26671, %int1_26672, %int128_26673 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29120 = torch.aten.view %29103, %29119 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29120, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26674 = torch.constant.int 4
%29121 = torch.aten.mul.int %int4_26674, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26675 = torch.constant.int 16
%int1_26676 = torch.constant.int 1
%int128_26677 = torch.constant.int 128
%29122 = torch.prim.ListConstruct %29121, %int16_26675, %int1_26676, %int128_26677 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29123 = torch.aten.view %29105, %29122 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29123, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26678 = torch.constant.int 4
%29124 = torch.aten.mul.int %int4_26678, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26679 = torch.constant.int 16
%int1_26680 = torch.constant.int 1
%int128_26681 = torch.constant.int 128
%29125 = torch.prim.ListConstruct %29124, %int16_26679, %int1_26680, %int128_26681 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29126 = torch.aten.view %29107, %29125 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29126, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26682 = torch.constant.int 4
%29127 = torch.aten.mul.int %int4_26682, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26683 = torch.constant.int 16
%int1_26684 = torch.constant.int 1
%int128_26685 = torch.constant.int 128
%29128 = torch.prim.ListConstruct %29127, %int16_26683, %int1_26684, %int128_26685 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29129 = torch.aten.view %29109, %29128 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29129, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26686 = torch.constant.int 4
%29130 = torch.aten.mul.int %int4_26686, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26687 = torch.constant.int 16
%int1_26688 = torch.constant.int 1
%int128_26689 = torch.constant.int 128
%29131 = torch.prim.ListConstruct %29130, %int16_26687, %int1_26688, %int128_26689 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29132 = torch.aten.view %29111, %29131 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29132, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26690 = torch.constant.int 4
%29133 = torch.aten.mul.int %int4_26690, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26691 = torch.constant.int 16
%int1_26692 = torch.constant.int 1
%int128_26693 = torch.constant.int 128
%29134 = torch.prim.ListConstruct %29133, %int16_26691, %int1_26692, %int128_26693 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29135 = torch.aten.view %29113, %29134 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29135, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26694 = torch.constant.int 4
%29136 = torch.aten.mul.int %int4_26694, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26695 = torch.constant.int 16
%int1_26696 = torch.constant.int 1
%int128_26697 = torch.constant.int 128
%29137 = torch.prim.ListConstruct %29136, %int16_26695, %int1_26696, %int128_26697 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29138 = torch.aten.view %29115, %29137 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29138, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26698 = torch.constant.int 4
%29139 = torch.aten.mul.int %int4_26698, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26699 = torch.constant.int 16
%int1_26700 = torch.constant.int 1
%int128_26701 = torch.constant.int 128
%29140 = torch.prim.ListConstruct %29139, %int16_26699, %int1_26700, %int128_26701 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29141 = torch.aten.view %29117, %29140 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29141, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26702 = torch.constant.int 4
%29142 = torch.aten.mul.int %int4_26702, %3095 : !torch.int, !torch.int -> !torch.int
%29143 = torch.prim.ListConstruct %29142 : (!torch.int) -> !torch.list<int>
%29144 = torch.aten.view %29094, %29143 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29144, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26703 = torch.constant.int 4
%29145 = torch.aten.mul.int %int4_26703, %3095 : !torch.int, !torch.int -> !torch.int
%29146 = torch.prim.ListConstruct %29145 : (!torch.int) -> !torch.list<int>
%29147 = torch.aten.view %29095, %29146 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29147, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26704 = torch.constant.int 4
%29148 = torch.aten.mul.int %int4_26704, %3095 : !torch.int, !torch.int -> !torch.int
%29149 = torch.prim.ListConstruct %29148 : (!torch.int) -> !torch.list<int>
%29150 = torch.aten.view %29096, %29149 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29150, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26705 = torch.constant.int 4
%29151 = torch.aten.mul.int %int4_26705, %3095 : !torch.int, !torch.int -> !torch.int
%29152 = torch.prim.ListConstruct %29151 : (!torch.int) -> !torch.list<int>
%29153 = torch.aten.view %29097, %29152 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29153, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26706 = torch.constant.int 4
%29154 = torch.aten.mul.int %int4_26706, %3095 : !torch.int, !torch.int -> !torch.int
%29155 = torch.prim.ListConstruct %29154 : (!torch.int) -> !torch.list<int>
%29156 = torch.aten.view %29098, %29155 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29156, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26707 = torch.constant.int 4
%29157 = torch.aten.mul.int %int4_26707, %3095 : !torch.int, !torch.int -> !torch.int
%29158 = torch.prim.ListConstruct %29157 : (!torch.int) -> !torch.list<int>
%29159 = torch.aten.view %29099, %29158 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29159, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26708 = torch.constant.int 4
%29160 = torch.aten.mul.int %int4_26708, %3095 : !torch.int, !torch.int -> !torch.int
%29161 = torch.prim.ListConstruct %29160 : (!torch.int) -> !torch.list<int>
%29162 = torch.aten.view %29100, %29161 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29162, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26709 = torch.constant.int 4
%29163 = torch.aten.mul.int %int4_26709, %3095 : !torch.int, !torch.int -> !torch.int
%29164 = torch.prim.ListConstruct %29163 : (!torch.int) -> !torch.list<int>
%29165 = torch.aten.view %29101, %29164 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29165, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26710 = torch.constant.int 4
%int16_26711 = torch.constant.int 16
%int1_26712 = torch.constant.int 1
%int128_26713 = torch.constant.int 128
%29166 = torch.prim.ListConstruct %int4_26710, %3095, %int16_26711, %int1_26712, %int128_26713 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29167 = torch.aten.view %28755, %29166 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29167, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26714 = torch.constant.int 4
%int16_26715 = torch.constant.int 16
%int1_26716 = torch.constant.int 1
%int128_26717 = torch.constant.int 128
%29168 = torch.prim.ListConstruct %int4_26714, %3095, %int16_26715, %int1_26716, %int128_26717 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29169 = torch.aten.view %28757, %29168 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29169, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26718 = torch.constant.int 4
%int16_26719 = torch.constant.int 16
%int1_26720 = torch.constant.int 1
%int128_26721 = torch.constant.int 128
%29170 = torch.prim.ListConstruct %int4_26718, %3095, %int16_26719, %int1_26720, %int128_26721 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29171 = torch.aten.view %28759, %29170 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29171, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26722 = torch.constant.int 4
%int16_26723 = torch.constant.int 16
%int1_26724 = torch.constant.int 1
%int128_26725 = torch.constant.int 128
%29172 = torch.prim.ListConstruct %int4_26722, %3095, %int16_26723, %int1_26724, %int128_26725 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29173 = torch.aten.view %28761, %29172 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29173, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26726 = torch.constant.int 4
%int16_26727 = torch.constant.int 16
%int1_26728 = torch.constant.int 1
%int128_26729 = torch.constant.int 128
%29174 = torch.prim.ListConstruct %int4_26726, %3095, %int16_26727, %int1_26728, %int128_26729 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29175 = torch.aten.view %28763, %29174 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29175, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26730 = torch.constant.int 4
%int16_26731 = torch.constant.int 16
%int1_26732 = torch.constant.int 1
%int128_26733 = torch.constant.int 128
%29176 = torch.prim.ListConstruct %int4_26730, %3095, %int16_26731, %int1_26732, %int128_26733 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29177 = torch.aten.view %28765, %29176 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29177, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26734 = torch.constant.int 4
%int16_26735 = torch.constant.int 16
%int1_26736 = torch.constant.int 1
%int128_26737 = torch.constant.int 128
%29178 = torch.prim.ListConstruct %int4_26734, %3095, %int16_26735, %int1_26736, %int128_26737 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29179 = torch.aten.view %28767, %29178 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29179, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26738 = torch.constant.int 4
%int16_26739 = torch.constant.int 16
%int1_26740 = torch.constant.int 1
%int128_26741 = torch.constant.int 128
%29180 = torch.prim.ListConstruct %int4_26738, %3095, %int16_26739, %int1_26740, %int128_26741 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29181 = torch.aten.view %28769, %29180 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %29181, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_26742 = torch.constant.int 4
%29182 = torch.aten.mul.int %int4_26742, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26743 = torch.constant.int 16
%int1_26744 = torch.constant.int 1
%int128_26745 = torch.constant.int 128
%29183 = torch.prim.ListConstruct %29182, %int16_26743, %int1_26744, %int128_26745 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29184 = torch.aten.view %29167, %29183 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29184, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26746 = torch.constant.int 4
%29185 = torch.aten.mul.int %int4_26746, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26747 = torch.constant.int 16
%int1_26748 = torch.constant.int 1
%int128_26749 = torch.constant.int 128
%29186 = torch.prim.ListConstruct %29185, %int16_26747, %int1_26748, %int128_26749 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29187 = torch.aten.view %29169, %29186 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29187, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26750 = torch.constant.int 4
%29188 = torch.aten.mul.int %int4_26750, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26751 = torch.constant.int 16
%int1_26752 = torch.constant.int 1
%int128_26753 = torch.constant.int 128
%29189 = torch.prim.ListConstruct %29188, %int16_26751, %int1_26752, %int128_26753 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29190 = torch.aten.view %29171, %29189 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29190, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26754 = torch.constant.int 4
%29191 = torch.aten.mul.int %int4_26754, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26755 = torch.constant.int 16
%int1_26756 = torch.constant.int 1
%int128_26757 = torch.constant.int 128
%29192 = torch.prim.ListConstruct %29191, %int16_26755, %int1_26756, %int128_26757 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29193 = torch.aten.view %29173, %29192 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29193, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26758 = torch.constant.int 4
%29194 = torch.aten.mul.int %int4_26758, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26759 = torch.constant.int 16
%int1_26760 = torch.constant.int 1
%int128_26761 = torch.constant.int 128
%29195 = torch.prim.ListConstruct %29194, %int16_26759, %int1_26760, %int128_26761 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29196 = torch.aten.view %29175, %29195 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29196, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26762 = torch.constant.int 4
%29197 = torch.aten.mul.int %int4_26762, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26763 = torch.constant.int 16
%int1_26764 = torch.constant.int 1
%int128_26765 = torch.constant.int 128
%29198 = torch.prim.ListConstruct %29197, %int16_26763, %int1_26764, %int128_26765 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29199 = torch.aten.view %29177, %29198 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29199, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26766 = torch.constant.int 4
%29200 = torch.aten.mul.int %int4_26766, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26767 = torch.constant.int 16
%int1_26768 = torch.constant.int 1
%int128_26769 = torch.constant.int 128
%29201 = torch.prim.ListConstruct %29200, %int16_26767, %int1_26768, %int128_26769 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29202 = torch.aten.view %29179, %29201 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29202, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_26770 = torch.constant.int 4
%29203 = torch.aten.mul.int %int4_26770, %3095 : !torch.int, !torch.int -> !torch.int
%int16_26771 = torch.constant.int 16
%int1_26772 = torch.constant.int 1
%int128_26773 = torch.constant.int 128
%29204 = torch.prim.ListConstruct %29203, %int16_26771, %int1_26772, %int128_26773 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29205 = torch.aten.view %29181, %29204 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29205, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_26774 = torch.constant.int 1
%int1_26775 = torch.constant.int 1
%29206 = torch.aten.add.Scalar %29094, %int1_26774, %int1_26775 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29206, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26776 = torch.constant.int 1
%int1_26777 = torch.constant.int 1
%29207 = torch.aten.add.Scalar %29095, %int1_26776, %int1_26777 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29207, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26778 = torch.constant.int 1
%int1_26779 = torch.constant.int 1
%29208 = torch.aten.add.Scalar %29096, %int1_26778, %int1_26779 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29208, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26780 = torch.constant.int 1
%int1_26781 = torch.constant.int 1
%29209 = torch.aten.add.Scalar %29097, %int1_26780, %int1_26781 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29209, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26782 = torch.constant.int 1
%int1_26783 = torch.constant.int 1
%29210 = torch.aten.add.Scalar %29098, %int1_26782, %int1_26783 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29210, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26784 = torch.constant.int 1
%int1_26785 = torch.constant.int 1
%29211 = torch.aten.add.Scalar %29099, %int1_26784, %int1_26785 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29211, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26786 = torch.constant.int 1
%int1_26787 = torch.constant.int 1
%29212 = torch.aten.add.Scalar %29100, %int1_26786, %int1_26787 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29212, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_26788 = torch.constant.int 1
%int1_26789 = torch.constant.int 1
%29213 = torch.aten.add.Scalar %29101, %int1_26788, %int1_26789 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %29213, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_26790 = torch.constant.int 4
%29214 = torch.aten.mul.int %int4_26790, %3095 : !torch.int, !torch.int -> !torch.int
%29215 = torch.prim.ListConstruct %29214 : (!torch.int) -> !torch.list<int>
%29216 = torch.aten.view %29206, %29215 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29216, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26791 = torch.constant.int 4
%29217 = torch.aten.mul.int %int4_26791, %3095 : !torch.int, !torch.int -> !torch.int
%29218 = torch.prim.ListConstruct %29217 : (!torch.int) -> !torch.list<int>
%29219 = torch.aten.view %29207, %29218 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29219, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26792 = torch.constant.int 4
%29220 = torch.aten.mul.int %int4_26792, %3095 : !torch.int, !torch.int -> !torch.int
%29221 = torch.prim.ListConstruct %29220 : (!torch.int) -> !torch.list<int>
%29222 = torch.aten.view %29208, %29221 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29222, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26793 = torch.constant.int 4
%29223 = torch.aten.mul.int %int4_26793, %3095 : !torch.int, !torch.int -> !torch.int
%29224 = torch.prim.ListConstruct %29223 : (!torch.int) -> !torch.list<int>
%29225 = torch.aten.view %29209, %29224 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29225, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26794 = torch.constant.int 4
%29226 = torch.aten.mul.int %int4_26794, %3095 : !torch.int, !torch.int -> !torch.int
%29227 = torch.prim.ListConstruct %29226 : (!torch.int) -> !torch.list<int>
%29228 = torch.aten.view %29210, %29227 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29228, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26795 = torch.constant.int 4
%29229 = torch.aten.mul.int %int4_26795, %3095 : !torch.int, !torch.int -> !torch.int
%29230 = torch.prim.ListConstruct %29229 : (!torch.int) -> !torch.list<int>
%29231 = torch.aten.view %29211, %29230 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29231, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26796 = torch.constant.int 4
%29232 = torch.aten.mul.int %int4_26796, %3095 : !torch.int, !torch.int -> !torch.int
%29233 = torch.prim.ListConstruct %29232 : (!torch.int) -> !torch.list<int>
%29234 = torch.aten.view %29212, %29233 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29234, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_26797 = torch.constant.int 4
%29235 = torch.aten.mul.int %int4_26797, %3095 : !torch.int, !torch.int -> !torch.int
%29236 = torch.prim.ListConstruct %29235 : (!torch.int) -> !torch.list<int>
%29237 = torch.aten.view %29213, %29236 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29237, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%29238 = torch.prim.ListConstruct %29144, %29216 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26798 = torch.constant.int 0
%29239 = torch.aten.cat %29238, %int0_26798 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29239, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29240 = torch.prim.ListConstruct %29147, %29219 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26799 = torch.constant.int 0
%29241 = torch.aten.cat %29240, %int0_26799 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29241, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29242 = torch.prim.ListConstruct %29150, %29222 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26800 = torch.constant.int 0
%29243 = torch.aten.cat %29242, %int0_26800 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29243, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29244 = torch.prim.ListConstruct %29153, %29225 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26801 = torch.constant.int 0
%29245 = torch.aten.cat %29244, %int0_26801 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29245, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29246 = torch.prim.ListConstruct %29156, %29228 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26802 = torch.constant.int 0
%29247 = torch.aten.cat %29246, %int0_26802 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29247, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29248 = torch.prim.ListConstruct %29159, %29231 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26803 = torch.constant.int 0
%29249 = torch.aten.cat %29248, %int0_26803 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29249, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29250 = torch.prim.ListConstruct %29162, %29234 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26804 = torch.constant.int 0
%29251 = torch.aten.cat %29250, %int0_26804 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29251, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29252 = torch.prim.ListConstruct %29165, %29237 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_26805 = torch.constant.int 0
%29253 = torch.aten.cat %29252, %int0_26805 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %29253, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%29254 = torch.prim.ListConstruct %29120, %29184 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26806 = torch.constant.int 0
%29255 = torch.aten.cat %29254, %int0_26806 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29255, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29256 = torch.prim.ListConstruct %29123, %29187 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26807 = torch.constant.int 0
%29257 = torch.aten.cat %29256, %int0_26807 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29257, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29258 = torch.prim.ListConstruct %29126, %29190 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26808 = torch.constant.int 0
%29259 = torch.aten.cat %29258, %int0_26808 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29259, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29260 = torch.prim.ListConstruct %29129, %29193 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26809 = torch.constant.int 0
%29261 = torch.aten.cat %29260, %int0_26809 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29261, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29262 = torch.prim.ListConstruct %29132, %29196 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26810 = torch.constant.int 0
%29263 = torch.aten.cat %29262, %int0_26810 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29263, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29264 = torch.prim.ListConstruct %29135, %29199 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26811 = torch.constant.int 0
%29265 = torch.aten.cat %29264, %int0_26811 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29265, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29266 = torch.prim.ListConstruct %29138, %29202 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26812 = torch.constant.int 0
%29267 = torch.aten.cat %29266, %int0_26812 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29267, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29268 = torch.prim.ListConstruct %29141, %29205 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_26813 = torch.constant.int 0
%29269 = torch.aten.cat %29268, %int0_26813 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29269, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26814 = torch.constant.int 32
%int2_26815 = torch.constant.int 2
%int16_26816 = torch.constant.int 16
%int1_26817 = torch.constant.int 1
%int128_26818 = torch.constant.int 128
%29270 = torch.prim.ListConstruct %3023, %int32_26814, %int2_26815, %int16_26816, %int1_26817, %int128_26818 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29271 = torch.aten.view %27420, %29270 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29271, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26819 = torch.constant.int 32
%29272 = torch.aten.mul.int %3023, %int32_26819 : !torch.int, !torch.int -> !torch.int
%int2_26820 = torch.constant.int 2
%29273 = torch.aten.mul.int %29272, %int2_26820 : !torch.int, !torch.int -> !torch.int
%int16_26821 = torch.constant.int 16
%int1_26822 = torch.constant.int 1
%int128_26823 = torch.constant.int 128
%29274 = torch.prim.ListConstruct %29273, %int16_26821, %int1_26822, %int128_26823 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29275 = torch.aten.view %29271, %29274 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29275, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29276 = torch.prim.ListConstruct %29239 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26824 = torch.constant.bool false
%29277 = torch.aten.index_put %29275, %29276, %29255, %false_26824 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29277, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26825 = torch.constant.int 32
%int2_26826 = torch.constant.int 2
%int16_26827 = torch.constant.int 16
%int1_26828 = torch.constant.int 1
%int128_26829 = torch.constant.int 128
%29278 = torch.prim.ListConstruct %3023, %int32_26825, %int2_26826, %int16_26827, %int1_26828, %int128_26829 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29279 = torch.aten.view %29277, %29278 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29279, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26830 = torch.constant.int 131072
%29280 = torch.prim.ListConstruct %3023, %int131072_26830 : (!torch.int, !torch.int) -> !torch.list<int>
%29281 = torch.aten.view %29279, %29280 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29281, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26831 = torch.constant.int 32
%int2_26832 = torch.constant.int 2
%int16_26833 = torch.constant.int 16
%int1_26834 = torch.constant.int 1
%int128_26835 = torch.constant.int 128
%29282 = torch.prim.ListConstruct %3026, %int32_26831, %int2_26832, %int16_26833, %int1_26834, %int128_26835 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29283 = torch.aten.view %27432, %29282 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29283, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26836 = torch.constant.int 32
%29284 = torch.aten.mul.int %3026, %int32_26836 : !torch.int, !torch.int -> !torch.int
%int2_26837 = torch.constant.int 2
%29285 = torch.aten.mul.int %29284, %int2_26837 : !torch.int, !torch.int -> !torch.int
%int16_26838 = torch.constant.int 16
%int1_26839 = torch.constant.int 1
%int128_26840 = torch.constant.int 128
%29286 = torch.prim.ListConstruct %29285, %int16_26838, %int1_26839, %int128_26840 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29287 = torch.aten.view %29283, %29286 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29287, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29288 = torch.prim.ListConstruct %29241 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26841 = torch.constant.bool false
%29289 = torch.aten.index_put %29287, %29288, %29257, %false_26841 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29289, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26842 = torch.constant.int 32
%int2_26843 = torch.constant.int 2
%int16_26844 = torch.constant.int 16
%int1_26845 = torch.constant.int 1
%int128_26846 = torch.constant.int 128
%29290 = torch.prim.ListConstruct %3026, %int32_26842, %int2_26843, %int16_26844, %int1_26845, %int128_26846 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29291 = torch.aten.view %29289, %29290 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29291, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26847 = torch.constant.int 131072
%29292 = torch.prim.ListConstruct %3026, %int131072_26847 : (!torch.int, !torch.int) -> !torch.list<int>
%29293 = torch.aten.view %29291, %29292 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29293, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26848 = torch.constant.int 32
%int2_26849 = torch.constant.int 2
%int16_26850 = torch.constant.int 16
%int1_26851 = torch.constant.int 1
%int128_26852 = torch.constant.int 128
%29294 = torch.prim.ListConstruct %3029, %int32_26848, %int2_26849, %int16_26850, %int1_26851, %int128_26852 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29295 = torch.aten.view %27444, %29294 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29295, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26853 = torch.constant.int 32
%29296 = torch.aten.mul.int %3029, %int32_26853 : !torch.int, !torch.int -> !torch.int
%int2_26854 = torch.constant.int 2
%29297 = torch.aten.mul.int %29296, %int2_26854 : !torch.int, !torch.int -> !torch.int
%int16_26855 = torch.constant.int 16
%int1_26856 = torch.constant.int 1
%int128_26857 = torch.constant.int 128
%29298 = torch.prim.ListConstruct %29297, %int16_26855, %int1_26856, %int128_26857 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29299 = torch.aten.view %29295, %29298 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29299, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29300 = torch.prim.ListConstruct %29243 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26858 = torch.constant.bool false
%29301 = torch.aten.index_put %29299, %29300, %29259, %false_26858 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29301, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26859 = torch.constant.int 32
%int2_26860 = torch.constant.int 2
%int16_26861 = torch.constant.int 16
%int1_26862 = torch.constant.int 1
%int128_26863 = torch.constant.int 128
%29302 = torch.prim.ListConstruct %3029, %int32_26859, %int2_26860, %int16_26861, %int1_26862, %int128_26863 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29303 = torch.aten.view %29301, %29302 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29303, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26864 = torch.constant.int 131072
%29304 = torch.prim.ListConstruct %3029, %int131072_26864 : (!torch.int, !torch.int) -> !torch.list<int>
%29305 = torch.aten.view %29303, %29304 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29305, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26865 = torch.constant.int 32
%int2_26866 = torch.constant.int 2
%int16_26867 = torch.constant.int 16
%int1_26868 = torch.constant.int 1
%int128_26869 = torch.constant.int 128
%29306 = torch.prim.ListConstruct %3032, %int32_26865, %int2_26866, %int16_26867, %int1_26868, %int128_26869 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29307 = torch.aten.view %27456, %29306 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29307, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26870 = torch.constant.int 32
%29308 = torch.aten.mul.int %3032, %int32_26870 : !torch.int, !torch.int -> !torch.int
%int2_26871 = torch.constant.int 2
%29309 = torch.aten.mul.int %29308, %int2_26871 : !torch.int, !torch.int -> !torch.int
%int16_26872 = torch.constant.int 16
%int1_26873 = torch.constant.int 1
%int128_26874 = torch.constant.int 128
%29310 = torch.prim.ListConstruct %29309, %int16_26872, %int1_26873, %int128_26874 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29311 = torch.aten.view %29307, %29310 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29311, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29312 = torch.prim.ListConstruct %29245 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26875 = torch.constant.bool false
%29313 = torch.aten.index_put %29311, %29312, %29261, %false_26875 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29313, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26876 = torch.constant.int 32
%int2_26877 = torch.constant.int 2
%int16_26878 = torch.constant.int 16
%int1_26879 = torch.constant.int 1
%int128_26880 = torch.constant.int 128
%29314 = torch.prim.ListConstruct %3032, %int32_26876, %int2_26877, %int16_26878, %int1_26879, %int128_26880 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29315 = torch.aten.view %29313, %29314 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29315, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26881 = torch.constant.int 131072
%29316 = torch.prim.ListConstruct %3032, %int131072_26881 : (!torch.int, !torch.int) -> !torch.list<int>
%29317 = torch.aten.view %29315, %29316 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29317, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26882 = torch.constant.int 32
%int2_26883 = torch.constant.int 2
%int16_26884 = torch.constant.int 16
%int1_26885 = torch.constant.int 1
%int128_26886 = torch.constant.int 128
%29318 = torch.prim.ListConstruct %3035, %int32_26882, %int2_26883, %int16_26884, %int1_26885, %int128_26886 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29319 = torch.aten.view %27468, %29318 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29319, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26887 = torch.constant.int 32
%29320 = torch.aten.mul.int %3035, %int32_26887 : !torch.int, !torch.int -> !torch.int
%int2_26888 = torch.constant.int 2
%29321 = torch.aten.mul.int %29320, %int2_26888 : !torch.int, !torch.int -> !torch.int
%int16_26889 = torch.constant.int 16
%int1_26890 = torch.constant.int 1
%int128_26891 = torch.constant.int 128
%29322 = torch.prim.ListConstruct %29321, %int16_26889, %int1_26890, %int128_26891 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29323 = torch.aten.view %29319, %29322 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29323, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29324 = torch.prim.ListConstruct %29247 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26892 = torch.constant.bool false
%29325 = torch.aten.index_put %29323, %29324, %29263, %false_26892 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29325, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26893 = torch.constant.int 32
%int2_26894 = torch.constant.int 2
%int16_26895 = torch.constant.int 16
%int1_26896 = torch.constant.int 1
%int128_26897 = torch.constant.int 128
%29326 = torch.prim.ListConstruct %3035, %int32_26893, %int2_26894, %int16_26895, %int1_26896, %int128_26897 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29327 = torch.aten.view %29325, %29326 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29327, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26898 = torch.constant.int 131072
%29328 = torch.prim.ListConstruct %3035, %int131072_26898 : (!torch.int, !torch.int) -> !torch.list<int>
%29329 = torch.aten.view %29327, %29328 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29329, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26899 = torch.constant.int 32
%int2_26900 = torch.constant.int 2
%int16_26901 = torch.constant.int 16
%int1_26902 = torch.constant.int 1
%int128_26903 = torch.constant.int 128
%29330 = torch.prim.ListConstruct %3038, %int32_26899, %int2_26900, %int16_26901, %int1_26902, %int128_26903 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29331 = torch.aten.view %27480, %29330 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29331, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26904 = torch.constant.int 32
%29332 = torch.aten.mul.int %3038, %int32_26904 : !torch.int, !torch.int -> !torch.int
%int2_26905 = torch.constant.int 2
%29333 = torch.aten.mul.int %29332, %int2_26905 : !torch.int, !torch.int -> !torch.int
%int16_26906 = torch.constant.int 16
%int1_26907 = torch.constant.int 1
%int128_26908 = torch.constant.int 128
%29334 = torch.prim.ListConstruct %29333, %int16_26906, %int1_26907, %int128_26908 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29335 = torch.aten.view %29331, %29334 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29335, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29336 = torch.prim.ListConstruct %29249 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26909 = torch.constant.bool false
%29337 = torch.aten.index_put %29335, %29336, %29265, %false_26909 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29337, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26910 = torch.constant.int 32
%int2_26911 = torch.constant.int 2
%int16_26912 = torch.constant.int 16
%int1_26913 = torch.constant.int 1
%int128_26914 = torch.constant.int 128
%29338 = torch.prim.ListConstruct %3038, %int32_26910, %int2_26911, %int16_26912, %int1_26913, %int128_26914 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29339 = torch.aten.view %29337, %29338 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29339, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26915 = torch.constant.int 131072
%29340 = torch.prim.ListConstruct %3038, %int131072_26915 : (!torch.int, !torch.int) -> !torch.list<int>
%29341 = torch.aten.view %29339, %29340 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29341, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26916 = torch.constant.int 32
%int2_26917 = torch.constant.int 2
%int16_26918 = torch.constant.int 16
%int1_26919 = torch.constant.int 1
%int128_26920 = torch.constant.int 128
%29342 = torch.prim.ListConstruct %3041, %int32_26916, %int2_26917, %int16_26918, %int1_26919, %int128_26920 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29343 = torch.aten.view %27492, %29342 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29343, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26921 = torch.constant.int 32
%29344 = torch.aten.mul.int %3041, %int32_26921 : !torch.int, !torch.int -> !torch.int
%int2_26922 = torch.constant.int 2
%29345 = torch.aten.mul.int %29344, %int2_26922 : !torch.int, !torch.int -> !torch.int
%int16_26923 = torch.constant.int 16
%int1_26924 = torch.constant.int 1
%int128_26925 = torch.constant.int 128
%29346 = torch.prim.ListConstruct %29345, %int16_26923, %int1_26924, %int128_26925 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29347 = torch.aten.view %29343, %29346 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29347, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29348 = torch.prim.ListConstruct %29251 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26926 = torch.constant.bool false
%29349 = torch.aten.index_put %29347, %29348, %29267, %false_26926 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29349, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26927 = torch.constant.int 32
%int2_26928 = torch.constant.int 2
%int16_26929 = torch.constant.int 16
%int1_26930 = torch.constant.int 1
%int128_26931 = torch.constant.int 128
%29350 = torch.prim.ListConstruct %3041, %int32_26927, %int2_26928, %int16_26929, %int1_26930, %int128_26931 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29351 = torch.aten.view %29349, %29350 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29351, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26932 = torch.constant.int 131072
%29352 = torch.prim.ListConstruct %3041, %int131072_26932 : (!torch.int, !torch.int) -> !torch.list<int>
%29353 = torch.aten.view %29351, %29352 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29353, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_26933 = torch.constant.int 32
%int2_26934 = torch.constant.int 2
%int16_26935 = torch.constant.int 16
%int1_26936 = torch.constant.int 1
%int128_26937 = torch.constant.int 128
%29354 = torch.prim.ListConstruct %3044, %int32_26933, %int2_26934, %int16_26935, %int1_26936, %int128_26937 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29355 = torch.aten.view %27504, %29354 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29355, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_26938 = torch.constant.int 32
%29356 = torch.aten.mul.int %3044, %int32_26938 : !torch.int, !torch.int -> !torch.int
%int2_26939 = torch.constant.int 2
%29357 = torch.aten.mul.int %29356, %int2_26939 : !torch.int, !torch.int -> !torch.int
%int16_26940 = torch.constant.int 16
%int1_26941 = torch.constant.int 1
%int128_26942 = torch.constant.int 128
%29358 = torch.prim.ListConstruct %29357, %int16_26940, %int1_26941, %int128_26942 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29359 = torch.aten.view %29355, %29358 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29359, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%29360 = torch.prim.ListConstruct %29253 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_26943 = torch.constant.bool false
%29361 = torch.aten.index_put %29359, %29360, %29269, %false_26943 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %29361, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_26944 = torch.constant.int 32
%int2_26945 = torch.constant.int 2
%int16_26946 = torch.constant.int 16
%int1_26947 = torch.constant.int 1
%int128_26948 = torch.constant.int 128
%29362 = torch.prim.ListConstruct %3044, %int32_26944, %int2_26945, %int16_26946, %int1_26947, %int128_26948 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29363 = torch.aten.view %29361, %29362 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %29363, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_26949 = torch.constant.int 131072
%29364 = torch.prim.ListConstruct %3044, %int131072_26949 : (!torch.int, !torch.int) -> !torch.list<int>
%29365 = torch.aten.view %29363, %29364 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %29365, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_26950 = torch.constant.int -2
%29366 = torch.aten.unsqueeze %28980, %int-2_26950 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26951 = torch.constant.int -2
%29367 = torch.aten.unsqueeze %28995, %int-2_26951 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26952 = torch.constant.int -2
%29368 = torch.aten.unsqueeze %29010, %int-2_26952 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26953 = torch.constant.int -2
%29369 = torch.aten.unsqueeze %29025, %int-2_26953 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26954 = torch.constant.int -2
%29370 = torch.aten.unsqueeze %29040, %int-2_26954 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26955 = torch.constant.int -2
%29371 = torch.aten.unsqueeze %29055, %int-2_26955 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26956 = torch.constant.int -2
%29372 = torch.aten.unsqueeze %29070, %int-2_26956 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_26957 = torch.constant.int -2
%29373 = torch.aten.unsqueeze %29085, %int-2_26957 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_26958 = torch.constant.int 4
%int1_26959 = torch.constant.int 1
%int4_26960 = torch.constant.int 4
%int128_26961 = torch.constant.int 128
%29374 = torch.prim.ListConstruct %int4_26958, %28966, %int1_26959, %int4_26960, %int128_26961 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26962 = torch.constant.bool false
%29375 = torch.aten.expand %29366, %29374, %false_26962 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26963 = torch.constant.int 4
%int1_26964 = torch.constant.int 1
%int4_26965 = torch.constant.int 4
%int128_26966 = torch.constant.int 128
%29376 = torch.prim.ListConstruct %int4_26963, %28966, %int1_26964, %int4_26965, %int128_26966 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26967 = torch.constant.bool false
%29377 = torch.aten.expand %29367, %29376, %false_26967 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26968 = torch.constant.int 4
%int1_26969 = torch.constant.int 1
%int4_26970 = torch.constant.int 4
%int128_26971 = torch.constant.int 128
%29378 = torch.prim.ListConstruct %int4_26968, %28966, %int1_26969, %int4_26970, %int128_26971 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26972 = torch.constant.bool false
%29379 = torch.aten.expand %29368, %29378, %false_26972 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26973 = torch.constant.int 4
%int1_26974 = torch.constant.int 1
%int4_26975 = torch.constant.int 4
%int128_26976 = torch.constant.int 128
%29380 = torch.prim.ListConstruct %int4_26973, %28966, %int1_26974, %int4_26975, %int128_26976 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26977 = torch.constant.bool false
%29381 = torch.aten.expand %29369, %29380, %false_26977 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26978 = torch.constant.int 4
%int1_26979 = torch.constant.int 1
%int4_26980 = torch.constant.int 4
%int128_26981 = torch.constant.int 128
%29382 = torch.prim.ListConstruct %int4_26978, %28966, %int1_26979, %int4_26980, %int128_26981 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26982 = torch.constant.bool false
%29383 = torch.aten.expand %29370, %29382, %false_26982 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26983 = torch.constant.int 4
%int1_26984 = torch.constant.int 1
%int4_26985 = torch.constant.int 4
%int128_26986 = torch.constant.int 128
%29384 = torch.prim.ListConstruct %int4_26983, %28966, %int1_26984, %int4_26985, %int128_26986 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26987 = torch.constant.bool false
%29385 = torch.aten.expand %29371, %29384, %false_26987 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26988 = torch.constant.int 4
%int1_26989 = torch.constant.int 1
%int4_26990 = torch.constant.int 4
%int128_26991 = torch.constant.int 128
%29386 = torch.prim.ListConstruct %int4_26988, %28966, %int1_26989, %int4_26990, %int128_26991 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26992 = torch.constant.bool false
%29387 = torch.aten.expand %29372, %29386, %false_26992 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26993 = torch.constant.int 4
%int1_26994 = torch.constant.int 1
%int4_26995 = torch.constant.int 4
%int128_26996 = torch.constant.int 128
%29388 = torch.prim.ListConstruct %int4_26993, %28966, %int1_26994, %int4_26995, %int128_26996 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_26997 = torch.constant.bool false
%29389 = torch.aten.expand %29373, %29388, %false_26997 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_26998 = torch.constant.int 4
%int4_26999 = torch.constant.int 4
%int128_27000 = torch.constant.int 128
%29390 = torch.prim.ListConstruct %int4_26998, %28966, %int4_26999, %int128_27000 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29391 = torch.aten.view %29375, %29390 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27001 = torch.constant.int 4
%int4_27002 = torch.constant.int 4
%int128_27003 = torch.constant.int 128
%29392 = torch.prim.ListConstruct %int4_27001, %28966, %int4_27002, %int128_27003 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29393 = torch.aten.view %29377, %29392 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27004 = torch.constant.int 4
%int4_27005 = torch.constant.int 4
%int128_27006 = torch.constant.int 128
%29394 = torch.prim.ListConstruct %int4_27004, %28966, %int4_27005, %int128_27006 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29395 = torch.aten.view %29379, %29394 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27007 = torch.constant.int 4
%int4_27008 = torch.constant.int 4
%int128_27009 = torch.constant.int 128
%29396 = torch.prim.ListConstruct %int4_27007, %28966, %int4_27008, %int128_27009 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29397 = torch.aten.view %29381, %29396 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27010 = torch.constant.int 4
%int4_27011 = torch.constant.int 4
%int128_27012 = torch.constant.int 128
%29398 = torch.prim.ListConstruct %int4_27010, %28966, %int4_27011, %int128_27012 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29399 = torch.aten.view %29383, %29398 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27013 = torch.constant.int 4
%int4_27014 = torch.constant.int 4
%int128_27015 = torch.constant.int 128
%29400 = torch.prim.ListConstruct %int4_27013, %28966, %int4_27014, %int128_27015 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29401 = torch.aten.view %29385, %29400 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27016 = torch.constant.int 4
%int4_27017 = torch.constant.int 4
%int128_27018 = torch.constant.int 128
%29402 = torch.prim.ListConstruct %int4_27016, %28966, %int4_27017, %int128_27018 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29403 = torch.aten.view %29387, %29402 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27019 = torch.constant.int 4
%int4_27020 = torch.constant.int 4
%int128_27021 = torch.constant.int 128
%29404 = torch.prim.ListConstruct %int4_27019, %28966, %int4_27020, %int128_27021 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29405 = torch.aten.view %29389, %29404 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_27022 = torch.constant.int -2
%29406 = torch.aten.unsqueeze %28755, %int-2_27022 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27023 = torch.constant.int -2
%29407 = torch.aten.unsqueeze %28757, %int-2_27023 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27024 = torch.constant.int -2
%29408 = torch.aten.unsqueeze %28759, %int-2_27024 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27025 = torch.constant.int -2
%29409 = torch.aten.unsqueeze %28761, %int-2_27025 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27026 = torch.constant.int -2
%29410 = torch.aten.unsqueeze %28763, %int-2_27026 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29410, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27027 = torch.constant.int -2
%29411 = torch.aten.unsqueeze %28765, %int-2_27027 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27028 = torch.constant.int -2
%29412 = torch.aten.unsqueeze %28767, %int-2_27028 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_27029 = torch.constant.int -2
%29413 = torch.aten.unsqueeze %28769, %int-2_27029 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %29413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_27030 = torch.constant.int 1
%29414 = torch.aten.size.int %28679, %int1_27030 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_27031 = torch.constant.int 4
%int1_27032 = torch.constant.int 1
%int4_27033 = torch.constant.int 4
%int128_27034 = torch.constant.int 128
%29415 = torch.prim.ListConstruct %int4_27031, %29414, %int1_27032, %int4_27033, %int128_27034 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27035 = torch.constant.bool false
%29416 = torch.aten.expand %29406, %29415, %false_27035 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27036 = torch.constant.int 4
%int1_27037 = torch.constant.int 1
%int4_27038 = torch.constant.int 4
%int128_27039 = torch.constant.int 128
%29417 = torch.prim.ListConstruct %int4_27036, %29414, %int1_27037, %int4_27038, %int128_27039 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27040 = torch.constant.bool false
%29418 = torch.aten.expand %29407, %29417, %false_27040 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27041 = torch.constant.int 4
%int1_27042 = torch.constant.int 1
%int4_27043 = torch.constant.int 4
%int128_27044 = torch.constant.int 128
%29419 = torch.prim.ListConstruct %int4_27041, %29414, %int1_27042, %int4_27043, %int128_27044 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27045 = torch.constant.bool false
%29420 = torch.aten.expand %29408, %29419, %false_27045 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27046 = torch.constant.int 4
%int1_27047 = torch.constant.int 1
%int4_27048 = torch.constant.int 4
%int128_27049 = torch.constant.int 128
%29421 = torch.prim.ListConstruct %int4_27046, %29414, %int1_27047, %int4_27048, %int128_27049 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27050 = torch.constant.bool false
%29422 = torch.aten.expand %29409, %29421, %false_27050 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27051 = torch.constant.int 4
%int1_27052 = torch.constant.int 1
%int4_27053 = torch.constant.int 4
%int128_27054 = torch.constant.int 128
%29423 = torch.prim.ListConstruct %int4_27051, %29414, %int1_27052, %int4_27053, %int128_27054 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27055 = torch.constant.bool false
%29424 = torch.aten.expand %29410, %29423, %false_27055 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27056 = torch.constant.int 4
%int1_27057 = torch.constant.int 1
%int4_27058 = torch.constant.int 4
%int128_27059 = torch.constant.int 128
%29425 = torch.prim.ListConstruct %int4_27056, %29414, %int1_27057, %int4_27058, %int128_27059 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27060 = torch.constant.bool false
%29426 = torch.aten.expand %29411, %29425, %false_27060 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27061 = torch.constant.int 4
%int1_27062 = torch.constant.int 1
%int4_27063 = torch.constant.int 4
%int128_27064 = torch.constant.int 128
%29427 = torch.prim.ListConstruct %int4_27061, %29414, %int1_27062, %int4_27063, %int128_27064 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27065 = torch.constant.bool false
%29428 = torch.aten.expand %29412, %29427, %false_27065 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27066 = torch.constant.int 4
%int1_27067 = torch.constant.int 1
%int4_27068 = torch.constant.int 4
%int128_27069 = torch.constant.int 128
%29429 = torch.prim.ListConstruct %int4_27066, %29414, %int1_27067, %int4_27068, %int128_27069 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_27070 = torch.constant.bool false
%29430 = torch.aten.expand %29413, %29429, %false_27070 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %29430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_27071 = torch.constant.int 4
%int4_27072 = torch.constant.int 4
%int128_27073 = torch.constant.int 128
%29431 = torch.prim.ListConstruct %int4_27071, %29414, %int4_27072, %int128_27073 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29432 = torch.aten.view %29416, %29431 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27074 = torch.constant.int 4
%int4_27075 = torch.constant.int 4
%int128_27076 = torch.constant.int 128
%29433 = torch.prim.ListConstruct %int4_27074, %29414, %int4_27075, %int128_27076 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29434 = torch.aten.view %29418, %29433 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27077 = torch.constant.int 4
%int4_27078 = torch.constant.int 4
%int128_27079 = torch.constant.int 128
%29435 = torch.prim.ListConstruct %int4_27077, %29414, %int4_27078, %int128_27079 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29436 = torch.aten.view %29420, %29435 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27080 = torch.constant.int 4
%int4_27081 = torch.constant.int 4
%int128_27082 = torch.constant.int 128
%29437 = torch.prim.ListConstruct %int4_27080, %29414, %int4_27081, %int128_27082 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29438 = torch.aten.view %29422, %29437 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27083 = torch.constant.int 4
%int4_27084 = torch.constant.int 4
%int128_27085 = torch.constant.int 128
%29439 = torch.prim.ListConstruct %int4_27083, %29414, %int4_27084, %int128_27085 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29440 = torch.aten.view %29424, %29439 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27086 = torch.constant.int 4
%int4_27087 = torch.constant.int 4
%int128_27088 = torch.constant.int 128
%29441 = torch.prim.ListConstruct %int4_27086, %29414, %int4_27087, %int128_27088 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29442 = torch.aten.view %29426, %29441 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27089 = torch.constant.int 4
%int4_27090 = torch.constant.int 4
%int128_27091 = torch.constant.int 128
%29443 = torch.prim.ListConstruct %int4_27089, %29414, %int4_27090, %int128_27091 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29444 = torch.aten.view %29428, %29443 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27092 = torch.constant.int 4
%int4_27093 = torch.constant.int 4
%int128_27094 = torch.constant.int 128
%29445 = torch.prim.ListConstruct %int4_27092, %29414, %int4_27093, %int128_27094 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29446 = torch.aten.view %29430, %29445 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27095 = torch.constant.int 1
%int2_27096 = torch.constant.int 2
%29447 = torch.aten.transpose.int %28822, %int1_27095, %int2_27096 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29447, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27097 = torch.constant.int 1
%int2_27098 = torch.constant.int 2
%29448 = torch.aten.transpose.int %28837, %int1_27097, %int2_27098 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29448, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27099 = torch.constant.int 1
%int2_27100 = torch.constant.int 2
%29449 = torch.aten.transpose.int %28852, %int1_27099, %int2_27100 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29449, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27101 = torch.constant.int 1
%int2_27102 = torch.constant.int 2
%29450 = torch.aten.transpose.int %28867, %int1_27101, %int2_27102 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29450, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27103 = torch.constant.int 1
%int2_27104 = torch.constant.int 2
%29451 = torch.aten.transpose.int %28882, %int1_27103, %int2_27104 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29451, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27105 = torch.constant.int 1
%int2_27106 = torch.constant.int 2
%29452 = torch.aten.transpose.int %28897, %int1_27105, %int2_27106 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29452, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27107 = torch.constant.int 1
%int2_27108 = torch.constant.int 2
%29453 = torch.aten.transpose.int %28912, %int1_27107, %int2_27108 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29453, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27109 = torch.constant.int 1
%int2_27110 = torch.constant.int 2
%29454 = torch.aten.transpose.int %28927, %int1_27109, %int2_27110 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29454, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27111 = torch.constant.int 1
%int2_27112 = torch.constant.int 2
%29455 = torch.aten.transpose.int %29391, %int1_27111, %int2_27112 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29455, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27113 = torch.constant.int 1
%int2_27114 = torch.constant.int 2
%29456 = torch.aten.transpose.int %29393, %int1_27113, %int2_27114 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29456, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27115 = torch.constant.int 1
%int2_27116 = torch.constant.int 2
%29457 = torch.aten.transpose.int %29395, %int1_27115, %int2_27116 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29457, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27117 = torch.constant.int 1
%int2_27118 = torch.constant.int 2
%29458 = torch.aten.transpose.int %29397, %int1_27117, %int2_27118 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29458, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27119 = torch.constant.int 1
%int2_27120 = torch.constant.int 2
%29459 = torch.aten.transpose.int %29399, %int1_27119, %int2_27120 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29459, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27121 = torch.constant.int 1
%int2_27122 = torch.constant.int 2
%29460 = torch.aten.transpose.int %29401, %int1_27121, %int2_27122 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29460, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27123 = torch.constant.int 1
%int2_27124 = torch.constant.int 2
%29461 = torch.aten.transpose.int %29403, %int1_27123, %int2_27124 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29461, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27125 = torch.constant.int 1
%int2_27126 = torch.constant.int 2
%29462 = torch.aten.transpose.int %29405, %int1_27125, %int2_27126 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29462, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27127 = torch.constant.int 1
%int2_27128 = torch.constant.int 2
%29463 = torch.aten.transpose.int %29432, %int1_27127, %int2_27128 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29463, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27129 = torch.constant.int 1
%int2_27130 = torch.constant.int 2
%29464 = torch.aten.transpose.int %29434, %int1_27129, %int2_27130 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29464, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27131 = torch.constant.int 1
%int2_27132 = torch.constant.int 2
%29465 = torch.aten.transpose.int %29436, %int1_27131, %int2_27132 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29465, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27133 = torch.constant.int 1
%int2_27134 = torch.constant.int 2
%29466 = torch.aten.transpose.int %29438, %int1_27133, %int2_27134 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29466, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27135 = torch.constant.int 1
%int2_27136 = torch.constant.int 2
%29467 = torch.aten.transpose.int %29440, %int1_27135, %int2_27136 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29467, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27137 = torch.constant.int 1
%int2_27138 = torch.constant.int 2
%29468 = torch.aten.transpose.int %29442, %int1_27137, %int2_27138 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29468, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27139 = torch.constant.int 1
%int2_27140 = torch.constant.int 2
%29469 = torch.aten.transpose.int %29444, %int1_27139, %int2_27140 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29469, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27141 = torch.constant.int 1
%int2_27142 = torch.constant.int 2
%29470 = torch.aten.transpose.int %29446, %int1_27141, %int2_27142 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %29470, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27143 = torch.constant.float 0.000000e+00
%true_27144 = torch.constant.bool true
%none_27145 = torch.constant.none
%none_27146 = torch.constant.none
%29471:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29447, %29455, %29463, %float0.000000e00_27143, %true_27144, %none_27145, %none_27146) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29471#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27147 = torch.constant.float 0.000000e+00
%true_27148 = torch.constant.bool true
%none_27149 = torch.constant.none
%none_27150 = torch.constant.none
%29472:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29448, %29456, %29464, %float0.000000e00_27147, %true_27148, %none_27149, %none_27150) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29472#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27151 = torch.constant.float 0.000000e+00
%true_27152 = torch.constant.bool true
%none_27153 = torch.constant.none
%none_27154 = torch.constant.none
%29473:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29449, %29457, %29465, %float0.000000e00_27151, %true_27152, %none_27153, %none_27154) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29473#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27155 = torch.constant.float 0.000000e+00
%true_27156 = torch.constant.bool true
%none_27157 = torch.constant.none
%none_27158 = torch.constant.none
%29474:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29450, %29458, %29466, %float0.000000e00_27155, %true_27156, %none_27157, %none_27158) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29474#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27159 = torch.constant.float 0.000000e+00
%true_27160 = torch.constant.bool true
%none_27161 = torch.constant.none
%none_27162 = torch.constant.none
%29475:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29451, %29459, %29467, %float0.000000e00_27159, %true_27160, %none_27161, %none_27162) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29475#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27163 = torch.constant.float 0.000000e+00
%true_27164 = torch.constant.bool true
%none_27165 = torch.constant.none
%none_27166 = torch.constant.none
%29476:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29452, %29460, %29468, %float0.000000e00_27163, %true_27164, %none_27165, %none_27166) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29476#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27167 = torch.constant.float 0.000000e+00
%true_27168 = torch.constant.bool true
%none_27169 = torch.constant.none
%none_27170 = torch.constant.none
%29477:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29453, %29461, %29469, %float0.000000e00_27167, %true_27168, %none_27169, %none_27170) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29477#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_27171 = torch.constant.float 0.000000e+00
%true_27172 = torch.constant.bool true
%none_27173 = torch.constant.none
%none_27174 = torch.constant.none
%29478:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%29454, %29462, %29470, %float0.000000e00_27171, %true_27172, %none_27173, %none_27174) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %29478#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_27175 = torch.constant.int 1
%int2_27176 = torch.constant.int 2
%29479 = torch.aten.transpose.int %29471#0, %int1_27175, %int2_27176 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27177 = torch.constant.int 1
%int2_27178 = torch.constant.int 2
%29480 = torch.aten.transpose.int %29472#0, %int1_27177, %int2_27178 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27179 = torch.constant.int 1
%int2_27180 = torch.constant.int 2
%29481 = torch.aten.transpose.int %29473#0, %int1_27179, %int2_27180 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27181 = torch.constant.int 1
%int2_27182 = torch.constant.int 2
%29482 = torch.aten.transpose.int %29474#0, %int1_27181, %int2_27182 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27183 = torch.constant.int 1
%int2_27184 = torch.constant.int 2
%29483 = torch.aten.transpose.int %29475#0, %int1_27183, %int2_27184 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27185 = torch.constant.int 1
%int2_27186 = torch.constant.int 2
%29484 = torch.aten.transpose.int %29476#0, %int1_27185, %int2_27186 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27187 = torch.constant.int 1
%int2_27188 = torch.constant.int 2
%29485 = torch.aten.transpose.int %29477#0, %int1_27187, %int2_27188 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_27189 = torch.constant.int 1
%int2_27190 = torch.constant.int 2
%29486 = torch.aten.transpose.int %29478#0, %int1_27189, %int2_27190 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %29486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_27191 = torch.constant.int 4
%int512_27192 = torch.constant.int 512
%29487 = torch.prim.ListConstruct %int4_27191, %28808, %int512_27192 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29488 = torch.aten.view %29479, %29487 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27193 = torch.constant.int 4
%int512_27194 = torch.constant.int 512
%29489 = torch.prim.ListConstruct %int4_27193, %28823, %int512_27194 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29490 = torch.aten.view %29480, %29489 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27195 = torch.constant.int 4
%int512_27196 = torch.constant.int 512
%29491 = torch.prim.ListConstruct %int4_27195, %28838, %int512_27196 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29492 = torch.aten.view %29481, %29491 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27197 = torch.constant.int 4
%int512_27198 = torch.constant.int 512
%29493 = torch.prim.ListConstruct %int4_27197, %28853, %int512_27198 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29494 = torch.aten.view %29482, %29493 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27199 = torch.constant.int 4
%int512_27200 = torch.constant.int 512
%29495 = torch.prim.ListConstruct %int4_27199, %28868, %int512_27200 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29496 = torch.aten.view %29483, %29495 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27201 = torch.constant.int 4
%int512_27202 = torch.constant.int 512
%29497 = torch.prim.ListConstruct %int4_27201, %28883, %int512_27202 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29498 = torch.aten.view %29484, %29497 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27203 = torch.constant.int 4
%int512_27204 = torch.constant.int 512
%29499 = torch.prim.ListConstruct %int4_27203, %28898, %int512_27204 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29500 = torch.aten.view %29485, %29499 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27205 = torch.constant.int 4
%int512_27206 = torch.constant.int 512
%29501 = torch.prim.ListConstruct %int4_27205, %28913, %int512_27206 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29502 = torch.aten.view %29486, %29501 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %29502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_27207 = torch.constant.int 1
%int0_27208 = torch.constant.int 0
%29503 = torch.prim.ListConstruct %int1_27207, %int0_27208 : (!torch.int, !torch.int) -> !torch.list<int>
%29504 = torch.aten.permute %1048, %29503 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27209 = torch.constant.int 1
%int0_27210 = torch.constant.int 0
%29505 = torch.prim.ListConstruct %int1_27209, %int0_27210 : (!torch.int, !torch.int) -> !torch.list<int>
%29506 = torch.aten.permute %1049, %29505 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27211 = torch.constant.int 1
%int0_27212 = torch.constant.int 0
%29507 = torch.prim.ListConstruct %int1_27211, %int0_27212 : (!torch.int, !torch.int) -> !torch.list<int>
%29508 = torch.aten.permute %1050, %29507 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27213 = torch.constant.int 1
%int0_27214 = torch.constant.int 0
%29509 = torch.prim.ListConstruct %int1_27213, %int0_27214 : (!torch.int, !torch.int) -> !torch.list<int>
%29510 = torch.aten.permute %1051, %29509 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27215 = torch.constant.int 1
%int0_27216 = torch.constant.int 0
%29511 = torch.prim.ListConstruct %int1_27215, %int0_27216 : (!torch.int, !torch.int) -> !torch.list<int>
%29512 = torch.aten.permute %1052, %29511 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27217 = torch.constant.int 1
%int0_27218 = torch.constant.int 0
%29513 = torch.prim.ListConstruct %int1_27217, %int0_27218 : (!torch.int, !torch.int) -> !torch.list<int>
%29514 = torch.aten.permute %1053, %29513 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27219 = torch.constant.int 1
%int0_27220 = torch.constant.int 0
%29515 = torch.prim.ListConstruct %int1_27219, %int0_27220 : (!torch.int, !torch.int) -> !torch.list<int>
%29516 = torch.aten.permute %1054, %29515 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_27221 = torch.constant.int 1
%int0_27222 = torch.constant.int 0
%29517 = torch.prim.ListConstruct %int1_27221, %int0_27222 : (!torch.int, !torch.int) -> !torch.list<int>
%29518 = torch.aten.permute %1055, %29517 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_27223 = torch.constant.int 4
%29519 = torch.aten.mul.int %int4_27223, %28808 : !torch.int, !torch.int -> !torch.int
%int512_27224 = torch.constant.int 512
%29520 = torch.prim.ListConstruct %29519, %int512_27224 : (!torch.int, !torch.int) -> !torch.list<int>
%29521 = torch.aten.view %29488, %29520 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29521, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29522 = torch.aten.mm %29521, %29504 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29522, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27225 = torch.constant.int 4
%int4096_27226 = torch.constant.int 4096
%29523 = torch.prim.ListConstruct %int4_27225, %28808, %int4096_27226 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29524 = torch.aten.view %29522, %29523 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27227 = torch.constant.int 4
%29525 = torch.aten.mul.int %int4_27227, %28823 : !torch.int, !torch.int -> !torch.int
%int512_27228 = torch.constant.int 512
%29526 = torch.prim.ListConstruct %29525, %int512_27228 : (!torch.int, !torch.int) -> !torch.list<int>
%29527 = torch.aten.view %29490, %29526 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29527, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29528 = torch.aten.mm %29527, %29506 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29528, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27229 = torch.constant.int 4
%int4096_27230 = torch.constant.int 4096
%29529 = torch.prim.ListConstruct %int4_27229, %28823, %int4096_27230 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29530 = torch.aten.view %29528, %29529 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27231 = torch.constant.int 4
%29531 = torch.aten.mul.int %int4_27231, %28838 : !torch.int, !torch.int -> !torch.int
%int512_27232 = torch.constant.int 512
%29532 = torch.prim.ListConstruct %29531, %int512_27232 : (!torch.int, !torch.int) -> !torch.list<int>
%29533 = torch.aten.view %29492, %29532 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29533, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29534 = torch.aten.mm %29533, %29508 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29534, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27233 = torch.constant.int 4
%int4096_27234 = torch.constant.int 4096
%29535 = torch.prim.ListConstruct %int4_27233, %28838, %int4096_27234 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29536 = torch.aten.view %29534, %29535 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27235 = torch.constant.int 4
%29537 = torch.aten.mul.int %int4_27235, %28853 : !torch.int, !torch.int -> !torch.int
%int512_27236 = torch.constant.int 512
%29538 = torch.prim.ListConstruct %29537, %int512_27236 : (!torch.int, !torch.int) -> !torch.list<int>
%29539 = torch.aten.view %29494, %29538 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29539, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29540 = torch.aten.mm %29539, %29510 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29540, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27237 = torch.constant.int 4
%int4096_27238 = torch.constant.int 4096
%29541 = torch.prim.ListConstruct %int4_27237, %28853, %int4096_27238 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29542 = torch.aten.view %29540, %29541 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27239 = torch.constant.int 4
%29543 = torch.aten.mul.int %int4_27239, %28868 : !torch.int, !torch.int -> !torch.int
%int512_27240 = torch.constant.int 512
%29544 = torch.prim.ListConstruct %29543, %int512_27240 : (!torch.int, !torch.int) -> !torch.list<int>
%29545 = torch.aten.view %29496, %29544 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29545, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29546 = torch.aten.mm %29545, %29512 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29546, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27241 = torch.constant.int 4
%int4096_27242 = torch.constant.int 4096
%29547 = torch.prim.ListConstruct %int4_27241, %28868, %int4096_27242 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29548 = torch.aten.view %29546, %29547 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27243 = torch.constant.int 4
%29549 = torch.aten.mul.int %int4_27243, %28883 : !torch.int, !torch.int -> !torch.int
%int512_27244 = torch.constant.int 512
%29550 = torch.prim.ListConstruct %29549, %int512_27244 : (!torch.int, !torch.int) -> !torch.list<int>
%29551 = torch.aten.view %29498, %29550 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29551, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29552 = torch.aten.mm %29551, %29514 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29552, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27245 = torch.constant.int 4
%int4096_27246 = torch.constant.int 4096
%29553 = torch.prim.ListConstruct %int4_27245, %28883, %int4096_27246 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29554 = torch.aten.view %29552, %29553 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27247 = torch.constant.int 4
%29555 = torch.aten.mul.int %int4_27247, %28898 : !torch.int, !torch.int -> !torch.int
%int512_27248 = torch.constant.int 512
%29556 = torch.prim.ListConstruct %29555, %int512_27248 : (!torch.int, !torch.int) -> !torch.list<int>
%29557 = torch.aten.view %29500, %29556 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29557, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29558 = torch.aten.mm %29557, %29516 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29558, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27249 = torch.constant.int 4
%int4096_27250 = torch.constant.int 4096
%29559 = torch.prim.ListConstruct %int4_27249, %28898, %int4096_27250 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29560 = torch.aten.view %29558, %29559 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_27251 = torch.constant.int 4
%29561 = torch.aten.mul.int %int4_27251, %28913 : !torch.int, !torch.int -> !torch.int
%int512_27252 = torch.constant.int 512
%29562 = torch.prim.ListConstruct %29561, %int512_27252 : (!torch.int, !torch.int) -> !torch.list<int>
%29563 = torch.aten.view %29502, %29562 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %29563, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%29564 = torch.aten.mm %29563, %29518 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29564, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27253 = torch.constant.int 4
%int4096_27254 = torch.constant.int 4096
%29565 = torch.prim.ListConstruct %int4_27253, %28913, %int4096_27254 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29566 = torch.aten.view %29564, %29565 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29567 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27255 = arith.constant 1 : index
%dim_27256 = tensor.dim %29567, %c1_27255 : tensor<4x?x4096xf16>
%29568 = flow.tensor.transfer %29567 : tensor<4x?x4096xf16>{%dim_27256} to #hal.device.promise<@__device_0>
%29569 = torch_c.from_builtin_tensor %29568 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29570 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27257 = arith.constant 1 : index
%dim_27258 = tensor.dim %29570, %c1_27257 : tensor<4x?x4096xf16>
%29571 = flow.tensor.transfer %29570 : tensor<4x?x4096xf16>{%dim_27258} to #hal.device.promise<@__device_0>
%29572 = torch_c.from_builtin_tensor %29571 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29573 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27259 = arith.constant 1 : index
%dim_27260 = tensor.dim %29573, %c1_27259 : tensor<4x?x4096xf16>
%29574 = flow.tensor.transfer %29573 : tensor<4x?x4096xf16>{%dim_27260} to #hal.device.promise<@__device_0>
%29575 = torch_c.from_builtin_tensor %29574 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29576 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27261 = arith.constant 1 : index
%dim_27262 = tensor.dim %29576, %c1_27261 : tensor<4x?x4096xf16>
%29577 = flow.tensor.transfer %29576 : tensor<4x?x4096xf16>{%dim_27262} to #hal.device.promise<@__device_0>
%29578 = torch_c.from_builtin_tensor %29577 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29579 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27263 = arith.constant 1 : index
%dim_27264 = tensor.dim %29579, %c1_27263 : tensor<4x?x4096xf16>
%29580 = flow.tensor.transfer %29579 : tensor<4x?x4096xf16>{%dim_27264} to #hal.device.promise<@__device_0>
%29581 = torch_c.from_builtin_tensor %29580 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29582 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27265 = arith.constant 1 : index
%dim_27266 = tensor.dim %29582, %c1_27265 : tensor<4x?x4096xf16>
%29583 = flow.tensor.transfer %29582 : tensor<4x?x4096xf16>{%dim_27266} to #hal.device.promise<@__device_0>
%29584 = torch_c.from_builtin_tensor %29583 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29585 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27267 = arith.constant 1 : index
%dim_27268 = tensor.dim %29585, %c1_27267 : tensor<4x?x4096xf16>
%29586 = flow.tensor.transfer %29585 : tensor<4x?x4096xf16>{%dim_27268} to #hal.device.promise<@__device_0>
%29587 = torch_c.from_builtin_tensor %29586 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27269 = torch.constant.int 1
%29588 = torch.aten.add.Tensor %29524, %29569, %int1_27269 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27270 = torch.constant.int 1
%29589 = torch.aten.add.Tensor %29588, %29572, %int1_27270 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27271 = torch.constant.int 1
%29590 = torch.aten.add.Tensor %29589, %29575, %int1_27271 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27272 = torch.constant.int 1
%29591 = torch.aten.add.Tensor %29590, %29578, %int1_27272 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27273 = torch.constant.int 1
%29592 = torch.aten.add.Tensor %29591, %29581, %int1_27273 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27274 = torch.constant.int 1
%29593 = torch.aten.add.Tensor %29592, %29584, %int1_27274 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27275 = torch.constant.int 1
%29594 = torch.aten.add.Tensor %29593, %29587, %int1_27275 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29595 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27276 = arith.constant 1 : index
%dim_27277 = tensor.dim %29595, %c1_27276 : tensor<4x?x4096xf16>
%29596 = flow.tensor.transfer %29595 : tensor<4x?x4096xf16>{%dim_27277} to #hal.device.promise<@__device_1>
%29597 = torch_c.from_builtin_tensor %29596 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29598 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27278 = arith.constant 1 : index
%dim_27279 = tensor.dim %29598, %c1_27278 : tensor<4x?x4096xf16>
%29599 = flow.tensor.transfer %29598 : tensor<4x?x4096xf16>{%dim_27279} to #hal.device.promise<@__device_1>
%29600 = torch_c.from_builtin_tensor %29599 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29601 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27280 = arith.constant 1 : index
%dim_27281 = tensor.dim %29601, %c1_27280 : tensor<4x?x4096xf16>
%29602 = flow.tensor.transfer %29601 : tensor<4x?x4096xf16>{%dim_27281} to #hal.device.promise<@__device_1>
%29603 = torch_c.from_builtin_tensor %29602 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29604 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27282 = arith.constant 1 : index
%dim_27283 = tensor.dim %29604, %c1_27282 : tensor<4x?x4096xf16>
%29605 = flow.tensor.transfer %29604 : tensor<4x?x4096xf16>{%dim_27283} to #hal.device.promise<@__device_1>
%29606 = torch_c.from_builtin_tensor %29605 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29607 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27284 = arith.constant 1 : index
%dim_27285 = tensor.dim %29607, %c1_27284 : tensor<4x?x4096xf16>
%29608 = flow.tensor.transfer %29607 : tensor<4x?x4096xf16>{%dim_27285} to #hal.device.promise<@__device_1>
%29609 = torch_c.from_builtin_tensor %29608 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29610 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27286 = arith.constant 1 : index
%dim_27287 = tensor.dim %29610, %c1_27286 : tensor<4x?x4096xf16>
%29611 = flow.tensor.transfer %29610 : tensor<4x?x4096xf16>{%dim_27287} to #hal.device.promise<@__device_1>
%29612 = torch_c.from_builtin_tensor %29611 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29613 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27288 = arith.constant 1 : index
%dim_27289 = tensor.dim %29613, %c1_27288 : tensor<4x?x4096xf16>
%29614 = flow.tensor.transfer %29613 : tensor<4x?x4096xf16>{%dim_27289} to #hal.device.promise<@__device_1>
%29615 = torch_c.from_builtin_tensor %29614 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27290 = torch.constant.int 1
%29616 = torch.aten.add.Tensor %29597, %29530, %int1_27290 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27291 = torch.constant.int 1
%29617 = torch.aten.add.Tensor %29616, %29600, %int1_27291 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27292 = torch.constant.int 1
%29618 = torch.aten.add.Tensor %29617, %29603, %int1_27292 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27293 = torch.constant.int 1
%29619 = torch.aten.add.Tensor %29618, %29606, %int1_27293 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27294 = torch.constant.int 1
%29620 = torch.aten.add.Tensor %29619, %29609, %int1_27294 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27295 = torch.constant.int 1
%29621 = torch.aten.add.Tensor %29620, %29612, %int1_27295 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27296 = torch.constant.int 1
%29622 = torch.aten.add.Tensor %29621, %29615, %int1_27296 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29623 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27297 = arith.constant 1 : index
%dim_27298 = tensor.dim %29623, %c1_27297 : tensor<4x?x4096xf16>
%29624 = flow.tensor.transfer %29623 : tensor<4x?x4096xf16>{%dim_27298} to #hal.device.promise<@__device_2>
%29625 = torch_c.from_builtin_tensor %29624 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29626 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27299 = arith.constant 1 : index
%dim_27300 = tensor.dim %29626, %c1_27299 : tensor<4x?x4096xf16>
%29627 = flow.tensor.transfer %29626 : tensor<4x?x4096xf16>{%dim_27300} to #hal.device.promise<@__device_2>
%29628 = torch_c.from_builtin_tensor %29627 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29629 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27301 = arith.constant 1 : index
%dim_27302 = tensor.dim %29629, %c1_27301 : tensor<4x?x4096xf16>
%29630 = flow.tensor.transfer %29629 : tensor<4x?x4096xf16>{%dim_27302} to #hal.device.promise<@__device_2>
%29631 = torch_c.from_builtin_tensor %29630 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29632 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27303 = arith.constant 1 : index
%dim_27304 = tensor.dim %29632, %c1_27303 : tensor<4x?x4096xf16>
%29633 = flow.tensor.transfer %29632 : tensor<4x?x4096xf16>{%dim_27304} to #hal.device.promise<@__device_2>
%29634 = torch_c.from_builtin_tensor %29633 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29635 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27305 = arith.constant 1 : index
%dim_27306 = tensor.dim %29635, %c1_27305 : tensor<4x?x4096xf16>
%29636 = flow.tensor.transfer %29635 : tensor<4x?x4096xf16>{%dim_27306} to #hal.device.promise<@__device_2>
%29637 = torch_c.from_builtin_tensor %29636 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29638 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27307 = arith.constant 1 : index
%dim_27308 = tensor.dim %29638, %c1_27307 : tensor<4x?x4096xf16>
%29639 = flow.tensor.transfer %29638 : tensor<4x?x4096xf16>{%dim_27308} to #hal.device.promise<@__device_2>
%29640 = torch_c.from_builtin_tensor %29639 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29641 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27309 = arith.constant 1 : index
%dim_27310 = tensor.dim %29641, %c1_27309 : tensor<4x?x4096xf16>
%29642 = flow.tensor.transfer %29641 : tensor<4x?x4096xf16>{%dim_27310} to #hal.device.promise<@__device_2>
%29643 = torch_c.from_builtin_tensor %29642 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27311 = torch.constant.int 1
%29644 = torch.aten.add.Tensor %29625, %29628, %int1_27311 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27312 = torch.constant.int 1
%29645 = torch.aten.add.Tensor %29644, %29536, %int1_27312 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27313 = torch.constant.int 1
%29646 = torch.aten.add.Tensor %29645, %29631, %int1_27313 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27314 = torch.constant.int 1
%29647 = torch.aten.add.Tensor %29646, %29634, %int1_27314 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27315 = torch.constant.int 1
%29648 = torch.aten.add.Tensor %29647, %29637, %int1_27315 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27316 = torch.constant.int 1
%29649 = torch.aten.add.Tensor %29648, %29640, %int1_27316 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27317 = torch.constant.int 1
%29650 = torch.aten.add.Tensor %29649, %29643, %int1_27317 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29651 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27318 = arith.constant 1 : index
%dim_27319 = tensor.dim %29651, %c1_27318 : tensor<4x?x4096xf16>
%29652 = flow.tensor.transfer %29651 : tensor<4x?x4096xf16>{%dim_27319} to #hal.device.promise<@__device_3>
%29653 = torch_c.from_builtin_tensor %29652 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29654 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27320 = arith.constant 1 : index
%dim_27321 = tensor.dim %29654, %c1_27320 : tensor<4x?x4096xf16>
%29655 = flow.tensor.transfer %29654 : tensor<4x?x4096xf16>{%dim_27321} to #hal.device.promise<@__device_3>
%29656 = torch_c.from_builtin_tensor %29655 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29657 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27322 = arith.constant 1 : index
%dim_27323 = tensor.dim %29657, %c1_27322 : tensor<4x?x4096xf16>
%29658 = flow.tensor.transfer %29657 : tensor<4x?x4096xf16>{%dim_27323} to #hal.device.promise<@__device_3>
%29659 = torch_c.from_builtin_tensor %29658 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29660 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27324 = arith.constant 1 : index
%dim_27325 = tensor.dim %29660, %c1_27324 : tensor<4x?x4096xf16>
%29661 = flow.tensor.transfer %29660 : tensor<4x?x4096xf16>{%dim_27325} to #hal.device.promise<@__device_3>
%29662 = torch_c.from_builtin_tensor %29661 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29663 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27326 = arith.constant 1 : index
%dim_27327 = tensor.dim %29663, %c1_27326 : tensor<4x?x4096xf16>
%29664 = flow.tensor.transfer %29663 : tensor<4x?x4096xf16>{%dim_27327} to #hal.device.promise<@__device_3>
%29665 = torch_c.from_builtin_tensor %29664 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29666 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27328 = arith.constant 1 : index
%dim_27329 = tensor.dim %29666, %c1_27328 : tensor<4x?x4096xf16>
%29667 = flow.tensor.transfer %29666 : tensor<4x?x4096xf16>{%dim_27329} to #hal.device.promise<@__device_3>
%29668 = torch_c.from_builtin_tensor %29667 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29669 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27330 = arith.constant 1 : index
%dim_27331 = tensor.dim %29669, %c1_27330 : tensor<4x?x4096xf16>
%29670 = flow.tensor.transfer %29669 : tensor<4x?x4096xf16>{%dim_27331} to #hal.device.promise<@__device_3>
%29671 = torch_c.from_builtin_tensor %29670 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27332 = torch.constant.int 1
%29672 = torch.aten.add.Tensor %29653, %29656, %int1_27332 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27333 = torch.constant.int 1
%29673 = torch.aten.add.Tensor %29672, %29659, %int1_27333 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27334 = torch.constant.int 1
%29674 = torch.aten.add.Tensor %29673, %29542, %int1_27334 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27335 = torch.constant.int 1
%29675 = torch.aten.add.Tensor %29674, %29662, %int1_27335 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27336 = torch.constant.int 1
%29676 = torch.aten.add.Tensor %29675, %29665, %int1_27336 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27337 = torch.constant.int 1
%29677 = torch.aten.add.Tensor %29676, %29668, %int1_27337 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27338 = torch.constant.int 1
%29678 = torch.aten.add.Tensor %29677, %29671, %int1_27338 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29679 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27339 = arith.constant 1 : index
%dim_27340 = tensor.dim %29679, %c1_27339 : tensor<4x?x4096xf16>
%29680 = flow.tensor.transfer %29679 : tensor<4x?x4096xf16>{%dim_27340} to #hal.device.promise<@__device_4>
%29681 = torch_c.from_builtin_tensor %29680 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29682 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27341 = arith.constant 1 : index
%dim_27342 = tensor.dim %29682, %c1_27341 : tensor<4x?x4096xf16>
%29683 = flow.tensor.transfer %29682 : tensor<4x?x4096xf16>{%dim_27342} to #hal.device.promise<@__device_4>
%29684 = torch_c.from_builtin_tensor %29683 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29685 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27343 = arith.constant 1 : index
%dim_27344 = tensor.dim %29685, %c1_27343 : tensor<4x?x4096xf16>
%29686 = flow.tensor.transfer %29685 : tensor<4x?x4096xf16>{%dim_27344} to #hal.device.promise<@__device_4>
%29687 = torch_c.from_builtin_tensor %29686 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29688 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27345 = arith.constant 1 : index
%dim_27346 = tensor.dim %29688, %c1_27345 : tensor<4x?x4096xf16>
%29689 = flow.tensor.transfer %29688 : tensor<4x?x4096xf16>{%dim_27346} to #hal.device.promise<@__device_4>
%29690 = torch_c.from_builtin_tensor %29689 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29691 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27347 = arith.constant 1 : index
%dim_27348 = tensor.dim %29691, %c1_27347 : tensor<4x?x4096xf16>
%29692 = flow.tensor.transfer %29691 : tensor<4x?x4096xf16>{%dim_27348} to #hal.device.promise<@__device_4>
%29693 = torch_c.from_builtin_tensor %29692 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29694 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27349 = arith.constant 1 : index
%dim_27350 = tensor.dim %29694, %c1_27349 : tensor<4x?x4096xf16>
%29695 = flow.tensor.transfer %29694 : tensor<4x?x4096xf16>{%dim_27350} to #hal.device.promise<@__device_4>
%29696 = torch_c.from_builtin_tensor %29695 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29697 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27351 = arith.constant 1 : index
%dim_27352 = tensor.dim %29697, %c1_27351 : tensor<4x?x4096xf16>
%29698 = flow.tensor.transfer %29697 : tensor<4x?x4096xf16>{%dim_27352} to #hal.device.promise<@__device_4>
%29699 = torch_c.from_builtin_tensor %29698 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27353 = torch.constant.int 1
%29700 = torch.aten.add.Tensor %29681, %29684, %int1_27353 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27354 = torch.constant.int 1
%29701 = torch.aten.add.Tensor %29700, %29687, %int1_27354 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27355 = torch.constant.int 1
%29702 = torch.aten.add.Tensor %29701, %29690, %int1_27355 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27356 = torch.constant.int 1
%29703 = torch.aten.add.Tensor %29702, %29548, %int1_27356 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27357 = torch.constant.int 1
%29704 = torch.aten.add.Tensor %29703, %29693, %int1_27357 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27358 = torch.constant.int 1
%29705 = torch.aten.add.Tensor %29704, %29696, %int1_27358 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27359 = torch.constant.int 1
%29706 = torch.aten.add.Tensor %29705, %29699, %int1_27359 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29707 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27360 = arith.constant 1 : index
%dim_27361 = tensor.dim %29707, %c1_27360 : tensor<4x?x4096xf16>
%29708 = flow.tensor.transfer %29707 : tensor<4x?x4096xf16>{%dim_27361} to #hal.device.promise<@__device_5>
%29709 = torch_c.from_builtin_tensor %29708 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29710 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27362 = arith.constant 1 : index
%dim_27363 = tensor.dim %29710, %c1_27362 : tensor<4x?x4096xf16>
%29711 = flow.tensor.transfer %29710 : tensor<4x?x4096xf16>{%dim_27363} to #hal.device.promise<@__device_5>
%29712 = torch_c.from_builtin_tensor %29711 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29713 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27364 = arith.constant 1 : index
%dim_27365 = tensor.dim %29713, %c1_27364 : tensor<4x?x4096xf16>
%29714 = flow.tensor.transfer %29713 : tensor<4x?x4096xf16>{%dim_27365} to #hal.device.promise<@__device_5>
%29715 = torch_c.from_builtin_tensor %29714 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29716 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27366 = arith.constant 1 : index
%dim_27367 = tensor.dim %29716, %c1_27366 : tensor<4x?x4096xf16>
%29717 = flow.tensor.transfer %29716 : tensor<4x?x4096xf16>{%dim_27367} to #hal.device.promise<@__device_5>
%29718 = torch_c.from_builtin_tensor %29717 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29719 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27368 = arith.constant 1 : index
%dim_27369 = tensor.dim %29719, %c1_27368 : tensor<4x?x4096xf16>
%29720 = flow.tensor.transfer %29719 : tensor<4x?x4096xf16>{%dim_27369} to #hal.device.promise<@__device_5>
%29721 = torch_c.from_builtin_tensor %29720 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29722 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27370 = arith.constant 1 : index
%dim_27371 = tensor.dim %29722, %c1_27370 : tensor<4x?x4096xf16>
%29723 = flow.tensor.transfer %29722 : tensor<4x?x4096xf16>{%dim_27371} to #hal.device.promise<@__device_5>
%29724 = torch_c.from_builtin_tensor %29723 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29725 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27372 = arith.constant 1 : index
%dim_27373 = tensor.dim %29725, %c1_27372 : tensor<4x?x4096xf16>
%29726 = flow.tensor.transfer %29725 : tensor<4x?x4096xf16>{%dim_27373} to #hal.device.promise<@__device_5>
%29727 = torch_c.from_builtin_tensor %29726 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27374 = torch.constant.int 1
%29728 = torch.aten.add.Tensor %29709, %29712, %int1_27374 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27375 = torch.constant.int 1
%29729 = torch.aten.add.Tensor %29728, %29715, %int1_27375 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27376 = torch.constant.int 1
%29730 = torch.aten.add.Tensor %29729, %29718, %int1_27376 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27377 = torch.constant.int 1
%29731 = torch.aten.add.Tensor %29730, %29721, %int1_27377 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27378 = torch.constant.int 1
%29732 = torch.aten.add.Tensor %29731, %29554, %int1_27378 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27379 = torch.constant.int 1
%29733 = torch.aten.add.Tensor %29732, %29724, %int1_27379 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27380 = torch.constant.int 1
%29734 = torch.aten.add.Tensor %29733, %29727, %int1_27380 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29735 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27381 = arith.constant 1 : index
%dim_27382 = tensor.dim %29735, %c1_27381 : tensor<4x?x4096xf16>
%29736 = flow.tensor.transfer %29735 : tensor<4x?x4096xf16>{%dim_27382} to #hal.device.promise<@__device_6>
%29737 = torch_c.from_builtin_tensor %29736 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29738 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27383 = arith.constant 1 : index
%dim_27384 = tensor.dim %29738, %c1_27383 : tensor<4x?x4096xf16>
%29739 = flow.tensor.transfer %29738 : tensor<4x?x4096xf16>{%dim_27384} to #hal.device.promise<@__device_6>
%29740 = torch_c.from_builtin_tensor %29739 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29741 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27385 = arith.constant 1 : index
%dim_27386 = tensor.dim %29741, %c1_27385 : tensor<4x?x4096xf16>
%29742 = flow.tensor.transfer %29741 : tensor<4x?x4096xf16>{%dim_27386} to #hal.device.promise<@__device_6>
%29743 = torch_c.from_builtin_tensor %29742 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29744 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27387 = arith.constant 1 : index
%dim_27388 = tensor.dim %29744, %c1_27387 : tensor<4x?x4096xf16>
%29745 = flow.tensor.transfer %29744 : tensor<4x?x4096xf16>{%dim_27388} to #hal.device.promise<@__device_6>
%29746 = torch_c.from_builtin_tensor %29745 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29747 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27389 = arith.constant 1 : index
%dim_27390 = tensor.dim %29747, %c1_27389 : tensor<4x?x4096xf16>
%29748 = flow.tensor.transfer %29747 : tensor<4x?x4096xf16>{%dim_27390} to #hal.device.promise<@__device_6>
%29749 = torch_c.from_builtin_tensor %29748 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29750 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27391 = arith.constant 1 : index
%dim_27392 = tensor.dim %29750, %c1_27391 : tensor<4x?x4096xf16>
%29751 = flow.tensor.transfer %29750 : tensor<4x?x4096xf16>{%dim_27392} to #hal.device.promise<@__device_6>
%29752 = torch_c.from_builtin_tensor %29751 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29753 = torch_c.to_builtin_tensor %29566 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27393 = arith.constant 1 : index
%dim_27394 = tensor.dim %29753, %c1_27393 : tensor<4x?x4096xf16>
%29754 = flow.tensor.transfer %29753 : tensor<4x?x4096xf16>{%dim_27394} to #hal.device.promise<@__device_6>
%29755 = torch_c.from_builtin_tensor %29754 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27395 = torch.constant.int 1
%29756 = torch.aten.add.Tensor %29737, %29740, %int1_27395 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27396 = torch.constant.int 1
%29757 = torch.aten.add.Tensor %29756, %29743, %int1_27396 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27397 = torch.constant.int 1
%29758 = torch.aten.add.Tensor %29757, %29746, %int1_27397 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27398 = torch.constant.int 1
%29759 = torch.aten.add.Tensor %29758, %29749, %int1_27398 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27399 = torch.constant.int 1
%29760 = torch.aten.add.Tensor %29759, %29752, %int1_27399 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27400 = torch.constant.int 1
%29761 = torch.aten.add.Tensor %29760, %29560, %int1_27400 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27401 = torch.constant.int 1
%29762 = torch.aten.add.Tensor %29761, %29755, %int1_27401 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29763 = torch_c.to_builtin_tensor %29524 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27402 = arith.constant 1 : index
%dim_27403 = tensor.dim %29763, %c1_27402 : tensor<4x?x4096xf16>
%29764 = flow.tensor.transfer %29763 : tensor<4x?x4096xf16>{%dim_27403} to #hal.device.promise<@__device_7>
%29765 = torch_c.from_builtin_tensor %29764 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29766 = torch_c.to_builtin_tensor %29530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27404 = arith.constant 1 : index
%dim_27405 = tensor.dim %29766, %c1_27404 : tensor<4x?x4096xf16>
%29767 = flow.tensor.transfer %29766 : tensor<4x?x4096xf16>{%dim_27405} to #hal.device.promise<@__device_7>
%29768 = torch_c.from_builtin_tensor %29767 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29769 = torch_c.to_builtin_tensor %29536 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27406 = arith.constant 1 : index
%dim_27407 = tensor.dim %29769, %c1_27406 : tensor<4x?x4096xf16>
%29770 = flow.tensor.transfer %29769 : tensor<4x?x4096xf16>{%dim_27407} to #hal.device.promise<@__device_7>
%29771 = torch_c.from_builtin_tensor %29770 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29772 = torch_c.to_builtin_tensor %29542 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27408 = arith.constant 1 : index
%dim_27409 = tensor.dim %29772, %c1_27408 : tensor<4x?x4096xf16>
%29773 = flow.tensor.transfer %29772 : tensor<4x?x4096xf16>{%dim_27409} to #hal.device.promise<@__device_7>
%29774 = torch_c.from_builtin_tensor %29773 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29775 = torch_c.to_builtin_tensor %29548 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27410 = arith.constant 1 : index
%dim_27411 = tensor.dim %29775, %c1_27410 : tensor<4x?x4096xf16>
%29776 = flow.tensor.transfer %29775 : tensor<4x?x4096xf16>{%dim_27411} to #hal.device.promise<@__device_7>
%29777 = torch_c.from_builtin_tensor %29776 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29778 = torch_c.to_builtin_tensor %29554 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27412 = arith.constant 1 : index
%dim_27413 = tensor.dim %29778, %c1_27412 : tensor<4x?x4096xf16>
%29779 = flow.tensor.transfer %29778 : tensor<4x?x4096xf16>{%dim_27413} to #hal.device.promise<@__device_7>
%29780 = torch_c.from_builtin_tensor %29779 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%29781 = torch_c.to_builtin_tensor %29560 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27414 = arith.constant 1 : index
%dim_27415 = tensor.dim %29781, %c1_27414 : tensor<4x?x4096xf16>
%29782 = flow.tensor.transfer %29781 : tensor<4x?x4096xf16>{%dim_27415} to #hal.device.promise<@__device_7>
%29783 = torch_c.from_builtin_tensor %29782 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27416 = torch.constant.int 1
%29784 = torch.aten.add.Tensor %29765, %29768, %int1_27416 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27417 = torch.constant.int 1
%29785 = torch.aten.add.Tensor %29784, %29771, %int1_27417 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27418 = torch.constant.int 1
%29786 = torch.aten.add.Tensor %29785, %29774, %int1_27418 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27419 = torch.constant.int 1
%29787 = torch.aten.add.Tensor %29786, %29777, %int1_27419 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27420 = torch.constant.int 1
%29788 = torch.aten.add.Tensor %29787, %29780, %int1_27420 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27421 = torch.constant.int 1
%29789 = torch.aten.add.Tensor %29788, %29783, %int1_27421 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27422 = torch.constant.int 1
%29790 = torch.aten.add.Tensor %29789, %29566, %int1_27422 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27423 = torch.constant.int 1
%29791 = torch.aten.add.Tensor %28450, %29594, %int1_27423 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27424 = torch.constant.int 1
%29792 = torch.aten.add.Tensor %28451, %29622, %int1_27424 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27425 = torch.constant.int 1
%29793 = torch.aten.add.Tensor %28452, %29650, %int1_27425 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27426 = torch.constant.int 1
%29794 = torch.aten.add.Tensor %28453, %29678, %int1_27426 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27427 = torch.constant.int 1
%29795 = torch.aten.add.Tensor %28454, %29706, %int1_27427 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27428 = torch.constant.int 1
%29796 = torch.aten.add.Tensor %28455, %29734, %int1_27428 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27429 = torch.constant.int 1
%29797 = torch.aten.add.Tensor %28456, %29762, %int1_27429 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27430 = torch.constant.int 1
%29798 = torch.aten.add.Tensor %28457, %29790, %int1_27430 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_27431 = torch.constant.int 6
%29799 = torch.prims.convert_element_type %29791, %int6_27431 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27432 = torch.constant.int 6
%29800 = torch.prims.convert_element_type %29792, %int6_27432 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27433 = torch.constant.int 6
%29801 = torch.prims.convert_element_type %29793, %int6_27433 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27434 = torch.constant.int 6
%29802 = torch.prims.convert_element_type %29794, %int6_27434 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27435 = torch.constant.int 6
%29803 = torch.prims.convert_element_type %29795, %int6_27435 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27436 = torch.constant.int 6
%29804 = torch.prims.convert_element_type %29796, %int6_27436 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27437 = torch.constant.int 6
%29805 = torch.prims.convert_element_type %29797, %int6_27437 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27438 = torch.constant.int 6
%29806 = torch.prims.convert_element_type %29798, %int6_27438 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27439 = torch.constant.int 2
%29807 = torch.aten.pow.Tensor_Scalar %29799, %int2_27439 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27440 = torch.constant.int 2
%29808 = torch.aten.pow.Tensor_Scalar %29800, %int2_27440 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27441 = torch.constant.int 2
%29809 = torch.aten.pow.Tensor_Scalar %29801, %int2_27441 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27442 = torch.constant.int 2
%29810 = torch.aten.pow.Tensor_Scalar %29802, %int2_27442 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27443 = torch.constant.int 2
%29811 = torch.aten.pow.Tensor_Scalar %29803, %int2_27443 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27444 = torch.constant.int 2
%29812 = torch.aten.pow.Tensor_Scalar %29804, %int2_27444 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27445 = torch.constant.int 2
%29813 = torch.aten.pow.Tensor_Scalar %29805, %int2_27445 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27446 = torch.constant.int 2
%29814 = torch.aten.pow.Tensor_Scalar %29806, %int2_27446 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_27447 = torch.constant.int -1
%29815 = torch.prim.ListConstruct %int-1_27447 : (!torch.int) -> !torch.list<int>
%true_27448 = torch.constant.bool true
%none_27449 = torch.constant.none
%29816 = torch.aten.mean.dim %29807, %29815, %true_27448, %none_27449 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27450 = torch.constant.int -1
%29817 = torch.prim.ListConstruct %int-1_27450 : (!torch.int) -> !torch.list<int>
%true_27451 = torch.constant.bool true
%none_27452 = torch.constant.none
%29818 = torch.aten.mean.dim %29808, %29817, %true_27451, %none_27452 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27453 = torch.constant.int -1
%29819 = torch.prim.ListConstruct %int-1_27453 : (!torch.int) -> !torch.list<int>
%true_27454 = torch.constant.bool true
%none_27455 = torch.constant.none
%29820 = torch.aten.mean.dim %29809, %29819, %true_27454, %none_27455 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27456 = torch.constant.int -1
%29821 = torch.prim.ListConstruct %int-1_27456 : (!torch.int) -> !torch.list<int>
%true_27457 = torch.constant.bool true
%none_27458 = torch.constant.none
%29822 = torch.aten.mean.dim %29810, %29821, %true_27457, %none_27458 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27459 = torch.constant.int -1
%29823 = torch.prim.ListConstruct %int-1_27459 : (!torch.int) -> !torch.list<int>
%true_27460 = torch.constant.bool true
%none_27461 = torch.constant.none
%29824 = torch.aten.mean.dim %29811, %29823, %true_27460, %none_27461 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27462 = torch.constant.int -1
%29825 = torch.prim.ListConstruct %int-1_27462 : (!torch.int) -> !torch.list<int>
%true_27463 = torch.constant.bool true
%none_27464 = torch.constant.none
%29826 = torch.aten.mean.dim %29812, %29825, %true_27463, %none_27464 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27465 = torch.constant.int -1
%29827 = torch.prim.ListConstruct %int-1_27465 : (!torch.int) -> !torch.list<int>
%true_27466 = torch.constant.bool true
%none_27467 = torch.constant.none
%29828 = torch.aten.mean.dim %29813, %29827, %true_27466, %none_27467 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27468 = torch.constant.int -1
%29829 = torch.prim.ListConstruct %int-1_27468 : (!torch.int) -> !torch.list<int>
%true_27469 = torch.constant.bool true
%none_27470 = torch.constant.none
%29830 = torch.aten.mean.dim %29814, %29829, %true_27469, %none_27470 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27471 = torch.constant.float 9.9999997473787516E-6
%int1_27472 = torch.constant.int 1
%29831 = torch.aten.add.Scalar %29816, %float9.999990e-06_27471, %int1_27472 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27473 = torch.constant.float 9.9999997473787516E-6
%int1_27474 = torch.constant.int 1
%29832 = torch.aten.add.Scalar %29818, %float9.999990e-06_27473, %int1_27474 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27475 = torch.constant.float 9.9999997473787516E-6
%int1_27476 = torch.constant.int 1
%29833 = torch.aten.add.Scalar %29820, %float9.999990e-06_27475, %int1_27476 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27477 = torch.constant.float 9.9999997473787516E-6
%int1_27478 = torch.constant.int 1
%29834 = torch.aten.add.Scalar %29822, %float9.999990e-06_27477, %int1_27478 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27479 = torch.constant.float 9.9999997473787516E-6
%int1_27480 = torch.constant.int 1
%29835 = torch.aten.add.Scalar %29824, %float9.999990e-06_27479, %int1_27480 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27481 = torch.constant.float 9.9999997473787516E-6
%int1_27482 = torch.constant.int 1
%29836 = torch.aten.add.Scalar %29826, %float9.999990e-06_27481, %int1_27482 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27483 = torch.constant.float 9.9999997473787516E-6
%int1_27484 = torch.constant.int 1
%29837 = torch.aten.add.Scalar %29828, %float9.999990e-06_27483, %int1_27484 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27485 = torch.constant.float 9.9999997473787516E-6
%int1_27486 = torch.constant.int 1
%29838 = torch.aten.add.Scalar %29830, %float9.999990e-06_27485, %int1_27486 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29838, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29839 = torch.aten.rsqrt %29831 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29840 = torch.aten.rsqrt %29832 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29841 = torch.aten.rsqrt %29833 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29842 = torch.aten.rsqrt %29834 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29843 = torch.aten.rsqrt %29835 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29844 = torch.aten.rsqrt %29836 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29844, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29845 = torch.aten.rsqrt %29837 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29846 = torch.aten.rsqrt %29838 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %29846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%29847 = torch.aten.mul.Tensor %29799, %29839 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29848 = torch.aten.mul.Tensor %29800, %29840 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29849 = torch.aten.mul.Tensor %29801, %29841 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29850 = torch.aten.mul.Tensor %29802, %29842 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29851 = torch.aten.mul.Tensor %29803, %29843 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29852 = torch.aten.mul.Tensor %29804, %29844 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29853 = torch.aten.mul.Tensor %29805, %29845 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29854 = torch.aten.mul.Tensor %29806, %29846 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29855 = torch.aten.mul.Tensor %1056, %29847 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29856 = torch.aten.mul.Tensor %1057, %29848 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29857 = torch.aten.mul.Tensor %1058, %29849 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29858 = torch.aten.mul.Tensor %1059, %29850 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29859 = torch.aten.mul.Tensor %1060, %29851 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29860 = torch.aten.mul.Tensor %1061, %29852 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29861 = torch.aten.mul.Tensor %1062, %29853 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%29862 = torch.aten.mul.Tensor %1063, %29854 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %29862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_27487 = torch.constant.int 5
%29863 = torch.prims.convert_element_type %29855, %int5_27487 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27488 = torch.constant.int 5
%29864 = torch.prims.convert_element_type %29856, %int5_27488 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27489 = torch.constant.int 5
%29865 = torch.prims.convert_element_type %29857, %int5_27489 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27490 = torch.constant.int 5
%29866 = torch.prims.convert_element_type %29858, %int5_27490 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27491 = torch.constant.int 5
%29867 = torch.prims.convert_element_type %29859, %int5_27491 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27492 = torch.constant.int 5
%29868 = torch.prims.convert_element_type %29860, %int5_27492 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27493 = torch.constant.int 5
%29869 = torch.prims.convert_element_type %29861, %int5_27493 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27494 = torch.constant.int 5
%29870 = torch.prims.convert_element_type %29862, %int5_27494 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %29870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27495 = torch.constant.int 1
%int0_27496 = torch.constant.int 0
%29871 = torch.prim.ListConstruct %int1_27495, %int0_27496 : (!torch.int, !torch.int) -> !torch.list<int>
%29872 = torch.aten.permute %1064, %29871 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27497 = torch.constant.int 1
%int0_27498 = torch.constant.int 0
%29873 = torch.prim.ListConstruct %int1_27497, %int0_27498 : (!torch.int, !torch.int) -> !torch.list<int>
%29874 = torch.aten.permute %1065, %29873 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27499 = torch.constant.int 1
%int0_27500 = torch.constant.int 0
%29875 = torch.prim.ListConstruct %int1_27499, %int0_27500 : (!torch.int, !torch.int) -> !torch.list<int>
%29876 = torch.aten.permute %1066, %29875 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27501 = torch.constant.int 1
%int0_27502 = torch.constant.int 0
%29877 = torch.prim.ListConstruct %int1_27501, %int0_27502 : (!torch.int, !torch.int) -> !torch.list<int>
%29878 = torch.aten.permute %1067, %29877 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27503 = torch.constant.int 1
%int0_27504 = torch.constant.int 0
%29879 = torch.prim.ListConstruct %int1_27503, %int0_27504 : (!torch.int, !torch.int) -> !torch.list<int>
%29880 = torch.aten.permute %1068, %29879 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27505 = torch.constant.int 1
%int0_27506 = torch.constant.int 0
%29881 = torch.prim.ListConstruct %int1_27505, %int0_27506 : (!torch.int, !torch.int) -> !torch.list<int>
%29882 = torch.aten.permute %1069, %29881 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27507 = torch.constant.int 1
%int0_27508 = torch.constant.int 0
%29883 = torch.prim.ListConstruct %int1_27507, %int0_27508 : (!torch.int, !torch.int) -> !torch.list<int>
%29884 = torch.aten.permute %1070, %29883 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27509 = torch.constant.int 1
%int0_27510 = torch.constant.int 0
%29885 = torch.prim.ListConstruct %int1_27509, %int0_27510 : (!torch.int, !torch.int) -> !torch.list<int>
%29886 = torch.aten.permute %1071, %29885 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_27511 = torch.constant.int 4
%29887 = torch.aten.mul.int %int4_27511, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27512 = torch.constant.int 4096
%29888 = torch.prim.ListConstruct %29887, %int4096_27512 : (!torch.int, !torch.int) -> !torch.list<int>
%29889 = torch.aten.view %29863, %29888 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29889, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29890 = torch.aten.mm %29889, %29872 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29890, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27513 = torch.constant.int 4
%int1792_27514 = torch.constant.int 1792
%29891 = torch.prim.ListConstruct %int4_27513, %2482, %int1792_27514 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29892 = torch.aten.view %29890, %29891 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27515 = torch.constant.int 4
%29893 = torch.aten.mul.int %int4_27515, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27516 = torch.constant.int 4096
%29894 = torch.prim.ListConstruct %29893, %int4096_27516 : (!torch.int, !torch.int) -> !torch.list<int>
%29895 = torch.aten.view %29864, %29894 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29895, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29896 = torch.aten.mm %29895, %29874 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29896, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27517 = torch.constant.int 4
%int1792_27518 = torch.constant.int 1792
%29897 = torch.prim.ListConstruct %int4_27517, %2482, %int1792_27518 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29898 = torch.aten.view %29896, %29897 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27519 = torch.constant.int 4
%29899 = torch.aten.mul.int %int4_27519, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27520 = torch.constant.int 4096
%29900 = torch.prim.ListConstruct %29899, %int4096_27520 : (!torch.int, !torch.int) -> !torch.list<int>
%29901 = torch.aten.view %29865, %29900 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29901, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29902 = torch.aten.mm %29901, %29876 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29902, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27521 = torch.constant.int 4
%int1792_27522 = torch.constant.int 1792
%29903 = torch.prim.ListConstruct %int4_27521, %2482, %int1792_27522 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29904 = torch.aten.view %29902, %29903 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27523 = torch.constant.int 4
%29905 = torch.aten.mul.int %int4_27523, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27524 = torch.constant.int 4096
%29906 = torch.prim.ListConstruct %29905, %int4096_27524 : (!torch.int, !torch.int) -> !torch.list<int>
%29907 = torch.aten.view %29866, %29906 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29907, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29908 = torch.aten.mm %29907, %29878 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29908, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27525 = torch.constant.int 4
%int1792_27526 = torch.constant.int 1792
%29909 = torch.prim.ListConstruct %int4_27525, %2482, %int1792_27526 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29910 = torch.aten.view %29908, %29909 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27527 = torch.constant.int 4
%29911 = torch.aten.mul.int %int4_27527, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27528 = torch.constant.int 4096
%29912 = torch.prim.ListConstruct %29911, %int4096_27528 : (!torch.int, !torch.int) -> !torch.list<int>
%29913 = torch.aten.view %29867, %29912 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29913, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29914 = torch.aten.mm %29913, %29880 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29914, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27529 = torch.constant.int 4
%int1792_27530 = torch.constant.int 1792
%29915 = torch.prim.ListConstruct %int4_27529, %2482, %int1792_27530 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29916 = torch.aten.view %29914, %29915 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27531 = torch.constant.int 4
%29917 = torch.aten.mul.int %int4_27531, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27532 = torch.constant.int 4096
%29918 = torch.prim.ListConstruct %29917, %int4096_27532 : (!torch.int, !torch.int) -> !torch.list<int>
%29919 = torch.aten.view %29868, %29918 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29919, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29920 = torch.aten.mm %29919, %29882 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29920, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27533 = torch.constant.int 4
%int1792_27534 = torch.constant.int 1792
%29921 = torch.prim.ListConstruct %int4_27533, %2482, %int1792_27534 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29922 = torch.aten.view %29920, %29921 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27535 = torch.constant.int 4
%29923 = torch.aten.mul.int %int4_27535, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27536 = torch.constant.int 4096
%29924 = torch.prim.ListConstruct %29923, %int4096_27536 : (!torch.int, !torch.int) -> !torch.list<int>
%29925 = torch.aten.view %29869, %29924 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29925, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29926 = torch.aten.mm %29925, %29884 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29926, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27537 = torch.constant.int 4
%int1792_27538 = torch.constant.int 1792
%29927 = torch.prim.ListConstruct %int4_27537, %2482, %int1792_27538 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29928 = torch.aten.view %29926, %29927 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27539 = torch.constant.int 4
%29929 = torch.aten.mul.int %int4_27539, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27540 = torch.constant.int 4096
%29930 = torch.prim.ListConstruct %29929, %int4096_27540 : (!torch.int, !torch.int) -> !torch.list<int>
%29931 = torch.aten.view %29870, %29930 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29931, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29932 = torch.aten.mm %29931, %29886 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29932, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27541 = torch.constant.int 4
%int1792_27542 = torch.constant.int 1792
%29933 = torch.prim.ListConstruct %int4_27541, %2482, %int1792_27542 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29934 = torch.aten.view %29932, %29933 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29935 = torch.aten.silu %29892 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29936 = torch.aten.silu %29898 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29937 = torch.aten.silu %29904 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29938 = torch.aten.silu %29910 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29939 = torch.aten.silu %29916 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29940 = torch.aten.silu %29922 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29941 = torch.aten.silu %29928 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%29942 = torch.aten.silu %29934 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_27543 = torch.constant.int 1
%int0_27544 = torch.constant.int 0
%29943 = torch.prim.ListConstruct %int1_27543, %int0_27544 : (!torch.int, !torch.int) -> !torch.list<int>
%29944 = torch.aten.permute %1072, %29943 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27545 = torch.constant.int 1
%int0_27546 = torch.constant.int 0
%29945 = torch.prim.ListConstruct %int1_27545, %int0_27546 : (!torch.int, !torch.int) -> !torch.list<int>
%29946 = torch.aten.permute %1073, %29945 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27547 = torch.constant.int 1
%int0_27548 = torch.constant.int 0
%29947 = torch.prim.ListConstruct %int1_27547, %int0_27548 : (!torch.int, !torch.int) -> !torch.list<int>
%29948 = torch.aten.permute %1074, %29947 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27549 = torch.constant.int 1
%int0_27550 = torch.constant.int 0
%29949 = torch.prim.ListConstruct %int1_27549, %int0_27550 : (!torch.int, !torch.int) -> !torch.list<int>
%29950 = torch.aten.permute %1075, %29949 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27551 = torch.constant.int 1
%int0_27552 = torch.constant.int 0
%29951 = torch.prim.ListConstruct %int1_27551, %int0_27552 : (!torch.int, !torch.int) -> !torch.list<int>
%29952 = torch.aten.permute %1076, %29951 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27553 = torch.constant.int 1
%int0_27554 = torch.constant.int 0
%29953 = torch.prim.ListConstruct %int1_27553, %int0_27554 : (!torch.int, !torch.int) -> !torch.list<int>
%29954 = torch.aten.permute %1077, %29953 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27555 = torch.constant.int 1
%int0_27556 = torch.constant.int 0
%29955 = torch.prim.ListConstruct %int1_27555, %int0_27556 : (!torch.int, !torch.int) -> !torch.list<int>
%29956 = torch.aten.permute %1078, %29955 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_27557 = torch.constant.int 1
%int0_27558 = torch.constant.int 0
%29957 = torch.prim.ListConstruct %int1_27557, %int0_27558 : (!torch.int, !torch.int) -> !torch.list<int>
%29958 = torch.aten.permute %1079, %29957 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_27559 = torch.constant.int 4
%29959 = torch.aten.mul.int %int4_27559, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27560 = torch.constant.int 4096
%29960 = torch.prim.ListConstruct %29959, %int4096_27560 : (!torch.int, !torch.int) -> !torch.list<int>
%29961 = torch.aten.view %29863, %29960 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29961, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29962 = torch.aten.mm %29961, %29944 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29962, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27561 = torch.constant.int 4
%int1792_27562 = torch.constant.int 1792
%29963 = torch.prim.ListConstruct %int4_27561, %2482, %int1792_27562 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29964 = torch.aten.view %29962, %29963 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27563 = torch.constant.int 4
%29965 = torch.aten.mul.int %int4_27563, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27564 = torch.constant.int 4096
%29966 = torch.prim.ListConstruct %29965, %int4096_27564 : (!torch.int, !torch.int) -> !torch.list<int>
%29967 = torch.aten.view %29864, %29966 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29967, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29968 = torch.aten.mm %29967, %29946 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29968, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27565 = torch.constant.int 4
%int1792_27566 = torch.constant.int 1792
%29969 = torch.prim.ListConstruct %int4_27565, %2482, %int1792_27566 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29970 = torch.aten.view %29968, %29969 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27567 = torch.constant.int 4
%29971 = torch.aten.mul.int %int4_27567, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27568 = torch.constant.int 4096
%29972 = torch.prim.ListConstruct %29971, %int4096_27568 : (!torch.int, !torch.int) -> !torch.list<int>
%29973 = torch.aten.view %29865, %29972 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29973, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29974 = torch.aten.mm %29973, %29948 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29974, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27569 = torch.constant.int 4
%int1792_27570 = torch.constant.int 1792
%29975 = torch.prim.ListConstruct %int4_27569, %2482, %int1792_27570 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29976 = torch.aten.view %29974, %29975 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27571 = torch.constant.int 4
%29977 = torch.aten.mul.int %int4_27571, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27572 = torch.constant.int 4096
%29978 = torch.prim.ListConstruct %29977, %int4096_27572 : (!torch.int, !torch.int) -> !torch.list<int>
%29979 = torch.aten.view %29866, %29978 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29979, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29980 = torch.aten.mm %29979, %29950 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29980, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27573 = torch.constant.int 4
%int1792_27574 = torch.constant.int 1792
%29981 = torch.prim.ListConstruct %int4_27573, %2482, %int1792_27574 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29982 = torch.aten.view %29980, %29981 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27575 = torch.constant.int 4
%29983 = torch.aten.mul.int %int4_27575, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27576 = torch.constant.int 4096
%29984 = torch.prim.ListConstruct %29983, %int4096_27576 : (!torch.int, !torch.int) -> !torch.list<int>
%29985 = torch.aten.view %29867, %29984 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29985, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29986 = torch.aten.mm %29985, %29952 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29986, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27577 = torch.constant.int 4
%int1792_27578 = torch.constant.int 1792
%29987 = torch.prim.ListConstruct %int4_27577, %2482, %int1792_27578 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29988 = torch.aten.view %29986, %29987 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27579 = torch.constant.int 4
%29989 = torch.aten.mul.int %int4_27579, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27580 = torch.constant.int 4096
%29990 = torch.prim.ListConstruct %29989, %int4096_27580 : (!torch.int, !torch.int) -> !torch.list<int>
%29991 = torch.aten.view %29868, %29990 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29991, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29992 = torch.aten.mm %29991, %29954 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29992, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27581 = torch.constant.int 4
%int1792_27582 = torch.constant.int 1792
%29993 = torch.prim.ListConstruct %int4_27581, %2482, %int1792_27582 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%29994 = torch.aten.view %29992, %29993 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %29994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27583 = torch.constant.int 4
%29995 = torch.aten.mul.int %int4_27583, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27584 = torch.constant.int 4096
%29996 = torch.prim.ListConstruct %29995, %int4096_27584 : (!torch.int, !torch.int) -> !torch.list<int>
%29997 = torch.aten.view %29869, %29996 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %29997, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%29998 = torch.aten.mm %29997, %29956 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %29998, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27585 = torch.constant.int 4
%int1792_27586 = torch.constant.int 1792
%29999 = torch.prim.ListConstruct %int4_27585, %2482, %int1792_27586 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30000 = torch.aten.view %29998, %29999 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_27587 = torch.constant.int 4
%30001 = torch.aten.mul.int %int4_27587, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27588 = torch.constant.int 4096
%30002 = torch.prim.ListConstruct %30001, %int4096_27588 : (!torch.int, !torch.int) -> !torch.list<int>
%30003 = torch.aten.view %29870, %30002 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30003, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30004 = torch.aten.mm %30003, %29958 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30004, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_27589 = torch.constant.int 4
%int1792_27590 = torch.constant.int 1792
%30005 = torch.prim.ListConstruct %int4_27589, %2482, %int1792_27590 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30006 = torch.aten.view %30004, %30005 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30007 = torch.aten.mul.Tensor %29935, %29964 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30008 = torch.aten.mul.Tensor %29936, %29970 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30009 = torch.aten.mul.Tensor %29937, %29976 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30010 = torch.aten.mul.Tensor %29938, %29982 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30011 = torch.aten.mul.Tensor %29939, %29988 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30012 = torch.aten.mul.Tensor %29940, %29994 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30013 = torch.aten.mul.Tensor %29941, %30000 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%30014 = torch.aten.mul.Tensor %29942, %30006 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %30014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_27591 = torch.constant.int 1
%int0_27592 = torch.constant.int 0
%30015 = torch.prim.ListConstruct %int1_27591, %int0_27592 : (!torch.int, !torch.int) -> !torch.list<int>
%30016 = torch.aten.permute %1080, %30015 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27593 = torch.constant.int 1
%int0_27594 = torch.constant.int 0
%30017 = torch.prim.ListConstruct %int1_27593, %int0_27594 : (!torch.int, !torch.int) -> !torch.list<int>
%30018 = torch.aten.permute %1081, %30017 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27595 = torch.constant.int 1
%int0_27596 = torch.constant.int 0
%30019 = torch.prim.ListConstruct %int1_27595, %int0_27596 : (!torch.int, !torch.int) -> !torch.list<int>
%30020 = torch.aten.permute %1082, %30019 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27597 = torch.constant.int 1
%int0_27598 = torch.constant.int 0
%30021 = torch.prim.ListConstruct %int1_27597, %int0_27598 : (!torch.int, !torch.int) -> !torch.list<int>
%30022 = torch.aten.permute %1083, %30021 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27599 = torch.constant.int 1
%int0_27600 = torch.constant.int 0
%30023 = torch.prim.ListConstruct %int1_27599, %int0_27600 : (!torch.int, !torch.int) -> !torch.list<int>
%30024 = torch.aten.permute %1084, %30023 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27601 = torch.constant.int 1
%int0_27602 = torch.constant.int 0
%30025 = torch.prim.ListConstruct %int1_27601, %int0_27602 : (!torch.int, !torch.int) -> !torch.list<int>
%30026 = torch.aten.permute %1085, %30025 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27603 = torch.constant.int 1
%int0_27604 = torch.constant.int 0
%30027 = torch.prim.ListConstruct %int1_27603, %int0_27604 : (!torch.int, !torch.int) -> !torch.list<int>
%30028 = torch.aten.permute %1086, %30027 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27605 = torch.constant.int 1
%int0_27606 = torch.constant.int 0
%30029 = torch.prim.ListConstruct %int1_27605, %int0_27606 : (!torch.int, !torch.int) -> !torch.list<int>
%30030 = torch.aten.permute %1087, %30029 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_27607 = torch.constant.int 1
%30031 = torch.aten.size.int %29892, %int1_27607 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27608 = torch.constant.int 4
%30032 = torch.aten.mul.int %int4_27608, %30031 : !torch.int, !torch.int -> !torch.int
%int1792_27609 = torch.constant.int 1792
%30033 = torch.prim.ListConstruct %30032, %int1792_27609 : (!torch.int, !torch.int) -> !torch.list<int>
%30034 = torch.aten.view %30007, %30033 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30034, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30035 = torch.aten.mm %30034, %30016 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30035, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27610 = torch.constant.int 4
%int4096_27611 = torch.constant.int 4096
%30036 = torch.prim.ListConstruct %int4_27610, %30031, %int4096_27611 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30037 = torch.aten.view %30035, %30036 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27612 = torch.constant.int 1
%30038 = torch.aten.size.int %29898, %int1_27612 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27613 = torch.constant.int 4
%30039 = torch.aten.mul.int %int4_27613, %30038 : !torch.int, !torch.int -> !torch.int
%int1792_27614 = torch.constant.int 1792
%30040 = torch.prim.ListConstruct %30039, %int1792_27614 : (!torch.int, !torch.int) -> !torch.list<int>
%30041 = torch.aten.view %30008, %30040 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30041, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30042 = torch.aten.mm %30041, %30018 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30042, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27615 = torch.constant.int 4
%int4096_27616 = torch.constant.int 4096
%30043 = torch.prim.ListConstruct %int4_27615, %30038, %int4096_27616 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30044 = torch.aten.view %30042, %30043 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27617 = torch.constant.int 1
%30045 = torch.aten.size.int %29904, %int1_27617 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27618 = torch.constant.int 4
%30046 = torch.aten.mul.int %int4_27618, %30045 : !torch.int, !torch.int -> !torch.int
%int1792_27619 = torch.constant.int 1792
%30047 = torch.prim.ListConstruct %30046, %int1792_27619 : (!torch.int, !torch.int) -> !torch.list<int>
%30048 = torch.aten.view %30009, %30047 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30048, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30049 = torch.aten.mm %30048, %30020 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30049, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27620 = torch.constant.int 4
%int4096_27621 = torch.constant.int 4096
%30050 = torch.prim.ListConstruct %int4_27620, %30045, %int4096_27621 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30051 = torch.aten.view %30049, %30050 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27622 = torch.constant.int 1
%30052 = torch.aten.size.int %29910, %int1_27622 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27623 = torch.constant.int 4
%30053 = torch.aten.mul.int %int4_27623, %30052 : !torch.int, !torch.int -> !torch.int
%int1792_27624 = torch.constant.int 1792
%30054 = torch.prim.ListConstruct %30053, %int1792_27624 : (!torch.int, !torch.int) -> !torch.list<int>
%30055 = torch.aten.view %30010, %30054 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30055, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30056 = torch.aten.mm %30055, %30022 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30056, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27625 = torch.constant.int 4
%int4096_27626 = torch.constant.int 4096
%30057 = torch.prim.ListConstruct %int4_27625, %30052, %int4096_27626 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30058 = torch.aten.view %30056, %30057 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27627 = torch.constant.int 1
%30059 = torch.aten.size.int %29916, %int1_27627 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27628 = torch.constant.int 4
%30060 = torch.aten.mul.int %int4_27628, %30059 : !torch.int, !torch.int -> !torch.int
%int1792_27629 = torch.constant.int 1792
%30061 = torch.prim.ListConstruct %30060, %int1792_27629 : (!torch.int, !torch.int) -> !torch.list<int>
%30062 = torch.aten.view %30011, %30061 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30062, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30063 = torch.aten.mm %30062, %30024 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30063, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27630 = torch.constant.int 4
%int4096_27631 = torch.constant.int 4096
%30064 = torch.prim.ListConstruct %int4_27630, %30059, %int4096_27631 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30065 = torch.aten.view %30063, %30064 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27632 = torch.constant.int 1
%30066 = torch.aten.size.int %29922, %int1_27632 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27633 = torch.constant.int 4
%30067 = torch.aten.mul.int %int4_27633, %30066 : !torch.int, !torch.int -> !torch.int
%int1792_27634 = torch.constant.int 1792
%30068 = torch.prim.ListConstruct %30067, %int1792_27634 : (!torch.int, !torch.int) -> !torch.list<int>
%30069 = torch.aten.view %30012, %30068 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30069, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30070 = torch.aten.mm %30069, %30026 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30070, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27635 = torch.constant.int 4
%int4096_27636 = torch.constant.int 4096
%30071 = torch.prim.ListConstruct %int4_27635, %30066, %int4096_27636 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30072 = torch.aten.view %30070, %30071 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27637 = torch.constant.int 1
%30073 = torch.aten.size.int %29928, %int1_27637 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27638 = torch.constant.int 4
%30074 = torch.aten.mul.int %int4_27638, %30073 : !torch.int, !torch.int -> !torch.int
%int1792_27639 = torch.constant.int 1792
%30075 = torch.prim.ListConstruct %30074, %int1792_27639 : (!torch.int, !torch.int) -> !torch.list<int>
%30076 = torch.aten.view %30013, %30075 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30076, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30077 = torch.aten.mm %30076, %30028 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30077, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27640 = torch.constant.int 4
%int4096_27641 = torch.constant.int 4096
%30078 = torch.prim.ListConstruct %int4_27640, %30073, %int4096_27641 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30079 = torch.aten.view %30077, %30078 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27642 = torch.constant.int 1
%30080 = torch.aten.size.int %29934, %int1_27642 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_27643 = torch.constant.int 4
%30081 = torch.aten.mul.int %int4_27643, %30080 : !torch.int, !torch.int -> !torch.int
%int1792_27644 = torch.constant.int 1792
%30082 = torch.prim.ListConstruct %30081, %int1792_27644 : (!torch.int, !torch.int) -> !torch.list<int>
%30083 = torch.aten.view %30014, %30082 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %30083, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%30084 = torch.aten.mm %30083, %30030 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30084, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_27645 = torch.constant.int 4
%int4096_27646 = torch.constant.int 4096
%30085 = torch.prim.ListConstruct %int4_27645, %30080, %int4096_27646 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30086 = torch.aten.view %30084, %30085 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30087 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27647 = arith.constant 1 : index
%dim_27648 = tensor.dim %30087, %c1_27647 : tensor<4x?x4096xf16>
%30088 = flow.tensor.transfer %30087 : tensor<4x?x4096xf16>{%dim_27648} to #hal.device.promise<@__device_0>
%30089 = torch_c.from_builtin_tensor %30088 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30090 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27649 = arith.constant 1 : index
%dim_27650 = tensor.dim %30090, %c1_27649 : tensor<4x?x4096xf16>
%30091 = flow.tensor.transfer %30090 : tensor<4x?x4096xf16>{%dim_27650} to #hal.device.promise<@__device_0>
%30092 = torch_c.from_builtin_tensor %30091 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30093 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27651 = arith.constant 1 : index
%dim_27652 = tensor.dim %30093, %c1_27651 : tensor<4x?x4096xf16>
%30094 = flow.tensor.transfer %30093 : tensor<4x?x4096xf16>{%dim_27652} to #hal.device.promise<@__device_0>
%30095 = torch_c.from_builtin_tensor %30094 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30096 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27653 = arith.constant 1 : index
%dim_27654 = tensor.dim %30096, %c1_27653 : tensor<4x?x4096xf16>
%30097 = flow.tensor.transfer %30096 : tensor<4x?x4096xf16>{%dim_27654} to #hal.device.promise<@__device_0>
%30098 = torch_c.from_builtin_tensor %30097 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30099 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27655 = arith.constant 1 : index
%dim_27656 = tensor.dim %30099, %c1_27655 : tensor<4x?x4096xf16>
%30100 = flow.tensor.transfer %30099 : tensor<4x?x4096xf16>{%dim_27656} to #hal.device.promise<@__device_0>
%30101 = torch_c.from_builtin_tensor %30100 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30102 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27657 = arith.constant 1 : index
%dim_27658 = tensor.dim %30102, %c1_27657 : tensor<4x?x4096xf16>
%30103 = flow.tensor.transfer %30102 : tensor<4x?x4096xf16>{%dim_27658} to #hal.device.promise<@__device_0>
%30104 = torch_c.from_builtin_tensor %30103 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30105 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27659 = arith.constant 1 : index
%dim_27660 = tensor.dim %30105, %c1_27659 : tensor<4x?x4096xf16>
%30106 = flow.tensor.transfer %30105 : tensor<4x?x4096xf16>{%dim_27660} to #hal.device.promise<@__device_0>
%30107 = torch_c.from_builtin_tensor %30106 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27661 = torch.constant.int 1
%30108 = torch.aten.add.Tensor %30037, %30089, %int1_27661 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27662 = torch.constant.int 1
%30109 = torch.aten.add.Tensor %30108, %30092, %int1_27662 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27663 = torch.constant.int 1
%30110 = torch.aten.add.Tensor %30109, %30095, %int1_27663 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27664 = torch.constant.int 1
%30111 = torch.aten.add.Tensor %30110, %30098, %int1_27664 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27665 = torch.constant.int 1
%30112 = torch.aten.add.Tensor %30111, %30101, %int1_27665 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27666 = torch.constant.int 1
%30113 = torch.aten.add.Tensor %30112, %30104, %int1_27666 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27667 = torch.constant.int 1
%30114 = torch.aten.add.Tensor %30113, %30107, %int1_27667 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30115 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27668 = arith.constant 1 : index
%dim_27669 = tensor.dim %30115, %c1_27668 : tensor<4x?x4096xf16>
%30116 = flow.tensor.transfer %30115 : tensor<4x?x4096xf16>{%dim_27669} to #hal.device.promise<@__device_1>
%30117 = torch_c.from_builtin_tensor %30116 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30118 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27670 = arith.constant 1 : index
%dim_27671 = tensor.dim %30118, %c1_27670 : tensor<4x?x4096xf16>
%30119 = flow.tensor.transfer %30118 : tensor<4x?x4096xf16>{%dim_27671} to #hal.device.promise<@__device_1>
%30120 = torch_c.from_builtin_tensor %30119 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30121 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27672 = arith.constant 1 : index
%dim_27673 = tensor.dim %30121, %c1_27672 : tensor<4x?x4096xf16>
%30122 = flow.tensor.transfer %30121 : tensor<4x?x4096xf16>{%dim_27673} to #hal.device.promise<@__device_1>
%30123 = torch_c.from_builtin_tensor %30122 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30124 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27674 = arith.constant 1 : index
%dim_27675 = tensor.dim %30124, %c1_27674 : tensor<4x?x4096xf16>
%30125 = flow.tensor.transfer %30124 : tensor<4x?x4096xf16>{%dim_27675} to #hal.device.promise<@__device_1>
%30126 = torch_c.from_builtin_tensor %30125 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30127 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27676 = arith.constant 1 : index
%dim_27677 = tensor.dim %30127, %c1_27676 : tensor<4x?x4096xf16>
%30128 = flow.tensor.transfer %30127 : tensor<4x?x4096xf16>{%dim_27677} to #hal.device.promise<@__device_1>
%30129 = torch_c.from_builtin_tensor %30128 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30130 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27678 = arith.constant 1 : index
%dim_27679 = tensor.dim %30130, %c1_27678 : tensor<4x?x4096xf16>
%30131 = flow.tensor.transfer %30130 : tensor<4x?x4096xf16>{%dim_27679} to #hal.device.promise<@__device_1>
%30132 = torch_c.from_builtin_tensor %30131 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30133 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27680 = arith.constant 1 : index
%dim_27681 = tensor.dim %30133, %c1_27680 : tensor<4x?x4096xf16>
%30134 = flow.tensor.transfer %30133 : tensor<4x?x4096xf16>{%dim_27681} to #hal.device.promise<@__device_1>
%30135 = torch_c.from_builtin_tensor %30134 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27682 = torch.constant.int 1
%30136 = torch.aten.add.Tensor %30117, %30044, %int1_27682 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27683 = torch.constant.int 1
%30137 = torch.aten.add.Tensor %30136, %30120, %int1_27683 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27684 = torch.constant.int 1
%30138 = torch.aten.add.Tensor %30137, %30123, %int1_27684 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27685 = torch.constant.int 1
%30139 = torch.aten.add.Tensor %30138, %30126, %int1_27685 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27686 = torch.constant.int 1
%30140 = torch.aten.add.Tensor %30139, %30129, %int1_27686 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27687 = torch.constant.int 1
%30141 = torch.aten.add.Tensor %30140, %30132, %int1_27687 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27688 = torch.constant.int 1
%30142 = torch.aten.add.Tensor %30141, %30135, %int1_27688 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30143 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27689 = arith.constant 1 : index
%dim_27690 = tensor.dim %30143, %c1_27689 : tensor<4x?x4096xf16>
%30144 = flow.tensor.transfer %30143 : tensor<4x?x4096xf16>{%dim_27690} to #hal.device.promise<@__device_2>
%30145 = torch_c.from_builtin_tensor %30144 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30146 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27691 = arith.constant 1 : index
%dim_27692 = tensor.dim %30146, %c1_27691 : tensor<4x?x4096xf16>
%30147 = flow.tensor.transfer %30146 : tensor<4x?x4096xf16>{%dim_27692} to #hal.device.promise<@__device_2>
%30148 = torch_c.from_builtin_tensor %30147 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30149 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27693 = arith.constant 1 : index
%dim_27694 = tensor.dim %30149, %c1_27693 : tensor<4x?x4096xf16>
%30150 = flow.tensor.transfer %30149 : tensor<4x?x4096xf16>{%dim_27694} to #hal.device.promise<@__device_2>
%30151 = torch_c.from_builtin_tensor %30150 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30152 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27695 = arith.constant 1 : index
%dim_27696 = tensor.dim %30152, %c1_27695 : tensor<4x?x4096xf16>
%30153 = flow.tensor.transfer %30152 : tensor<4x?x4096xf16>{%dim_27696} to #hal.device.promise<@__device_2>
%30154 = torch_c.from_builtin_tensor %30153 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30155 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27697 = arith.constant 1 : index
%dim_27698 = tensor.dim %30155, %c1_27697 : tensor<4x?x4096xf16>
%30156 = flow.tensor.transfer %30155 : tensor<4x?x4096xf16>{%dim_27698} to #hal.device.promise<@__device_2>
%30157 = torch_c.from_builtin_tensor %30156 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30158 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27699 = arith.constant 1 : index
%dim_27700 = tensor.dim %30158, %c1_27699 : tensor<4x?x4096xf16>
%30159 = flow.tensor.transfer %30158 : tensor<4x?x4096xf16>{%dim_27700} to #hal.device.promise<@__device_2>
%30160 = torch_c.from_builtin_tensor %30159 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30161 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27701 = arith.constant 1 : index
%dim_27702 = tensor.dim %30161, %c1_27701 : tensor<4x?x4096xf16>
%30162 = flow.tensor.transfer %30161 : tensor<4x?x4096xf16>{%dim_27702} to #hal.device.promise<@__device_2>
%30163 = torch_c.from_builtin_tensor %30162 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27703 = torch.constant.int 1
%30164 = torch.aten.add.Tensor %30145, %30148, %int1_27703 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27704 = torch.constant.int 1
%30165 = torch.aten.add.Tensor %30164, %30051, %int1_27704 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27705 = torch.constant.int 1
%30166 = torch.aten.add.Tensor %30165, %30151, %int1_27705 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27706 = torch.constant.int 1
%30167 = torch.aten.add.Tensor %30166, %30154, %int1_27706 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27707 = torch.constant.int 1
%30168 = torch.aten.add.Tensor %30167, %30157, %int1_27707 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27708 = torch.constant.int 1
%30169 = torch.aten.add.Tensor %30168, %30160, %int1_27708 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27709 = torch.constant.int 1
%30170 = torch.aten.add.Tensor %30169, %30163, %int1_27709 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30171 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27710 = arith.constant 1 : index
%dim_27711 = tensor.dim %30171, %c1_27710 : tensor<4x?x4096xf16>
%30172 = flow.tensor.transfer %30171 : tensor<4x?x4096xf16>{%dim_27711} to #hal.device.promise<@__device_3>
%30173 = torch_c.from_builtin_tensor %30172 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30174 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27712 = arith.constant 1 : index
%dim_27713 = tensor.dim %30174, %c1_27712 : tensor<4x?x4096xf16>
%30175 = flow.tensor.transfer %30174 : tensor<4x?x4096xf16>{%dim_27713} to #hal.device.promise<@__device_3>
%30176 = torch_c.from_builtin_tensor %30175 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30177 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27714 = arith.constant 1 : index
%dim_27715 = tensor.dim %30177, %c1_27714 : tensor<4x?x4096xf16>
%30178 = flow.tensor.transfer %30177 : tensor<4x?x4096xf16>{%dim_27715} to #hal.device.promise<@__device_3>
%30179 = torch_c.from_builtin_tensor %30178 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30180 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27716 = arith.constant 1 : index
%dim_27717 = tensor.dim %30180, %c1_27716 : tensor<4x?x4096xf16>
%30181 = flow.tensor.transfer %30180 : tensor<4x?x4096xf16>{%dim_27717} to #hal.device.promise<@__device_3>
%30182 = torch_c.from_builtin_tensor %30181 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30183 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27718 = arith.constant 1 : index
%dim_27719 = tensor.dim %30183, %c1_27718 : tensor<4x?x4096xf16>
%30184 = flow.tensor.transfer %30183 : tensor<4x?x4096xf16>{%dim_27719} to #hal.device.promise<@__device_3>
%30185 = torch_c.from_builtin_tensor %30184 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30186 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27720 = arith.constant 1 : index
%dim_27721 = tensor.dim %30186, %c1_27720 : tensor<4x?x4096xf16>
%30187 = flow.tensor.transfer %30186 : tensor<4x?x4096xf16>{%dim_27721} to #hal.device.promise<@__device_3>
%30188 = torch_c.from_builtin_tensor %30187 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30189 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27722 = arith.constant 1 : index
%dim_27723 = tensor.dim %30189, %c1_27722 : tensor<4x?x4096xf16>
%30190 = flow.tensor.transfer %30189 : tensor<4x?x4096xf16>{%dim_27723} to #hal.device.promise<@__device_3>
%30191 = torch_c.from_builtin_tensor %30190 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27724 = torch.constant.int 1
%30192 = torch.aten.add.Tensor %30173, %30176, %int1_27724 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27725 = torch.constant.int 1
%30193 = torch.aten.add.Tensor %30192, %30179, %int1_27725 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27726 = torch.constant.int 1
%30194 = torch.aten.add.Tensor %30193, %30058, %int1_27726 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27727 = torch.constant.int 1
%30195 = torch.aten.add.Tensor %30194, %30182, %int1_27727 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27728 = torch.constant.int 1
%30196 = torch.aten.add.Tensor %30195, %30185, %int1_27728 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27729 = torch.constant.int 1
%30197 = torch.aten.add.Tensor %30196, %30188, %int1_27729 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27730 = torch.constant.int 1
%30198 = torch.aten.add.Tensor %30197, %30191, %int1_27730 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30199 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27731 = arith.constant 1 : index
%dim_27732 = tensor.dim %30199, %c1_27731 : tensor<4x?x4096xf16>
%30200 = flow.tensor.transfer %30199 : tensor<4x?x4096xf16>{%dim_27732} to #hal.device.promise<@__device_4>
%30201 = torch_c.from_builtin_tensor %30200 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30202 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27733 = arith.constant 1 : index
%dim_27734 = tensor.dim %30202, %c1_27733 : tensor<4x?x4096xf16>
%30203 = flow.tensor.transfer %30202 : tensor<4x?x4096xf16>{%dim_27734} to #hal.device.promise<@__device_4>
%30204 = torch_c.from_builtin_tensor %30203 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30205 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27735 = arith.constant 1 : index
%dim_27736 = tensor.dim %30205, %c1_27735 : tensor<4x?x4096xf16>
%30206 = flow.tensor.transfer %30205 : tensor<4x?x4096xf16>{%dim_27736} to #hal.device.promise<@__device_4>
%30207 = torch_c.from_builtin_tensor %30206 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30208 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27737 = arith.constant 1 : index
%dim_27738 = tensor.dim %30208, %c1_27737 : tensor<4x?x4096xf16>
%30209 = flow.tensor.transfer %30208 : tensor<4x?x4096xf16>{%dim_27738} to #hal.device.promise<@__device_4>
%30210 = torch_c.from_builtin_tensor %30209 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30211 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27739 = arith.constant 1 : index
%dim_27740 = tensor.dim %30211, %c1_27739 : tensor<4x?x4096xf16>
%30212 = flow.tensor.transfer %30211 : tensor<4x?x4096xf16>{%dim_27740} to #hal.device.promise<@__device_4>
%30213 = torch_c.from_builtin_tensor %30212 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30214 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27741 = arith.constant 1 : index
%dim_27742 = tensor.dim %30214, %c1_27741 : tensor<4x?x4096xf16>
%30215 = flow.tensor.transfer %30214 : tensor<4x?x4096xf16>{%dim_27742} to #hal.device.promise<@__device_4>
%30216 = torch_c.from_builtin_tensor %30215 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30217 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27743 = arith.constant 1 : index
%dim_27744 = tensor.dim %30217, %c1_27743 : tensor<4x?x4096xf16>
%30218 = flow.tensor.transfer %30217 : tensor<4x?x4096xf16>{%dim_27744} to #hal.device.promise<@__device_4>
%30219 = torch_c.from_builtin_tensor %30218 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27745 = torch.constant.int 1
%30220 = torch.aten.add.Tensor %30201, %30204, %int1_27745 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27746 = torch.constant.int 1
%30221 = torch.aten.add.Tensor %30220, %30207, %int1_27746 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27747 = torch.constant.int 1
%30222 = torch.aten.add.Tensor %30221, %30210, %int1_27747 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27748 = torch.constant.int 1
%30223 = torch.aten.add.Tensor %30222, %30065, %int1_27748 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27749 = torch.constant.int 1
%30224 = torch.aten.add.Tensor %30223, %30213, %int1_27749 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27750 = torch.constant.int 1
%30225 = torch.aten.add.Tensor %30224, %30216, %int1_27750 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27751 = torch.constant.int 1
%30226 = torch.aten.add.Tensor %30225, %30219, %int1_27751 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30227 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27752 = arith.constant 1 : index
%dim_27753 = tensor.dim %30227, %c1_27752 : tensor<4x?x4096xf16>
%30228 = flow.tensor.transfer %30227 : tensor<4x?x4096xf16>{%dim_27753} to #hal.device.promise<@__device_5>
%30229 = torch_c.from_builtin_tensor %30228 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30230 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27754 = arith.constant 1 : index
%dim_27755 = tensor.dim %30230, %c1_27754 : tensor<4x?x4096xf16>
%30231 = flow.tensor.transfer %30230 : tensor<4x?x4096xf16>{%dim_27755} to #hal.device.promise<@__device_5>
%30232 = torch_c.from_builtin_tensor %30231 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30233 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27756 = arith.constant 1 : index
%dim_27757 = tensor.dim %30233, %c1_27756 : tensor<4x?x4096xf16>
%30234 = flow.tensor.transfer %30233 : tensor<4x?x4096xf16>{%dim_27757} to #hal.device.promise<@__device_5>
%30235 = torch_c.from_builtin_tensor %30234 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30236 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27758 = arith.constant 1 : index
%dim_27759 = tensor.dim %30236, %c1_27758 : tensor<4x?x4096xf16>
%30237 = flow.tensor.transfer %30236 : tensor<4x?x4096xf16>{%dim_27759} to #hal.device.promise<@__device_5>
%30238 = torch_c.from_builtin_tensor %30237 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30239 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27760 = arith.constant 1 : index
%dim_27761 = tensor.dim %30239, %c1_27760 : tensor<4x?x4096xf16>
%30240 = flow.tensor.transfer %30239 : tensor<4x?x4096xf16>{%dim_27761} to #hal.device.promise<@__device_5>
%30241 = torch_c.from_builtin_tensor %30240 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30242 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27762 = arith.constant 1 : index
%dim_27763 = tensor.dim %30242, %c1_27762 : tensor<4x?x4096xf16>
%30243 = flow.tensor.transfer %30242 : tensor<4x?x4096xf16>{%dim_27763} to #hal.device.promise<@__device_5>
%30244 = torch_c.from_builtin_tensor %30243 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30245 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27764 = arith.constant 1 : index
%dim_27765 = tensor.dim %30245, %c1_27764 : tensor<4x?x4096xf16>
%30246 = flow.tensor.transfer %30245 : tensor<4x?x4096xf16>{%dim_27765} to #hal.device.promise<@__device_5>
%30247 = torch_c.from_builtin_tensor %30246 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27766 = torch.constant.int 1
%30248 = torch.aten.add.Tensor %30229, %30232, %int1_27766 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27767 = torch.constant.int 1
%30249 = torch.aten.add.Tensor %30248, %30235, %int1_27767 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27768 = torch.constant.int 1
%30250 = torch.aten.add.Tensor %30249, %30238, %int1_27768 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27769 = torch.constant.int 1
%30251 = torch.aten.add.Tensor %30250, %30241, %int1_27769 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27770 = torch.constant.int 1
%30252 = torch.aten.add.Tensor %30251, %30072, %int1_27770 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27771 = torch.constant.int 1
%30253 = torch.aten.add.Tensor %30252, %30244, %int1_27771 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27772 = torch.constant.int 1
%30254 = torch.aten.add.Tensor %30253, %30247, %int1_27772 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30255 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27773 = arith.constant 1 : index
%dim_27774 = tensor.dim %30255, %c1_27773 : tensor<4x?x4096xf16>
%30256 = flow.tensor.transfer %30255 : tensor<4x?x4096xf16>{%dim_27774} to #hal.device.promise<@__device_6>
%30257 = torch_c.from_builtin_tensor %30256 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30258 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27775 = arith.constant 1 : index
%dim_27776 = tensor.dim %30258, %c1_27775 : tensor<4x?x4096xf16>
%30259 = flow.tensor.transfer %30258 : tensor<4x?x4096xf16>{%dim_27776} to #hal.device.promise<@__device_6>
%30260 = torch_c.from_builtin_tensor %30259 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30261 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27777 = arith.constant 1 : index
%dim_27778 = tensor.dim %30261, %c1_27777 : tensor<4x?x4096xf16>
%30262 = flow.tensor.transfer %30261 : tensor<4x?x4096xf16>{%dim_27778} to #hal.device.promise<@__device_6>
%30263 = torch_c.from_builtin_tensor %30262 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30263, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30264 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27779 = arith.constant 1 : index
%dim_27780 = tensor.dim %30264, %c1_27779 : tensor<4x?x4096xf16>
%30265 = flow.tensor.transfer %30264 : tensor<4x?x4096xf16>{%dim_27780} to #hal.device.promise<@__device_6>
%30266 = torch_c.from_builtin_tensor %30265 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30267 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27781 = arith.constant 1 : index
%dim_27782 = tensor.dim %30267, %c1_27781 : tensor<4x?x4096xf16>
%30268 = flow.tensor.transfer %30267 : tensor<4x?x4096xf16>{%dim_27782} to #hal.device.promise<@__device_6>
%30269 = torch_c.from_builtin_tensor %30268 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30270 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27783 = arith.constant 1 : index
%dim_27784 = tensor.dim %30270, %c1_27783 : tensor<4x?x4096xf16>
%30271 = flow.tensor.transfer %30270 : tensor<4x?x4096xf16>{%dim_27784} to #hal.device.promise<@__device_6>
%30272 = torch_c.from_builtin_tensor %30271 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30273 = torch_c.to_builtin_tensor %30086 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27785 = arith.constant 1 : index
%dim_27786 = tensor.dim %30273, %c1_27785 : tensor<4x?x4096xf16>
%30274 = flow.tensor.transfer %30273 : tensor<4x?x4096xf16>{%dim_27786} to #hal.device.promise<@__device_6>
%30275 = torch_c.from_builtin_tensor %30274 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27787 = torch.constant.int 1
%30276 = torch.aten.add.Tensor %30257, %30260, %int1_27787 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27788 = torch.constant.int 1
%30277 = torch.aten.add.Tensor %30276, %30263, %int1_27788 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27789 = torch.constant.int 1
%30278 = torch.aten.add.Tensor %30277, %30266, %int1_27789 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27790 = torch.constant.int 1
%30279 = torch.aten.add.Tensor %30278, %30269, %int1_27790 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27791 = torch.constant.int 1
%30280 = torch.aten.add.Tensor %30279, %30272, %int1_27791 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27792 = torch.constant.int 1
%30281 = torch.aten.add.Tensor %30280, %30079, %int1_27792 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27793 = torch.constant.int 1
%30282 = torch.aten.add.Tensor %30281, %30275, %int1_27793 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30283 = torch_c.to_builtin_tensor %30037 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27794 = arith.constant 1 : index
%dim_27795 = tensor.dim %30283, %c1_27794 : tensor<4x?x4096xf16>
%30284 = flow.tensor.transfer %30283 : tensor<4x?x4096xf16>{%dim_27795} to #hal.device.promise<@__device_7>
%30285 = torch_c.from_builtin_tensor %30284 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30286 = torch_c.to_builtin_tensor %30044 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27796 = arith.constant 1 : index
%dim_27797 = tensor.dim %30286, %c1_27796 : tensor<4x?x4096xf16>
%30287 = flow.tensor.transfer %30286 : tensor<4x?x4096xf16>{%dim_27797} to #hal.device.promise<@__device_7>
%30288 = torch_c.from_builtin_tensor %30287 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30289 = torch_c.to_builtin_tensor %30051 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27798 = arith.constant 1 : index
%dim_27799 = tensor.dim %30289, %c1_27798 : tensor<4x?x4096xf16>
%30290 = flow.tensor.transfer %30289 : tensor<4x?x4096xf16>{%dim_27799} to #hal.device.promise<@__device_7>
%30291 = torch_c.from_builtin_tensor %30290 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30292 = torch_c.to_builtin_tensor %30058 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27800 = arith.constant 1 : index
%dim_27801 = tensor.dim %30292, %c1_27800 : tensor<4x?x4096xf16>
%30293 = flow.tensor.transfer %30292 : tensor<4x?x4096xf16>{%dim_27801} to #hal.device.promise<@__device_7>
%30294 = torch_c.from_builtin_tensor %30293 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30295 = torch_c.to_builtin_tensor %30065 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27802 = arith.constant 1 : index
%dim_27803 = tensor.dim %30295, %c1_27802 : tensor<4x?x4096xf16>
%30296 = flow.tensor.transfer %30295 : tensor<4x?x4096xf16>{%dim_27803} to #hal.device.promise<@__device_7>
%30297 = torch_c.from_builtin_tensor %30296 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30298 = torch_c.to_builtin_tensor %30072 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27804 = arith.constant 1 : index
%dim_27805 = tensor.dim %30298, %c1_27804 : tensor<4x?x4096xf16>
%30299 = flow.tensor.transfer %30298 : tensor<4x?x4096xf16>{%dim_27805} to #hal.device.promise<@__device_7>
%30300 = torch_c.from_builtin_tensor %30299 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%30301 = torch_c.to_builtin_tensor %30079 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_27806 = arith.constant 1 : index
%dim_27807 = tensor.dim %30301, %c1_27806 : tensor<4x?x4096xf16>
%30302 = flow.tensor.transfer %30301 : tensor<4x?x4096xf16>{%dim_27807} to #hal.device.promise<@__device_7>
%30303 = torch_c.from_builtin_tensor %30302 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27808 = torch.constant.int 1
%30304 = torch.aten.add.Tensor %30285, %30288, %int1_27808 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27809 = torch.constant.int 1
%30305 = torch.aten.add.Tensor %30304, %30291, %int1_27809 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27810 = torch.constant.int 1
%30306 = torch.aten.add.Tensor %30305, %30294, %int1_27810 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27811 = torch.constant.int 1
%30307 = torch.aten.add.Tensor %30306, %30297, %int1_27811 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27812 = torch.constant.int 1
%30308 = torch.aten.add.Tensor %30307, %30300, %int1_27812 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27813 = torch.constant.int 1
%30309 = torch.aten.add.Tensor %30308, %30303, %int1_27813 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27814 = torch.constant.int 1
%30310 = torch.aten.add.Tensor %30309, %30086, %int1_27814 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27815 = torch.constant.int 1
%30311 = torch.aten.add.Tensor %29791, %30114, %int1_27815 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27816 = torch.constant.int 1
%30312 = torch.aten.add.Tensor %29792, %30142, %int1_27816 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27817 = torch.constant.int 1
%30313 = torch.aten.add.Tensor %29793, %30170, %int1_27817 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27818 = torch.constant.int 1
%30314 = torch.aten.add.Tensor %29794, %30198, %int1_27818 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27819 = torch.constant.int 1
%30315 = torch.aten.add.Tensor %29795, %30226, %int1_27819 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27820 = torch.constant.int 1
%30316 = torch.aten.add.Tensor %29796, %30254, %int1_27820 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27821 = torch.constant.int 1
%30317 = torch.aten.add.Tensor %29797, %30282, %int1_27821 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27822 = torch.constant.int 1
%30318 = torch.aten.add.Tensor %29798, %30310, %int1_27822 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_27823 = torch.constant.int 6
%30319 = torch.prims.convert_element_type %30311, %int6_27823 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27824 = torch.constant.int 6
%30320 = torch.prims.convert_element_type %30312, %int6_27824 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27825 = torch.constant.int 6
%30321 = torch.prims.convert_element_type %30313, %int6_27825 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27826 = torch.constant.int 6
%30322 = torch.prims.convert_element_type %30314, %int6_27826 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27827 = torch.constant.int 6
%30323 = torch.prims.convert_element_type %30315, %int6_27827 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27828 = torch.constant.int 6
%30324 = torch.prims.convert_element_type %30316, %int6_27828 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27829 = torch.constant.int 6
%30325 = torch.prims.convert_element_type %30317, %int6_27829 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_27830 = torch.constant.int 6
%30326 = torch.prims.convert_element_type %30318, %int6_27830 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27831 = torch.constant.int 2
%30327 = torch.aten.pow.Tensor_Scalar %30319, %int2_27831 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30327, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27832 = torch.constant.int 2
%30328 = torch.aten.pow.Tensor_Scalar %30320, %int2_27832 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27833 = torch.constant.int 2
%30329 = torch.aten.pow.Tensor_Scalar %30321, %int2_27833 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27834 = torch.constant.int 2
%30330 = torch.aten.pow.Tensor_Scalar %30322, %int2_27834 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27835 = torch.constant.int 2
%30331 = torch.aten.pow.Tensor_Scalar %30323, %int2_27835 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27836 = torch.constant.int 2
%30332 = torch.aten.pow.Tensor_Scalar %30324, %int2_27836 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27837 = torch.constant.int 2
%30333 = torch.aten.pow.Tensor_Scalar %30325, %int2_27837 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30333, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_27838 = torch.constant.int 2
%30334 = torch.aten.pow.Tensor_Scalar %30326, %int2_27838 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_27839 = torch.constant.int -1
%30335 = torch.prim.ListConstruct %int-1_27839 : (!torch.int) -> !torch.list<int>
%true_27840 = torch.constant.bool true
%none_27841 = torch.constant.none
%30336 = torch.aten.mean.dim %30327, %30335, %true_27840, %none_27841 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27842 = torch.constant.int -1
%30337 = torch.prim.ListConstruct %int-1_27842 : (!torch.int) -> !torch.list<int>
%true_27843 = torch.constant.bool true
%none_27844 = torch.constant.none
%30338 = torch.aten.mean.dim %30328, %30337, %true_27843, %none_27844 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27845 = torch.constant.int -1
%30339 = torch.prim.ListConstruct %int-1_27845 : (!torch.int) -> !torch.list<int>
%true_27846 = torch.constant.bool true
%none_27847 = torch.constant.none
%30340 = torch.aten.mean.dim %30329, %30339, %true_27846, %none_27847 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27848 = torch.constant.int -1
%30341 = torch.prim.ListConstruct %int-1_27848 : (!torch.int) -> !torch.list<int>
%true_27849 = torch.constant.bool true
%none_27850 = torch.constant.none
%30342 = torch.aten.mean.dim %30330, %30341, %true_27849, %none_27850 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27851 = torch.constant.int -1
%30343 = torch.prim.ListConstruct %int-1_27851 : (!torch.int) -> !torch.list<int>
%true_27852 = torch.constant.bool true
%none_27853 = torch.constant.none
%30344 = torch.aten.mean.dim %30331, %30343, %true_27852, %none_27853 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27854 = torch.constant.int -1
%30345 = torch.prim.ListConstruct %int-1_27854 : (!torch.int) -> !torch.list<int>
%true_27855 = torch.constant.bool true
%none_27856 = torch.constant.none
%30346 = torch.aten.mean.dim %30332, %30345, %true_27855, %none_27856 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27857 = torch.constant.int -1
%30347 = torch.prim.ListConstruct %int-1_27857 : (!torch.int) -> !torch.list<int>
%true_27858 = torch.constant.bool true
%none_27859 = torch.constant.none
%30348 = torch.aten.mean.dim %30333, %30347, %true_27858, %none_27859 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_27860 = torch.constant.int -1
%30349 = torch.prim.ListConstruct %int-1_27860 : (!torch.int) -> !torch.list<int>
%true_27861 = torch.constant.bool true
%none_27862 = torch.constant.none
%30350 = torch.aten.mean.dim %30334, %30349, %true_27861, %none_27862 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27863 = torch.constant.float 9.9999997473787516E-6
%int1_27864 = torch.constant.int 1
%30351 = torch.aten.add.Scalar %30336, %float9.999990e-06_27863, %int1_27864 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27865 = torch.constant.float 9.9999997473787516E-6
%int1_27866 = torch.constant.int 1
%30352 = torch.aten.add.Scalar %30338, %float9.999990e-06_27865, %int1_27866 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27867 = torch.constant.float 9.9999997473787516E-6
%int1_27868 = torch.constant.int 1
%30353 = torch.aten.add.Scalar %30340, %float9.999990e-06_27867, %int1_27868 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27869 = torch.constant.float 9.9999997473787516E-6
%int1_27870 = torch.constant.int 1
%30354 = torch.aten.add.Scalar %30342, %float9.999990e-06_27869, %int1_27870 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27871 = torch.constant.float 9.9999997473787516E-6
%int1_27872 = torch.constant.int 1
%30355 = torch.aten.add.Scalar %30344, %float9.999990e-06_27871, %int1_27872 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27873 = torch.constant.float 9.9999997473787516E-6
%int1_27874 = torch.constant.int 1
%30356 = torch.aten.add.Scalar %30346, %float9.999990e-06_27873, %int1_27874 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27875 = torch.constant.float 9.9999997473787516E-6
%int1_27876 = torch.constant.int 1
%30357 = torch.aten.add.Scalar %30348, %float9.999990e-06_27875, %int1_27876 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_27877 = torch.constant.float 9.9999997473787516E-6
%int1_27878 = torch.constant.int 1
%30358 = torch.aten.add.Scalar %30350, %float9.999990e-06_27877, %int1_27878 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30358, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30359 = torch.aten.rsqrt %30351 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30360 = torch.aten.rsqrt %30352 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30361 = torch.aten.rsqrt %30353 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30362 = torch.aten.rsqrt %30354 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30363 = torch.aten.rsqrt %30355 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30364 = torch.aten.rsqrt %30356 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30364, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30365 = torch.aten.rsqrt %30357 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30366 = torch.aten.rsqrt %30358 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %30366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%30367 = torch.aten.mul.Tensor %30319, %30359 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30368 = torch.aten.mul.Tensor %30320, %30360 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30369 = torch.aten.mul.Tensor %30321, %30361 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30370 = torch.aten.mul.Tensor %30322, %30362 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30371 = torch.aten.mul.Tensor %30323, %30363 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30372 = torch.aten.mul.Tensor %30324, %30364 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30373 = torch.aten.mul.Tensor %30325, %30365 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30374 = torch.aten.mul.Tensor %30326, %30366 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30375 = torch.aten.mul.Tensor %1088, %30367 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30376 = torch.aten.mul.Tensor %1089, %30368 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30377 = torch.aten.mul.Tensor %1090, %30369 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30378 = torch.aten.mul.Tensor %1091, %30370 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30379 = torch.aten.mul.Tensor %1092, %30371 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30380 = torch.aten.mul.Tensor %1093, %30372 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30381 = torch.aten.mul.Tensor %1094, %30373 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%30382 = torch.aten.mul.Tensor %1095, %30374 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %30382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_27879 = torch.constant.int 5
%30383 = torch.prims.convert_element_type %30375, %int5_27879 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27880 = torch.constant.int 5
%30384 = torch.prims.convert_element_type %30376, %int5_27880 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27881 = torch.constant.int 5
%30385 = torch.prims.convert_element_type %30377, %int5_27881 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27882 = torch.constant.int 5
%30386 = torch.prims.convert_element_type %30378, %int5_27882 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27883 = torch.constant.int 5
%30387 = torch.prims.convert_element_type %30379, %int5_27883 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27884 = torch.constant.int 5
%30388 = torch.prims.convert_element_type %30380, %int5_27884 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27885 = torch.constant.int 5
%30389 = torch.prims.convert_element_type %30381, %int5_27885 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_27886 = torch.constant.int 5
%30390 = torch.prims.convert_element_type %30382, %int5_27886 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %30390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_27887 = torch.constant.int 1
%int0_27888 = torch.constant.int 0
%30391 = torch.prim.ListConstruct %int1_27887, %int0_27888 : (!torch.int, !torch.int) -> !torch.list<int>
%30392 = torch.aten.permute %1096, %30391 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27889 = torch.constant.int 1
%int0_27890 = torch.constant.int 0
%30393 = torch.prim.ListConstruct %int1_27889, %int0_27890 : (!torch.int, !torch.int) -> !torch.list<int>
%30394 = torch.aten.permute %1097, %30393 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27891 = torch.constant.int 1
%int0_27892 = torch.constant.int 0
%30395 = torch.prim.ListConstruct %int1_27891, %int0_27892 : (!torch.int, !torch.int) -> !torch.list<int>
%30396 = torch.aten.permute %1098, %30395 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27893 = torch.constant.int 1
%int0_27894 = torch.constant.int 0
%30397 = torch.prim.ListConstruct %int1_27893, %int0_27894 : (!torch.int, !torch.int) -> !torch.list<int>
%30398 = torch.aten.permute %1099, %30397 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27895 = torch.constant.int 1
%int0_27896 = torch.constant.int 0
%30399 = torch.prim.ListConstruct %int1_27895, %int0_27896 : (!torch.int, !torch.int) -> !torch.list<int>
%30400 = torch.aten.permute %1100, %30399 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27897 = torch.constant.int 1
%int0_27898 = torch.constant.int 0
%30401 = torch.prim.ListConstruct %int1_27897, %int0_27898 : (!torch.int, !torch.int) -> !torch.list<int>
%30402 = torch.aten.permute %1101, %30401 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27899 = torch.constant.int 1
%int0_27900 = torch.constant.int 0
%30403 = torch.prim.ListConstruct %int1_27899, %int0_27900 : (!torch.int, !torch.int) -> !torch.list<int>
%30404 = torch.aten.permute %1102, %30403 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_27901 = torch.constant.int 1
%int0_27902 = torch.constant.int 0
%30405 = torch.prim.ListConstruct %int1_27901, %int0_27902 : (!torch.int, !torch.int) -> !torch.list<int>
%30406 = torch.aten.permute %1103, %30405 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_27903 = torch.constant.int 4
%30407 = torch.aten.mul.int %int4_27903, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27904 = torch.constant.int 4096
%30408 = torch.prim.ListConstruct %30407, %int4096_27904 : (!torch.int, !torch.int) -> !torch.list<int>
%30409 = torch.aten.view %30383, %30408 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30409, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30410 = torch.aten.mm %30409, %30392 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30410, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27905 = torch.constant.int 4
%int512_27906 = torch.constant.int 512
%30411 = torch.prim.ListConstruct %int4_27905, %2482, %int512_27906 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30412 = torch.aten.view %30410, %30411 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27907 = torch.constant.int 4
%30413 = torch.aten.mul.int %int4_27907, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27908 = torch.constant.int 4096
%30414 = torch.prim.ListConstruct %30413, %int4096_27908 : (!torch.int, !torch.int) -> !torch.list<int>
%30415 = torch.aten.view %30384, %30414 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30415, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30416 = torch.aten.mm %30415, %30394 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30416, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27909 = torch.constant.int 4
%int512_27910 = torch.constant.int 512
%30417 = torch.prim.ListConstruct %int4_27909, %2482, %int512_27910 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30418 = torch.aten.view %30416, %30417 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27911 = torch.constant.int 4
%30419 = torch.aten.mul.int %int4_27911, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27912 = torch.constant.int 4096
%30420 = torch.prim.ListConstruct %30419, %int4096_27912 : (!torch.int, !torch.int) -> !torch.list<int>
%30421 = torch.aten.view %30385, %30420 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30421, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30422 = torch.aten.mm %30421, %30396 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30422, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27913 = torch.constant.int 4
%int512_27914 = torch.constant.int 512
%30423 = torch.prim.ListConstruct %int4_27913, %2482, %int512_27914 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30424 = torch.aten.view %30422, %30423 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27915 = torch.constant.int 4
%30425 = torch.aten.mul.int %int4_27915, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27916 = torch.constant.int 4096
%30426 = torch.prim.ListConstruct %30425, %int4096_27916 : (!torch.int, !torch.int) -> !torch.list<int>
%30427 = torch.aten.view %30386, %30426 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30427, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30428 = torch.aten.mm %30427, %30398 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30428, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27917 = torch.constant.int 4
%int512_27918 = torch.constant.int 512
%30429 = torch.prim.ListConstruct %int4_27917, %2482, %int512_27918 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30430 = torch.aten.view %30428, %30429 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27919 = torch.constant.int 4
%30431 = torch.aten.mul.int %int4_27919, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27920 = torch.constant.int 4096
%30432 = torch.prim.ListConstruct %30431, %int4096_27920 : (!torch.int, !torch.int) -> !torch.list<int>
%30433 = torch.aten.view %30387, %30432 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30433, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30434 = torch.aten.mm %30433, %30400 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30434, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27921 = torch.constant.int 4
%int512_27922 = torch.constant.int 512
%30435 = torch.prim.ListConstruct %int4_27921, %2482, %int512_27922 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30436 = torch.aten.view %30434, %30435 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27923 = torch.constant.int 4
%30437 = torch.aten.mul.int %int4_27923, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27924 = torch.constant.int 4096
%30438 = torch.prim.ListConstruct %30437, %int4096_27924 : (!torch.int, !torch.int) -> !torch.list<int>
%30439 = torch.aten.view %30388, %30438 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30439, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30440 = torch.aten.mm %30439, %30402 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30440, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27925 = torch.constant.int 4
%int512_27926 = torch.constant.int 512
%30441 = torch.prim.ListConstruct %int4_27925, %2482, %int512_27926 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30442 = torch.aten.view %30440, %30441 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27927 = torch.constant.int 4
%30443 = torch.aten.mul.int %int4_27927, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27928 = torch.constant.int 4096
%30444 = torch.prim.ListConstruct %30443, %int4096_27928 : (!torch.int, !torch.int) -> !torch.list<int>
%30445 = torch.aten.view %30389, %30444 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30445, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30446 = torch.aten.mm %30445, %30404 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30446, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27929 = torch.constant.int 4
%int512_27930 = torch.constant.int 512
%30447 = torch.prim.ListConstruct %int4_27929, %2482, %int512_27930 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30448 = torch.aten.view %30446, %30447 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_27931 = torch.constant.int 4
%30449 = torch.aten.mul.int %int4_27931, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27932 = torch.constant.int 4096
%30450 = torch.prim.ListConstruct %30449, %int4096_27932 : (!torch.int, !torch.int) -> !torch.list<int>
%30451 = torch.aten.view %30390, %30450 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30451, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30452 = torch.aten.mm %30451, %30406 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %30452, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_27933 = torch.constant.int 4
%int512_27934 = torch.constant.int 512
%30453 = torch.prim.ListConstruct %int4_27933, %2482, %int512_27934 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30454 = torch.aten.view %30452, %30453 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %30454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_27935 = torch.constant.int 1
%int0_27936 = torch.constant.int 0
%30455 = torch.prim.ListConstruct %int1_27935, %int0_27936 : (!torch.int, !torch.int) -> !torch.list<int>
%30456 = torch.aten.permute %1104, %30455 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27937 = torch.constant.int 1
%int0_27938 = torch.constant.int 0
%30457 = torch.prim.ListConstruct %int1_27937, %int0_27938 : (!torch.int, !torch.int) -> !torch.list<int>
%30458 = torch.aten.permute %1105, %30457 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27939 = torch.constant.int 1
%int0_27940 = torch.constant.int 0
%30459 = torch.prim.ListConstruct %int1_27939, %int0_27940 : (!torch.int, !torch.int) -> !torch.list<int>
%30460 = torch.aten.permute %1106, %30459 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27941 = torch.constant.int 1
%int0_27942 = torch.constant.int 0
%30461 = torch.prim.ListConstruct %int1_27941, %int0_27942 : (!torch.int, !torch.int) -> !torch.list<int>
%30462 = torch.aten.permute %1107, %30461 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27943 = torch.constant.int 1
%int0_27944 = torch.constant.int 0
%30463 = torch.prim.ListConstruct %int1_27943, %int0_27944 : (!torch.int, !torch.int) -> !torch.list<int>
%30464 = torch.aten.permute %1108, %30463 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27945 = torch.constant.int 1
%int0_27946 = torch.constant.int 0
%30465 = torch.prim.ListConstruct %int1_27945, %int0_27946 : (!torch.int, !torch.int) -> !torch.list<int>
%30466 = torch.aten.permute %1109, %30465 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27947 = torch.constant.int 1
%int0_27948 = torch.constant.int 0
%30467 = torch.prim.ListConstruct %int1_27947, %int0_27948 : (!torch.int, !torch.int) -> !torch.list<int>
%30468 = torch.aten.permute %1110, %30467 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27949 = torch.constant.int 1
%int0_27950 = torch.constant.int 0
%30469 = torch.prim.ListConstruct %int1_27949, %int0_27950 : (!torch.int, !torch.int) -> !torch.list<int>
%30470 = torch.aten.permute %1111, %30469 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_27951 = torch.constant.int 4
%30471 = torch.aten.mul.int %int4_27951, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27952 = torch.constant.int 4096
%30472 = torch.prim.ListConstruct %30471, %int4096_27952 : (!torch.int, !torch.int) -> !torch.list<int>
%30473 = torch.aten.view %30383, %30472 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30473, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30474 = torch.aten.mm %30473, %30456 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30474, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27953 = torch.constant.int 4
%int128_27954 = torch.constant.int 128
%30475 = torch.prim.ListConstruct %int4_27953, %2482, %int128_27954 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30476 = torch.aten.view %30474, %30475 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27955 = torch.constant.int 4
%30477 = torch.aten.mul.int %int4_27955, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27956 = torch.constant.int 4096
%30478 = torch.prim.ListConstruct %30477, %int4096_27956 : (!torch.int, !torch.int) -> !torch.list<int>
%30479 = torch.aten.view %30384, %30478 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30479, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30480 = torch.aten.mm %30479, %30458 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30480, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27957 = torch.constant.int 4
%int128_27958 = torch.constant.int 128
%30481 = torch.prim.ListConstruct %int4_27957, %2482, %int128_27958 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30482 = torch.aten.view %30480, %30481 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27959 = torch.constant.int 4
%30483 = torch.aten.mul.int %int4_27959, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27960 = torch.constant.int 4096
%30484 = torch.prim.ListConstruct %30483, %int4096_27960 : (!torch.int, !torch.int) -> !torch.list<int>
%30485 = torch.aten.view %30385, %30484 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30485, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30486 = torch.aten.mm %30485, %30460 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30486, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27961 = torch.constant.int 4
%int128_27962 = torch.constant.int 128
%30487 = torch.prim.ListConstruct %int4_27961, %2482, %int128_27962 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30488 = torch.aten.view %30486, %30487 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27963 = torch.constant.int 4
%30489 = torch.aten.mul.int %int4_27963, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27964 = torch.constant.int 4096
%30490 = torch.prim.ListConstruct %30489, %int4096_27964 : (!torch.int, !torch.int) -> !torch.list<int>
%30491 = torch.aten.view %30386, %30490 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30491, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30492 = torch.aten.mm %30491, %30462 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30492, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27965 = torch.constant.int 4
%int128_27966 = torch.constant.int 128
%30493 = torch.prim.ListConstruct %int4_27965, %2482, %int128_27966 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30494 = torch.aten.view %30492, %30493 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27967 = torch.constant.int 4
%30495 = torch.aten.mul.int %int4_27967, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27968 = torch.constant.int 4096
%30496 = torch.prim.ListConstruct %30495, %int4096_27968 : (!torch.int, !torch.int) -> !torch.list<int>
%30497 = torch.aten.view %30387, %30496 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30497, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30498 = torch.aten.mm %30497, %30464 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30498, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27969 = torch.constant.int 4
%int128_27970 = torch.constant.int 128
%30499 = torch.prim.ListConstruct %int4_27969, %2482, %int128_27970 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30500 = torch.aten.view %30498, %30499 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27971 = torch.constant.int 4
%30501 = torch.aten.mul.int %int4_27971, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27972 = torch.constant.int 4096
%30502 = torch.prim.ListConstruct %30501, %int4096_27972 : (!torch.int, !torch.int) -> !torch.list<int>
%30503 = torch.aten.view %30388, %30502 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30503, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30504 = torch.aten.mm %30503, %30466 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30504, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27973 = torch.constant.int 4
%int128_27974 = torch.constant.int 128
%30505 = torch.prim.ListConstruct %int4_27973, %2482, %int128_27974 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30506 = torch.aten.view %30504, %30505 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27975 = torch.constant.int 4
%30507 = torch.aten.mul.int %int4_27975, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27976 = torch.constant.int 4096
%30508 = torch.prim.ListConstruct %30507, %int4096_27976 : (!torch.int, !torch.int) -> !torch.list<int>
%30509 = torch.aten.view %30389, %30508 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30509, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30510 = torch.aten.mm %30509, %30468 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30510, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27977 = torch.constant.int 4
%int128_27978 = torch.constant.int 128
%30511 = torch.prim.ListConstruct %int4_27977, %2482, %int128_27978 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30512 = torch.aten.view %30510, %30511 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_27979 = torch.constant.int 4
%30513 = torch.aten.mul.int %int4_27979, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_27980 = torch.constant.int 4096
%30514 = torch.prim.ListConstruct %30513, %int4096_27980 : (!torch.int, !torch.int) -> !torch.list<int>
%30515 = torch.aten.view %30390, %30514 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30515, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30516 = torch.aten.mm %30515, %30470 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30516, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_27981 = torch.constant.int 4
%int128_27982 = torch.constant.int 128
%30517 = torch.prim.ListConstruct %int4_27981, %2482, %int128_27982 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30518 = torch.aten.view %30516, %30517 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_27983 = torch.constant.int 1
%int0_27984 = torch.constant.int 0
%30519 = torch.prim.ListConstruct %int1_27983, %int0_27984 : (!torch.int, !torch.int) -> !torch.list<int>
%30520 = torch.aten.permute %1112, %30519 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27985 = torch.constant.int 1
%int0_27986 = torch.constant.int 0
%30521 = torch.prim.ListConstruct %int1_27985, %int0_27986 : (!torch.int, !torch.int) -> !torch.list<int>
%30522 = torch.aten.permute %1113, %30521 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27987 = torch.constant.int 1
%int0_27988 = torch.constant.int 0
%30523 = torch.prim.ListConstruct %int1_27987, %int0_27988 : (!torch.int, !torch.int) -> !torch.list<int>
%30524 = torch.aten.permute %1114, %30523 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27989 = torch.constant.int 1
%int0_27990 = torch.constant.int 0
%30525 = torch.prim.ListConstruct %int1_27989, %int0_27990 : (!torch.int, !torch.int) -> !torch.list<int>
%30526 = torch.aten.permute %1115, %30525 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27991 = torch.constant.int 1
%int0_27992 = torch.constant.int 0
%30527 = torch.prim.ListConstruct %int1_27991, %int0_27992 : (!torch.int, !torch.int) -> !torch.list<int>
%30528 = torch.aten.permute %1116, %30527 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27993 = torch.constant.int 1
%int0_27994 = torch.constant.int 0
%30529 = torch.prim.ListConstruct %int1_27993, %int0_27994 : (!torch.int, !torch.int) -> !torch.list<int>
%30530 = torch.aten.permute %1117, %30529 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27995 = torch.constant.int 1
%int0_27996 = torch.constant.int 0
%30531 = torch.prim.ListConstruct %int1_27995, %int0_27996 : (!torch.int, !torch.int) -> !torch.list<int>
%30532 = torch.aten.permute %1118, %30531 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_27997 = torch.constant.int 1
%int0_27998 = torch.constant.int 0
%30533 = torch.prim.ListConstruct %int1_27997, %int0_27998 : (!torch.int, !torch.int) -> !torch.list<int>
%30534 = torch.aten.permute %1119, %30533 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_27999 = torch.constant.int 4
%30535 = torch.aten.mul.int %int4_27999, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28000 = torch.constant.int 4096
%30536 = torch.prim.ListConstruct %30535, %int4096_28000 : (!torch.int, !torch.int) -> !torch.list<int>
%30537 = torch.aten.view %30383, %30536 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30537, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30538 = torch.aten.mm %30537, %30520 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30538, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28001 = torch.constant.int 4
%int128_28002 = torch.constant.int 128
%30539 = torch.prim.ListConstruct %int4_28001, %2482, %int128_28002 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30540 = torch.aten.view %30538, %30539 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28003 = torch.constant.int 4
%30541 = torch.aten.mul.int %int4_28003, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28004 = torch.constant.int 4096
%30542 = torch.prim.ListConstruct %30541, %int4096_28004 : (!torch.int, !torch.int) -> !torch.list<int>
%30543 = torch.aten.view %30384, %30542 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30543, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30544 = torch.aten.mm %30543, %30522 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30544, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28005 = torch.constant.int 4
%int128_28006 = torch.constant.int 128
%30545 = torch.prim.ListConstruct %int4_28005, %2482, %int128_28006 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30546 = torch.aten.view %30544, %30545 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28007 = torch.constant.int 4
%30547 = torch.aten.mul.int %int4_28007, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28008 = torch.constant.int 4096
%30548 = torch.prim.ListConstruct %30547, %int4096_28008 : (!torch.int, !torch.int) -> !torch.list<int>
%30549 = torch.aten.view %30385, %30548 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30549, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30550 = torch.aten.mm %30549, %30524 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30550, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28009 = torch.constant.int 4
%int128_28010 = torch.constant.int 128
%30551 = torch.prim.ListConstruct %int4_28009, %2482, %int128_28010 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30552 = torch.aten.view %30550, %30551 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28011 = torch.constant.int 4
%30553 = torch.aten.mul.int %int4_28011, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28012 = torch.constant.int 4096
%30554 = torch.prim.ListConstruct %30553, %int4096_28012 : (!torch.int, !torch.int) -> !torch.list<int>
%30555 = torch.aten.view %30386, %30554 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30555, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30556 = torch.aten.mm %30555, %30526 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30556, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28013 = torch.constant.int 4
%int128_28014 = torch.constant.int 128
%30557 = torch.prim.ListConstruct %int4_28013, %2482, %int128_28014 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30558 = torch.aten.view %30556, %30557 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28015 = torch.constant.int 4
%30559 = torch.aten.mul.int %int4_28015, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28016 = torch.constant.int 4096
%30560 = torch.prim.ListConstruct %30559, %int4096_28016 : (!torch.int, !torch.int) -> !torch.list<int>
%30561 = torch.aten.view %30387, %30560 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30561, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30562 = torch.aten.mm %30561, %30528 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30562, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28017 = torch.constant.int 4
%int128_28018 = torch.constant.int 128
%30563 = torch.prim.ListConstruct %int4_28017, %2482, %int128_28018 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30564 = torch.aten.view %30562, %30563 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28019 = torch.constant.int 4
%30565 = torch.aten.mul.int %int4_28019, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28020 = torch.constant.int 4096
%30566 = torch.prim.ListConstruct %30565, %int4096_28020 : (!torch.int, !torch.int) -> !torch.list<int>
%30567 = torch.aten.view %30388, %30566 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30567, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30568 = torch.aten.mm %30567, %30530 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30568, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28021 = torch.constant.int 4
%int128_28022 = torch.constant.int 128
%30569 = torch.prim.ListConstruct %int4_28021, %2482, %int128_28022 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30570 = torch.aten.view %30568, %30569 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28023 = torch.constant.int 4
%30571 = torch.aten.mul.int %int4_28023, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28024 = torch.constant.int 4096
%30572 = torch.prim.ListConstruct %30571, %int4096_28024 : (!torch.int, !torch.int) -> !torch.list<int>
%30573 = torch.aten.view %30389, %30572 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30573, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30574 = torch.aten.mm %30573, %30532 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30574, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28025 = torch.constant.int 4
%int128_28026 = torch.constant.int 128
%30575 = torch.prim.ListConstruct %int4_28025, %2482, %int128_28026 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30576 = torch.aten.view %30574, %30575 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28027 = torch.constant.int 4
%30577 = torch.aten.mul.int %int4_28027, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_28028 = torch.constant.int 4096
%30578 = torch.prim.ListConstruct %30577, %int4096_28028 : (!torch.int, !torch.int) -> !torch.list<int>
%30579 = torch.aten.view %30390, %30578 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %30579, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%30580 = torch.aten.mm %30579, %30534 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %30580, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_28029 = torch.constant.int 4
%int128_28030 = torch.constant.int 128
%30581 = torch.prim.ListConstruct %int4_28029, %2482, %int128_28030 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30582 = torch.aten.view %30580, %30581 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %30582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_28031 = torch.constant.int 4
%int4_28032 = torch.constant.int 4
%int128_28033 = torch.constant.int 128
%30583 = torch.prim.ListConstruct %int4_28031, %2482, %int4_28032, %int128_28033 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30584 = torch.aten.view %30412, %30583 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28034 = torch.constant.int 4
%int4_28035 = torch.constant.int 4
%int128_28036 = torch.constant.int 128
%30585 = torch.prim.ListConstruct %int4_28034, %2482, %int4_28035, %int128_28036 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30586 = torch.aten.view %30418, %30585 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28037 = torch.constant.int 4
%int4_28038 = torch.constant.int 4
%int128_28039 = torch.constant.int 128
%30587 = torch.prim.ListConstruct %int4_28037, %2482, %int4_28038, %int128_28039 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30588 = torch.aten.view %30424, %30587 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28040 = torch.constant.int 4
%int4_28041 = torch.constant.int 4
%int128_28042 = torch.constant.int 128
%30589 = torch.prim.ListConstruct %int4_28040, %2482, %int4_28041, %int128_28042 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30590 = torch.aten.view %30430, %30589 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28043 = torch.constant.int 4
%int4_28044 = torch.constant.int 4
%int128_28045 = torch.constant.int 128
%30591 = torch.prim.ListConstruct %int4_28043, %2482, %int4_28044, %int128_28045 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30592 = torch.aten.view %30436, %30591 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28046 = torch.constant.int 4
%int4_28047 = torch.constant.int 4
%int128_28048 = torch.constant.int 128
%30593 = torch.prim.ListConstruct %int4_28046, %2482, %int4_28047, %int128_28048 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30594 = torch.aten.view %30442, %30593 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28049 = torch.constant.int 4
%int4_28050 = torch.constant.int 4
%int128_28051 = torch.constant.int 128
%30595 = torch.prim.ListConstruct %int4_28049, %2482, %int4_28050, %int128_28051 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30596 = torch.aten.view %30448, %30595 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28052 = torch.constant.int 4
%int4_28053 = torch.constant.int 4
%int128_28054 = torch.constant.int 128
%30597 = torch.prim.ListConstruct %int4_28052, %2482, %int4_28053, %int128_28054 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30598 = torch.aten.view %30454, %30597 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28055 = torch.constant.int 4
%int1_28056 = torch.constant.int 1
%int128_28057 = torch.constant.int 128
%30599 = torch.prim.ListConstruct %int4_28055, %2482, %int1_28056, %int128_28057 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30600 = torch.aten.view %30476, %30599 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28058 = torch.constant.int 4
%int1_28059 = torch.constant.int 1
%int128_28060 = torch.constant.int 128
%30601 = torch.prim.ListConstruct %int4_28058, %2482, %int1_28059, %int128_28060 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30602 = torch.aten.view %30482, %30601 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28061 = torch.constant.int 4
%int1_28062 = torch.constant.int 1
%int128_28063 = torch.constant.int 128
%30603 = torch.prim.ListConstruct %int4_28061, %2482, %int1_28062, %int128_28063 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30604 = torch.aten.view %30488, %30603 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28064 = torch.constant.int 4
%int1_28065 = torch.constant.int 1
%int128_28066 = torch.constant.int 128
%30605 = torch.prim.ListConstruct %int4_28064, %2482, %int1_28065, %int128_28066 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30606 = torch.aten.view %30494, %30605 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30606, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28067 = torch.constant.int 4
%int1_28068 = torch.constant.int 1
%int128_28069 = torch.constant.int 128
%30607 = torch.prim.ListConstruct %int4_28067, %2482, %int1_28068, %int128_28069 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30608 = torch.aten.view %30500, %30607 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28070 = torch.constant.int 4
%int1_28071 = torch.constant.int 1
%int128_28072 = torch.constant.int 128
%30609 = torch.prim.ListConstruct %int4_28070, %2482, %int1_28071, %int128_28072 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30610 = torch.aten.view %30506, %30609 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28073 = torch.constant.int 4
%int1_28074 = torch.constant.int 1
%int128_28075 = torch.constant.int 128
%30611 = torch.prim.ListConstruct %int4_28073, %2482, %int1_28074, %int128_28075 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30612 = torch.aten.view %30512, %30611 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28076 = torch.constant.int 4
%int1_28077 = torch.constant.int 1
%int128_28078 = torch.constant.int 128
%30613 = torch.prim.ListConstruct %int4_28076, %2482, %int1_28077, %int128_28078 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30614 = torch.aten.view %30518, %30613 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28079 = torch.constant.int 4
%int1_28080 = torch.constant.int 1
%int128_28081 = torch.constant.int 128
%30615 = torch.prim.ListConstruct %int4_28079, %2482, %int1_28080, %int128_28081 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30616 = torch.aten.view %30540, %30615 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28082 = torch.constant.int 4
%int1_28083 = torch.constant.int 1
%int128_28084 = torch.constant.int 128
%30617 = torch.prim.ListConstruct %int4_28082, %2482, %int1_28083, %int128_28084 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30618 = torch.aten.view %30546, %30617 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28085 = torch.constant.int 4
%int1_28086 = torch.constant.int 1
%int128_28087 = torch.constant.int 128
%30619 = torch.prim.ListConstruct %int4_28085, %2482, %int1_28086, %int128_28087 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30620 = torch.aten.view %30552, %30619 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28088 = torch.constant.int 4
%int1_28089 = torch.constant.int 1
%int128_28090 = torch.constant.int 128
%30621 = torch.prim.ListConstruct %int4_28088, %2482, %int1_28089, %int128_28090 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30622 = torch.aten.view %30558, %30621 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28091 = torch.constant.int 4
%int1_28092 = torch.constant.int 1
%int128_28093 = torch.constant.int 128
%30623 = torch.prim.ListConstruct %int4_28091, %2482, %int1_28092, %int128_28093 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30624 = torch.aten.view %30564, %30623 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28094 = torch.constant.int 4
%int1_28095 = torch.constant.int 1
%int128_28096 = torch.constant.int 128
%30625 = torch.prim.ListConstruct %int4_28094, %2482, %int1_28095, %int128_28096 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30626 = torch.aten.view %30570, %30625 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28097 = torch.constant.int 4
%int1_28098 = torch.constant.int 1
%int128_28099 = torch.constant.int 128
%30627 = torch.prim.ListConstruct %int4_28097, %2482, %int1_28098, %int128_28099 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30628 = torch.aten.view %30576, %30627 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_28100 = torch.constant.int 4
%int1_28101 = torch.constant.int 1
%int128_28102 = torch.constant.int 128
%30629 = torch.prim.ListConstruct %int4_28100, %2482, %int1_28101, %int128_28102 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30630 = torch.aten.view %30582, %30629 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_28103 = torch.constant.int 131072
%none_28104 = torch.constant.none
%none_28105 = torch.constant.none
%cpu_28106 = torch.constant.device "cpu"
%false_28107 = torch.constant.bool false
%30631 = torch.aten.arange %int131072_28103, %none_28104, %none_28105, %cpu_28106, %false_28107 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_28108 = torch.constant.int 0
%int128_28109 = torch.constant.int 128
%int2_28110 = torch.constant.int 2
%none_28111 = torch.constant.none
%none_28112 = torch.constant.none
%cpu_28113 = torch.constant.device "cpu"
%false_28114 = torch.constant.bool false
%30632 = torch.aten.arange.start_step %int0_28108, %int128_28109, %int2_28110, %none_28111, %none_28112, %cpu_28113, %false_28114 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_28115 = torch.constant.int 0
%int0_28116 = torch.constant.int 0
%int64_28117 = torch.constant.int 64
%int1_28118 = torch.constant.int 1
%30633 = torch.aten.slice.Tensor %30632, %int0_28115, %int0_28116, %int64_28117, %int1_28118 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_28119 = torch.constant.int 6
%30634 = torch.prims.convert_element_type %30633, %int6_28119 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_28120 = torch.constant.int 128
%30635 = torch.aten.div.Scalar %30634, %int128_28120 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_28121 = torch.constant.float 5.000000e+05
%30636 = torch.aten.pow.Scalar %float5.000000e05_28121, %30635 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%30637 = torch.aten.reciprocal %30636 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_28122 = torch.constant.float 1.000000e+00
%30638 = torch.aten.mul.Scalar %30637, %float1.000000e00_28122 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_28123 = torch.constant.int 131072
%int1_28124 = torch.constant.int 1
%30639 = torch.prim.ListConstruct %int131072_28123, %int1_28124 : (!torch.int, !torch.int) -> !torch.list<int>
%30640 = torch.aten.view %30631, %30639 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%30641 = torch.aten.mul.Tensor %30640, %30638 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%30642 = torch.aten.cos %30641 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%30643 = torch.aten.sin %30641 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%30644 = torch.aten.complex %30642, %30643 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%30645 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30646 = flow.tensor.transfer %30645 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%30647 = torch_c.from_builtin_tensor %30646 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30648 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30649 = flow.tensor.transfer %30648 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%30650 = torch_c.from_builtin_tensor %30649 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30651 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30652 = flow.tensor.transfer %30651 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%30653 = torch_c.from_builtin_tensor %30652 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30654 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30655 = flow.tensor.transfer %30654 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%30656 = torch_c.from_builtin_tensor %30655 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30657 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30658 = flow.tensor.transfer %30657 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%30659 = torch_c.from_builtin_tensor %30658 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30660 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30661 = flow.tensor.transfer %30660 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%30662 = torch_c.from_builtin_tensor %30661 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30663 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30664 = flow.tensor.transfer %30663 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%30665 = torch_c.from_builtin_tensor %30664 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30666 = torch_c.to_builtin_tensor %30644 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30667 = flow.tensor.transfer %30666 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%30668 = torch_c.from_builtin_tensor %30667 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_28125 = torch.constant.int 1
%30669 = torch.aten.size.int %30412, %int1_28125 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28126 = torch.constant.int 0
%30670 = torch.aten.add.int %int0_28126, %30669 : !torch.int, !torch.int -> !torch.int
%int0_28127 = torch.constant.int 0
%int0_28128 = torch.constant.int 0
%int1_28129 = torch.constant.int 1
%30671 = torch.aten.slice.Tensor %30647, %int0_28127, %int0_28128, %30670, %int1_28129 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30671, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28130 = torch.constant.int 1
%int0_28131 = torch.constant.int 0
%int9223372036854775807_28132 = torch.constant.int 9223372036854775807
%int1_28133 = torch.constant.int 1
%30672 = torch.aten.slice.Tensor %30671, %int1_28130, %int0_28131, %int9223372036854775807_28132, %int1_28133 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30672, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28134 = torch.constant.int 0
%30673 = torch.aten.unsqueeze %30672, %int0_28134 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30673, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28135 = torch.constant.int 2
%30674 = torch.aten.unsqueeze %30673, %int2_28135 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30674, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28136 = torch.constant.int 3
%int0_28137 = torch.constant.int 0
%int9223372036854775807_28138 = torch.constant.int 9223372036854775807
%int1_28139 = torch.constant.int 1
%30675 = torch.aten.slice.Tensor %30674, %int3_28136, %int0_28137, %int9223372036854775807_28138, %int1_28139 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30675, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30676 = torch_c.to_builtin_tensor %30584 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28140 = arith.constant 1 : index
%dim_28141 = tensor.dim %30676, %c1_28140 : tensor<4x?x4x128xf16>
%30677 = flow.tensor.bitcast %30676 : tensor<4x?x4x128xf16>{%dim_28141} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28141}
%30678 = torch_c.from_builtin_tensor %30677 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30679 = torch.aten.mul.Tensor %30678, %30675 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30680 = torch_c.to_builtin_tensor %30679 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28142 = arith.constant 1 : index
%dim_28143 = tensor.dim %30680, %c1_28142 : tensor<4x?x4x64xcomplex<f32>>
%30681 = flow.tensor.bitcast %30680 : tensor<4x?x4x64xcomplex<f32>>{%dim_28143} -> tensor<4x?x4x128xf32>{%dim_28143}
%30682 = torch_c.from_builtin_tensor %30681 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28144 = torch.constant.int 5
%30683 = torch.prims.convert_element_type %30682, %int5_28144 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28145 = torch.constant.int 1
%30684 = torch.aten.size.int %30418, %int1_28145 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28146 = torch.constant.int 0
%30685 = torch.aten.add.int %int0_28146, %30684 : !torch.int, !torch.int -> !torch.int
%int0_28147 = torch.constant.int 0
%int0_28148 = torch.constant.int 0
%int1_28149 = torch.constant.int 1
%30686 = torch.aten.slice.Tensor %30650, %int0_28147, %int0_28148, %30685, %int1_28149 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30686, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28150 = torch.constant.int 1
%int0_28151 = torch.constant.int 0
%int9223372036854775807_28152 = torch.constant.int 9223372036854775807
%int1_28153 = torch.constant.int 1
%30687 = torch.aten.slice.Tensor %30686, %int1_28150, %int0_28151, %int9223372036854775807_28152, %int1_28153 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30687, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28154 = torch.constant.int 0
%30688 = torch.aten.unsqueeze %30687, %int0_28154 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30688, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28155 = torch.constant.int 2
%30689 = torch.aten.unsqueeze %30688, %int2_28155 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30689, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28156 = torch.constant.int 3
%int0_28157 = torch.constant.int 0
%int9223372036854775807_28158 = torch.constant.int 9223372036854775807
%int1_28159 = torch.constant.int 1
%30690 = torch.aten.slice.Tensor %30689, %int3_28156, %int0_28157, %int9223372036854775807_28158, %int1_28159 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30690, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30691 = torch_c.to_builtin_tensor %30586 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28160 = arith.constant 1 : index
%dim_28161 = tensor.dim %30691, %c1_28160 : tensor<4x?x4x128xf16>
%30692 = flow.tensor.bitcast %30691 : tensor<4x?x4x128xf16>{%dim_28161} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28161}
%30693 = torch_c.from_builtin_tensor %30692 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30694 = torch.aten.mul.Tensor %30693, %30690 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30695 = torch_c.to_builtin_tensor %30694 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28162 = arith.constant 1 : index
%dim_28163 = tensor.dim %30695, %c1_28162 : tensor<4x?x4x64xcomplex<f32>>
%30696 = flow.tensor.bitcast %30695 : tensor<4x?x4x64xcomplex<f32>>{%dim_28163} -> tensor<4x?x4x128xf32>{%dim_28163}
%30697 = torch_c.from_builtin_tensor %30696 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28164 = torch.constant.int 5
%30698 = torch.prims.convert_element_type %30697, %int5_28164 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28165 = torch.constant.int 1
%30699 = torch.aten.size.int %30424, %int1_28165 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28166 = torch.constant.int 0
%30700 = torch.aten.add.int %int0_28166, %30699 : !torch.int, !torch.int -> !torch.int
%int0_28167 = torch.constant.int 0
%int0_28168 = torch.constant.int 0
%int1_28169 = torch.constant.int 1
%30701 = torch.aten.slice.Tensor %30653, %int0_28167, %int0_28168, %30700, %int1_28169 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30701, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28170 = torch.constant.int 1
%int0_28171 = torch.constant.int 0
%int9223372036854775807_28172 = torch.constant.int 9223372036854775807
%int1_28173 = torch.constant.int 1
%30702 = torch.aten.slice.Tensor %30701, %int1_28170, %int0_28171, %int9223372036854775807_28172, %int1_28173 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30702, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28174 = torch.constant.int 0
%30703 = torch.aten.unsqueeze %30702, %int0_28174 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30703, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28175 = torch.constant.int 2
%30704 = torch.aten.unsqueeze %30703, %int2_28175 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30704, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28176 = torch.constant.int 3
%int0_28177 = torch.constant.int 0
%int9223372036854775807_28178 = torch.constant.int 9223372036854775807
%int1_28179 = torch.constant.int 1
%30705 = torch.aten.slice.Tensor %30704, %int3_28176, %int0_28177, %int9223372036854775807_28178, %int1_28179 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30705, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30706 = torch_c.to_builtin_tensor %30588 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28180 = arith.constant 1 : index
%dim_28181 = tensor.dim %30706, %c1_28180 : tensor<4x?x4x128xf16>
%30707 = flow.tensor.bitcast %30706 : tensor<4x?x4x128xf16>{%dim_28181} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28181}
%30708 = torch_c.from_builtin_tensor %30707 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30709 = torch.aten.mul.Tensor %30708, %30705 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30710 = torch_c.to_builtin_tensor %30709 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28182 = arith.constant 1 : index
%dim_28183 = tensor.dim %30710, %c1_28182 : tensor<4x?x4x64xcomplex<f32>>
%30711 = flow.tensor.bitcast %30710 : tensor<4x?x4x64xcomplex<f32>>{%dim_28183} -> tensor<4x?x4x128xf32>{%dim_28183}
%30712 = torch_c.from_builtin_tensor %30711 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28184 = torch.constant.int 5
%30713 = torch.prims.convert_element_type %30712, %int5_28184 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28185 = torch.constant.int 1
%30714 = torch.aten.size.int %30430, %int1_28185 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28186 = torch.constant.int 0
%30715 = torch.aten.add.int %int0_28186, %30714 : !torch.int, !torch.int -> !torch.int
%int0_28187 = torch.constant.int 0
%int0_28188 = torch.constant.int 0
%int1_28189 = torch.constant.int 1
%30716 = torch.aten.slice.Tensor %30656, %int0_28187, %int0_28188, %30715, %int1_28189 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30716, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28190 = torch.constant.int 1
%int0_28191 = torch.constant.int 0
%int9223372036854775807_28192 = torch.constant.int 9223372036854775807
%int1_28193 = torch.constant.int 1
%30717 = torch.aten.slice.Tensor %30716, %int1_28190, %int0_28191, %int9223372036854775807_28192, %int1_28193 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30717, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28194 = torch.constant.int 0
%30718 = torch.aten.unsqueeze %30717, %int0_28194 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30718, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28195 = torch.constant.int 2
%30719 = torch.aten.unsqueeze %30718, %int2_28195 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30719, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28196 = torch.constant.int 3
%int0_28197 = torch.constant.int 0
%int9223372036854775807_28198 = torch.constant.int 9223372036854775807
%int1_28199 = torch.constant.int 1
%30720 = torch.aten.slice.Tensor %30719, %int3_28196, %int0_28197, %int9223372036854775807_28198, %int1_28199 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30720, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30721 = torch_c.to_builtin_tensor %30590 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28200 = arith.constant 1 : index
%dim_28201 = tensor.dim %30721, %c1_28200 : tensor<4x?x4x128xf16>
%30722 = flow.tensor.bitcast %30721 : tensor<4x?x4x128xf16>{%dim_28201} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28201}
%30723 = torch_c.from_builtin_tensor %30722 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30724 = torch.aten.mul.Tensor %30723, %30720 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30725 = torch_c.to_builtin_tensor %30724 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28202 = arith.constant 1 : index
%dim_28203 = tensor.dim %30725, %c1_28202 : tensor<4x?x4x64xcomplex<f32>>
%30726 = flow.tensor.bitcast %30725 : tensor<4x?x4x64xcomplex<f32>>{%dim_28203} -> tensor<4x?x4x128xf32>{%dim_28203}
%30727 = torch_c.from_builtin_tensor %30726 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28204 = torch.constant.int 5
%30728 = torch.prims.convert_element_type %30727, %int5_28204 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28205 = torch.constant.int 1
%30729 = torch.aten.size.int %30436, %int1_28205 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28206 = torch.constant.int 0
%30730 = torch.aten.add.int %int0_28206, %30729 : !torch.int, !torch.int -> !torch.int
%int0_28207 = torch.constant.int 0
%int0_28208 = torch.constant.int 0
%int1_28209 = torch.constant.int 1
%30731 = torch.aten.slice.Tensor %30659, %int0_28207, %int0_28208, %30730, %int1_28209 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30731, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28210 = torch.constant.int 1
%int0_28211 = torch.constant.int 0
%int9223372036854775807_28212 = torch.constant.int 9223372036854775807
%int1_28213 = torch.constant.int 1
%30732 = torch.aten.slice.Tensor %30731, %int1_28210, %int0_28211, %int9223372036854775807_28212, %int1_28213 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30732, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28214 = torch.constant.int 0
%30733 = torch.aten.unsqueeze %30732, %int0_28214 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30733, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28215 = torch.constant.int 2
%30734 = torch.aten.unsqueeze %30733, %int2_28215 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30734, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28216 = torch.constant.int 3
%int0_28217 = torch.constant.int 0
%int9223372036854775807_28218 = torch.constant.int 9223372036854775807
%int1_28219 = torch.constant.int 1
%30735 = torch.aten.slice.Tensor %30734, %int3_28216, %int0_28217, %int9223372036854775807_28218, %int1_28219 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30735, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30736 = torch_c.to_builtin_tensor %30592 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28220 = arith.constant 1 : index
%dim_28221 = tensor.dim %30736, %c1_28220 : tensor<4x?x4x128xf16>
%30737 = flow.tensor.bitcast %30736 : tensor<4x?x4x128xf16>{%dim_28221} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28221}
%30738 = torch_c.from_builtin_tensor %30737 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30739 = torch.aten.mul.Tensor %30738, %30735 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30740 = torch_c.to_builtin_tensor %30739 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28222 = arith.constant 1 : index
%dim_28223 = tensor.dim %30740, %c1_28222 : tensor<4x?x4x64xcomplex<f32>>
%30741 = flow.tensor.bitcast %30740 : tensor<4x?x4x64xcomplex<f32>>{%dim_28223} -> tensor<4x?x4x128xf32>{%dim_28223}
%30742 = torch_c.from_builtin_tensor %30741 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28224 = torch.constant.int 5
%30743 = torch.prims.convert_element_type %30742, %int5_28224 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28225 = torch.constant.int 1
%30744 = torch.aten.size.int %30442, %int1_28225 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28226 = torch.constant.int 0
%30745 = torch.aten.add.int %int0_28226, %30744 : !torch.int, !torch.int -> !torch.int
%int0_28227 = torch.constant.int 0
%int0_28228 = torch.constant.int 0
%int1_28229 = torch.constant.int 1
%30746 = torch.aten.slice.Tensor %30662, %int0_28227, %int0_28228, %30745, %int1_28229 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30746, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28230 = torch.constant.int 1
%int0_28231 = torch.constant.int 0
%int9223372036854775807_28232 = torch.constant.int 9223372036854775807
%int1_28233 = torch.constant.int 1
%30747 = torch.aten.slice.Tensor %30746, %int1_28230, %int0_28231, %int9223372036854775807_28232, %int1_28233 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30747, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28234 = torch.constant.int 0
%30748 = torch.aten.unsqueeze %30747, %int0_28234 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30748, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28235 = torch.constant.int 2
%30749 = torch.aten.unsqueeze %30748, %int2_28235 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30749, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28236 = torch.constant.int 3
%int0_28237 = torch.constant.int 0
%int9223372036854775807_28238 = torch.constant.int 9223372036854775807
%int1_28239 = torch.constant.int 1
%30750 = torch.aten.slice.Tensor %30749, %int3_28236, %int0_28237, %int9223372036854775807_28238, %int1_28239 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30750, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30751 = torch_c.to_builtin_tensor %30594 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28240 = arith.constant 1 : index
%dim_28241 = tensor.dim %30751, %c1_28240 : tensor<4x?x4x128xf16>
%30752 = flow.tensor.bitcast %30751 : tensor<4x?x4x128xf16>{%dim_28241} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28241}
%30753 = torch_c.from_builtin_tensor %30752 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30754 = torch.aten.mul.Tensor %30753, %30750 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30755 = torch_c.to_builtin_tensor %30754 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28242 = arith.constant 1 : index
%dim_28243 = tensor.dim %30755, %c1_28242 : tensor<4x?x4x64xcomplex<f32>>
%30756 = flow.tensor.bitcast %30755 : tensor<4x?x4x64xcomplex<f32>>{%dim_28243} -> tensor<4x?x4x128xf32>{%dim_28243}
%30757 = torch_c.from_builtin_tensor %30756 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28244 = torch.constant.int 5
%30758 = torch.prims.convert_element_type %30757, %int5_28244 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28245 = torch.constant.int 1
%30759 = torch.aten.size.int %30448, %int1_28245 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28246 = torch.constant.int 0
%30760 = torch.aten.add.int %int0_28246, %30759 : !torch.int, !torch.int -> !torch.int
%int0_28247 = torch.constant.int 0
%int0_28248 = torch.constant.int 0
%int1_28249 = torch.constant.int 1
%30761 = torch.aten.slice.Tensor %30665, %int0_28247, %int0_28248, %30760, %int1_28249 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30761, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28250 = torch.constant.int 1
%int0_28251 = torch.constant.int 0
%int9223372036854775807_28252 = torch.constant.int 9223372036854775807
%int1_28253 = torch.constant.int 1
%30762 = torch.aten.slice.Tensor %30761, %int1_28250, %int0_28251, %int9223372036854775807_28252, %int1_28253 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30762, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28254 = torch.constant.int 0
%30763 = torch.aten.unsqueeze %30762, %int0_28254 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30763, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28255 = torch.constant.int 2
%30764 = torch.aten.unsqueeze %30763, %int2_28255 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30764, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28256 = torch.constant.int 3
%int0_28257 = torch.constant.int 0
%int9223372036854775807_28258 = torch.constant.int 9223372036854775807
%int1_28259 = torch.constant.int 1
%30765 = torch.aten.slice.Tensor %30764, %int3_28256, %int0_28257, %int9223372036854775807_28258, %int1_28259 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30765, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30766 = torch_c.to_builtin_tensor %30596 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28260 = arith.constant 1 : index
%dim_28261 = tensor.dim %30766, %c1_28260 : tensor<4x?x4x128xf16>
%30767 = flow.tensor.bitcast %30766 : tensor<4x?x4x128xf16>{%dim_28261} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28261}
%30768 = torch_c.from_builtin_tensor %30767 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30769 = torch.aten.mul.Tensor %30768, %30765 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30770 = torch_c.to_builtin_tensor %30769 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28262 = arith.constant 1 : index
%dim_28263 = tensor.dim %30770, %c1_28262 : tensor<4x?x4x64xcomplex<f32>>
%30771 = flow.tensor.bitcast %30770 : tensor<4x?x4x64xcomplex<f32>>{%dim_28263} -> tensor<4x?x4x128xf32>{%dim_28263}
%30772 = torch_c.from_builtin_tensor %30771 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28264 = torch.constant.int 5
%30773 = torch.prims.convert_element_type %30772, %int5_28264 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28265 = torch.constant.int 1
%30774 = torch.aten.size.int %30454, %int1_28265 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_28266 = torch.constant.int 0
%30775 = torch.aten.add.int %int0_28266, %30774 : !torch.int, !torch.int -> !torch.int
%int0_28267 = torch.constant.int 0
%int0_28268 = torch.constant.int 0
%int1_28269 = torch.constant.int 1
%30776 = torch.aten.slice.Tensor %30668, %int0_28267, %int0_28268, %30775, %int1_28269 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30776, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28270 = torch.constant.int 1
%int0_28271 = torch.constant.int 0
%int9223372036854775807_28272 = torch.constant.int 9223372036854775807
%int1_28273 = torch.constant.int 1
%30777 = torch.aten.slice.Tensor %30776, %int1_28270, %int0_28271, %int9223372036854775807_28272, %int1_28273 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30777, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28274 = torch.constant.int 0
%30778 = torch.aten.unsqueeze %30777, %int0_28274 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30778, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28275 = torch.constant.int 2
%30779 = torch.aten.unsqueeze %30778, %int2_28275 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30779, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28276 = torch.constant.int 3
%int0_28277 = torch.constant.int 0
%int9223372036854775807_28278 = torch.constant.int 9223372036854775807
%int1_28279 = torch.constant.int 1
%30780 = torch.aten.slice.Tensor %30779, %int3_28276, %int0_28277, %int9223372036854775807_28278, %int1_28279 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30780, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30781 = torch_c.to_builtin_tensor %30598 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_28280 = arith.constant 1 : index
%dim_28281 = tensor.dim %30781, %c1_28280 : tensor<4x?x4x128xf16>
%30782 = flow.tensor.bitcast %30781 : tensor<4x?x4x128xf16>{%dim_28281} -> tensor<4x?x4x64xcomplex<f16>>{%dim_28281}
%30783 = torch_c.from_builtin_tensor %30782 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %30783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%30784 = torch.aten.mul.Tensor %30783, %30780 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %30784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%30785 = torch_c.to_builtin_tensor %30784 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_28282 = arith.constant 1 : index
%dim_28283 = tensor.dim %30785, %c1_28282 : tensor<4x?x4x64xcomplex<f32>>
%30786 = flow.tensor.bitcast %30785 : tensor<4x?x4x64xcomplex<f32>>{%dim_28283} -> tensor<4x?x4x128xf32>{%dim_28283}
%30787 = torch_c.from_builtin_tensor %30786 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %30787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_28284 = torch.constant.int 5
%30788 = torch.prims.convert_element_type %30787, %int5_28284 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %30788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_28285 = torch.constant.int 131072
%none_28286 = torch.constant.none
%none_28287 = torch.constant.none
%cpu_28288 = torch.constant.device "cpu"
%false_28289 = torch.constant.bool false
%30789 = torch.aten.arange %int131072_28285, %none_28286, %none_28287, %cpu_28288, %false_28289 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_28290 = torch.constant.int 0
%int128_28291 = torch.constant.int 128
%int2_28292 = torch.constant.int 2
%none_28293 = torch.constant.none
%none_28294 = torch.constant.none
%cpu_28295 = torch.constant.device "cpu"
%false_28296 = torch.constant.bool false
%30790 = torch.aten.arange.start_step %int0_28290, %int128_28291, %int2_28292, %none_28293, %none_28294, %cpu_28295, %false_28296 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_28297 = torch.constant.int 0
%int0_28298 = torch.constant.int 0
%int64_28299 = torch.constant.int 64
%int1_28300 = torch.constant.int 1
%30791 = torch.aten.slice.Tensor %30790, %int0_28297, %int0_28298, %int64_28299, %int1_28300 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_28301 = torch.constant.int 6
%30792 = torch.prims.convert_element_type %30791, %int6_28301 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_28302 = torch.constant.int 128
%30793 = torch.aten.div.Scalar %30792, %int128_28302 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_28303 = torch.constant.float 5.000000e+05
%30794 = torch.aten.pow.Scalar %float5.000000e05_28303, %30793 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%30795 = torch.aten.reciprocal %30794 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_28304 = torch.constant.float 1.000000e+00
%30796 = torch.aten.mul.Scalar %30795, %float1.000000e00_28304 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_28305 = torch.constant.int 131072
%int1_28306 = torch.constant.int 1
%30797 = torch.prim.ListConstruct %int131072_28305, %int1_28306 : (!torch.int, !torch.int) -> !torch.list<int>
%30798 = torch.aten.view %30789, %30797 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%30799 = torch.aten.mul.Tensor %30798, %30796 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%30800 = torch.aten.cos %30799 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%30801 = torch.aten.sin %30799 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%30802 = torch.aten.complex %30800, %30801 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%30803 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30804 = flow.tensor.transfer %30803 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%30805 = torch_c.from_builtin_tensor %30804 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30806 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30807 = flow.tensor.transfer %30806 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%30808 = torch_c.from_builtin_tensor %30807 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30809 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30810 = flow.tensor.transfer %30809 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%30811 = torch_c.from_builtin_tensor %30810 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30812 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30813 = flow.tensor.transfer %30812 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%30814 = torch_c.from_builtin_tensor %30813 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30815 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30816 = flow.tensor.transfer %30815 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%30817 = torch_c.from_builtin_tensor %30816 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30818 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30819 = flow.tensor.transfer %30818 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%30820 = torch_c.from_builtin_tensor %30819 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30821 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30822 = flow.tensor.transfer %30821 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%30823 = torch_c.from_builtin_tensor %30822 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%30824 = torch_c.to_builtin_tensor %30802 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%30825 = flow.tensor.transfer %30824 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%30826 = torch_c.from_builtin_tensor %30825 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_28307 = torch.constant.int 1
%30827 = torch.aten.size.int %30476, %int1_28307 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28308 = torch.constant.int 0
%30828 = torch.aten.add.int %int0_28308, %30827 : !torch.int, !torch.int -> !torch.int
%int0_28309 = torch.constant.int 0
%int0_28310 = torch.constant.int 0
%int1_28311 = torch.constant.int 1
%30829 = torch.aten.slice.Tensor %30805, %int0_28309, %int0_28310, %30828, %int1_28311 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30829, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28312 = torch.constant.int 1
%int0_28313 = torch.constant.int 0
%int9223372036854775807_28314 = torch.constant.int 9223372036854775807
%int1_28315 = torch.constant.int 1
%30830 = torch.aten.slice.Tensor %30829, %int1_28312, %int0_28313, %int9223372036854775807_28314, %int1_28315 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30830, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28316 = torch.constant.int 0
%30831 = torch.aten.unsqueeze %30830, %int0_28316 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30831, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28317 = torch.constant.int 2
%30832 = torch.aten.unsqueeze %30831, %int2_28317 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30832, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28318 = torch.constant.int 3
%int0_28319 = torch.constant.int 0
%int9223372036854775807_28320 = torch.constant.int 9223372036854775807
%int1_28321 = torch.constant.int 1
%30833 = torch.aten.slice.Tensor %30832, %int3_28318, %int0_28319, %int9223372036854775807_28320, %int1_28321 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30833, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30834 = torch_c.to_builtin_tensor %30600 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28322 = arith.constant 1 : index
%dim_28323 = tensor.dim %30834, %c1_28322 : tensor<4x?x1x128xf16>
%30835 = flow.tensor.bitcast %30834 : tensor<4x?x1x128xf16>{%dim_28323} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28323}
%30836 = torch_c.from_builtin_tensor %30835 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30837 = torch.aten.mul.Tensor %30836, %30833 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30838 = torch_c.to_builtin_tensor %30837 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28324 = arith.constant 1 : index
%dim_28325 = tensor.dim %30838, %c1_28324 : tensor<4x?x1x64xcomplex<f32>>
%30839 = flow.tensor.bitcast %30838 : tensor<4x?x1x64xcomplex<f32>>{%dim_28325} -> tensor<4x?x1x128xf32>{%dim_28325}
%30840 = torch_c.from_builtin_tensor %30839 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28326 = torch.constant.int 5
%30841 = torch.prims.convert_element_type %30840, %int5_28326 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28327 = torch.constant.int 1
%30842 = torch.aten.size.int %30482, %int1_28327 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28328 = torch.constant.int 0
%30843 = torch.aten.add.int %int0_28328, %30842 : !torch.int, !torch.int -> !torch.int
%int0_28329 = torch.constant.int 0
%int0_28330 = torch.constant.int 0
%int1_28331 = torch.constant.int 1
%30844 = torch.aten.slice.Tensor %30808, %int0_28329, %int0_28330, %30843, %int1_28331 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30844, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28332 = torch.constant.int 1
%int0_28333 = torch.constant.int 0
%int9223372036854775807_28334 = torch.constant.int 9223372036854775807
%int1_28335 = torch.constant.int 1
%30845 = torch.aten.slice.Tensor %30844, %int1_28332, %int0_28333, %int9223372036854775807_28334, %int1_28335 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30845, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28336 = torch.constant.int 0
%30846 = torch.aten.unsqueeze %30845, %int0_28336 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30846, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28337 = torch.constant.int 2
%30847 = torch.aten.unsqueeze %30846, %int2_28337 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30847, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28338 = torch.constant.int 3
%int0_28339 = torch.constant.int 0
%int9223372036854775807_28340 = torch.constant.int 9223372036854775807
%int1_28341 = torch.constant.int 1
%30848 = torch.aten.slice.Tensor %30847, %int3_28338, %int0_28339, %int9223372036854775807_28340, %int1_28341 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30848, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30849 = torch_c.to_builtin_tensor %30602 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28342 = arith.constant 1 : index
%dim_28343 = tensor.dim %30849, %c1_28342 : tensor<4x?x1x128xf16>
%30850 = flow.tensor.bitcast %30849 : tensor<4x?x1x128xf16>{%dim_28343} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28343}
%30851 = torch_c.from_builtin_tensor %30850 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30852 = torch.aten.mul.Tensor %30851, %30848 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30853 = torch_c.to_builtin_tensor %30852 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28344 = arith.constant 1 : index
%dim_28345 = tensor.dim %30853, %c1_28344 : tensor<4x?x1x64xcomplex<f32>>
%30854 = flow.tensor.bitcast %30853 : tensor<4x?x1x64xcomplex<f32>>{%dim_28345} -> tensor<4x?x1x128xf32>{%dim_28345}
%30855 = torch_c.from_builtin_tensor %30854 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28346 = torch.constant.int 5
%30856 = torch.prims.convert_element_type %30855, %int5_28346 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28347 = torch.constant.int 1
%30857 = torch.aten.size.int %30488, %int1_28347 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28348 = torch.constant.int 0
%30858 = torch.aten.add.int %int0_28348, %30857 : !torch.int, !torch.int -> !torch.int
%int0_28349 = torch.constant.int 0
%int0_28350 = torch.constant.int 0
%int1_28351 = torch.constant.int 1
%30859 = torch.aten.slice.Tensor %30811, %int0_28349, %int0_28350, %30858, %int1_28351 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30859, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28352 = torch.constant.int 1
%int0_28353 = torch.constant.int 0
%int9223372036854775807_28354 = torch.constant.int 9223372036854775807
%int1_28355 = torch.constant.int 1
%30860 = torch.aten.slice.Tensor %30859, %int1_28352, %int0_28353, %int9223372036854775807_28354, %int1_28355 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30860, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28356 = torch.constant.int 0
%30861 = torch.aten.unsqueeze %30860, %int0_28356 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30861, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28357 = torch.constant.int 2
%30862 = torch.aten.unsqueeze %30861, %int2_28357 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30862, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28358 = torch.constant.int 3
%int0_28359 = torch.constant.int 0
%int9223372036854775807_28360 = torch.constant.int 9223372036854775807
%int1_28361 = torch.constant.int 1
%30863 = torch.aten.slice.Tensor %30862, %int3_28358, %int0_28359, %int9223372036854775807_28360, %int1_28361 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30863, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30864 = torch_c.to_builtin_tensor %30604 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28362 = arith.constant 1 : index
%dim_28363 = tensor.dim %30864, %c1_28362 : tensor<4x?x1x128xf16>
%30865 = flow.tensor.bitcast %30864 : tensor<4x?x1x128xf16>{%dim_28363} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28363}
%30866 = torch_c.from_builtin_tensor %30865 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30867 = torch.aten.mul.Tensor %30866, %30863 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30868 = torch_c.to_builtin_tensor %30867 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28364 = arith.constant 1 : index
%dim_28365 = tensor.dim %30868, %c1_28364 : tensor<4x?x1x64xcomplex<f32>>
%30869 = flow.tensor.bitcast %30868 : tensor<4x?x1x64xcomplex<f32>>{%dim_28365} -> tensor<4x?x1x128xf32>{%dim_28365}
%30870 = torch_c.from_builtin_tensor %30869 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28366 = torch.constant.int 5
%30871 = torch.prims.convert_element_type %30870, %int5_28366 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28367 = torch.constant.int 1
%30872 = torch.aten.size.int %30494, %int1_28367 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28368 = torch.constant.int 0
%30873 = torch.aten.add.int %int0_28368, %30872 : !torch.int, !torch.int -> !torch.int
%int0_28369 = torch.constant.int 0
%int0_28370 = torch.constant.int 0
%int1_28371 = torch.constant.int 1
%30874 = torch.aten.slice.Tensor %30814, %int0_28369, %int0_28370, %30873, %int1_28371 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30874, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28372 = torch.constant.int 1
%int0_28373 = torch.constant.int 0
%int9223372036854775807_28374 = torch.constant.int 9223372036854775807
%int1_28375 = torch.constant.int 1
%30875 = torch.aten.slice.Tensor %30874, %int1_28372, %int0_28373, %int9223372036854775807_28374, %int1_28375 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30875, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28376 = torch.constant.int 0
%30876 = torch.aten.unsqueeze %30875, %int0_28376 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30876, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28377 = torch.constant.int 2
%30877 = torch.aten.unsqueeze %30876, %int2_28377 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30877, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28378 = torch.constant.int 3
%int0_28379 = torch.constant.int 0
%int9223372036854775807_28380 = torch.constant.int 9223372036854775807
%int1_28381 = torch.constant.int 1
%30878 = torch.aten.slice.Tensor %30877, %int3_28378, %int0_28379, %int9223372036854775807_28380, %int1_28381 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30878, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30879 = torch_c.to_builtin_tensor %30606 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28382 = arith.constant 1 : index
%dim_28383 = tensor.dim %30879, %c1_28382 : tensor<4x?x1x128xf16>
%30880 = flow.tensor.bitcast %30879 : tensor<4x?x1x128xf16>{%dim_28383} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28383}
%30881 = torch_c.from_builtin_tensor %30880 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30882 = torch.aten.mul.Tensor %30881, %30878 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30883 = torch_c.to_builtin_tensor %30882 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28384 = arith.constant 1 : index
%dim_28385 = tensor.dim %30883, %c1_28384 : tensor<4x?x1x64xcomplex<f32>>
%30884 = flow.tensor.bitcast %30883 : tensor<4x?x1x64xcomplex<f32>>{%dim_28385} -> tensor<4x?x1x128xf32>{%dim_28385}
%30885 = torch_c.from_builtin_tensor %30884 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28386 = torch.constant.int 5
%30886 = torch.prims.convert_element_type %30885, %int5_28386 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28387 = torch.constant.int 1
%30887 = torch.aten.size.int %30500, %int1_28387 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28388 = torch.constant.int 0
%30888 = torch.aten.add.int %int0_28388, %30887 : !torch.int, !torch.int -> !torch.int
%int0_28389 = torch.constant.int 0
%int0_28390 = torch.constant.int 0
%int1_28391 = torch.constant.int 1
%30889 = torch.aten.slice.Tensor %30817, %int0_28389, %int0_28390, %30888, %int1_28391 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30889, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28392 = torch.constant.int 1
%int0_28393 = torch.constant.int 0
%int9223372036854775807_28394 = torch.constant.int 9223372036854775807
%int1_28395 = torch.constant.int 1
%30890 = torch.aten.slice.Tensor %30889, %int1_28392, %int0_28393, %int9223372036854775807_28394, %int1_28395 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30890, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28396 = torch.constant.int 0
%30891 = torch.aten.unsqueeze %30890, %int0_28396 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30891, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28397 = torch.constant.int 2
%30892 = torch.aten.unsqueeze %30891, %int2_28397 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30892, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28398 = torch.constant.int 3
%int0_28399 = torch.constant.int 0
%int9223372036854775807_28400 = torch.constant.int 9223372036854775807
%int1_28401 = torch.constant.int 1
%30893 = torch.aten.slice.Tensor %30892, %int3_28398, %int0_28399, %int9223372036854775807_28400, %int1_28401 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30893, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30894 = torch_c.to_builtin_tensor %30608 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28402 = arith.constant 1 : index
%dim_28403 = tensor.dim %30894, %c1_28402 : tensor<4x?x1x128xf16>
%30895 = flow.tensor.bitcast %30894 : tensor<4x?x1x128xf16>{%dim_28403} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28403}
%30896 = torch_c.from_builtin_tensor %30895 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30897 = torch.aten.mul.Tensor %30896, %30893 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30898 = torch_c.to_builtin_tensor %30897 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28404 = arith.constant 1 : index
%dim_28405 = tensor.dim %30898, %c1_28404 : tensor<4x?x1x64xcomplex<f32>>
%30899 = flow.tensor.bitcast %30898 : tensor<4x?x1x64xcomplex<f32>>{%dim_28405} -> tensor<4x?x1x128xf32>{%dim_28405}
%30900 = torch_c.from_builtin_tensor %30899 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28406 = torch.constant.int 5
%30901 = torch.prims.convert_element_type %30900, %int5_28406 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28407 = torch.constant.int 1
%30902 = torch.aten.size.int %30506, %int1_28407 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28408 = torch.constant.int 0
%30903 = torch.aten.add.int %int0_28408, %30902 : !torch.int, !torch.int -> !torch.int
%int0_28409 = torch.constant.int 0
%int0_28410 = torch.constant.int 0
%int1_28411 = torch.constant.int 1
%30904 = torch.aten.slice.Tensor %30820, %int0_28409, %int0_28410, %30903, %int1_28411 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30904, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28412 = torch.constant.int 1
%int0_28413 = torch.constant.int 0
%int9223372036854775807_28414 = torch.constant.int 9223372036854775807
%int1_28415 = torch.constant.int 1
%30905 = torch.aten.slice.Tensor %30904, %int1_28412, %int0_28413, %int9223372036854775807_28414, %int1_28415 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30905, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28416 = torch.constant.int 0
%30906 = torch.aten.unsqueeze %30905, %int0_28416 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30906, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28417 = torch.constant.int 2
%30907 = torch.aten.unsqueeze %30906, %int2_28417 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30907, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28418 = torch.constant.int 3
%int0_28419 = torch.constant.int 0
%int9223372036854775807_28420 = torch.constant.int 9223372036854775807
%int1_28421 = torch.constant.int 1
%30908 = torch.aten.slice.Tensor %30907, %int3_28418, %int0_28419, %int9223372036854775807_28420, %int1_28421 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30908, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30909 = torch_c.to_builtin_tensor %30610 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28422 = arith.constant 1 : index
%dim_28423 = tensor.dim %30909, %c1_28422 : tensor<4x?x1x128xf16>
%30910 = flow.tensor.bitcast %30909 : tensor<4x?x1x128xf16>{%dim_28423} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28423}
%30911 = torch_c.from_builtin_tensor %30910 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30912 = torch.aten.mul.Tensor %30911, %30908 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30913 = torch_c.to_builtin_tensor %30912 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28424 = arith.constant 1 : index
%dim_28425 = tensor.dim %30913, %c1_28424 : tensor<4x?x1x64xcomplex<f32>>
%30914 = flow.tensor.bitcast %30913 : tensor<4x?x1x64xcomplex<f32>>{%dim_28425} -> tensor<4x?x1x128xf32>{%dim_28425}
%30915 = torch_c.from_builtin_tensor %30914 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28426 = torch.constant.int 5
%30916 = torch.prims.convert_element_type %30915, %int5_28426 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28427 = torch.constant.int 1
%30917 = torch.aten.size.int %30512, %int1_28427 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28428 = torch.constant.int 0
%30918 = torch.aten.add.int %int0_28428, %30917 : !torch.int, !torch.int -> !torch.int
%int0_28429 = torch.constant.int 0
%int0_28430 = torch.constant.int 0
%int1_28431 = torch.constant.int 1
%30919 = torch.aten.slice.Tensor %30823, %int0_28429, %int0_28430, %30918, %int1_28431 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30919, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28432 = torch.constant.int 1
%int0_28433 = torch.constant.int 0
%int9223372036854775807_28434 = torch.constant.int 9223372036854775807
%int1_28435 = torch.constant.int 1
%30920 = torch.aten.slice.Tensor %30919, %int1_28432, %int0_28433, %int9223372036854775807_28434, %int1_28435 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30920, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28436 = torch.constant.int 0
%30921 = torch.aten.unsqueeze %30920, %int0_28436 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30921, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28437 = torch.constant.int 2
%30922 = torch.aten.unsqueeze %30921, %int2_28437 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30922, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28438 = torch.constant.int 3
%int0_28439 = torch.constant.int 0
%int9223372036854775807_28440 = torch.constant.int 9223372036854775807
%int1_28441 = torch.constant.int 1
%30923 = torch.aten.slice.Tensor %30922, %int3_28438, %int0_28439, %int9223372036854775807_28440, %int1_28441 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30923, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30924 = torch_c.to_builtin_tensor %30612 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28442 = arith.constant 1 : index
%dim_28443 = tensor.dim %30924, %c1_28442 : tensor<4x?x1x128xf16>
%30925 = flow.tensor.bitcast %30924 : tensor<4x?x1x128xf16>{%dim_28443} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28443}
%30926 = torch_c.from_builtin_tensor %30925 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30927 = torch.aten.mul.Tensor %30926, %30923 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30928 = torch_c.to_builtin_tensor %30927 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28444 = arith.constant 1 : index
%dim_28445 = tensor.dim %30928, %c1_28444 : tensor<4x?x1x64xcomplex<f32>>
%30929 = flow.tensor.bitcast %30928 : tensor<4x?x1x64xcomplex<f32>>{%dim_28445} -> tensor<4x?x1x128xf32>{%dim_28445}
%30930 = torch_c.from_builtin_tensor %30929 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28446 = torch.constant.int 5
%30931 = torch.prims.convert_element_type %30930, %int5_28446 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_28447 = torch.constant.int 1
%30932 = torch.aten.size.int %30518, %int1_28447 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_28448 = torch.constant.int 0
%30933 = torch.aten.add.int %int0_28448, %30932 : !torch.int, !torch.int -> !torch.int
%int0_28449 = torch.constant.int 0
%int0_28450 = torch.constant.int 0
%int1_28451 = torch.constant.int 1
%30934 = torch.aten.slice.Tensor %30826, %int0_28449, %int0_28450, %30933, %int1_28451 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30934, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_28452 = torch.constant.int 1
%int0_28453 = torch.constant.int 0
%int9223372036854775807_28454 = torch.constant.int 9223372036854775807
%int1_28455 = torch.constant.int 1
%30935 = torch.aten.slice.Tensor %30934, %int1_28452, %int0_28453, %int9223372036854775807_28454, %int1_28455 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %30935, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_28456 = torch.constant.int 0
%30936 = torch.aten.unsqueeze %30935, %int0_28456 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %30936, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_28457 = torch.constant.int 2
%30937 = torch.aten.unsqueeze %30936, %int2_28457 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30937, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_28458 = torch.constant.int 3
%int0_28459 = torch.constant.int 0
%int9223372036854775807_28460 = torch.constant.int 9223372036854775807
%int1_28461 = torch.constant.int 1
%30938 = torch.aten.slice.Tensor %30937, %int3_28458, %int0_28459, %int9223372036854775807_28460, %int1_28461 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30938, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%30939 = torch_c.to_builtin_tensor %30614 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_28462 = arith.constant 1 : index
%dim_28463 = tensor.dim %30939, %c1_28462 : tensor<4x?x1x128xf16>
%30940 = flow.tensor.bitcast %30939 : tensor<4x?x1x128xf16>{%dim_28463} -> tensor<4x?x1x64xcomplex<f16>>{%dim_28463}
%30941 = torch_c.from_builtin_tensor %30940 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %30941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%30942 = torch.aten.mul.Tensor %30941, %30938 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %30942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%30943 = torch_c.to_builtin_tensor %30942 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_28464 = arith.constant 1 : index
%dim_28465 = tensor.dim %30943, %c1_28464 : tensor<4x?x1x64xcomplex<f32>>
%30944 = flow.tensor.bitcast %30943 : tensor<4x?x1x64xcomplex<f32>>{%dim_28465} -> tensor<4x?x1x128xf32>{%dim_28465}
%30945 = torch_c.from_builtin_tensor %30944 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %30945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_28466 = torch.constant.int 5
%30946 = torch.prims.convert_element_type %30945, %int5_28466 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %30946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_28467 = torch.constant.int 64
%30947 = torch.aten.mul.Scalar %2364, %int64_28467 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30947, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28468 = torch.constant.int 64
%30948 = torch.aten.mul.Scalar %2367, %int64_28468 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30948, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28469 = torch.constant.int 64
%30949 = torch.aten.mul.Scalar %2370, %int64_28469 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30949, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28470 = torch.constant.int 64
%30950 = torch.aten.mul.Scalar %2373, %int64_28470 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30950, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28471 = torch.constant.int 64
%30951 = torch.aten.mul.Scalar %2376, %int64_28471 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30951, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28472 = torch.constant.int 64
%30952 = torch.aten.mul.Scalar %2379, %int64_28472 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30952, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28473 = torch.constant.int 64
%30953 = torch.aten.mul.Scalar %2382, %int64_28473 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30953, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_28474 = torch.constant.int 64
%30954 = torch.aten.mul.Scalar %2385, %int64_28474 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30954, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30 = torch.constant.int 30
%int1_28475 = torch.constant.int 1
%30955 = torch.aten.add.Scalar %30947, %int30, %int1_28475 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30955, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28476 = torch.constant.int 30
%int1_28477 = torch.constant.int 1
%30956 = torch.aten.add.Scalar %30948, %int30_28476, %int1_28477 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30956, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28478 = torch.constant.int 30
%int1_28479 = torch.constant.int 1
%30957 = torch.aten.add.Scalar %30949, %int30_28478, %int1_28479 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30957, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28480 = torch.constant.int 30
%int1_28481 = torch.constant.int 1
%30958 = torch.aten.add.Scalar %30950, %int30_28480, %int1_28481 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30958, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28482 = torch.constant.int 30
%int1_28483 = torch.constant.int 1
%30959 = torch.aten.add.Scalar %30951, %int30_28482, %int1_28483 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30959, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28484 = torch.constant.int 30
%int1_28485 = torch.constant.int 1
%30960 = torch.aten.add.Scalar %30952, %int30_28484, %int1_28485 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30960, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28486 = torch.constant.int 30
%int1_28487 = torch.constant.int 1
%30961 = torch.aten.add.Scalar %30953, %int30_28486, %int1_28487 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30961, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int30_28488 = torch.constant.int 30
%int1_28489 = torch.constant.int 1
%30962 = torch.aten.add.Scalar %30954, %int30_28488, %int1_28489 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %30962, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_28490 = torch.constant.int 4
%int16_28491 = torch.constant.int 16
%int1_28492 = torch.constant.int 1
%int128_28493 = torch.constant.int 128
%30963 = torch.prim.ListConstruct %int4_28490, %3095, %int16_28491, %int1_28492, %int128_28493 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30964 = torch.aten.view %30841, %30963 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30964, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28494 = torch.constant.int 4
%int16_28495 = torch.constant.int 16
%int1_28496 = torch.constant.int 1
%int128_28497 = torch.constant.int 128
%30965 = torch.prim.ListConstruct %int4_28494, %3095, %int16_28495, %int1_28496, %int128_28497 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30966 = torch.aten.view %30856, %30965 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30966, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28498 = torch.constant.int 4
%int16_28499 = torch.constant.int 16
%int1_28500 = torch.constant.int 1
%int128_28501 = torch.constant.int 128
%30967 = torch.prim.ListConstruct %int4_28498, %3095, %int16_28499, %int1_28500, %int128_28501 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30968 = torch.aten.view %30871, %30967 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30968, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28502 = torch.constant.int 4
%int16_28503 = torch.constant.int 16
%int1_28504 = torch.constant.int 1
%int128_28505 = torch.constant.int 128
%30969 = torch.prim.ListConstruct %int4_28502, %3095, %int16_28503, %int1_28504, %int128_28505 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30970 = torch.aten.view %30886, %30969 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30970, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28506 = torch.constant.int 4
%int16_28507 = torch.constant.int 16
%int1_28508 = torch.constant.int 1
%int128_28509 = torch.constant.int 128
%30971 = torch.prim.ListConstruct %int4_28506, %3095, %int16_28507, %int1_28508, %int128_28509 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30972 = torch.aten.view %30901, %30971 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30972, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28510 = torch.constant.int 4
%int16_28511 = torch.constant.int 16
%int1_28512 = torch.constant.int 1
%int128_28513 = torch.constant.int 128
%30973 = torch.prim.ListConstruct %int4_28510, %3095, %int16_28511, %int1_28512, %int128_28513 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30974 = torch.aten.view %30916, %30973 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30974, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28514 = torch.constant.int 4
%int16_28515 = torch.constant.int 16
%int1_28516 = torch.constant.int 1
%int128_28517 = torch.constant.int 128
%30975 = torch.prim.ListConstruct %int4_28514, %3095, %int16_28515, %int1_28516, %int128_28517 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30976 = torch.aten.view %30931, %30975 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30976, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28518 = torch.constant.int 4
%int16_28519 = torch.constant.int 16
%int1_28520 = torch.constant.int 1
%int128_28521 = torch.constant.int 128
%30977 = torch.prim.ListConstruct %int4_28518, %3095, %int16_28519, %int1_28520, %int128_28521 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30978 = torch.aten.view %30946, %30977 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %30978, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28522 = torch.constant.int 4
%30979 = torch.aten.mul.int %int4_28522, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28523 = torch.constant.int 16
%int1_28524 = torch.constant.int 1
%int128_28525 = torch.constant.int 128
%30980 = torch.prim.ListConstruct %30979, %int16_28523, %int1_28524, %int128_28525 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30981 = torch.aten.view %30964, %30980 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30981, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28526 = torch.constant.int 4
%30982 = torch.aten.mul.int %int4_28526, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28527 = torch.constant.int 16
%int1_28528 = torch.constant.int 1
%int128_28529 = torch.constant.int 128
%30983 = torch.prim.ListConstruct %30982, %int16_28527, %int1_28528, %int128_28529 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30984 = torch.aten.view %30966, %30983 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30984, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28530 = torch.constant.int 4
%30985 = torch.aten.mul.int %int4_28530, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28531 = torch.constant.int 16
%int1_28532 = torch.constant.int 1
%int128_28533 = torch.constant.int 128
%30986 = torch.prim.ListConstruct %30985, %int16_28531, %int1_28532, %int128_28533 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30987 = torch.aten.view %30968, %30986 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30987, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28534 = torch.constant.int 4
%30988 = torch.aten.mul.int %int4_28534, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28535 = torch.constant.int 16
%int1_28536 = torch.constant.int 1
%int128_28537 = torch.constant.int 128
%30989 = torch.prim.ListConstruct %30988, %int16_28535, %int1_28536, %int128_28537 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30990 = torch.aten.view %30970, %30989 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30990, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28538 = torch.constant.int 4
%30991 = torch.aten.mul.int %int4_28538, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28539 = torch.constant.int 16
%int1_28540 = torch.constant.int 1
%int128_28541 = torch.constant.int 128
%30992 = torch.prim.ListConstruct %30991, %int16_28539, %int1_28540, %int128_28541 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30993 = torch.aten.view %30972, %30992 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30993, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28542 = torch.constant.int 4
%30994 = torch.aten.mul.int %int4_28542, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28543 = torch.constant.int 16
%int1_28544 = torch.constant.int 1
%int128_28545 = torch.constant.int 128
%30995 = torch.prim.ListConstruct %30994, %int16_28543, %int1_28544, %int128_28545 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30996 = torch.aten.view %30974, %30995 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30996, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28546 = torch.constant.int 4
%30997 = torch.aten.mul.int %int4_28546, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28547 = torch.constant.int 16
%int1_28548 = torch.constant.int 1
%int128_28549 = torch.constant.int 128
%30998 = torch.prim.ListConstruct %30997, %int16_28547, %int1_28548, %int128_28549 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%30999 = torch.aten.view %30976, %30998 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %30999, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28550 = torch.constant.int 4
%31000 = torch.aten.mul.int %int4_28550, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28551 = torch.constant.int 16
%int1_28552 = torch.constant.int 1
%int128_28553 = torch.constant.int 128
%31001 = torch.prim.ListConstruct %31000, %int16_28551, %int1_28552, %int128_28553 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31002 = torch.aten.view %30978, %31001 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31002, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28554 = torch.constant.int 4
%31003 = torch.aten.mul.int %int4_28554, %3095 : !torch.int, !torch.int -> !torch.int
%31004 = torch.prim.ListConstruct %31003 : (!torch.int) -> !torch.list<int>
%31005 = torch.aten.view %30955, %31004 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31005, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28555 = torch.constant.int 4
%31006 = torch.aten.mul.int %int4_28555, %3095 : !torch.int, !torch.int -> !torch.int
%31007 = torch.prim.ListConstruct %31006 : (!torch.int) -> !torch.list<int>
%31008 = torch.aten.view %30956, %31007 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31008, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28556 = torch.constant.int 4
%31009 = torch.aten.mul.int %int4_28556, %3095 : !torch.int, !torch.int -> !torch.int
%31010 = torch.prim.ListConstruct %31009 : (!torch.int) -> !torch.list<int>
%31011 = torch.aten.view %30957, %31010 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31011, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28557 = torch.constant.int 4
%31012 = torch.aten.mul.int %int4_28557, %3095 : !torch.int, !torch.int -> !torch.int
%31013 = torch.prim.ListConstruct %31012 : (!torch.int) -> !torch.list<int>
%31014 = torch.aten.view %30958, %31013 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31014, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28558 = torch.constant.int 4
%31015 = torch.aten.mul.int %int4_28558, %3095 : !torch.int, !torch.int -> !torch.int
%31016 = torch.prim.ListConstruct %31015 : (!torch.int) -> !torch.list<int>
%31017 = torch.aten.view %30959, %31016 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31017, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28559 = torch.constant.int 4
%31018 = torch.aten.mul.int %int4_28559, %3095 : !torch.int, !torch.int -> !torch.int
%31019 = torch.prim.ListConstruct %31018 : (!torch.int) -> !torch.list<int>
%31020 = torch.aten.view %30960, %31019 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31020, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28560 = torch.constant.int 4
%31021 = torch.aten.mul.int %int4_28560, %3095 : !torch.int, !torch.int -> !torch.int
%31022 = torch.prim.ListConstruct %31021 : (!torch.int) -> !torch.list<int>
%31023 = torch.aten.view %30961, %31022 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31023, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28561 = torch.constant.int 4
%31024 = torch.aten.mul.int %int4_28561, %3095 : !torch.int, !torch.int -> !torch.int
%31025 = torch.prim.ListConstruct %31024 : (!torch.int) -> !torch.list<int>
%31026 = torch.aten.view %30962, %31025 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31026, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28562 = torch.constant.int 4
%int16_28563 = torch.constant.int 16
%int1_28564 = torch.constant.int 1
%int128_28565 = torch.constant.int 128
%31027 = torch.prim.ListConstruct %int4_28562, %3095, %int16_28563, %int1_28564, %int128_28565 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31028 = torch.aten.view %30616, %31027 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31028, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28566 = torch.constant.int 4
%int16_28567 = torch.constant.int 16
%int1_28568 = torch.constant.int 1
%int128_28569 = torch.constant.int 128
%31029 = torch.prim.ListConstruct %int4_28566, %3095, %int16_28567, %int1_28568, %int128_28569 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31030 = torch.aten.view %30618, %31029 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31030, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28570 = torch.constant.int 4
%int16_28571 = torch.constant.int 16
%int1_28572 = torch.constant.int 1
%int128_28573 = torch.constant.int 128
%31031 = torch.prim.ListConstruct %int4_28570, %3095, %int16_28571, %int1_28572, %int128_28573 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31032 = torch.aten.view %30620, %31031 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31032, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28574 = torch.constant.int 4
%int16_28575 = torch.constant.int 16
%int1_28576 = torch.constant.int 1
%int128_28577 = torch.constant.int 128
%31033 = torch.prim.ListConstruct %int4_28574, %3095, %int16_28575, %int1_28576, %int128_28577 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31034 = torch.aten.view %30622, %31033 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31034, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28578 = torch.constant.int 4
%int16_28579 = torch.constant.int 16
%int1_28580 = torch.constant.int 1
%int128_28581 = torch.constant.int 128
%31035 = torch.prim.ListConstruct %int4_28578, %3095, %int16_28579, %int1_28580, %int128_28581 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31036 = torch.aten.view %30624, %31035 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31036, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28582 = torch.constant.int 4
%int16_28583 = torch.constant.int 16
%int1_28584 = torch.constant.int 1
%int128_28585 = torch.constant.int 128
%31037 = torch.prim.ListConstruct %int4_28582, %3095, %int16_28583, %int1_28584, %int128_28585 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31038 = torch.aten.view %30626, %31037 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31038, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28586 = torch.constant.int 4
%int16_28587 = torch.constant.int 16
%int1_28588 = torch.constant.int 1
%int128_28589 = torch.constant.int 128
%31039 = torch.prim.ListConstruct %int4_28586, %3095, %int16_28587, %int1_28588, %int128_28589 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31040 = torch.aten.view %30628, %31039 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31040, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28590 = torch.constant.int 4
%int16_28591 = torch.constant.int 16
%int1_28592 = torch.constant.int 1
%int128_28593 = torch.constant.int 128
%31041 = torch.prim.ListConstruct %int4_28590, %3095, %int16_28591, %int1_28592, %int128_28593 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31042 = torch.aten.view %30630, %31041 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %31042, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_28594 = torch.constant.int 4
%31043 = torch.aten.mul.int %int4_28594, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28595 = torch.constant.int 16
%int1_28596 = torch.constant.int 1
%int128_28597 = torch.constant.int 128
%31044 = torch.prim.ListConstruct %31043, %int16_28595, %int1_28596, %int128_28597 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31045 = torch.aten.view %31028, %31044 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31045, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28598 = torch.constant.int 4
%31046 = torch.aten.mul.int %int4_28598, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28599 = torch.constant.int 16
%int1_28600 = torch.constant.int 1
%int128_28601 = torch.constant.int 128
%31047 = torch.prim.ListConstruct %31046, %int16_28599, %int1_28600, %int128_28601 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31048 = torch.aten.view %31030, %31047 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31048, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28602 = torch.constant.int 4
%31049 = torch.aten.mul.int %int4_28602, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28603 = torch.constant.int 16
%int1_28604 = torch.constant.int 1
%int128_28605 = torch.constant.int 128
%31050 = torch.prim.ListConstruct %31049, %int16_28603, %int1_28604, %int128_28605 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31051 = torch.aten.view %31032, %31050 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31051, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28606 = torch.constant.int 4
%31052 = torch.aten.mul.int %int4_28606, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28607 = torch.constant.int 16
%int1_28608 = torch.constant.int 1
%int128_28609 = torch.constant.int 128
%31053 = torch.prim.ListConstruct %31052, %int16_28607, %int1_28608, %int128_28609 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31054 = torch.aten.view %31034, %31053 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31054, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28610 = torch.constant.int 4
%31055 = torch.aten.mul.int %int4_28610, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28611 = torch.constant.int 16
%int1_28612 = torch.constant.int 1
%int128_28613 = torch.constant.int 128
%31056 = torch.prim.ListConstruct %31055, %int16_28611, %int1_28612, %int128_28613 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31057 = torch.aten.view %31036, %31056 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31057, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28614 = torch.constant.int 4
%31058 = torch.aten.mul.int %int4_28614, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28615 = torch.constant.int 16
%int1_28616 = torch.constant.int 1
%int128_28617 = torch.constant.int 128
%31059 = torch.prim.ListConstruct %31058, %int16_28615, %int1_28616, %int128_28617 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31060 = torch.aten.view %31038, %31059 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31060, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28618 = torch.constant.int 4
%31061 = torch.aten.mul.int %int4_28618, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28619 = torch.constant.int 16
%int1_28620 = torch.constant.int 1
%int128_28621 = torch.constant.int 128
%31062 = torch.prim.ListConstruct %31061, %int16_28619, %int1_28620, %int128_28621 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31063 = torch.aten.view %31040, %31062 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31063, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_28622 = torch.constant.int 4
%31064 = torch.aten.mul.int %int4_28622, %3095 : !torch.int, !torch.int -> !torch.int
%int16_28623 = torch.constant.int 16
%int1_28624 = torch.constant.int 1
%int128_28625 = torch.constant.int 128
%31065 = torch.prim.ListConstruct %31064, %int16_28623, %int1_28624, %int128_28625 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31066 = torch.aten.view %31042, %31065 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31066, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_28626 = torch.constant.int 1
%int1_28627 = torch.constant.int 1
%31067 = torch.aten.add.Scalar %30955, %int1_28626, %int1_28627 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31067, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28628 = torch.constant.int 1
%int1_28629 = torch.constant.int 1
%31068 = torch.aten.add.Scalar %30956, %int1_28628, %int1_28629 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31068, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28630 = torch.constant.int 1
%int1_28631 = torch.constant.int 1
%31069 = torch.aten.add.Scalar %30957, %int1_28630, %int1_28631 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31069, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28632 = torch.constant.int 1
%int1_28633 = torch.constant.int 1
%31070 = torch.aten.add.Scalar %30958, %int1_28632, %int1_28633 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31070, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28634 = torch.constant.int 1
%int1_28635 = torch.constant.int 1
%31071 = torch.aten.add.Scalar %30959, %int1_28634, %int1_28635 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31071, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28636 = torch.constant.int 1
%int1_28637 = torch.constant.int 1
%31072 = torch.aten.add.Scalar %30960, %int1_28636, %int1_28637 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31072, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28638 = torch.constant.int 1
%int1_28639 = torch.constant.int 1
%31073 = torch.aten.add.Scalar %30961, %int1_28638, %int1_28639 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31073, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_28640 = torch.constant.int 1
%int1_28641 = torch.constant.int 1
%31074 = torch.aten.add.Scalar %30962, %int1_28640, %int1_28641 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %31074, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_28642 = torch.constant.int 4
%31075 = torch.aten.mul.int %int4_28642, %3095 : !torch.int, !torch.int -> !torch.int
%31076 = torch.prim.ListConstruct %31075 : (!torch.int) -> !torch.list<int>
%31077 = torch.aten.view %31067, %31076 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31077, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28643 = torch.constant.int 4
%31078 = torch.aten.mul.int %int4_28643, %3095 : !torch.int, !torch.int -> !torch.int
%31079 = torch.prim.ListConstruct %31078 : (!torch.int) -> !torch.list<int>
%31080 = torch.aten.view %31068, %31079 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31080, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28644 = torch.constant.int 4
%31081 = torch.aten.mul.int %int4_28644, %3095 : !torch.int, !torch.int -> !torch.int
%31082 = torch.prim.ListConstruct %31081 : (!torch.int) -> !torch.list<int>
%31083 = torch.aten.view %31069, %31082 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31083, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28645 = torch.constant.int 4
%31084 = torch.aten.mul.int %int4_28645, %3095 : !torch.int, !torch.int -> !torch.int
%31085 = torch.prim.ListConstruct %31084 : (!torch.int) -> !torch.list<int>
%31086 = torch.aten.view %31070, %31085 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31086, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28646 = torch.constant.int 4
%31087 = torch.aten.mul.int %int4_28646, %3095 : !torch.int, !torch.int -> !torch.int
%31088 = torch.prim.ListConstruct %31087 : (!torch.int) -> !torch.list<int>
%31089 = torch.aten.view %31071, %31088 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31089, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28647 = torch.constant.int 4
%31090 = torch.aten.mul.int %int4_28647, %3095 : !torch.int, !torch.int -> !torch.int
%31091 = torch.prim.ListConstruct %31090 : (!torch.int) -> !torch.list<int>
%31092 = torch.aten.view %31072, %31091 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31092, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28648 = torch.constant.int 4
%31093 = torch.aten.mul.int %int4_28648, %3095 : !torch.int, !torch.int -> !torch.int
%31094 = torch.prim.ListConstruct %31093 : (!torch.int) -> !torch.list<int>
%31095 = torch.aten.view %31073, %31094 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31095, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_28649 = torch.constant.int 4
%31096 = torch.aten.mul.int %int4_28649, %3095 : !torch.int, !torch.int -> !torch.int
%31097 = torch.prim.ListConstruct %31096 : (!torch.int) -> !torch.list<int>
%31098 = torch.aten.view %31074, %31097 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31098, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%31099 = torch.prim.ListConstruct %31005, %31077 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28650 = torch.constant.int 0
%31100 = torch.aten.cat %31099, %int0_28650 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31100, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31101 = torch.prim.ListConstruct %31008, %31080 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28651 = torch.constant.int 0
%31102 = torch.aten.cat %31101, %int0_28651 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31102, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31103 = torch.prim.ListConstruct %31011, %31083 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28652 = torch.constant.int 0
%31104 = torch.aten.cat %31103, %int0_28652 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31104, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31105 = torch.prim.ListConstruct %31014, %31086 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28653 = torch.constant.int 0
%31106 = torch.aten.cat %31105, %int0_28653 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31106, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31107 = torch.prim.ListConstruct %31017, %31089 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28654 = torch.constant.int 0
%31108 = torch.aten.cat %31107, %int0_28654 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31108, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31109 = torch.prim.ListConstruct %31020, %31092 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28655 = torch.constant.int 0
%31110 = torch.aten.cat %31109, %int0_28655 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31110, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31111 = torch.prim.ListConstruct %31023, %31095 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28656 = torch.constant.int 0
%31112 = torch.aten.cat %31111, %int0_28656 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31112, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31113 = torch.prim.ListConstruct %31026, %31098 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_28657 = torch.constant.int 0
%31114 = torch.aten.cat %31113, %int0_28657 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %31114, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%31115 = torch.prim.ListConstruct %30981, %31045 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28658 = torch.constant.int 0
%31116 = torch.aten.cat %31115, %int0_28658 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31116, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31117 = torch.prim.ListConstruct %30984, %31048 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28659 = torch.constant.int 0
%31118 = torch.aten.cat %31117, %int0_28659 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31118, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31119 = torch.prim.ListConstruct %30987, %31051 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28660 = torch.constant.int 0
%31120 = torch.aten.cat %31119, %int0_28660 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31120, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31121 = torch.prim.ListConstruct %30990, %31054 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28661 = torch.constant.int 0
%31122 = torch.aten.cat %31121, %int0_28661 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31122, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31123 = torch.prim.ListConstruct %30993, %31057 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28662 = torch.constant.int 0
%31124 = torch.aten.cat %31123, %int0_28662 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31124, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31125 = torch.prim.ListConstruct %30996, %31060 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28663 = torch.constant.int 0
%31126 = torch.aten.cat %31125, %int0_28663 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31126, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31127 = torch.prim.ListConstruct %30999, %31063 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28664 = torch.constant.int 0
%31128 = torch.aten.cat %31127, %int0_28664 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31128, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31129 = torch.prim.ListConstruct %31002, %31066 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_28665 = torch.constant.int 0
%31130 = torch.aten.cat %31129, %int0_28665 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31130, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28666 = torch.constant.int 32
%int2_28667 = torch.constant.int 2
%int16_28668 = torch.constant.int 16
%int1_28669 = torch.constant.int 1
%int128_28670 = torch.constant.int 128
%31131 = torch.prim.ListConstruct %3023, %int32_28666, %int2_28667, %int16_28668, %int1_28669, %int128_28670 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31132 = torch.aten.view %29281, %31131 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31132, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28671 = torch.constant.int 32
%31133 = torch.aten.mul.int %3023, %int32_28671 : !torch.int, !torch.int -> !torch.int
%int2_28672 = torch.constant.int 2
%31134 = torch.aten.mul.int %31133, %int2_28672 : !torch.int, !torch.int -> !torch.int
%int16_28673 = torch.constant.int 16
%int1_28674 = torch.constant.int 1
%int128_28675 = torch.constant.int 128
%31135 = torch.prim.ListConstruct %31134, %int16_28673, %int1_28674, %int128_28675 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31136 = torch.aten.view %31132, %31135 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31136, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31137 = torch.prim.ListConstruct %31100 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28676 = torch.constant.bool false
%31138 = torch.aten.index_put %31136, %31137, %31116, %false_28676 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31138, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28677 = torch.constant.int 32
%int2_28678 = torch.constant.int 2
%int16_28679 = torch.constant.int 16
%int1_28680 = torch.constant.int 1
%int128_28681 = torch.constant.int 128
%31139 = torch.prim.ListConstruct %3023, %int32_28677, %int2_28678, %int16_28679, %int1_28680, %int128_28681 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31140 = torch.aten.view %31138, %31139 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31140, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28682 = torch.constant.int 131072
%31141 = torch.prim.ListConstruct %3023, %int131072_28682 : (!torch.int, !torch.int) -> !torch.list<int>
%31142 = torch.aten.view %31140, %31141 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31142, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28683 = torch.constant.int 32
%int2_28684 = torch.constant.int 2
%int16_28685 = torch.constant.int 16
%int1_28686 = torch.constant.int 1
%int128_28687 = torch.constant.int 128
%31143 = torch.prim.ListConstruct %3026, %int32_28683, %int2_28684, %int16_28685, %int1_28686, %int128_28687 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31144 = torch.aten.view %29293, %31143 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31144, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28688 = torch.constant.int 32
%31145 = torch.aten.mul.int %3026, %int32_28688 : !torch.int, !torch.int -> !torch.int
%int2_28689 = torch.constant.int 2
%31146 = torch.aten.mul.int %31145, %int2_28689 : !torch.int, !torch.int -> !torch.int
%int16_28690 = torch.constant.int 16
%int1_28691 = torch.constant.int 1
%int128_28692 = torch.constant.int 128
%31147 = torch.prim.ListConstruct %31146, %int16_28690, %int1_28691, %int128_28692 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31148 = torch.aten.view %31144, %31147 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31148, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31149 = torch.prim.ListConstruct %31102 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28693 = torch.constant.bool false
%31150 = torch.aten.index_put %31148, %31149, %31118, %false_28693 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31150, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28694 = torch.constant.int 32
%int2_28695 = torch.constant.int 2
%int16_28696 = torch.constant.int 16
%int1_28697 = torch.constant.int 1
%int128_28698 = torch.constant.int 128
%31151 = torch.prim.ListConstruct %3026, %int32_28694, %int2_28695, %int16_28696, %int1_28697, %int128_28698 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31152 = torch.aten.view %31150, %31151 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31152, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28699 = torch.constant.int 131072
%31153 = torch.prim.ListConstruct %3026, %int131072_28699 : (!torch.int, !torch.int) -> !torch.list<int>
%31154 = torch.aten.view %31152, %31153 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31154, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28700 = torch.constant.int 32
%int2_28701 = torch.constant.int 2
%int16_28702 = torch.constant.int 16
%int1_28703 = torch.constant.int 1
%int128_28704 = torch.constant.int 128
%31155 = torch.prim.ListConstruct %3029, %int32_28700, %int2_28701, %int16_28702, %int1_28703, %int128_28704 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31156 = torch.aten.view %29305, %31155 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31156, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28705 = torch.constant.int 32
%31157 = torch.aten.mul.int %3029, %int32_28705 : !torch.int, !torch.int -> !torch.int
%int2_28706 = torch.constant.int 2
%31158 = torch.aten.mul.int %31157, %int2_28706 : !torch.int, !torch.int -> !torch.int
%int16_28707 = torch.constant.int 16
%int1_28708 = torch.constant.int 1
%int128_28709 = torch.constant.int 128
%31159 = torch.prim.ListConstruct %31158, %int16_28707, %int1_28708, %int128_28709 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31160 = torch.aten.view %31156, %31159 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31160, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31161 = torch.prim.ListConstruct %31104 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28710 = torch.constant.bool false
%31162 = torch.aten.index_put %31160, %31161, %31120, %false_28710 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31162, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28711 = torch.constant.int 32
%int2_28712 = torch.constant.int 2
%int16_28713 = torch.constant.int 16
%int1_28714 = torch.constant.int 1
%int128_28715 = torch.constant.int 128
%31163 = torch.prim.ListConstruct %3029, %int32_28711, %int2_28712, %int16_28713, %int1_28714, %int128_28715 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31164 = torch.aten.view %31162, %31163 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31164, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28716 = torch.constant.int 131072
%31165 = torch.prim.ListConstruct %3029, %int131072_28716 : (!torch.int, !torch.int) -> !torch.list<int>
%31166 = torch.aten.view %31164, %31165 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31166, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28717 = torch.constant.int 32
%int2_28718 = torch.constant.int 2
%int16_28719 = torch.constant.int 16
%int1_28720 = torch.constant.int 1
%int128_28721 = torch.constant.int 128
%31167 = torch.prim.ListConstruct %3032, %int32_28717, %int2_28718, %int16_28719, %int1_28720, %int128_28721 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31168 = torch.aten.view %29317, %31167 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31168, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28722 = torch.constant.int 32
%31169 = torch.aten.mul.int %3032, %int32_28722 : !torch.int, !torch.int -> !torch.int
%int2_28723 = torch.constant.int 2
%31170 = torch.aten.mul.int %31169, %int2_28723 : !torch.int, !torch.int -> !torch.int
%int16_28724 = torch.constant.int 16
%int1_28725 = torch.constant.int 1
%int128_28726 = torch.constant.int 128
%31171 = torch.prim.ListConstruct %31170, %int16_28724, %int1_28725, %int128_28726 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31172 = torch.aten.view %31168, %31171 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31172, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31173 = torch.prim.ListConstruct %31106 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28727 = torch.constant.bool false
%31174 = torch.aten.index_put %31172, %31173, %31122, %false_28727 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31174, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28728 = torch.constant.int 32
%int2_28729 = torch.constant.int 2
%int16_28730 = torch.constant.int 16
%int1_28731 = torch.constant.int 1
%int128_28732 = torch.constant.int 128
%31175 = torch.prim.ListConstruct %3032, %int32_28728, %int2_28729, %int16_28730, %int1_28731, %int128_28732 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31176 = torch.aten.view %31174, %31175 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31176, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28733 = torch.constant.int 131072
%31177 = torch.prim.ListConstruct %3032, %int131072_28733 : (!torch.int, !torch.int) -> !torch.list<int>
%31178 = torch.aten.view %31176, %31177 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31178, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28734 = torch.constant.int 32
%int2_28735 = torch.constant.int 2
%int16_28736 = torch.constant.int 16
%int1_28737 = torch.constant.int 1
%int128_28738 = torch.constant.int 128
%31179 = torch.prim.ListConstruct %3035, %int32_28734, %int2_28735, %int16_28736, %int1_28737, %int128_28738 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31180 = torch.aten.view %29329, %31179 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31180, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28739 = torch.constant.int 32
%31181 = torch.aten.mul.int %3035, %int32_28739 : !torch.int, !torch.int -> !torch.int
%int2_28740 = torch.constant.int 2
%31182 = torch.aten.mul.int %31181, %int2_28740 : !torch.int, !torch.int -> !torch.int
%int16_28741 = torch.constant.int 16
%int1_28742 = torch.constant.int 1
%int128_28743 = torch.constant.int 128
%31183 = torch.prim.ListConstruct %31182, %int16_28741, %int1_28742, %int128_28743 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31184 = torch.aten.view %31180, %31183 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31184, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31185 = torch.prim.ListConstruct %31108 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28744 = torch.constant.bool false
%31186 = torch.aten.index_put %31184, %31185, %31124, %false_28744 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31186, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28745 = torch.constant.int 32
%int2_28746 = torch.constant.int 2
%int16_28747 = torch.constant.int 16
%int1_28748 = torch.constant.int 1
%int128_28749 = torch.constant.int 128
%31187 = torch.prim.ListConstruct %3035, %int32_28745, %int2_28746, %int16_28747, %int1_28748, %int128_28749 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31188 = torch.aten.view %31186, %31187 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31188, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28750 = torch.constant.int 131072
%31189 = torch.prim.ListConstruct %3035, %int131072_28750 : (!torch.int, !torch.int) -> !torch.list<int>
%31190 = torch.aten.view %31188, %31189 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31190, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28751 = torch.constant.int 32
%int2_28752 = torch.constant.int 2
%int16_28753 = torch.constant.int 16
%int1_28754 = torch.constant.int 1
%int128_28755 = torch.constant.int 128
%31191 = torch.prim.ListConstruct %3038, %int32_28751, %int2_28752, %int16_28753, %int1_28754, %int128_28755 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31192 = torch.aten.view %29341, %31191 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31192, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28756 = torch.constant.int 32
%31193 = torch.aten.mul.int %3038, %int32_28756 : !torch.int, !torch.int -> !torch.int
%int2_28757 = torch.constant.int 2
%31194 = torch.aten.mul.int %31193, %int2_28757 : !torch.int, !torch.int -> !torch.int
%int16_28758 = torch.constant.int 16
%int1_28759 = torch.constant.int 1
%int128_28760 = torch.constant.int 128
%31195 = torch.prim.ListConstruct %31194, %int16_28758, %int1_28759, %int128_28760 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31196 = torch.aten.view %31192, %31195 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31196, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31197 = torch.prim.ListConstruct %31110 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28761 = torch.constant.bool false
%31198 = torch.aten.index_put %31196, %31197, %31126, %false_28761 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31198, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28762 = torch.constant.int 32
%int2_28763 = torch.constant.int 2
%int16_28764 = torch.constant.int 16
%int1_28765 = torch.constant.int 1
%int128_28766 = torch.constant.int 128
%31199 = torch.prim.ListConstruct %3038, %int32_28762, %int2_28763, %int16_28764, %int1_28765, %int128_28766 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31200 = torch.aten.view %31198, %31199 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31200, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28767 = torch.constant.int 131072
%31201 = torch.prim.ListConstruct %3038, %int131072_28767 : (!torch.int, !torch.int) -> !torch.list<int>
%31202 = torch.aten.view %31200, %31201 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31202, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28768 = torch.constant.int 32
%int2_28769 = torch.constant.int 2
%int16_28770 = torch.constant.int 16
%int1_28771 = torch.constant.int 1
%int128_28772 = torch.constant.int 128
%31203 = torch.prim.ListConstruct %3041, %int32_28768, %int2_28769, %int16_28770, %int1_28771, %int128_28772 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31204 = torch.aten.view %29353, %31203 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31204, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28773 = torch.constant.int 32
%31205 = torch.aten.mul.int %3041, %int32_28773 : !torch.int, !torch.int -> !torch.int
%int2_28774 = torch.constant.int 2
%31206 = torch.aten.mul.int %31205, %int2_28774 : !torch.int, !torch.int -> !torch.int
%int16_28775 = torch.constant.int 16
%int1_28776 = torch.constant.int 1
%int128_28777 = torch.constant.int 128
%31207 = torch.prim.ListConstruct %31206, %int16_28775, %int1_28776, %int128_28777 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31208 = torch.aten.view %31204, %31207 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31208, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31209 = torch.prim.ListConstruct %31112 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28778 = torch.constant.bool false
%31210 = torch.aten.index_put %31208, %31209, %31128, %false_28778 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31210, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28779 = torch.constant.int 32
%int2_28780 = torch.constant.int 2
%int16_28781 = torch.constant.int 16
%int1_28782 = torch.constant.int 1
%int128_28783 = torch.constant.int 128
%31211 = torch.prim.ListConstruct %3041, %int32_28779, %int2_28780, %int16_28781, %int1_28782, %int128_28783 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31212 = torch.aten.view %31210, %31211 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31212, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28784 = torch.constant.int 131072
%31213 = torch.prim.ListConstruct %3041, %int131072_28784 : (!torch.int, !torch.int) -> !torch.list<int>
%31214 = torch.aten.view %31212, %31213 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31214, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_28785 = torch.constant.int 32
%int2_28786 = torch.constant.int 2
%int16_28787 = torch.constant.int 16
%int1_28788 = torch.constant.int 1
%int128_28789 = torch.constant.int 128
%31215 = torch.prim.ListConstruct %3044, %int32_28785, %int2_28786, %int16_28787, %int1_28788, %int128_28789 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31216 = torch.aten.view %29365, %31215 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31216, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_28790 = torch.constant.int 32
%31217 = torch.aten.mul.int %3044, %int32_28790 : !torch.int, !torch.int -> !torch.int
%int2_28791 = torch.constant.int 2
%31218 = torch.aten.mul.int %31217, %int2_28791 : !torch.int, !torch.int -> !torch.int
%int16_28792 = torch.constant.int 16
%int1_28793 = torch.constant.int 1
%int128_28794 = torch.constant.int 128
%31219 = torch.prim.ListConstruct %31218, %int16_28792, %int1_28793, %int128_28794 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31220 = torch.aten.view %31216, %31219 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31220, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%31221 = torch.prim.ListConstruct %31114 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_28795 = torch.constant.bool false
%31222 = torch.aten.index_put %31220, %31221, %31130, %false_28795 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %31222, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_28796 = torch.constant.int 32
%int2_28797 = torch.constant.int 2
%int16_28798 = torch.constant.int 16
%int1_28799 = torch.constant.int 1
%int128_28800 = torch.constant.int 128
%31223 = torch.prim.ListConstruct %3044, %int32_28796, %int2_28797, %int16_28798, %int1_28799, %int128_28800 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31224 = torch.aten.view %31222, %31223 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %31224, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_28801 = torch.constant.int 131072
%31225 = torch.prim.ListConstruct %3044, %int131072_28801 : (!torch.int, !torch.int) -> !torch.list<int>
%31226 = torch.aten.view %31224, %31225 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %31226, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_28802 = torch.constant.int -2
%31227 = torch.aten.unsqueeze %30841, %int-2_28802 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28803 = torch.constant.int -2
%31228 = torch.aten.unsqueeze %30856, %int-2_28803 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28804 = torch.constant.int -2
%31229 = torch.aten.unsqueeze %30871, %int-2_28804 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28805 = torch.constant.int -2
%31230 = torch.aten.unsqueeze %30886, %int-2_28805 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28806 = torch.constant.int -2
%31231 = torch.aten.unsqueeze %30901, %int-2_28806 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28807 = torch.constant.int -2
%31232 = torch.aten.unsqueeze %30916, %int-2_28807 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28808 = torch.constant.int -2
%31233 = torch.aten.unsqueeze %30931, %int-2_28808 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28809 = torch.constant.int -2
%31234 = torch.aten.unsqueeze %30946, %int-2_28809 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_28810 = torch.constant.int 4
%int1_28811 = torch.constant.int 1
%int4_28812 = torch.constant.int 4
%int128_28813 = torch.constant.int 128
%31235 = torch.prim.ListConstruct %int4_28810, %30827, %int1_28811, %int4_28812, %int128_28813 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28814 = torch.constant.bool false
%31236 = torch.aten.expand %31227, %31235, %false_28814 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31236, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28815 = torch.constant.int 4
%int1_28816 = torch.constant.int 1
%int4_28817 = torch.constant.int 4
%int128_28818 = torch.constant.int 128
%31237 = torch.prim.ListConstruct %int4_28815, %30827, %int1_28816, %int4_28817, %int128_28818 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28819 = torch.constant.bool false
%31238 = torch.aten.expand %31228, %31237, %false_28819 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28820 = torch.constant.int 4
%int1_28821 = torch.constant.int 1
%int4_28822 = torch.constant.int 4
%int128_28823 = torch.constant.int 128
%31239 = torch.prim.ListConstruct %int4_28820, %30827, %int1_28821, %int4_28822, %int128_28823 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28824 = torch.constant.bool false
%31240 = torch.aten.expand %31229, %31239, %false_28824 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28825 = torch.constant.int 4
%int1_28826 = torch.constant.int 1
%int4_28827 = torch.constant.int 4
%int128_28828 = torch.constant.int 128
%31241 = torch.prim.ListConstruct %int4_28825, %30827, %int1_28826, %int4_28827, %int128_28828 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28829 = torch.constant.bool false
%31242 = torch.aten.expand %31230, %31241, %false_28829 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28830 = torch.constant.int 4
%int1_28831 = torch.constant.int 1
%int4_28832 = torch.constant.int 4
%int128_28833 = torch.constant.int 128
%31243 = torch.prim.ListConstruct %int4_28830, %30827, %int1_28831, %int4_28832, %int128_28833 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28834 = torch.constant.bool false
%31244 = torch.aten.expand %31231, %31243, %false_28834 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28835 = torch.constant.int 4
%int1_28836 = torch.constant.int 1
%int4_28837 = torch.constant.int 4
%int128_28838 = torch.constant.int 128
%31245 = torch.prim.ListConstruct %int4_28835, %30827, %int1_28836, %int4_28837, %int128_28838 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28839 = torch.constant.bool false
%31246 = torch.aten.expand %31232, %31245, %false_28839 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28840 = torch.constant.int 4
%int1_28841 = torch.constant.int 1
%int4_28842 = torch.constant.int 4
%int128_28843 = torch.constant.int 128
%31247 = torch.prim.ListConstruct %int4_28840, %30827, %int1_28841, %int4_28842, %int128_28843 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28844 = torch.constant.bool false
%31248 = torch.aten.expand %31233, %31247, %false_28844 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28845 = torch.constant.int 4
%int1_28846 = torch.constant.int 1
%int4_28847 = torch.constant.int 4
%int128_28848 = torch.constant.int 128
%31249 = torch.prim.ListConstruct %int4_28845, %30827, %int1_28846, %int4_28847, %int128_28848 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28849 = torch.constant.bool false
%31250 = torch.aten.expand %31234, %31249, %false_28849 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28850 = torch.constant.int 4
%int4_28851 = torch.constant.int 4
%int128_28852 = torch.constant.int 128
%31251 = torch.prim.ListConstruct %int4_28850, %30827, %int4_28851, %int128_28852 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31252 = torch.aten.view %31236, %31251 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28853 = torch.constant.int 4
%int4_28854 = torch.constant.int 4
%int128_28855 = torch.constant.int 128
%31253 = torch.prim.ListConstruct %int4_28853, %30827, %int4_28854, %int128_28855 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31254 = torch.aten.view %31238, %31253 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28856 = torch.constant.int 4
%int4_28857 = torch.constant.int 4
%int128_28858 = torch.constant.int 128
%31255 = torch.prim.ListConstruct %int4_28856, %30827, %int4_28857, %int128_28858 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31256 = torch.aten.view %31240, %31255 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28859 = torch.constant.int 4
%int4_28860 = torch.constant.int 4
%int128_28861 = torch.constant.int 128
%31257 = torch.prim.ListConstruct %int4_28859, %30827, %int4_28860, %int128_28861 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31258 = torch.aten.view %31242, %31257 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28862 = torch.constant.int 4
%int4_28863 = torch.constant.int 4
%int128_28864 = torch.constant.int 128
%31259 = torch.prim.ListConstruct %int4_28862, %30827, %int4_28863, %int128_28864 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31260 = torch.aten.view %31244, %31259 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28865 = torch.constant.int 4
%int4_28866 = torch.constant.int 4
%int128_28867 = torch.constant.int 128
%31261 = torch.prim.ListConstruct %int4_28865, %30827, %int4_28866, %int128_28867 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31262 = torch.aten.view %31246, %31261 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28868 = torch.constant.int 4
%int4_28869 = torch.constant.int 4
%int128_28870 = torch.constant.int 128
%31263 = torch.prim.ListConstruct %int4_28868, %30827, %int4_28869, %int128_28870 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31264 = torch.aten.view %31248, %31263 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28871 = torch.constant.int 4
%int4_28872 = torch.constant.int 4
%int128_28873 = torch.constant.int 128
%31265 = torch.prim.ListConstruct %int4_28871, %30827, %int4_28872, %int128_28873 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31266 = torch.aten.view %31250, %31265 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_28874 = torch.constant.int -2
%31267 = torch.aten.unsqueeze %30616, %int-2_28874 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28875 = torch.constant.int -2
%31268 = torch.aten.unsqueeze %30618, %int-2_28875 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28876 = torch.constant.int -2
%31269 = torch.aten.unsqueeze %30620, %int-2_28876 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28877 = torch.constant.int -2
%31270 = torch.aten.unsqueeze %30622, %int-2_28877 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28878 = torch.constant.int -2
%31271 = torch.aten.unsqueeze %30624, %int-2_28878 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31271, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28879 = torch.constant.int -2
%31272 = torch.aten.unsqueeze %30626, %int-2_28879 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28880 = torch.constant.int -2
%31273 = torch.aten.unsqueeze %30628, %int-2_28880 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_28881 = torch.constant.int -2
%31274 = torch.aten.unsqueeze %30630, %int-2_28881 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %31274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_28882 = torch.constant.int 1
%31275 = torch.aten.size.int %30540, %int1_28882 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_28883 = torch.constant.int 4
%int1_28884 = torch.constant.int 1
%int4_28885 = torch.constant.int 4
%int128_28886 = torch.constant.int 128
%31276 = torch.prim.ListConstruct %int4_28883, %31275, %int1_28884, %int4_28885, %int128_28886 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28887 = torch.constant.bool false
%31277 = torch.aten.expand %31267, %31276, %false_28887 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28888 = torch.constant.int 4
%int1_28889 = torch.constant.int 1
%int4_28890 = torch.constant.int 4
%int128_28891 = torch.constant.int 128
%31278 = torch.prim.ListConstruct %int4_28888, %31275, %int1_28889, %int4_28890, %int128_28891 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28892 = torch.constant.bool false
%31279 = torch.aten.expand %31268, %31278, %false_28892 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28893 = torch.constant.int 4
%int1_28894 = torch.constant.int 1
%int4_28895 = torch.constant.int 4
%int128_28896 = torch.constant.int 128
%31280 = torch.prim.ListConstruct %int4_28893, %31275, %int1_28894, %int4_28895, %int128_28896 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28897 = torch.constant.bool false
%31281 = torch.aten.expand %31269, %31280, %false_28897 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28898 = torch.constant.int 4
%int1_28899 = torch.constant.int 1
%int4_28900 = torch.constant.int 4
%int128_28901 = torch.constant.int 128
%31282 = torch.prim.ListConstruct %int4_28898, %31275, %int1_28899, %int4_28900, %int128_28901 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28902 = torch.constant.bool false
%31283 = torch.aten.expand %31270, %31282, %false_28902 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28903 = torch.constant.int 4
%int1_28904 = torch.constant.int 1
%int4_28905 = torch.constant.int 4
%int128_28906 = torch.constant.int 128
%31284 = torch.prim.ListConstruct %int4_28903, %31275, %int1_28904, %int4_28905, %int128_28906 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28907 = torch.constant.bool false
%31285 = torch.aten.expand %31271, %31284, %false_28907 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28908 = torch.constant.int 4
%int1_28909 = torch.constant.int 1
%int4_28910 = torch.constant.int 4
%int128_28911 = torch.constant.int 128
%31286 = torch.prim.ListConstruct %int4_28908, %31275, %int1_28909, %int4_28910, %int128_28911 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28912 = torch.constant.bool false
%31287 = torch.aten.expand %31272, %31286, %false_28912 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28913 = torch.constant.int 4
%int1_28914 = torch.constant.int 1
%int4_28915 = torch.constant.int 4
%int128_28916 = torch.constant.int 128
%31288 = torch.prim.ListConstruct %int4_28913, %31275, %int1_28914, %int4_28915, %int128_28916 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28917 = torch.constant.bool false
%31289 = torch.aten.expand %31273, %31288, %false_28917 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28918 = torch.constant.int 4
%int1_28919 = torch.constant.int 1
%int4_28920 = torch.constant.int 4
%int128_28921 = torch.constant.int 128
%31290 = torch.prim.ListConstruct %int4_28918, %31275, %int1_28919, %int4_28920, %int128_28921 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_28922 = torch.constant.bool false
%31291 = torch.aten.expand %31274, %31290, %false_28922 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %31291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_28923 = torch.constant.int 4
%int4_28924 = torch.constant.int 4
%int128_28925 = torch.constant.int 128
%31292 = torch.prim.ListConstruct %int4_28923, %31275, %int4_28924, %int128_28925 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31293 = torch.aten.view %31277, %31292 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28926 = torch.constant.int 4
%int4_28927 = torch.constant.int 4
%int128_28928 = torch.constant.int 128
%31294 = torch.prim.ListConstruct %int4_28926, %31275, %int4_28927, %int128_28928 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31295 = torch.aten.view %31279, %31294 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28929 = torch.constant.int 4
%int4_28930 = torch.constant.int 4
%int128_28931 = torch.constant.int 128
%31296 = torch.prim.ListConstruct %int4_28929, %31275, %int4_28930, %int128_28931 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31297 = torch.aten.view %31281, %31296 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28932 = torch.constant.int 4
%int4_28933 = torch.constant.int 4
%int128_28934 = torch.constant.int 128
%31298 = torch.prim.ListConstruct %int4_28932, %31275, %int4_28933, %int128_28934 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31299 = torch.aten.view %31283, %31298 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28935 = torch.constant.int 4
%int4_28936 = torch.constant.int 4
%int128_28937 = torch.constant.int 128
%31300 = torch.prim.ListConstruct %int4_28935, %31275, %int4_28936, %int128_28937 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31301 = torch.aten.view %31285, %31300 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28938 = torch.constant.int 4
%int4_28939 = torch.constant.int 4
%int128_28940 = torch.constant.int 128
%31302 = torch.prim.ListConstruct %int4_28938, %31275, %int4_28939, %int128_28940 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31303 = torch.aten.view %31287, %31302 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28941 = torch.constant.int 4
%int4_28942 = torch.constant.int 4
%int128_28943 = torch.constant.int 128
%31304 = torch.prim.ListConstruct %int4_28941, %31275, %int4_28942, %int128_28943 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31305 = torch.aten.view %31289, %31304 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_28944 = torch.constant.int 4
%int4_28945 = torch.constant.int 4
%int128_28946 = torch.constant.int 128
%31306 = torch.prim.ListConstruct %int4_28944, %31275, %int4_28945, %int128_28946 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31307 = torch.aten.view %31291, %31306 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_28947 = torch.constant.int 1
%int2_28948 = torch.constant.int 2
%31308 = torch.aten.transpose.int %30683, %int1_28947, %int2_28948 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31308, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28949 = torch.constant.int 1
%int2_28950 = torch.constant.int 2
%31309 = torch.aten.transpose.int %30698, %int1_28949, %int2_28950 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31309, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28951 = torch.constant.int 1
%int2_28952 = torch.constant.int 2
%31310 = torch.aten.transpose.int %30713, %int1_28951, %int2_28952 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31310, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28953 = torch.constant.int 1
%int2_28954 = torch.constant.int 2
%31311 = torch.aten.transpose.int %30728, %int1_28953, %int2_28954 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31311, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28955 = torch.constant.int 1
%int2_28956 = torch.constant.int 2
%31312 = torch.aten.transpose.int %30743, %int1_28955, %int2_28956 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31312, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28957 = torch.constant.int 1
%int2_28958 = torch.constant.int 2
%31313 = torch.aten.transpose.int %30758, %int1_28957, %int2_28958 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31313, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28959 = torch.constant.int 1
%int2_28960 = torch.constant.int 2
%31314 = torch.aten.transpose.int %30773, %int1_28959, %int2_28960 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31314, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28961 = torch.constant.int 1
%int2_28962 = torch.constant.int 2
%31315 = torch.aten.transpose.int %30788, %int1_28961, %int2_28962 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31315, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28963 = torch.constant.int 1
%int2_28964 = torch.constant.int 2
%31316 = torch.aten.transpose.int %31252, %int1_28963, %int2_28964 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31316, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28965 = torch.constant.int 1
%int2_28966 = torch.constant.int 2
%31317 = torch.aten.transpose.int %31254, %int1_28965, %int2_28966 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31317, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28967 = torch.constant.int 1
%int2_28968 = torch.constant.int 2
%31318 = torch.aten.transpose.int %31256, %int1_28967, %int2_28968 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31318, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28969 = torch.constant.int 1
%int2_28970 = torch.constant.int 2
%31319 = torch.aten.transpose.int %31258, %int1_28969, %int2_28970 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31319, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28971 = torch.constant.int 1
%int2_28972 = torch.constant.int 2
%31320 = torch.aten.transpose.int %31260, %int1_28971, %int2_28972 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31320, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28973 = torch.constant.int 1
%int2_28974 = torch.constant.int 2
%31321 = torch.aten.transpose.int %31262, %int1_28973, %int2_28974 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31321, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28975 = torch.constant.int 1
%int2_28976 = torch.constant.int 2
%31322 = torch.aten.transpose.int %31264, %int1_28975, %int2_28976 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31322, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28977 = torch.constant.int 1
%int2_28978 = torch.constant.int 2
%31323 = torch.aten.transpose.int %31266, %int1_28977, %int2_28978 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31323, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28979 = torch.constant.int 1
%int2_28980 = torch.constant.int 2
%31324 = torch.aten.transpose.int %31293, %int1_28979, %int2_28980 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31324, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28981 = torch.constant.int 1
%int2_28982 = torch.constant.int 2
%31325 = torch.aten.transpose.int %31295, %int1_28981, %int2_28982 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31325, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28983 = torch.constant.int 1
%int2_28984 = torch.constant.int 2
%31326 = torch.aten.transpose.int %31297, %int1_28983, %int2_28984 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31326, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28985 = torch.constant.int 1
%int2_28986 = torch.constant.int 2
%31327 = torch.aten.transpose.int %31299, %int1_28985, %int2_28986 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31327, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28987 = torch.constant.int 1
%int2_28988 = torch.constant.int 2
%31328 = torch.aten.transpose.int %31301, %int1_28987, %int2_28988 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31328, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28989 = torch.constant.int 1
%int2_28990 = torch.constant.int 2
%31329 = torch.aten.transpose.int %31303, %int1_28989, %int2_28990 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31329, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28991 = torch.constant.int 1
%int2_28992 = torch.constant.int 2
%31330 = torch.aten.transpose.int %31305, %int1_28991, %int2_28992 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31330, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_28993 = torch.constant.int 1
%int2_28994 = torch.constant.int 2
%31331 = torch.aten.transpose.int %31307, %int1_28993, %int2_28994 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %31331, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_28995 = torch.constant.float 0.000000e+00
%true_28996 = torch.constant.bool true
%none_28997 = torch.constant.none
%none_28998 = torch.constant.none
%31332:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31308, %31316, %31324, %float0.000000e00_28995, %true_28996, %none_28997, %none_28998) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31332#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_28999 = torch.constant.float 0.000000e+00
%true_29000 = torch.constant.bool true
%none_29001 = torch.constant.none
%none_29002 = torch.constant.none
%31333:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31309, %31317, %31325, %float0.000000e00_28999, %true_29000, %none_29001, %none_29002) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31333#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_29003 = torch.constant.float 0.000000e+00
%true_29004 = torch.constant.bool true
%none_29005 = torch.constant.none
%none_29006 = torch.constant.none
%31334:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31310, %31318, %31326, %float0.000000e00_29003, %true_29004, %none_29005, %none_29006) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31334#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_29007 = torch.constant.float 0.000000e+00
%true_29008 = torch.constant.bool true
%none_29009 = torch.constant.none
%none_29010 = torch.constant.none
%31335:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31311, %31319, %31327, %float0.000000e00_29007, %true_29008, %none_29009, %none_29010) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31335#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_29011 = torch.constant.float 0.000000e+00
%true_29012 = torch.constant.bool true
%none_29013 = torch.constant.none
%none_29014 = torch.constant.none
%31336:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31312, %31320, %31328, %float0.000000e00_29011, %true_29012, %none_29013, %none_29014) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31336#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_29015 = torch.constant.float 0.000000e+00
%true_29016 = torch.constant.bool true
%none_29017 = torch.constant.none
%none_29018 = torch.constant.none
%31337:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31313, %31321, %31329, %float0.000000e00_29015, %true_29016, %none_29017, %none_29018) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31337#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_29019 = torch.constant.float 0.000000e+00
%true_29020 = torch.constant.bool true
%none_29021 = torch.constant.none
%none_29022 = torch.constant.none
%31338:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31314, %31322, %31330, %float0.000000e00_29019, %true_29020, %none_29021, %none_29022) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31338#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_29023 = torch.constant.float 0.000000e+00
%true_29024 = torch.constant.bool true
%none_29025 = torch.constant.none
%none_29026 = torch.constant.none
%31339:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%31315, %31323, %31331, %float0.000000e00_29023, %true_29024, %none_29025, %none_29026) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %31339#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_29027 = torch.constant.int 1
%int2_29028 = torch.constant.int 2
%31340 = torch.aten.transpose.int %31332#0, %int1_29027, %int2_29028 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29029 = torch.constant.int 1
%int2_29030 = torch.constant.int 2
%31341 = torch.aten.transpose.int %31333#0, %int1_29029, %int2_29030 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29031 = torch.constant.int 1
%int2_29032 = torch.constant.int 2
%31342 = torch.aten.transpose.int %31334#0, %int1_29031, %int2_29032 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29033 = torch.constant.int 1
%int2_29034 = torch.constant.int 2
%31343 = torch.aten.transpose.int %31335#0, %int1_29033, %int2_29034 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29035 = torch.constant.int 1
%int2_29036 = torch.constant.int 2
%31344 = torch.aten.transpose.int %31336#0, %int1_29035, %int2_29036 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29037 = torch.constant.int 1
%int2_29038 = torch.constant.int 2
%31345 = torch.aten.transpose.int %31337#0, %int1_29037, %int2_29038 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29039 = torch.constant.int 1
%int2_29040 = torch.constant.int 2
%31346 = torch.aten.transpose.int %31338#0, %int1_29039, %int2_29040 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29041 = torch.constant.int 1
%int2_29042 = torch.constant.int 2
%31347 = torch.aten.transpose.int %31339#0, %int1_29041, %int2_29042 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %31347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29043 = torch.constant.int 4
%int512_29044 = torch.constant.int 512
%31348 = torch.prim.ListConstruct %int4_29043, %30669, %int512_29044 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31349 = torch.aten.view %31340, %31348 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29045 = torch.constant.int 4
%int512_29046 = torch.constant.int 512
%31350 = torch.prim.ListConstruct %int4_29045, %30684, %int512_29046 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31351 = torch.aten.view %31341, %31350 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29047 = torch.constant.int 4
%int512_29048 = torch.constant.int 512
%31352 = torch.prim.ListConstruct %int4_29047, %30699, %int512_29048 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31353 = torch.aten.view %31342, %31352 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29049 = torch.constant.int 4
%int512_29050 = torch.constant.int 512
%31354 = torch.prim.ListConstruct %int4_29049, %30714, %int512_29050 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31355 = torch.aten.view %31343, %31354 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29051 = torch.constant.int 4
%int512_29052 = torch.constant.int 512
%31356 = torch.prim.ListConstruct %int4_29051, %30729, %int512_29052 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31357 = torch.aten.view %31344, %31356 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29053 = torch.constant.int 4
%int512_29054 = torch.constant.int 512
%31358 = torch.prim.ListConstruct %int4_29053, %30744, %int512_29054 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31359 = torch.aten.view %31345, %31358 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29055 = torch.constant.int 4
%int512_29056 = torch.constant.int 512
%31360 = torch.prim.ListConstruct %int4_29055, %30759, %int512_29056 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31361 = torch.aten.view %31346, %31360 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29057 = torch.constant.int 4
%int512_29058 = torch.constant.int 512
%31362 = torch.prim.ListConstruct %int4_29057, %30774, %int512_29058 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31363 = torch.aten.view %31347, %31362 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %31363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_29059 = torch.constant.int 1
%int0_29060 = torch.constant.int 0
%31364 = torch.prim.ListConstruct %int1_29059, %int0_29060 : (!torch.int, !torch.int) -> !torch.list<int>
%31365 = torch.aten.permute %1120, %31364 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29061 = torch.constant.int 1
%int0_29062 = torch.constant.int 0
%31366 = torch.prim.ListConstruct %int1_29061, %int0_29062 : (!torch.int, !torch.int) -> !torch.list<int>
%31367 = torch.aten.permute %1121, %31366 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29063 = torch.constant.int 1
%int0_29064 = torch.constant.int 0
%31368 = torch.prim.ListConstruct %int1_29063, %int0_29064 : (!torch.int, !torch.int) -> !torch.list<int>
%31369 = torch.aten.permute %1122, %31368 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29065 = torch.constant.int 1
%int0_29066 = torch.constant.int 0
%31370 = torch.prim.ListConstruct %int1_29065, %int0_29066 : (!torch.int, !torch.int) -> !torch.list<int>
%31371 = torch.aten.permute %1123, %31370 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29067 = torch.constant.int 1
%int0_29068 = torch.constant.int 0
%31372 = torch.prim.ListConstruct %int1_29067, %int0_29068 : (!torch.int, !torch.int) -> !torch.list<int>
%31373 = torch.aten.permute %1124, %31372 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29069 = torch.constant.int 1
%int0_29070 = torch.constant.int 0
%31374 = torch.prim.ListConstruct %int1_29069, %int0_29070 : (!torch.int, !torch.int) -> !torch.list<int>
%31375 = torch.aten.permute %1125, %31374 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29071 = torch.constant.int 1
%int0_29072 = torch.constant.int 0
%31376 = torch.prim.ListConstruct %int1_29071, %int0_29072 : (!torch.int, !torch.int) -> !torch.list<int>
%31377 = torch.aten.permute %1126, %31376 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_29073 = torch.constant.int 1
%int0_29074 = torch.constant.int 0
%31378 = torch.prim.ListConstruct %int1_29073, %int0_29074 : (!torch.int, !torch.int) -> !torch.list<int>
%31379 = torch.aten.permute %1127, %31378 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_29075 = torch.constant.int 4
%31380 = torch.aten.mul.int %int4_29075, %30669 : !torch.int, !torch.int -> !torch.int
%int512_29076 = torch.constant.int 512
%31381 = torch.prim.ListConstruct %31380, %int512_29076 : (!torch.int, !torch.int) -> !torch.list<int>
%31382 = torch.aten.view %31349, %31381 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31382, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31383 = torch.aten.mm %31382, %31365 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31383, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29077 = torch.constant.int 4
%int4096_29078 = torch.constant.int 4096
%31384 = torch.prim.ListConstruct %int4_29077, %30669, %int4096_29078 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31385 = torch.aten.view %31383, %31384 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29079 = torch.constant.int 4
%31386 = torch.aten.mul.int %int4_29079, %30684 : !torch.int, !torch.int -> !torch.int
%int512_29080 = torch.constant.int 512
%31387 = torch.prim.ListConstruct %31386, %int512_29080 : (!torch.int, !torch.int) -> !torch.list<int>
%31388 = torch.aten.view %31351, %31387 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31388, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31389 = torch.aten.mm %31388, %31367 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31389, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29081 = torch.constant.int 4
%int4096_29082 = torch.constant.int 4096
%31390 = torch.prim.ListConstruct %int4_29081, %30684, %int4096_29082 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31391 = torch.aten.view %31389, %31390 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29083 = torch.constant.int 4
%31392 = torch.aten.mul.int %int4_29083, %30699 : !torch.int, !torch.int -> !torch.int
%int512_29084 = torch.constant.int 512
%31393 = torch.prim.ListConstruct %31392, %int512_29084 : (!torch.int, !torch.int) -> !torch.list<int>
%31394 = torch.aten.view %31353, %31393 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31394, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31395 = torch.aten.mm %31394, %31369 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31395, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29085 = torch.constant.int 4
%int4096_29086 = torch.constant.int 4096
%31396 = torch.prim.ListConstruct %int4_29085, %30699, %int4096_29086 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31397 = torch.aten.view %31395, %31396 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29087 = torch.constant.int 4
%31398 = torch.aten.mul.int %int4_29087, %30714 : !torch.int, !torch.int -> !torch.int
%int512_29088 = torch.constant.int 512
%31399 = torch.prim.ListConstruct %31398, %int512_29088 : (!torch.int, !torch.int) -> !torch.list<int>
%31400 = torch.aten.view %31355, %31399 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31400, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31401 = torch.aten.mm %31400, %31371 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31401, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29089 = torch.constant.int 4
%int4096_29090 = torch.constant.int 4096
%31402 = torch.prim.ListConstruct %int4_29089, %30714, %int4096_29090 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31403 = torch.aten.view %31401, %31402 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29091 = torch.constant.int 4
%31404 = torch.aten.mul.int %int4_29091, %30729 : !torch.int, !torch.int -> !torch.int
%int512_29092 = torch.constant.int 512
%31405 = torch.prim.ListConstruct %31404, %int512_29092 : (!torch.int, !torch.int) -> !torch.list<int>
%31406 = torch.aten.view %31357, %31405 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31406, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31407 = torch.aten.mm %31406, %31373 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31407, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29093 = torch.constant.int 4
%int4096_29094 = torch.constant.int 4096
%31408 = torch.prim.ListConstruct %int4_29093, %30729, %int4096_29094 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31409 = torch.aten.view %31407, %31408 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29095 = torch.constant.int 4
%31410 = torch.aten.mul.int %int4_29095, %30744 : !torch.int, !torch.int -> !torch.int
%int512_29096 = torch.constant.int 512
%31411 = torch.prim.ListConstruct %31410, %int512_29096 : (!torch.int, !torch.int) -> !torch.list<int>
%31412 = torch.aten.view %31359, %31411 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31412, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31413 = torch.aten.mm %31412, %31375 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31413, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29097 = torch.constant.int 4
%int4096_29098 = torch.constant.int 4096
%31414 = torch.prim.ListConstruct %int4_29097, %30744, %int4096_29098 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31415 = torch.aten.view %31413, %31414 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29099 = torch.constant.int 4
%31416 = torch.aten.mul.int %int4_29099, %30759 : !torch.int, !torch.int -> !torch.int
%int512_29100 = torch.constant.int 512
%31417 = torch.prim.ListConstruct %31416, %int512_29100 : (!torch.int, !torch.int) -> !torch.list<int>
%31418 = torch.aten.view %31361, %31417 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31418, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31419 = torch.aten.mm %31418, %31377 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31419, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29101 = torch.constant.int 4
%int4096_29102 = torch.constant.int 4096
%31420 = torch.prim.ListConstruct %int4_29101, %30759, %int4096_29102 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31421 = torch.aten.view %31419, %31420 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_29103 = torch.constant.int 4
%31422 = torch.aten.mul.int %int4_29103, %30774 : !torch.int, !torch.int -> !torch.int
%int512_29104 = torch.constant.int 512
%31423 = torch.prim.ListConstruct %31422, %int512_29104 : (!torch.int, !torch.int) -> !torch.list<int>
%31424 = torch.aten.view %31363, %31423 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %31424, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%31425 = torch.aten.mm %31424, %31379 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31425, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29105 = torch.constant.int 4
%int4096_29106 = torch.constant.int 4096
%31426 = torch.prim.ListConstruct %int4_29105, %30774, %int4096_29106 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31427 = torch.aten.view %31425, %31426 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31428 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29107 = arith.constant 1 : index
%dim_29108 = tensor.dim %31428, %c1_29107 : tensor<4x?x4096xf16>
%31429 = flow.tensor.transfer %31428 : tensor<4x?x4096xf16>{%dim_29108} to #hal.device.promise<@__device_0>
%31430 = torch_c.from_builtin_tensor %31429 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31431 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29109 = arith.constant 1 : index
%dim_29110 = tensor.dim %31431, %c1_29109 : tensor<4x?x4096xf16>
%31432 = flow.tensor.transfer %31431 : tensor<4x?x4096xf16>{%dim_29110} to #hal.device.promise<@__device_0>
%31433 = torch_c.from_builtin_tensor %31432 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31434 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29111 = arith.constant 1 : index
%dim_29112 = tensor.dim %31434, %c1_29111 : tensor<4x?x4096xf16>
%31435 = flow.tensor.transfer %31434 : tensor<4x?x4096xf16>{%dim_29112} to #hal.device.promise<@__device_0>
%31436 = torch_c.from_builtin_tensor %31435 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31437 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29113 = arith.constant 1 : index
%dim_29114 = tensor.dim %31437, %c1_29113 : tensor<4x?x4096xf16>
%31438 = flow.tensor.transfer %31437 : tensor<4x?x4096xf16>{%dim_29114} to #hal.device.promise<@__device_0>
%31439 = torch_c.from_builtin_tensor %31438 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31440 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29115 = arith.constant 1 : index
%dim_29116 = tensor.dim %31440, %c1_29115 : tensor<4x?x4096xf16>
%31441 = flow.tensor.transfer %31440 : tensor<4x?x4096xf16>{%dim_29116} to #hal.device.promise<@__device_0>
%31442 = torch_c.from_builtin_tensor %31441 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31443 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29117 = arith.constant 1 : index
%dim_29118 = tensor.dim %31443, %c1_29117 : tensor<4x?x4096xf16>
%31444 = flow.tensor.transfer %31443 : tensor<4x?x4096xf16>{%dim_29118} to #hal.device.promise<@__device_0>
%31445 = torch_c.from_builtin_tensor %31444 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31446 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29119 = arith.constant 1 : index
%dim_29120 = tensor.dim %31446, %c1_29119 : tensor<4x?x4096xf16>
%31447 = flow.tensor.transfer %31446 : tensor<4x?x4096xf16>{%dim_29120} to #hal.device.promise<@__device_0>
%31448 = torch_c.from_builtin_tensor %31447 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29121 = torch.constant.int 1
%31449 = torch.aten.add.Tensor %31385, %31430, %int1_29121 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29122 = torch.constant.int 1
%31450 = torch.aten.add.Tensor %31449, %31433, %int1_29122 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29123 = torch.constant.int 1
%31451 = torch.aten.add.Tensor %31450, %31436, %int1_29123 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29124 = torch.constant.int 1
%31452 = torch.aten.add.Tensor %31451, %31439, %int1_29124 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29125 = torch.constant.int 1
%31453 = torch.aten.add.Tensor %31452, %31442, %int1_29125 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29126 = torch.constant.int 1
%31454 = torch.aten.add.Tensor %31453, %31445, %int1_29126 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29127 = torch.constant.int 1
%31455 = torch.aten.add.Tensor %31454, %31448, %int1_29127 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31456 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29128 = arith.constant 1 : index
%dim_29129 = tensor.dim %31456, %c1_29128 : tensor<4x?x4096xf16>
%31457 = flow.tensor.transfer %31456 : tensor<4x?x4096xf16>{%dim_29129} to #hal.device.promise<@__device_1>
%31458 = torch_c.from_builtin_tensor %31457 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31459 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29130 = arith.constant 1 : index
%dim_29131 = tensor.dim %31459, %c1_29130 : tensor<4x?x4096xf16>
%31460 = flow.tensor.transfer %31459 : tensor<4x?x4096xf16>{%dim_29131} to #hal.device.promise<@__device_1>
%31461 = torch_c.from_builtin_tensor %31460 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31462 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29132 = arith.constant 1 : index
%dim_29133 = tensor.dim %31462, %c1_29132 : tensor<4x?x4096xf16>
%31463 = flow.tensor.transfer %31462 : tensor<4x?x4096xf16>{%dim_29133} to #hal.device.promise<@__device_1>
%31464 = torch_c.from_builtin_tensor %31463 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31465 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29134 = arith.constant 1 : index
%dim_29135 = tensor.dim %31465, %c1_29134 : tensor<4x?x4096xf16>
%31466 = flow.tensor.transfer %31465 : tensor<4x?x4096xf16>{%dim_29135} to #hal.device.promise<@__device_1>
%31467 = torch_c.from_builtin_tensor %31466 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31468 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29136 = arith.constant 1 : index
%dim_29137 = tensor.dim %31468, %c1_29136 : tensor<4x?x4096xf16>
%31469 = flow.tensor.transfer %31468 : tensor<4x?x4096xf16>{%dim_29137} to #hal.device.promise<@__device_1>
%31470 = torch_c.from_builtin_tensor %31469 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31471 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29138 = arith.constant 1 : index
%dim_29139 = tensor.dim %31471, %c1_29138 : tensor<4x?x4096xf16>
%31472 = flow.tensor.transfer %31471 : tensor<4x?x4096xf16>{%dim_29139} to #hal.device.promise<@__device_1>
%31473 = torch_c.from_builtin_tensor %31472 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31474 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29140 = arith.constant 1 : index
%dim_29141 = tensor.dim %31474, %c1_29140 : tensor<4x?x4096xf16>
%31475 = flow.tensor.transfer %31474 : tensor<4x?x4096xf16>{%dim_29141} to #hal.device.promise<@__device_1>
%31476 = torch_c.from_builtin_tensor %31475 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29142 = torch.constant.int 1
%31477 = torch.aten.add.Tensor %31458, %31391, %int1_29142 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29143 = torch.constant.int 1
%31478 = torch.aten.add.Tensor %31477, %31461, %int1_29143 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29144 = torch.constant.int 1
%31479 = torch.aten.add.Tensor %31478, %31464, %int1_29144 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29145 = torch.constant.int 1
%31480 = torch.aten.add.Tensor %31479, %31467, %int1_29145 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29146 = torch.constant.int 1
%31481 = torch.aten.add.Tensor %31480, %31470, %int1_29146 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29147 = torch.constant.int 1
%31482 = torch.aten.add.Tensor %31481, %31473, %int1_29147 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29148 = torch.constant.int 1
%31483 = torch.aten.add.Tensor %31482, %31476, %int1_29148 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31484 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29149 = arith.constant 1 : index
%dim_29150 = tensor.dim %31484, %c1_29149 : tensor<4x?x4096xf16>
%31485 = flow.tensor.transfer %31484 : tensor<4x?x4096xf16>{%dim_29150} to #hal.device.promise<@__device_2>
%31486 = torch_c.from_builtin_tensor %31485 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31486, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31487 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29151 = arith.constant 1 : index
%dim_29152 = tensor.dim %31487, %c1_29151 : tensor<4x?x4096xf16>
%31488 = flow.tensor.transfer %31487 : tensor<4x?x4096xf16>{%dim_29152} to #hal.device.promise<@__device_2>
%31489 = torch_c.from_builtin_tensor %31488 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31490 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29153 = arith.constant 1 : index
%dim_29154 = tensor.dim %31490, %c1_29153 : tensor<4x?x4096xf16>
%31491 = flow.tensor.transfer %31490 : tensor<4x?x4096xf16>{%dim_29154} to #hal.device.promise<@__device_2>
%31492 = torch_c.from_builtin_tensor %31491 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31492, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31493 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29155 = arith.constant 1 : index
%dim_29156 = tensor.dim %31493, %c1_29155 : tensor<4x?x4096xf16>
%31494 = flow.tensor.transfer %31493 : tensor<4x?x4096xf16>{%dim_29156} to #hal.device.promise<@__device_2>
%31495 = torch_c.from_builtin_tensor %31494 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31496 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29157 = arith.constant 1 : index
%dim_29158 = tensor.dim %31496, %c1_29157 : tensor<4x?x4096xf16>
%31497 = flow.tensor.transfer %31496 : tensor<4x?x4096xf16>{%dim_29158} to #hal.device.promise<@__device_2>
%31498 = torch_c.from_builtin_tensor %31497 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31499 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29159 = arith.constant 1 : index
%dim_29160 = tensor.dim %31499, %c1_29159 : tensor<4x?x4096xf16>
%31500 = flow.tensor.transfer %31499 : tensor<4x?x4096xf16>{%dim_29160} to #hal.device.promise<@__device_2>
%31501 = torch_c.from_builtin_tensor %31500 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31502 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29161 = arith.constant 1 : index
%dim_29162 = tensor.dim %31502, %c1_29161 : tensor<4x?x4096xf16>
%31503 = flow.tensor.transfer %31502 : tensor<4x?x4096xf16>{%dim_29162} to #hal.device.promise<@__device_2>
%31504 = torch_c.from_builtin_tensor %31503 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31504, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29163 = torch.constant.int 1
%31505 = torch.aten.add.Tensor %31486, %31489, %int1_29163 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29164 = torch.constant.int 1
%31506 = torch.aten.add.Tensor %31505, %31397, %int1_29164 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29165 = torch.constant.int 1
%31507 = torch.aten.add.Tensor %31506, %31492, %int1_29165 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29166 = torch.constant.int 1
%31508 = torch.aten.add.Tensor %31507, %31495, %int1_29166 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29167 = torch.constant.int 1
%31509 = torch.aten.add.Tensor %31508, %31498, %int1_29167 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29168 = torch.constant.int 1
%31510 = torch.aten.add.Tensor %31509, %31501, %int1_29168 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29169 = torch.constant.int 1
%31511 = torch.aten.add.Tensor %31510, %31504, %int1_29169 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31512 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29170 = arith.constant 1 : index
%dim_29171 = tensor.dim %31512, %c1_29170 : tensor<4x?x4096xf16>
%31513 = flow.tensor.transfer %31512 : tensor<4x?x4096xf16>{%dim_29171} to #hal.device.promise<@__device_3>
%31514 = torch_c.from_builtin_tensor %31513 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31515 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29172 = arith.constant 1 : index
%dim_29173 = tensor.dim %31515, %c1_29172 : tensor<4x?x4096xf16>
%31516 = flow.tensor.transfer %31515 : tensor<4x?x4096xf16>{%dim_29173} to #hal.device.promise<@__device_3>
%31517 = torch_c.from_builtin_tensor %31516 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31518 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29174 = arith.constant 1 : index
%dim_29175 = tensor.dim %31518, %c1_29174 : tensor<4x?x4096xf16>
%31519 = flow.tensor.transfer %31518 : tensor<4x?x4096xf16>{%dim_29175} to #hal.device.promise<@__device_3>
%31520 = torch_c.from_builtin_tensor %31519 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31521 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29176 = arith.constant 1 : index
%dim_29177 = tensor.dim %31521, %c1_29176 : tensor<4x?x4096xf16>
%31522 = flow.tensor.transfer %31521 : tensor<4x?x4096xf16>{%dim_29177} to #hal.device.promise<@__device_3>
%31523 = torch_c.from_builtin_tensor %31522 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31524 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29178 = arith.constant 1 : index
%dim_29179 = tensor.dim %31524, %c1_29178 : tensor<4x?x4096xf16>
%31525 = flow.tensor.transfer %31524 : tensor<4x?x4096xf16>{%dim_29179} to #hal.device.promise<@__device_3>
%31526 = torch_c.from_builtin_tensor %31525 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31527 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29180 = arith.constant 1 : index
%dim_29181 = tensor.dim %31527, %c1_29180 : tensor<4x?x4096xf16>
%31528 = flow.tensor.transfer %31527 : tensor<4x?x4096xf16>{%dim_29181} to #hal.device.promise<@__device_3>
%31529 = torch_c.from_builtin_tensor %31528 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31530 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29182 = arith.constant 1 : index
%dim_29183 = tensor.dim %31530, %c1_29182 : tensor<4x?x4096xf16>
%31531 = flow.tensor.transfer %31530 : tensor<4x?x4096xf16>{%dim_29183} to #hal.device.promise<@__device_3>
%31532 = torch_c.from_builtin_tensor %31531 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29184 = torch.constant.int 1
%31533 = torch.aten.add.Tensor %31514, %31517, %int1_29184 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29185 = torch.constant.int 1
%31534 = torch.aten.add.Tensor %31533, %31520, %int1_29185 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29186 = torch.constant.int 1
%31535 = torch.aten.add.Tensor %31534, %31403, %int1_29186 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29187 = torch.constant.int 1
%31536 = torch.aten.add.Tensor %31535, %31523, %int1_29187 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29188 = torch.constant.int 1
%31537 = torch.aten.add.Tensor %31536, %31526, %int1_29188 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29189 = torch.constant.int 1
%31538 = torch.aten.add.Tensor %31537, %31529, %int1_29189 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29190 = torch.constant.int 1
%31539 = torch.aten.add.Tensor %31538, %31532, %int1_29190 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31540 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29191 = arith.constant 1 : index
%dim_29192 = tensor.dim %31540, %c1_29191 : tensor<4x?x4096xf16>
%31541 = flow.tensor.transfer %31540 : tensor<4x?x4096xf16>{%dim_29192} to #hal.device.promise<@__device_4>
%31542 = torch_c.from_builtin_tensor %31541 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31543 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29193 = arith.constant 1 : index
%dim_29194 = tensor.dim %31543, %c1_29193 : tensor<4x?x4096xf16>
%31544 = flow.tensor.transfer %31543 : tensor<4x?x4096xf16>{%dim_29194} to #hal.device.promise<@__device_4>
%31545 = torch_c.from_builtin_tensor %31544 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31546 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29195 = arith.constant 1 : index
%dim_29196 = tensor.dim %31546, %c1_29195 : tensor<4x?x4096xf16>
%31547 = flow.tensor.transfer %31546 : tensor<4x?x4096xf16>{%dim_29196} to #hal.device.promise<@__device_4>
%31548 = torch_c.from_builtin_tensor %31547 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31549 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29197 = arith.constant 1 : index
%dim_29198 = tensor.dim %31549, %c1_29197 : tensor<4x?x4096xf16>
%31550 = flow.tensor.transfer %31549 : tensor<4x?x4096xf16>{%dim_29198} to #hal.device.promise<@__device_4>
%31551 = torch_c.from_builtin_tensor %31550 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31552 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29199 = arith.constant 1 : index
%dim_29200 = tensor.dim %31552, %c1_29199 : tensor<4x?x4096xf16>
%31553 = flow.tensor.transfer %31552 : tensor<4x?x4096xf16>{%dim_29200} to #hal.device.promise<@__device_4>
%31554 = torch_c.from_builtin_tensor %31553 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31555 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29201 = arith.constant 1 : index
%dim_29202 = tensor.dim %31555, %c1_29201 : tensor<4x?x4096xf16>
%31556 = flow.tensor.transfer %31555 : tensor<4x?x4096xf16>{%dim_29202} to #hal.device.promise<@__device_4>
%31557 = torch_c.from_builtin_tensor %31556 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31558 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29203 = arith.constant 1 : index
%dim_29204 = tensor.dim %31558, %c1_29203 : tensor<4x?x4096xf16>
%31559 = flow.tensor.transfer %31558 : tensor<4x?x4096xf16>{%dim_29204} to #hal.device.promise<@__device_4>
%31560 = torch_c.from_builtin_tensor %31559 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29205 = torch.constant.int 1
%31561 = torch.aten.add.Tensor %31542, %31545, %int1_29205 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29206 = torch.constant.int 1
%31562 = torch.aten.add.Tensor %31561, %31548, %int1_29206 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29207 = torch.constant.int 1
%31563 = torch.aten.add.Tensor %31562, %31551, %int1_29207 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29208 = torch.constant.int 1
%31564 = torch.aten.add.Tensor %31563, %31409, %int1_29208 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29209 = torch.constant.int 1
%31565 = torch.aten.add.Tensor %31564, %31554, %int1_29209 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29210 = torch.constant.int 1
%31566 = torch.aten.add.Tensor %31565, %31557, %int1_29210 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29211 = torch.constant.int 1
%31567 = torch.aten.add.Tensor %31566, %31560, %int1_29211 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31568 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29212 = arith.constant 1 : index
%dim_29213 = tensor.dim %31568, %c1_29212 : tensor<4x?x4096xf16>
%31569 = flow.tensor.transfer %31568 : tensor<4x?x4096xf16>{%dim_29213} to #hal.device.promise<@__device_5>
%31570 = torch_c.from_builtin_tensor %31569 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31571 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29214 = arith.constant 1 : index
%dim_29215 = tensor.dim %31571, %c1_29214 : tensor<4x?x4096xf16>
%31572 = flow.tensor.transfer %31571 : tensor<4x?x4096xf16>{%dim_29215} to #hal.device.promise<@__device_5>
%31573 = torch_c.from_builtin_tensor %31572 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31574 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29216 = arith.constant 1 : index
%dim_29217 = tensor.dim %31574, %c1_29216 : tensor<4x?x4096xf16>
%31575 = flow.tensor.transfer %31574 : tensor<4x?x4096xf16>{%dim_29217} to #hal.device.promise<@__device_5>
%31576 = torch_c.from_builtin_tensor %31575 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31577 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29218 = arith.constant 1 : index
%dim_29219 = tensor.dim %31577, %c1_29218 : tensor<4x?x4096xf16>
%31578 = flow.tensor.transfer %31577 : tensor<4x?x4096xf16>{%dim_29219} to #hal.device.promise<@__device_5>
%31579 = torch_c.from_builtin_tensor %31578 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31580 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29220 = arith.constant 1 : index
%dim_29221 = tensor.dim %31580, %c1_29220 : tensor<4x?x4096xf16>
%31581 = flow.tensor.transfer %31580 : tensor<4x?x4096xf16>{%dim_29221} to #hal.device.promise<@__device_5>
%31582 = torch_c.from_builtin_tensor %31581 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31583 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29222 = arith.constant 1 : index
%dim_29223 = tensor.dim %31583, %c1_29222 : tensor<4x?x4096xf16>
%31584 = flow.tensor.transfer %31583 : tensor<4x?x4096xf16>{%dim_29223} to #hal.device.promise<@__device_5>
%31585 = torch_c.from_builtin_tensor %31584 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31586 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29224 = arith.constant 1 : index
%dim_29225 = tensor.dim %31586, %c1_29224 : tensor<4x?x4096xf16>
%31587 = flow.tensor.transfer %31586 : tensor<4x?x4096xf16>{%dim_29225} to #hal.device.promise<@__device_5>
%31588 = torch_c.from_builtin_tensor %31587 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29226 = torch.constant.int 1
%31589 = torch.aten.add.Tensor %31570, %31573, %int1_29226 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29227 = torch.constant.int 1
%31590 = torch.aten.add.Tensor %31589, %31576, %int1_29227 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29228 = torch.constant.int 1
%31591 = torch.aten.add.Tensor %31590, %31579, %int1_29228 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29229 = torch.constant.int 1
%31592 = torch.aten.add.Tensor %31591, %31582, %int1_29229 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29230 = torch.constant.int 1
%31593 = torch.aten.add.Tensor %31592, %31415, %int1_29230 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29231 = torch.constant.int 1
%31594 = torch.aten.add.Tensor %31593, %31585, %int1_29231 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29232 = torch.constant.int 1
%31595 = torch.aten.add.Tensor %31594, %31588, %int1_29232 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31596 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29233 = arith.constant 1 : index
%dim_29234 = tensor.dim %31596, %c1_29233 : tensor<4x?x4096xf16>
%31597 = flow.tensor.transfer %31596 : tensor<4x?x4096xf16>{%dim_29234} to #hal.device.promise<@__device_6>
%31598 = torch_c.from_builtin_tensor %31597 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31599 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29235 = arith.constant 1 : index
%dim_29236 = tensor.dim %31599, %c1_29235 : tensor<4x?x4096xf16>
%31600 = flow.tensor.transfer %31599 : tensor<4x?x4096xf16>{%dim_29236} to #hal.device.promise<@__device_6>
%31601 = torch_c.from_builtin_tensor %31600 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31602 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29237 = arith.constant 1 : index
%dim_29238 = tensor.dim %31602, %c1_29237 : tensor<4x?x4096xf16>
%31603 = flow.tensor.transfer %31602 : tensor<4x?x4096xf16>{%dim_29238} to #hal.device.promise<@__device_6>
%31604 = torch_c.from_builtin_tensor %31603 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31605 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29239 = arith.constant 1 : index
%dim_29240 = tensor.dim %31605, %c1_29239 : tensor<4x?x4096xf16>
%31606 = flow.tensor.transfer %31605 : tensor<4x?x4096xf16>{%dim_29240} to #hal.device.promise<@__device_6>
%31607 = torch_c.from_builtin_tensor %31606 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31608 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29241 = arith.constant 1 : index
%dim_29242 = tensor.dim %31608, %c1_29241 : tensor<4x?x4096xf16>
%31609 = flow.tensor.transfer %31608 : tensor<4x?x4096xf16>{%dim_29242} to #hal.device.promise<@__device_6>
%31610 = torch_c.from_builtin_tensor %31609 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31611 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29243 = arith.constant 1 : index
%dim_29244 = tensor.dim %31611, %c1_29243 : tensor<4x?x4096xf16>
%31612 = flow.tensor.transfer %31611 : tensor<4x?x4096xf16>{%dim_29244} to #hal.device.promise<@__device_6>
%31613 = torch_c.from_builtin_tensor %31612 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31614 = torch_c.to_builtin_tensor %31427 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29245 = arith.constant 1 : index
%dim_29246 = tensor.dim %31614, %c1_29245 : tensor<4x?x4096xf16>
%31615 = flow.tensor.transfer %31614 : tensor<4x?x4096xf16>{%dim_29246} to #hal.device.promise<@__device_6>
%31616 = torch_c.from_builtin_tensor %31615 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29247 = torch.constant.int 1
%31617 = torch.aten.add.Tensor %31598, %31601, %int1_29247 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29248 = torch.constant.int 1
%31618 = torch.aten.add.Tensor %31617, %31604, %int1_29248 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29249 = torch.constant.int 1
%31619 = torch.aten.add.Tensor %31618, %31607, %int1_29249 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29250 = torch.constant.int 1
%31620 = torch.aten.add.Tensor %31619, %31610, %int1_29250 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29251 = torch.constant.int 1
%31621 = torch.aten.add.Tensor %31620, %31613, %int1_29251 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29252 = torch.constant.int 1
%31622 = torch.aten.add.Tensor %31621, %31421, %int1_29252 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29253 = torch.constant.int 1
%31623 = torch.aten.add.Tensor %31622, %31616, %int1_29253 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31624 = torch_c.to_builtin_tensor %31385 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29254 = arith.constant 1 : index
%dim_29255 = tensor.dim %31624, %c1_29254 : tensor<4x?x4096xf16>
%31625 = flow.tensor.transfer %31624 : tensor<4x?x4096xf16>{%dim_29255} to #hal.device.promise<@__device_7>
%31626 = torch_c.from_builtin_tensor %31625 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31627 = torch_c.to_builtin_tensor %31391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29256 = arith.constant 1 : index
%dim_29257 = tensor.dim %31627, %c1_29256 : tensor<4x?x4096xf16>
%31628 = flow.tensor.transfer %31627 : tensor<4x?x4096xf16>{%dim_29257} to #hal.device.promise<@__device_7>
%31629 = torch_c.from_builtin_tensor %31628 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31630 = torch_c.to_builtin_tensor %31397 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29258 = arith.constant 1 : index
%dim_29259 = tensor.dim %31630, %c1_29258 : tensor<4x?x4096xf16>
%31631 = flow.tensor.transfer %31630 : tensor<4x?x4096xf16>{%dim_29259} to #hal.device.promise<@__device_7>
%31632 = torch_c.from_builtin_tensor %31631 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31633 = torch_c.to_builtin_tensor %31403 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29260 = arith.constant 1 : index
%dim_29261 = tensor.dim %31633, %c1_29260 : tensor<4x?x4096xf16>
%31634 = flow.tensor.transfer %31633 : tensor<4x?x4096xf16>{%dim_29261} to #hal.device.promise<@__device_7>
%31635 = torch_c.from_builtin_tensor %31634 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31636 = torch_c.to_builtin_tensor %31409 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29262 = arith.constant 1 : index
%dim_29263 = tensor.dim %31636, %c1_29262 : tensor<4x?x4096xf16>
%31637 = flow.tensor.transfer %31636 : tensor<4x?x4096xf16>{%dim_29263} to #hal.device.promise<@__device_7>
%31638 = torch_c.from_builtin_tensor %31637 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31639 = torch_c.to_builtin_tensor %31415 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29264 = arith.constant 1 : index
%dim_29265 = tensor.dim %31639, %c1_29264 : tensor<4x?x4096xf16>
%31640 = flow.tensor.transfer %31639 : tensor<4x?x4096xf16>{%dim_29265} to #hal.device.promise<@__device_7>
%31641 = torch_c.from_builtin_tensor %31640 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31642 = torch_c.to_builtin_tensor %31421 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29266 = arith.constant 1 : index
%dim_29267 = tensor.dim %31642, %c1_29266 : tensor<4x?x4096xf16>
%31643 = flow.tensor.transfer %31642 : tensor<4x?x4096xf16>{%dim_29267} to #hal.device.promise<@__device_7>
%31644 = torch_c.from_builtin_tensor %31643 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29268 = torch.constant.int 1
%31645 = torch.aten.add.Tensor %31626, %31629, %int1_29268 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29269 = torch.constant.int 1
%31646 = torch.aten.add.Tensor %31645, %31632, %int1_29269 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31646, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29270 = torch.constant.int 1
%31647 = torch.aten.add.Tensor %31646, %31635, %int1_29270 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29271 = torch.constant.int 1
%31648 = torch.aten.add.Tensor %31647, %31638, %int1_29271 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29272 = torch.constant.int 1
%31649 = torch.aten.add.Tensor %31648, %31641, %int1_29272 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29273 = torch.constant.int 1
%31650 = torch.aten.add.Tensor %31649, %31644, %int1_29273 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29274 = torch.constant.int 1
%31651 = torch.aten.add.Tensor %31650, %31427, %int1_29274 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29275 = torch.constant.int 1
%31652 = torch.aten.add.Tensor %30311, %31455, %int1_29275 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29276 = torch.constant.int 1
%31653 = torch.aten.add.Tensor %30312, %31483, %int1_29276 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29277 = torch.constant.int 1
%31654 = torch.aten.add.Tensor %30313, %31511, %int1_29277 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29278 = torch.constant.int 1
%31655 = torch.aten.add.Tensor %30314, %31539, %int1_29278 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29279 = torch.constant.int 1
%31656 = torch.aten.add.Tensor %30315, %31567, %int1_29279 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29280 = torch.constant.int 1
%31657 = torch.aten.add.Tensor %30316, %31595, %int1_29280 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29281 = torch.constant.int 1
%31658 = torch.aten.add.Tensor %30317, %31623, %int1_29281 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29282 = torch.constant.int 1
%31659 = torch.aten.add.Tensor %30318, %31651, %int1_29282 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_29283 = torch.constant.int 6
%31660 = torch.prims.convert_element_type %31652, %int6_29283 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29284 = torch.constant.int 6
%31661 = torch.prims.convert_element_type %31653, %int6_29284 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29285 = torch.constant.int 6
%31662 = torch.prims.convert_element_type %31654, %int6_29285 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29286 = torch.constant.int 6
%31663 = torch.prims.convert_element_type %31655, %int6_29286 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29287 = torch.constant.int 6
%31664 = torch.prims.convert_element_type %31656, %int6_29287 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29288 = torch.constant.int 6
%31665 = torch.prims.convert_element_type %31657, %int6_29288 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29289 = torch.constant.int 6
%31666 = torch.prims.convert_element_type %31658, %int6_29289 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29290 = torch.constant.int 6
%31667 = torch.prims.convert_element_type %31659, %int6_29290 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29291 = torch.constant.int 2
%31668 = torch.aten.pow.Tensor_Scalar %31660, %int2_29291 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29292 = torch.constant.int 2
%31669 = torch.aten.pow.Tensor_Scalar %31661, %int2_29292 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29293 = torch.constant.int 2
%31670 = torch.aten.pow.Tensor_Scalar %31662, %int2_29293 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29294 = torch.constant.int 2
%31671 = torch.aten.pow.Tensor_Scalar %31663, %int2_29294 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29295 = torch.constant.int 2
%31672 = torch.aten.pow.Tensor_Scalar %31664, %int2_29295 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29296 = torch.constant.int 2
%31673 = torch.aten.pow.Tensor_Scalar %31665, %int2_29296 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29297 = torch.constant.int 2
%31674 = torch.aten.pow.Tensor_Scalar %31666, %int2_29297 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29298 = torch.constant.int 2
%31675 = torch.aten.pow.Tensor_Scalar %31667, %int2_29298 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_29299 = torch.constant.int -1
%31676 = torch.prim.ListConstruct %int-1_29299 : (!torch.int) -> !torch.list<int>
%true_29300 = torch.constant.bool true
%none_29301 = torch.constant.none
%31677 = torch.aten.mean.dim %31668, %31676, %true_29300, %none_29301 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29302 = torch.constant.int -1
%31678 = torch.prim.ListConstruct %int-1_29302 : (!torch.int) -> !torch.list<int>
%true_29303 = torch.constant.bool true
%none_29304 = torch.constant.none
%31679 = torch.aten.mean.dim %31669, %31678, %true_29303, %none_29304 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29305 = torch.constant.int -1
%31680 = torch.prim.ListConstruct %int-1_29305 : (!torch.int) -> !torch.list<int>
%true_29306 = torch.constant.bool true
%none_29307 = torch.constant.none
%31681 = torch.aten.mean.dim %31670, %31680, %true_29306, %none_29307 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29308 = torch.constant.int -1
%31682 = torch.prim.ListConstruct %int-1_29308 : (!torch.int) -> !torch.list<int>
%true_29309 = torch.constant.bool true
%none_29310 = torch.constant.none
%31683 = torch.aten.mean.dim %31671, %31682, %true_29309, %none_29310 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29311 = torch.constant.int -1
%31684 = torch.prim.ListConstruct %int-1_29311 : (!torch.int) -> !torch.list<int>
%true_29312 = torch.constant.bool true
%none_29313 = torch.constant.none
%31685 = torch.aten.mean.dim %31672, %31684, %true_29312, %none_29313 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29314 = torch.constant.int -1
%31686 = torch.prim.ListConstruct %int-1_29314 : (!torch.int) -> !torch.list<int>
%true_29315 = torch.constant.bool true
%none_29316 = torch.constant.none
%31687 = torch.aten.mean.dim %31673, %31686, %true_29315, %none_29316 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29317 = torch.constant.int -1
%31688 = torch.prim.ListConstruct %int-1_29317 : (!torch.int) -> !torch.list<int>
%true_29318 = torch.constant.bool true
%none_29319 = torch.constant.none
%31689 = torch.aten.mean.dim %31674, %31688, %true_29318, %none_29319 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29320 = torch.constant.int -1
%31690 = torch.prim.ListConstruct %int-1_29320 : (!torch.int) -> !torch.list<int>
%true_29321 = torch.constant.bool true
%none_29322 = torch.constant.none
%31691 = torch.aten.mean.dim %31675, %31690, %true_29321, %none_29322 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29323 = torch.constant.float 9.9999997473787516E-6
%int1_29324 = torch.constant.int 1
%31692 = torch.aten.add.Scalar %31677, %float9.999990e-06_29323, %int1_29324 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29325 = torch.constant.float 9.9999997473787516E-6
%int1_29326 = torch.constant.int 1
%31693 = torch.aten.add.Scalar %31679, %float9.999990e-06_29325, %int1_29326 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29327 = torch.constant.float 9.9999997473787516E-6
%int1_29328 = torch.constant.int 1
%31694 = torch.aten.add.Scalar %31681, %float9.999990e-06_29327, %int1_29328 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29329 = torch.constant.float 9.9999997473787516E-6
%int1_29330 = torch.constant.int 1
%31695 = torch.aten.add.Scalar %31683, %float9.999990e-06_29329, %int1_29330 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29331 = torch.constant.float 9.9999997473787516E-6
%int1_29332 = torch.constant.int 1
%31696 = torch.aten.add.Scalar %31685, %float9.999990e-06_29331, %int1_29332 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29333 = torch.constant.float 9.9999997473787516E-6
%int1_29334 = torch.constant.int 1
%31697 = torch.aten.add.Scalar %31687, %float9.999990e-06_29333, %int1_29334 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29335 = torch.constant.float 9.9999997473787516E-6
%int1_29336 = torch.constant.int 1
%31698 = torch.aten.add.Scalar %31689, %float9.999990e-06_29335, %int1_29336 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29337 = torch.constant.float 9.9999997473787516E-6
%int1_29338 = torch.constant.int 1
%31699 = torch.aten.add.Scalar %31691, %float9.999990e-06_29337, %int1_29338 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31699, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31700 = torch.aten.rsqrt %31692 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31701 = torch.aten.rsqrt %31693 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31702 = torch.aten.rsqrt %31694 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31703 = torch.aten.rsqrt %31695 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31704 = torch.aten.rsqrt %31696 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31705 = torch.aten.rsqrt %31697 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31705, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31706 = torch.aten.rsqrt %31698 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31707 = torch.aten.rsqrt %31699 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %31707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%31708 = torch.aten.mul.Tensor %31660, %31700 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31709 = torch.aten.mul.Tensor %31661, %31701 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31710 = torch.aten.mul.Tensor %31662, %31702 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31711 = torch.aten.mul.Tensor %31663, %31703 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31712 = torch.aten.mul.Tensor %31664, %31704 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31713 = torch.aten.mul.Tensor %31665, %31705 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31714 = torch.aten.mul.Tensor %31666, %31706 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31715 = torch.aten.mul.Tensor %31667, %31707 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31716 = torch.aten.mul.Tensor %1128, %31708 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31717 = torch.aten.mul.Tensor %1129, %31709 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31718 = torch.aten.mul.Tensor %1130, %31710 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31719 = torch.aten.mul.Tensor %1131, %31711 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31720 = torch.aten.mul.Tensor %1132, %31712 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31721 = torch.aten.mul.Tensor %1133, %31713 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31722 = torch.aten.mul.Tensor %1134, %31714 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%31723 = torch.aten.mul.Tensor %1135, %31715 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %31723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_29339 = torch.constant.int 5
%31724 = torch.prims.convert_element_type %31716, %int5_29339 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29340 = torch.constant.int 5
%31725 = torch.prims.convert_element_type %31717, %int5_29340 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29341 = torch.constant.int 5
%31726 = torch.prims.convert_element_type %31718, %int5_29341 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29342 = torch.constant.int 5
%31727 = torch.prims.convert_element_type %31719, %int5_29342 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29343 = torch.constant.int 5
%31728 = torch.prims.convert_element_type %31720, %int5_29343 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29344 = torch.constant.int 5
%31729 = torch.prims.convert_element_type %31721, %int5_29344 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29345 = torch.constant.int 5
%31730 = torch.prims.convert_element_type %31722, %int5_29345 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29346 = torch.constant.int 5
%31731 = torch.prims.convert_element_type %31723, %int5_29346 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29347 = torch.constant.int 1
%int0_29348 = torch.constant.int 0
%31732 = torch.prim.ListConstruct %int1_29347, %int0_29348 : (!torch.int, !torch.int) -> !torch.list<int>
%31733 = torch.aten.permute %1136, %31732 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29349 = torch.constant.int 1
%int0_29350 = torch.constant.int 0
%31734 = torch.prim.ListConstruct %int1_29349, %int0_29350 : (!torch.int, !torch.int) -> !torch.list<int>
%31735 = torch.aten.permute %1137, %31734 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29351 = torch.constant.int 1
%int0_29352 = torch.constant.int 0
%31736 = torch.prim.ListConstruct %int1_29351, %int0_29352 : (!torch.int, !torch.int) -> !torch.list<int>
%31737 = torch.aten.permute %1138, %31736 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29353 = torch.constant.int 1
%int0_29354 = torch.constant.int 0
%31738 = torch.prim.ListConstruct %int1_29353, %int0_29354 : (!torch.int, !torch.int) -> !torch.list<int>
%31739 = torch.aten.permute %1139, %31738 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29355 = torch.constant.int 1
%int0_29356 = torch.constant.int 0
%31740 = torch.prim.ListConstruct %int1_29355, %int0_29356 : (!torch.int, !torch.int) -> !torch.list<int>
%31741 = torch.aten.permute %1140, %31740 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29357 = torch.constant.int 1
%int0_29358 = torch.constant.int 0
%31742 = torch.prim.ListConstruct %int1_29357, %int0_29358 : (!torch.int, !torch.int) -> !torch.list<int>
%31743 = torch.aten.permute %1141, %31742 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29359 = torch.constant.int 1
%int0_29360 = torch.constant.int 0
%31744 = torch.prim.ListConstruct %int1_29359, %int0_29360 : (!torch.int, !torch.int) -> !torch.list<int>
%31745 = torch.aten.permute %1142, %31744 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29361 = torch.constant.int 1
%int0_29362 = torch.constant.int 0
%31746 = torch.prim.ListConstruct %int1_29361, %int0_29362 : (!torch.int, !torch.int) -> !torch.list<int>
%31747 = torch.aten.permute %1143, %31746 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_29363 = torch.constant.int 4
%31748 = torch.aten.mul.int %int4_29363, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29364 = torch.constant.int 4096
%31749 = torch.prim.ListConstruct %31748, %int4096_29364 : (!torch.int, !torch.int) -> !torch.list<int>
%31750 = torch.aten.view %31724, %31749 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31750, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31751 = torch.aten.mm %31750, %31733 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31751, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29365 = torch.constant.int 4
%int1792_29366 = torch.constant.int 1792
%31752 = torch.prim.ListConstruct %int4_29365, %2482, %int1792_29366 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31753 = torch.aten.view %31751, %31752 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29367 = torch.constant.int 4
%31754 = torch.aten.mul.int %int4_29367, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29368 = torch.constant.int 4096
%31755 = torch.prim.ListConstruct %31754, %int4096_29368 : (!torch.int, !torch.int) -> !torch.list<int>
%31756 = torch.aten.view %31725, %31755 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31756, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31757 = torch.aten.mm %31756, %31735 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31757, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29369 = torch.constant.int 4
%int1792_29370 = torch.constant.int 1792
%31758 = torch.prim.ListConstruct %int4_29369, %2482, %int1792_29370 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31759 = torch.aten.view %31757, %31758 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29371 = torch.constant.int 4
%31760 = torch.aten.mul.int %int4_29371, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29372 = torch.constant.int 4096
%31761 = torch.prim.ListConstruct %31760, %int4096_29372 : (!torch.int, !torch.int) -> !torch.list<int>
%31762 = torch.aten.view %31726, %31761 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31762, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31763 = torch.aten.mm %31762, %31737 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31763, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29373 = torch.constant.int 4
%int1792_29374 = torch.constant.int 1792
%31764 = torch.prim.ListConstruct %int4_29373, %2482, %int1792_29374 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31765 = torch.aten.view %31763, %31764 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29375 = torch.constant.int 4
%31766 = torch.aten.mul.int %int4_29375, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29376 = torch.constant.int 4096
%31767 = torch.prim.ListConstruct %31766, %int4096_29376 : (!torch.int, !torch.int) -> !torch.list<int>
%31768 = torch.aten.view %31727, %31767 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31768, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31769 = torch.aten.mm %31768, %31739 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31769, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29377 = torch.constant.int 4
%int1792_29378 = torch.constant.int 1792
%31770 = torch.prim.ListConstruct %int4_29377, %2482, %int1792_29378 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31771 = torch.aten.view %31769, %31770 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29379 = torch.constant.int 4
%31772 = torch.aten.mul.int %int4_29379, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29380 = torch.constant.int 4096
%31773 = torch.prim.ListConstruct %31772, %int4096_29380 : (!torch.int, !torch.int) -> !torch.list<int>
%31774 = torch.aten.view %31728, %31773 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31774, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31775 = torch.aten.mm %31774, %31741 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31775, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29381 = torch.constant.int 4
%int1792_29382 = torch.constant.int 1792
%31776 = torch.prim.ListConstruct %int4_29381, %2482, %int1792_29382 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31777 = torch.aten.view %31775, %31776 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29383 = torch.constant.int 4
%31778 = torch.aten.mul.int %int4_29383, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29384 = torch.constant.int 4096
%31779 = torch.prim.ListConstruct %31778, %int4096_29384 : (!torch.int, !torch.int) -> !torch.list<int>
%31780 = torch.aten.view %31729, %31779 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31780, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31781 = torch.aten.mm %31780, %31743 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31781, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29385 = torch.constant.int 4
%int1792_29386 = torch.constant.int 1792
%31782 = torch.prim.ListConstruct %int4_29385, %2482, %int1792_29386 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31783 = torch.aten.view %31781, %31782 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31783, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29387 = torch.constant.int 4
%31784 = torch.aten.mul.int %int4_29387, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29388 = torch.constant.int 4096
%31785 = torch.prim.ListConstruct %31784, %int4096_29388 : (!torch.int, !torch.int) -> !torch.list<int>
%31786 = torch.aten.view %31730, %31785 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31786, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31787 = torch.aten.mm %31786, %31745 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31787, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29389 = torch.constant.int 4
%int1792_29390 = torch.constant.int 1792
%31788 = torch.prim.ListConstruct %int4_29389, %2482, %int1792_29390 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31789 = torch.aten.view %31787, %31788 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29391 = torch.constant.int 4
%31790 = torch.aten.mul.int %int4_29391, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29392 = torch.constant.int 4096
%31791 = torch.prim.ListConstruct %31790, %int4096_29392 : (!torch.int, !torch.int) -> !torch.list<int>
%31792 = torch.aten.view %31731, %31791 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31792, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31793 = torch.aten.mm %31792, %31747 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31793, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29393 = torch.constant.int 4
%int1792_29394 = torch.constant.int 1792
%31794 = torch.prim.ListConstruct %int4_29393, %2482, %int1792_29394 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31795 = torch.aten.view %31793, %31794 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31796 = torch.aten.silu %31753 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31797 = torch.aten.silu %31759 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31798 = torch.aten.silu %31765 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31799 = torch.aten.silu %31771 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31800 = torch.aten.silu %31777 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31801 = torch.aten.silu %31783 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31802 = torch.aten.silu %31789 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31803 = torch.aten.silu %31795 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_29395 = torch.constant.int 1
%int0_29396 = torch.constant.int 0
%31804 = torch.prim.ListConstruct %int1_29395, %int0_29396 : (!torch.int, !torch.int) -> !torch.list<int>
%31805 = torch.aten.permute %1144, %31804 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29397 = torch.constant.int 1
%int0_29398 = torch.constant.int 0
%31806 = torch.prim.ListConstruct %int1_29397, %int0_29398 : (!torch.int, !torch.int) -> !torch.list<int>
%31807 = torch.aten.permute %1145, %31806 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29399 = torch.constant.int 1
%int0_29400 = torch.constant.int 0
%31808 = torch.prim.ListConstruct %int1_29399, %int0_29400 : (!torch.int, !torch.int) -> !torch.list<int>
%31809 = torch.aten.permute %1146, %31808 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29401 = torch.constant.int 1
%int0_29402 = torch.constant.int 0
%31810 = torch.prim.ListConstruct %int1_29401, %int0_29402 : (!torch.int, !torch.int) -> !torch.list<int>
%31811 = torch.aten.permute %1147, %31810 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29403 = torch.constant.int 1
%int0_29404 = torch.constant.int 0
%31812 = torch.prim.ListConstruct %int1_29403, %int0_29404 : (!torch.int, !torch.int) -> !torch.list<int>
%31813 = torch.aten.permute %1148, %31812 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29405 = torch.constant.int 1
%int0_29406 = torch.constant.int 0
%31814 = torch.prim.ListConstruct %int1_29405, %int0_29406 : (!torch.int, !torch.int) -> !torch.list<int>
%31815 = torch.aten.permute %1149, %31814 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29407 = torch.constant.int 1
%int0_29408 = torch.constant.int 0
%31816 = torch.prim.ListConstruct %int1_29407, %int0_29408 : (!torch.int, !torch.int) -> !torch.list<int>
%31817 = torch.aten.permute %1150, %31816 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_29409 = torch.constant.int 1
%int0_29410 = torch.constant.int 0
%31818 = torch.prim.ListConstruct %int1_29409, %int0_29410 : (!torch.int, !torch.int) -> !torch.list<int>
%31819 = torch.aten.permute %1151, %31818 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_29411 = torch.constant.int 4
%31820 = torch.aten.mul.int %int4_29411, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29412 = torch.constant.int 4096
%31821 = torch.prim.ListConstruct %31820, %int4096_29412 : (!torch.int, !torch.int) -> !torch.list<int>
%31822 = torch.aten.view %31724, %31821 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31822, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31823 = torch.aten.mm %31822, %31805 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31823, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29413 = torch.constant.int 4
%int1792_29414 = torch.constant.int 1792
%31824 = torch.prim.ListConstruct %int4_29413, %2482, %int1792_29414 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31825 = torch.aten.view %31823, %31824 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29415 = torch.constant.int 4
%31826 = torch.aten.mul.int %int4_29415, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29416 = torch.constant.int 4096
%31827 = torch.prim.ListConstruct %31826, %int4096_29416 : (!torch.int, !torch.int) -> !torch.list<int>
%31828 = torch.aten.view %31725, %31827 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31828, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31829 = torch.aten.mm %31828, %31807 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31829, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29417 = torch.constant.int 4
%int1792_29418 = torch.constant.int 1792
%31830 = torch.prim.ListConstruct %int4_29417, %2482, %int1792_29418 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31831 = torch.aten.view %31829, %31830 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29419 = torch.constant.int 4
%31832 = torch.aten.mul.int %int4_29419, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29420 = torch.constant.int 4096
%31833 = torch.prim.ListConstruct %31832, %int4096_29420 : (!torch.int, !torch.int) -> !torch.list<int>
%31834 = torch.aten.view %31726, %31833 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31834, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31835 = torch.aten.mm %31834, %31809 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31835, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29421 = torch.constant.int 4
%int1792_29422 = torch.constant.int 1792
%31836 = torch.prim.ListConstruct %int4_29421, %2482, %int1792_29422 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31837 = torch.aten.view %31835, %31836 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29423 = torch.constant.int 4
%31838 = torch.aten.mul.int %int4_29423, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29424 = torch.constant.int 4096
%31839 = torch.prim.ListConstruct %31838, %int4096_29424 : (!torch.int, !torch.int) -> !torch.list<int>
%31840 = torch.aten.view %31727, %31839 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31840, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31841 = torch.aten.mm %31840, %31811 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31841, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29425 = torch.constant.int 4
%int1792_29426 = torch.constant.int 1792
%31842 = torch.prim.ListConstruct %int4_29425, %2482, %int1792_29426 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31843 = torch.aten.view %31841, %31842 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29427 = torch.constant.int 4
%31844 = torch.aten.mul.int %int4_29427, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29428 = torch.constant.int 4096
%31845 = torch.prim.ListConstruct %31844, %int4096_29428 : (!torch.int, !torch.int) -> !torch.list<int>
%31846 = torch.aten.view %31728, %31845 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31846, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31847 = torch.aten.mm %31846, %31813 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31847, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29429 = torch.constant.int 4
%int1792_29430 = torch.constant.int 1792
%31848 = torch.prim.ListConstruct %int4_29429, %2482, %int1792_29430 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31849 = torch.aten.view %31847, %31848 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29431 = torch.constant.int 4
%31850 = torch.aten.mul.int %int4_29431, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29432 = torch.constant.int 4096
%31851 = torch.prim.ListConstruct %31850, %int4096_29432 : (!torch.int, !torch.int) -> !torch.list<int>
%31852 = torch.aten.view %31729, %31851 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31852, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31853 = torch.aten.mm %31852, %31815 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31853, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29433 = torch.constant.int 4
%int1792_29434 = torch.constant.int 1792
%31854 = torch.prim.ListConstruct %int4_29433, %2482, %int1792_29434 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31855 = torch.aten.view %31853, %31854 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29435 = torch.constant.int 4
%31856 = torch.aten.mul.int %int4_29435, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29436 = torch.constant.int 4096
%31857 = torch.prim.ListConstruct %31856, %int4096_29436 : (!torch.int, !torch.int) -> !torch.list<int>
%31858 = torch.aten.view %31730, %31857 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31858, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31859 = torch.aten.mm %31858, %31817 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31859, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29437 = torch.constant.int 4
%int1792_29438 = torch.constant.int 1792
%31860 = torch.prim.ListConstruct %int4_29437, %2482, %int1792_29438 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31861 = torch.aten.view %31859, %31860 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_29439 = torch.constant.int 4
%31862 = torch.aten.mul.int %int4_29439, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29440 = torch.constant.int 4096
%31863 = torch.prim.ListConstruct %31862, %int4096_29440 : (!torch.int, !torch.int) -> !torch.list<int>
%31864 = torch.aten.view %31731, %31863 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31864, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%31865 = torch.aten.mm %31864, %31819 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31865, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_29441 = torch.constant.int 4
%int1792_29442 = torch.constant.int 1792
%31866 = torch.prim.ListConstruct %int4_29441, %2482, %int1792_29442 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31867 = torch.aten.view %31865, %31866 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31868 = torch.aten.mul.Tensor %31796, %31825 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31869 = torch.aten.mul.Tensor %31797, %31831 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31870 = torch.aten.mul.Tensor %31798, %31837 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31871 = torch.aten.mul.Tensor %31799, %31843 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31872 = torch.aten.mul.Tensor %31800, %31849 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31873 = torch.aten.mul.Tensor %31801, %31855 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31874 = torch.aten.mul.Tensor %31802, %31861 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%31875 = torch.aten.mul.Tensor %31803, %31867 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %31875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_29443 = torch.constant.int 1
%int0_29444 = torch.constant.int 0
%31876 = torch.prim.ListConstruct %int1_29443, %int0_29444 : (!torch.int, !torch.int) -> !torch.list<int>
%31877 = torch.aten.permute %1152, %31876 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29445 = torch.constant.int 1
%int0_29446 = torch.constant.int 0
%31878 = torch.prim.ListConstruct %int1_29445, %int0_29446 : (!torch.int, !torch.int) -> !torch.list<int>
%31879 = torch.aten.permute %1153, %31878 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29447 = torch.constant.int 1
%int0_29448 = torch.constant.int 0
%31880 = torch.prim.ListConstruct %int1_29447, %int0_29448 : (!torch.int, !torch.int) -> !torch.list<int>
%31881 = torch.aten.permute %1154, %31880 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29449 = torch.constant.int 1
%int0_29450 = torch.constant.int 0
%31882 = torch.prim.ListConstruct %int1_29449, %int0_29450 : (!torch.int, !torch.int) -> !torch.list<int>
%31883 = torch.aten.permute %1155, %31882 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29451 = torch.constant.int 1
%int0_29452 = torch.constant.int 0
%31884 = torch.prim.ListConstruct %int1_29451, %int0_29452 : (!torch.int, !torch.int) -> !torch.list<int>
%31885 = torch.aten.permute %1156, %31884 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29453 = torch.constant.int 1
%int0_29454 = torch.constant.int 0
%31886 = torch.prim.ListConstruct %int1_29453, %int0_29454 : (!torch.int, !torch.int) -> !torch.list<int>
%31887 = torch.aten.permute %1157, %31886 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29455 = torch.constant.int 1
%int0_29456 = torch.constant.int 0
%31888 = torch.prim.ListConstruct %int1_29455, %int0_29456 : (!torch.int, !torch.int) -> !torch.list<int>
%31889 = torch.aten.permute %1158, %31888 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29457 = torch.constant.int 1
%int0_29458 = torch.constant.int 0
%31890 = torch.prim.ListConstruct %int1_29457, %int0_29458 : (!torch.int, !torch.int) -> !torch.list<int>
%31891 = torch.aten.permute %1159, %31890 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_29459 = torch.constant.int 1
%31892 = torch.aten.size.int %31753, %int1_29459 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29460 = torch.constant.int 4
%31893 = torch.aten.mul.int %int4_29460, %31892 : !torch.int, !torch.int -> !torch.int
%int1792_29461 = torch.constant.int 1792
%31894 = torch.prim.ListConstruct %31893, %int1792_29461 : (!torch.int, !torch.int) -> !torch.list<int>
%31895 = torch.aten.view %31868, %31894 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31895, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31896 = torch.aten.mm %31895, %31877 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31896, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29462 = torch.constant.int 4
%int4096_29463 = torch.constant.int 4096
%31897 = torch.prim.ListConstruct %int4_29462, %31892, %int4096_29463 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31898 = torch.aten.view %31896, %31897 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29464 = torch.constant.int 1
%31899 = torch.aten.size.int %31759, %int1_29464 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29465 = torch.constant.int 4
%31900 = torch.aten.mul.int %int4_29465, %31899 : !torch.int, !torch.int -> !torch.int
%int1792_29466 = torch.constant.int 1792
%31901 = torch.prim.ListConstruct %31900, %int1792_29466 : (!torch.int, !torch.int) -> !torch.list<int>
%31902 = torch.aten.view %31869, %31901 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31902, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31903 = torch.aten.mm %31902, %31879 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31903, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29467 = torch.constant.int 4
%int4096_29468 = torch.constant.int 4096
%31904 = torch.prim.ListConstruct %int4_29467, %31899, %int4096_29468 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31905 = torch.aten.view %31903, %31904 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29469 = torch.constant.int 1
%31906 = torch.aten.size.int %31765, %int1_29469 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29470 = torch.constant.int 4
%31907 = torch.aten.mul.int %int4_29470, %31906 : !torch.int, !torch.int -> !torch.int
%int1792_29471 = torch.constant.int 1792
%31908 = torch.prim.ListConstruct %31907, %int1792_29471 : (!torch.int, !torch.int) -> !torch.list<int>
%31909 = torch.aten.view %31870, %31908 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31909, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31910 = torch.aten.mm %31909, %31881 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31910, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29472 = torch.constant.int 4
%int4096_29473 = torch.constant.int 4096
%31911 = torch.prim.ListConstruct %int4_29472, %31906, %int4096_29473 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31912 = torch.aten.view %31910, %31911 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29474 = torch.constant.int 1
%31913 = torch.aten.size.int %31771, %int1_29474 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29475 = torch.constant.int 4
%31914 = torch.aten.mul.int %int4_29475, %31913 : !torch.int, !torch.int -> !torch.int
%int1792_29476 = torch.constant.int 1792
%31915 = torch.prim.ListConstruct %31914, %int1792_29476 : (!torch.int, !torch.int) -> !torch.list<int>
%31916 = torch.aten.view %31871, %31915 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31916, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31917 = torch.aten.mm %31916, %31883 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31917, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29477 = torch.constant.int 4
%int4096_29478 = torch.constant.int 4096
%31918 = torch.prim.ListConstruct %int4_29477, %31913, %int4096_29478 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31919 = torch.aten.view %31917, %31918 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29479 = torch.constant.int 1
%31920 = torch.aten.size.int %31777, %int1_29479 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29480 = torch.constant.int 4
%31921 = torch.aten.mul.int %int4_29480, %31920 : !torch.int, !torch.int -> !torch.int
%int1792_29481 = torch.constant.int 1792
%31922 = torch.prim.ListConstruct %31921, %int1792_29481 : (!torch.int, !torch.int) -> !torch.list<int>
%31923 = torch.aten.view %31872, %31922 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31923, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31924 = torch.aten.mm %31923, %31885 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31924, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29482 = torch.constant.int 4
%int4096_29483 = torch.constant.int 4096
%31925 = torch.prim.ListConstruct %int4_29482, %31920, %int4096_29483 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31926 = torch.aten.view %31924, %31925 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29484 = torch.constant.int 1
%31927 = torch.aten.size.int %31783, %int1_29484 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29485 = torch.constant.int 4
%31928 = torch.aten.mul.int %int4_29485, %31927 : !torch.int, !torch.int -> !torch.int
%int1792_29486 = torch.constant.int 1792
%31929 = torch.prim.ListConstruct %31928, %int1792_29486 : (!torch.int, !torch.int) -> !torch.list<int>
%31930 = torch.aten.view %31873, %31929 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31930, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31931 = torch.aten.mm %31930, %31887 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31931, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29487 = torch.constant.int 4
%int4096_29488 = torch.constant.int 4096
%31932 = torch.prim.ListConstruct %int4_29487, %31927, %int4096_29488 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31933 = torch.aten.view %31931, %31932 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29489 = torch.constant.int 1
%31934 = torch.aten.size.int %31789, %int1_29489 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29490 = torch.constant.int 4
%31935 = torch.aten.mul.int %int4_29490, %31934 : !torch.int, !torch.int -> !torch.int
%int1792_29491 = torch.constant.int 1792
%31936 = torch.prim.ListConstruct %31935, %int1792_29491 : (!torch.int, !torch.int) -> !torch.list<int>
%31937 = torch.aten.view %31874, %31936 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31937, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31938 = torch.aten.mm %31937, %31889 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31938, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29492 = torch.constant.int 4
%int4096_29493 = torch.constant.int 4096
%31939 = torch.prim.ListConstruct %int4_29492, %31934, %int4096_29493 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31940 = torch.aten.view %31938, %31939 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29494 = torch.constant.int 1
%31941 = torch.aten.size.int %31795, %int1_29494 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_29495 = torch.constant.int 4
%31942 = torch.aten.mul.int %int4_29495, %31941 : !torch.int, !torch.int -> !torch.int
%int1792_29496 = torch.constant.int 1792
%31943 = torch.prim.ListConstruct %31942, %int1792_29496 : (!torch.int, !torch.int) -> !torch.list<int>
%31944 = torch.aten.view %31875, %31943 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %31944, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%31945 = torch.aten.mm %31944, %31891 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %31945, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_29497 = torch.constant.int 4
%int4096_29498 = torch.constant.int 4096
%31946 = torch.prim.ListConstruct %int4_29497, %31941, %int4096_29498 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%31947 = torch.aten.view %31945, %31946 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31948 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29499 = arith.constant 1 : index
%dim_29500 = tensor.dim %31948, %c1_29499 : tensor<4x?x4096xf16>
%31949 = flow.tensor.transfer %31948 : tensor<4x?x4096xf16>{%dim_29500} to #hal.device.promise<@__device_0>
%31950 = torch_c.from_builtin_tensor %31949 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31951 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29501 = arith.constant 1 : index
%dim_29502 = tensor.dim %31951, %c1_29501 : tensor<4x?x4096xf16>
%31952 = flow.tensor.transfer %31951 : tensor<4x?x4096xf16>{%dim_29502} to #hal.device.promise<@__device_0>
%31953 = torch_c.from_builtin_tensor %31952 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31954 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29503 = arith.constant 1 : index
%dim_29504 = tensor.dim %31954, %c1_29503 : tensor<4x?x4096xf16>
%31955 = flow.tensor.transfer %31954 : tensor<4x?x4096xf16>{%dim_29504} to #hal.device.promise<@__device_0>
%31956 = torch_c.from_builtin_tensor %31955 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31957 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29505 = arith.constant 1 : index
%dim_29506 = tensor.dim %31957, %c1_29505 : tensor<4x?x4096xf16>
%31958 = flow.tensor.transfer %31957 : tensor<4x?x4096xf16>{%dim_29506} to #hal.device.promise<@__device_0>
%31959 = torch_c.from_builtin_tensor %31958 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31960 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29507 = arith.constant 1 : index
%dim_29508 = tensor.dim %31960, %c1_29507 : tensor<4x?x4096xf16>
%31961 = flow.tensor.transfer %31960 : tensor<4x?x4096xf16>{%dim_29508} to #hal.device.promise<@__device_0>
%31962 = torch_c.from_builtin_tensor %31961 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31963 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29509 = arith.constant 1 : index
%dim_29510 = tensor.dim %31963, %c1_29509 : tensor<4x?x4096xf16>
%31964 = flow.tensor.transfer %31963 : tensor<4x?x4096xf16>{%dim_29510} to #hal.device.promise<@__device_0>
%31965 = torch_c.from_builtin_tensor %31964 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31966 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29511 = arith.constant 1 : index
%dim_29512 = tensor.dim %31966, %c1_29511 : tensor<4x?x4096xf16>
%31967 = flow.tensor.transfer %31966 : tensor<4x?x4096xf16>{%dim_29512} to #hal.device.promise<@__device_0>
%31968 = torch_c.from_builtin_tensor %31967 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29513 = torch.constant.int 1
%31969 = torch.aten.add.Tensor %31898, %31950, %int1_29513 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29514 = torch.constant.int 1
%31970 = torch.aten.add.Tensor %31969, %31953, %int1_29514 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29515 = torch.constant.int 1
%31971 = torch.aten.add.Tensor %31970, %31956, %int1_29515 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29516 = torch.constant.int 1
%31972 = torch.aten.add.Tensor %31971, %31959, %int1_29516 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29517 = torch.constant.int 1
%31973 = torch.aten.add.Tensor %31972, %31962, %int1_29517 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29518 = torch.constant.int 1
%31974 = torch.aten.add.Tensor %31973, %31965, %int1_29518 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29519 = torch.constant.int 1
%31975 = torch.aten.add.Tensor %31974, %31968, %int1_29519 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31976 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29520 = arith.constant 1 : index
%dim_29521 = tensor.dim %31976, %c1_29520 : tensor<4x?x4096xf16>
%31977 = flow.tensor.transfer %31976 : tensor<4x?x4096xf16>{%dim_29521} to #hal.device.promise<@__device_1>
%31978 = torch_c.from_builtin_tensor %31977 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31979 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29522 = arith.constant 1 : index
%dim_29523 = tensor.dim %31979, %c1_29522 : tensor<4x?x4096xf16>
%31980 = flow.tensor.transfer %31979 : tensor<4x?x4096xf16>{%dim_29523} to #hal.device.promise<@__device_1>
%31981 = torch_c.from_builtin_tensor %31980 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31982 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29524 = arith.constant 1 : index
%dim_29525 = tensor.dim %31982, %c1_29524 : tensor<4x?x4096xf16>
%31983 = flow.tensor.transfer %31982 : tensor<4x?x4096xf16>{%dim_29525} to #hal.device.promise<@__device_1>
%31984 = torch_c.from_builtin_tensor %31983 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31985 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29526 = arith.constant 1 : index
%dim_29527 = tensor.dim %31985, %c1_29526 : tensor<4x?x4096xf16>
%31986 = flow.tensor.transfer %31985 : tensor<4x?x4096xf16>{%dim_29527} to #hal.device.promise<@__device_1>
%31987 = torch_c.from_builtin_tensor %31986 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31988 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29528 = arith.constant 1 : index
%dim_29529 = tensor.dim %31988, %c1_29528 : tensor<4x?x4096xf16>
%31989 = flow.tensor.transfer %31988 : tensor<4x?x4096xf16>{%dim_29529} to #hal.device.promise<@__device_1>
%31990 = torch_c.from_builtin_tensor %31989 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31991 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29530 = arith.constant 1 : index
%dim_29531 = tensor.dim %31991, %c1_29530 : tensor<4x?x4096xf16>
%31992 = flow.tensor.transfer %31991 : tensor<4x?x4096xf16>{%dim_29531} to #hal.device.promise<@__device_1>
%31993 = torch_c.from_builtin_tensor %31992 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%31994 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29532 = arith.constant 1 : index
%dim_29533 = tensor.dim %31994, %c1_29532 : tensor<4x?x4096xf16>
%31995 = flow.tensor.transfer %31994 : tensor<4x?x4096xf16>{%dim_29533} to #hal.device.promise<@__device_1>
%31996 = torch_c.from_builtin_tensor %31995 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29534 = torch.constant.int 1
%31997 = torch.aten.add.Tensor %31978, %31905, %int1_29534 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29535 = torch.constant.int 1
%31998 = torch.aten.add.Tensor %31997, %31981, %int1_29535 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29536 = torch.constant.int 1
%31999 = torch.aten.add.Tensor %31998, %31984, %int1_29536 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %31999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29537 = torch.constant.int 1
%32000 = torch.aten.add.Tensor %31999, %31987, %int1_29537 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29538 = torch.constant.int 1
%32001 = torch.aten.add.Tensor %32000, %31990, %int1_29538 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29539 = torch.constant.int 1
%32002 = torch.aten.add.Tensor %32001, %31993, %int1_29539 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29540 = torch.constant.int 1
%32003 = torch.aten.add.Tensor %32002, %31996, %int1_29540 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32004 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29541 = arith.constant 1 : index
%dim_29542 = tensor.dim %32004, %c1_29541 : tensor<4x?x4096xf16>
%32005 = flow.tensor.transfer %32004 : tensor<4x?x4096xf16>{%dim_29542} to #hal.device.promise<@__device_2>
%32006 = torch_c.from_builtin_tensor %32005 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32007 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29543 = arith.constant 1 : index
%dim_29544 = tensor.dim %32007, %c1_29543 : tensor<4x?x4096xf16>
%32008 = flow.tensor.transfer %32007 : tensor<4x?x4096xf16>{%dim_29544} to #hal.device.promise<@__device_2>
%32009 = torch_c.from_builtin_tensor %32008 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32010 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29545 = arith.constant 1 : index
%dim_29546 = tensor.dim %32010, %c1_29545 : tensor<4x?x4096xf16>
%32011 = flow.tensor.transfer %32010 : tensor<4x?x4096xf16>{%dim_29546} to #hal.device.promise<@__device_2>
%32012 = torch_c.from_builtin_tensor %32011 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32012, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32013 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29547 = arith.constant 1 : index
%dim_29548 = tensor.dim %32013, %c1_29547 : tensor<4x?x4096xf16>
%32014 = flow.tensor.transfer %32013 : tensor<4x?x4096xf16>{%dim_29548} to #hal.device.promise<@__device_2>
%32015 = torch_c.from_builtin_tensor %32014 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32016 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29549 = arith.constant 1 : index
%dim_29550 = tensor.dim %32016, %c1_29549 : tensor<4x?x4096xf16>
%32017 = flow.tensor.transfer %32016 : tensor<4x?x4096xf16>{%dim_29550} to #hal.device.promise<@__device_2>
%32018 = torch_c.from_builtin_tensor %32017 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32019 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29551 = arith.constant 1 : index
%dim_29552 = tensor.dim %32019, %c1_29551 : tensor<4x?x4096xf16>
%32020 = flow.tensor.transfer %32019 : tensor<4x?x4096xf16>{%dim_29552} to #hal.device.promise<@__device_2>
%32021 = torch_c.from_builtin_tensor %32020 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32022 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29553 = arith.constant 1 : index
%dim_29554 = tensor.dim %32022, %c1_29553 : tensor<4x?x4096xf16>
%32023 = flow.tensor.transfer %32022 : tensor<4x?x4096xf16>{%dim_29554} to #hal.device.promise<@__device_2>
%32024 = torch_c.from_builtin_tensor %32023 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32024, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29555 = torch.constant.int 1
%32025 = torch.aten.add.Tensor %32006, %32009, %int1_29555 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29556 = torch.constant.int 1
%32026 = torch.aten.add.Tensor %32025, %31912, %int1_29556 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29557 = torch.constant.int 1
%32027 = torch.aten.add.Tensor %32026, %32012, %int1_29557 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29558 = torch.constant.int 1
%32028 = torch.aten.add.Tensor %32027, %32015, %int1_29558 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29559 = torch.constant.int 1
%32029 = torch.aten.add.Tensor %32028, %32018, %int1_29559 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29560 = torch.constant.int 1
%32030 = torch.aten.add.Tensor %32029, %32021, %int1_29560 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29561 = torch.constant.int 1
%32031 = torch.aten.add.Tensor %32030, %32024, %int1_29561 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32032 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29562 = arith.constant 1 : index
%dim_29563 = tensor.dim %32032, %c1_29562 : tensor<4x?x4096xf16>
%32033 = flow.tensor.transfer %32032 : tensor<4x?x4096xf16>{%dim_29563} to #hal.device.promise<@__device_3>
%32034 = torch_c.from_builtin_tensor %32033 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32035 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29564 = arith.constant 1 : index
%dim_29565 = tensor.dim %32035, %c1_29564 : tensor<4x?x4096xf16>
%32036 = flow.tensor.transfer %32035 : tensor<4x?x4096xf16>{%dim_29565} to #hal.device.promise<@__device_3>
%32037 = torch_c.from_builtin_tensor %32036 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32038 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29566 = arith.constant 1 : index
%dim_29567 = tensor.dim %32038, %c1_29566 : tensor<4x?x4096xf16>
%32039 = flow.tensor.transfer %32038 : tensor<4x?x4096xf16>{%dim_29567} to #hal.device.promise<@__device_3>
%32040 = torch_c.from_builtin_tensor %32039 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32041 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29568 = arith.constant 1 : index
%dim_29569 = tensor.dim %32041, %c1_29568 : tensor<4x?x4096xf16>
%32042 = flow.tensor.transfer %32041 : tensor<4x?x4096xf16>{%dim_29569} to #hal.device.promise<@__device_3>
%32043 = torch_c.from_builtin_tensor %32042 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32044 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29570 = arith.constant 1 : index
%dim_29571 = tensor.dim %32044, %c1_29570 : tensor<4x?x4096xf16>
%32045 = flow.tensor.transfer %32044 : tensor<4x?x4096xf16>{%dim_29571} to #hal.device.promise<@__device_3>
%32046 = torch_c.from_builtin_tensor %32045 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32047 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29572 = arith.constant 1 : index
%dim_29573 = tensor.dim %32047, %c1_29572 : tensor<4x?x4096xf16>
%32048 = flow.tensor.transfer %32047 : tensor<4x?x4096xf16>{%dim_29573} to #hal.device.promise<@__device_3>
%32049 = torch_c.from_builtin_tensor %32048 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32050 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29574 = arith.constant 1 : index
%dim_29575 = tensor.dim %32050, %c1_29574 : tensor<4x?x4096xf16>
%32051 = flow.tensor.transfer %32050 : tensor<4x?x4096xf16>{%dim_29575} to #hal.device.promise<@__device_3>
%32052 = torch_c.from_builtin_tensor %32051 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29576 = torch.constant.int 1
%32053 = torch.aten.add.Tensor %32034, %32037, %int1_29576 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29577 = torch.constant.int 1
%32054 = torch.aten.add.Tensor %32053, %32040, %int1_29577 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29578 = torch.constant.int 1
%32055 = torch.aten.add.Tensor %32054, %31919, %int1_29578 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29579 = torch.constant.int 1
%32056 = torch.aten.add.Tensor %32055, %32043, %int1_29579 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29580 = torch.constant.int 1
%32057 = torch.aten.add.Tensor %32056, %32046, %int1_29580 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29581 = torch.constant.int 1
%32058 = torch.aten.add.Tensor %32057, %32049, %int1_29581 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29582 = torch.constant.int 1
%32059 = torch.aten.add.Tensor %32058, %32052, %int1_29582 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32060 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29583 = arith.constant 1 : index
%dim_29584 = tensor.dim %32060, %c1_29583 : tensor<4x?x4096xf16>
%32061 = flow.tensor.transfer %32060 : tensor<4x?x4096xf16>{%dim_29584} to #hal.device.promise<@__device_4>
%32062 = torch_c.from_builtin_tensor %32061 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32063 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29585 = arith.constant 1 : index
%dim_29586 = tensor.dim %32063, %c1_29585 : tensor<4x?x4096xf16>
%32064 = flow.tensor.transfer %32063 : tensor<4x?x4096xf16>{%dim_29586} to #hal.device.promise<@__device_4>
%32065 = torch_c.from_builtin_tensor %32064 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32066 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29587 = arith.constant 1 : index
%dim_29588 = tensor.dim %32066, %c1_29587 : tensor<4x?x4096xf16>
%32067 = flow.tensor.transfer %32066 : tensor<4x?x4096xf16>{%dim_29588} to #hal.device.promise<@__device_4>
%32068 = torch_c.from_builtin_tensor %32067 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32069 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29589 = arith.constant 1 : index
%dim_29590 = tensor.dim %32069, %c1_29589 : tensor<4x?x4096xf16>
%32070 = flow.tensor.transfer %32069 : tensor<4x?x4096xf16>{%dim_29590} to #hal.device.promise<@__device_4>
%32071 = torch_c.from_builtin_tensor %32070 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32072 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29591 = arith.constant 1 : index
%dim_29592 = tensor.dim %32072, %c1_29591 : tensor<4x?x4096xf16>
%32073 = flow.tensor.transfer %32072 : tensor<4x?x4096xf16>{%dim_29592} to #hal.device.promise<@__device_4>
%32074 = torch_c.from_builtin_tensor %32073 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32075 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29593 = arith.constant 1 : index
%dim_29594 = tensor.dim %32075, %c1_29593 : tensor<4x?x4096xf16>
%32076 = flow.tensor.transfer %32075 : tensor<4x?x4096xf16>{%dim_29594} to #hal.device.promise<@__device_4>
%32077 = torch_c.from_builtin_tensor %32076 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32078 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29595 = arith.constant 1 : index
%dim_29596 = tensor.dim %32078, %c1_29595 : tensor<4x?x4096xf16>
%32079 = flow.tensor.transfer %32078 : tensor<4x?x4096xf16>{%dim_29596} to #hal.device.promise<@__device_4>
%32080 = torch_c.from_builtin_tensor %32079 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29597 = torch.constant.int 1
%32081 = torch.aten.add.Tensor %32062, %32065, %int1_29597 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29598 = torch.constant.int 1
%32082 = torch.aten.add.Tensor %32081, %32068, %int1_29598 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29599 = torch.constant.int 1
%32083 = torch.aten.add.Tensor %32082, %32071, %int1_29599 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29600 = torch.constant.int 1
%32084 = torch.aten.add.Tensor %32083, %31926, %int1_29600 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29601 = torch.constant.int 1
%32085 = torch.aten.add.Tensor %32084, %32074, %int1_29601 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29602 = torch.constant.int 1
%32086 = torch.aten.add.Tensor %32085, %32077, %int1_29602 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29603 = torch.constant.int 1
%32087 = torch.aten.add.Tensor %32086, %32080, %int1_29603 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32088 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29604 = arith.constant 1 : index
%dim_29605 = tensor.dim %32088, %c1_29604 : tensor<4x?x4096xf16>
%32089 = flow.tensor.transfer %32088 : tensor<4x?x4096xf16>{%dim_29605} to #hal.device.promise<@__device_5>
%32090 = torch_c.from_builtin_tensor %32089 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32091 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29606 = arith.constant 1 : index
%dim_29607 = tensor.dim %32091, %c1_29606 : tensor<4x?x4096xf16>
%32092 = flow.tensor.transfer %32091 : tensor<4x?x4096xf16>{%dim_29607} to #hal.device.promise<@__device_5>
%32093 = torch_c.from_builtin_tensor %32092 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32094 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29608 = arith.constant 1 : index
%dim_29609 = tensor.dim %32094, %c1_29608 : tensor<4x?x4096xf16>
%32095 = flow.tensor.transfer %32094 : tensor<4x?x4096xf16>{%dim_29609} to #hal.device.promise<@__device_5>
%32096 = torch_c.from_builtin_tensor %32095 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32097 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29610 = arith.constant 1 : index
%dim_29611 = tensor.dim %32097, %c1_29610 : tensor<4x?x4096xf16>
%32098 = flow.tensor.transfer %32097 : tensor<4x?x4096xf16>{%dim_29611} to #hal.device.promise<@__device_5>
%32099 = torch_c.from_builtin_tensor %32098 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32100 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29612 = arith.constant 1 : index
%dim_29613 = tensor.dim %32100, %c1_29612 : tensor<4x?x4096xf16>
%32101 = flow.tensor.transfer %32100 : tensor<4x?x4096xf16>{%dim_29613} to #hal.device.promise<@__device_5>
%32102 = torch_c.from_builtin_tensor %32101 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32103 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29614 = arith.constant 1 : index
%dim_29615 = tensor.dim %32103, %c1_29614 : tensor<4x?x4096xf16>
%32104 = flow.tensor.transfer %32103 : tensor<4x?x4096xf16>{%dim_29615} to #hal.device.promise<@__device_5>
%32105 = torch_c.from_builtin_tensor %32104 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32106 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29616 = arith.constant 1 : index
%dim_29617 = tensor.dim %32106, %c1_29616 : tensor<4x?x4096xf16>
%32107 = flow.tensor.transfer %32106 : tensor<4x?x4096xf16>{%dim_29617} to #hal.device.promise<@__device_5>
%32108 = torch_c.from_builtin_tensor %32107 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29618 = torch.constant.int 1
%32109 = torch.aten.add.Tensor %32090, %32093, %int1_29618 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29619 = torch.constant.int 1
%32110 = torch.aten.add.Tensor %32109, %32096, %int1_29619 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29620 = torch.constant.int 1
%32111 = torch.aten.add.Tensor %32110, %32099, %int1_29620 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29621 = torch.constant.int 1
%32112 = torch.aten.add.Tensor %32111, %32102, %int1_29621 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29622 = torch.constant.int 1
%32113 = torch.aten.add.Tensor %32112, %31933, %int1_29622 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29623 = torch.constant.int 1
%32114 = torch.aten.add.Tensor %32113, %32105, %int1_29623 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29624 = torch.constant.int 1
%32115 = torch.aten.add.Tensor %32114, %32108, %int1_29624 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32116 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29625 = arith.constant 1 : index
%dim_29626 = tensor.dim %32116, %c1_29625 : tensor<4x?x4096xf16>
%32117 = flow.tensor.transfer %32116 : tensor<4x?x4096xf16>{%dim_29626} to #hal.device.promise<@__device_6>
%32118 = torch_c.from_builtin_tensor %32117 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32119 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29627 = arith.constant 1 : index
%dim_29628 = tensor.dim %32119, %c1_29627 : tensor<4x?x4096xf16>
%32120 = flow.tensor.transfer %32119 : tensor<4x?x4096xf16>{%dim_29628} to #hal.device.promise<@__device_6>
%32121 = torch_c.from_builtin_tensor %32120 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32122 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29629 = arith.constant 1 : index
%dim_29630 = tensor.dim %32122, %c1_29629 : tensor<4x?x4096xf16>
%32123 = flow.tensor.transfer %32122 : tensor<4x?x4096xf16>{%dim_29630} to #hal.device.promise<@__device_6>
%32124 = torch_c.from_builtin_tensor %32123 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32124, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32125 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29631 = arith.constant 1 : index
%dim_29632 = tensor.dim %32125, %c1_29631 : tensor<4x?x4096xf16>
%32126 = flow.tensor.transfer %32125 : tensor<4x?x4096xf16>{%dim_29632} to #hal.device.promise<@__device_6>
%32127 = torch_c.from_builtin_tensor %32126 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32128 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29633 = arith.constant 1 : index
%dim_29634 = tensor.dim %32128, %c1_29633 : tensor<4x?x4096xf16>
%32129 = flow.tensor.transfer %32128 : tensor<4x?x4096xf16>{%dim_29634} to #hal.device.promise<@__device_6>
%32130 = torch_c.from_builtin_tensor %32129 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32131 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29635 = arith.constant 1 : index
%dim_29636 = tensor.dim %32131, %c1_29635 : tensor<4x?x4096xf16>
%32132 = flow.tensor.transfer %32131 : tensor<4x?x4096xf16>{%dim_29636} to #hal.device.promise<@__device_6>
%32133 = torch_c.from_builtin_tensor %32132 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32134 = torch_c.to_builtin_tensor %31947 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29637 = arith.constant 1 : index
%dim_29638 = tensor.dim %32134, %c1_29637 : tensor<4x?x4096xf16>
%32135 = flow.tensor.transfer %32134 : tensor<4x?x4096xf16>{%dim_29638} to #hal.device.promise<@__device_6>
%32136 = torch_c.from_builtin_tensor %32135 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29639 = torch.constant.int 1
%32137 = torch.aten.add.Tensor %32118, %32121, %int1_29639 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29640 = torch.constant.int 1
%32138 = torch.aten.add.Tensor %32137, %32124, %int1_29640 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29641 = torch.constant.int 1
%32139 = torch.aten.add.Tensor %32138, %32127, %int1_29641 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29642 = torch.constant.int 1
%32140 = torch.aten.add.Tensor %32139, %32130, %int1_29642 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29643 = torch.constant.int 1
%32141 = torch.aten.add.Tensor %32140, %32133, %int1_29643 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29644 = torch.constant.int 1
%32142 = torch.aten.add.Tensor %32141, %31940, %int1_29644 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29645 = torch.constant.int 1
%32143 = torch.aten.add.Tensor %32142, %32136, %int1_29645 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32144 = torch_c.to_builtin_tensor %31898 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29646 = arith.constant 1 : index
%dim_29647 = tensor.dim %32144, %c1_29646 : tensor<4x?x4096xf16>
%32145 = flow.tensor.transfer %32144 : tensor<4x?x4096xf16>{%dim_29647} to #hal.device.promise<@__device_7>
%32146 = torch_c.from_builtin_tensor %32145 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32147 = torch_c.to_builtin_tensor %31905 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29648 = arith.constant 1 : index
%dim_29649 = tensor.dim %32147, %c1_29648 : tensor<4x?x4096xf16>
%32148 = flow.tensor.transfer %32147 : tensor<4x?x4096xf16>{%dim_29649} to #hal.device.promise<@__device_7>
%32149 = torch_c.from_builtin_tensor %32148 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32150 = torch_c.to_builtin_tensor %31912 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29650 = arith.constant 1 : index
%dim_29651 = tensor.dim %32150, %c1_29650 : tensor<4x?x4096xf16>
%32151 = flow.tensor.transfer %32150 : tensor<4x?x4096xf16>{%dim_29651} to #hal.device.promise<@__device_7>
%32152 = torch_c.from_builtin_tensor %32151 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32153 = torch_c.to_builtin_tensor %31919 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29652 = arith.constant 1 : index
%dim_29653 = tensor.dim %32153, %c1_29652 : tensor<4x?x4096xf16>
%32154 = flow.tensor.transfer %32153 : tensor<4x?x4096xf16>{%dim_29653} to #hal.device.promise<@__device_7>
%32155 = torch_c.from_builtin_tensor %32154 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32156 = torch_c.to_builtin_tensor %31926 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29654 = arith.constant 1 : index
%dim_29655 = tensor.dim %32156, %c1_29654 : tensor<4x?x4096xf16>
%32157 = flow.tensor.transfer %32156 : tensor<4x?x4096xf16>{%dim_29655} to #hal.device.promise<@__device_7>
%32158 = torch_c.from_builtin_tensor %32157 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32159 = torch_c.to_builtin_tensor %31933 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29656 = arith.constant 1 : index
%dim_29657 = tensor.dim %32159, %c1_29656 : tensor<4x?x4096xf16>
%32160 = flow.tensor.transfer %32159 : tensor<4x?x4096xf16>{%dim_29657} to #hal.device.promise<@__device_7>
%32161 = torch_c.from_builtin_tensor %32160 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%32162 = torch_c.to_builtin_tensor %31940 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_29658 = arith.constant 1 : index
%dim_29659 = tensor.dim %32162, %c1_29658 : tensor<4x?x4096xf16>
%32163 = flow.tensor.transfer %32162 : tensor<4x?x4096xf16>{%dim_29659} to #hal.device.promise<@__device_7>
%32164 = torch_c.from_builtin_tensor %32163 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29660 = torch.constant.int 1
%32165 = torch.aten.add.Tensor %32146, %32149, %int1_29660 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29661 = torch.constant.int 1
%32166 = torch.aten.add.Tensor %32165, %32152, %int1_29661 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29662 = torch.constant.int 1
%32167 = torch.aten.add.Tensor %32166, %32155, %int1_29662 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29663 = torch.constant.int 1
%32168 = torch.aten.add.Tensor %32167, %32158, %int1_29663 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29664 = torch.constant.int 1
%32169 = torch.aten.add.Tensor %32168, %32161, %int1_29664 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29665 = torch.constant.int 1
%32170 = torch.aten.add.Tensor %32169, %32164, %int1_29665 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29666 = torch.constant.int 1
%32171 = torch.aten.add.Tensor %32170, %31947, %int1_29666 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29667 = torch.constant.int 1
%32172 = torch.aten.add.Tensor %31652, %31975, %int1_29667 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29668 = torch.constant.int 1
%32173 = torch.aten.add.Tensor %31653, %32003, %int1_29668 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29669 = torch.constant.int 1
%32174 = torch.aten.add.Tensor %31654, %32031, %int1_29669 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29670 = torch.constant.int 1
%32175 = torch.aten.add.Tensor %31655, %32059, %int1_29670 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29671 = torch.constant.int 1
%32176 = torch.aten.add.Tensor %31656, %32087, %int1_29671 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29672 = torch.constant.int 1
%32177 = torch.aten.add.Tensor %31657, %32115, %int1_29672 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29673 = torch.constant.int 1
%32178 = torch.aten.add.Tensor %31658, %32143, %int1_29673 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29674 = torch.constant.int 1
%32179 = torch.aten.add.Tensor %31659, %32171, %int1_29674 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_29675 = torch.constant.int 6
%32180 = torch.prims.convert_element_type %32172, %int6_29675 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29676 = torch.constant.int 6
%32181 = torch.prims.convert_element_type %32173, %int6_29676 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29677 = torch.constant.int 6
%32182 = torch.prims.convert_element_type %32174, %int6_29677 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29678 = torch.constant.int 6
%32183 = torch.prims.convert_element_type %32175, %int6_29678 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29679 = torch.constant.int 6
%32184 = torch.prims.convert_element_type %32176, %int6_29679 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29680 = torch.constant.int 6
%32185 = torch.prims.convert_element_type %32177, %int6_29680 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29681 = torch.constant.int 6
%32186 = torch.prims.convert_element_type %32178, %int6_29681 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_29682 = torch.constant.int 6
%32187 = torch.prims.convert_element_type %32179, %int6_29682 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29683 = torch.constant.int 2
%32188 = torch.aten.pow.Tensor_Scalar %32180, %int2_29683 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32188, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29684 = torch.constant.int 2
%32189 = torch.aten.pow.Tensor_Scalar %32181, %int2_29684 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29685 = torch.constant.int 2
%32190 = torch.aten.pow.Tensor_Scalar %32182, %int2_29685 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29686 = torch.constant.int 2
%32191 = torch.aten.pow.Tensor_Scalar %32183, %int2_29686 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29687 = torch.constant.int 2
%32192 = torch.aten.pow.Tensor_Scalar %32184, %int2_29687 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29688 = torch.constant.int 2
%32193 = torch.aten.pow.Tensor_Scalar %32185, %int2_29688 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29689 = torch.constant.int 2
%32194 = torch.aten.pow.Tensor_Scalar %32186, %int2_29689 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32194, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_29690 = torch.constant.int 2
%32195 = torch.aten.pow.Tensor_Scalar %32187, %int2_29690 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_29691 = torch.constant.int -1
%32196 = torch.prim.ListConstruct %int-1_29691 : (!torch.int) -> !torch.list<int>
%true_29692 = torch.constant.bool true
%none_29693 = torch.constant.none
%32197 = torch.aten.mean.dim %32188, %32196, %true_29692, %none_29693 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29694 = torch.constant.int -1
%32198 = torch.prim.ListConstruct %int-1_29694 : (!torch.int) -> !torch.list<int>
%true_29695 = torch.constant.bool true
%none_29696 = torch.constant.none
%32199 = torch.aten.mean.dim %32189, %32198, %true_29695, %none_29696 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29697 = torch.constant.int -1
%32200 = torch.prim.ListConstruct %int-1_29697 : (!torch.int) -> !torch.list<int>
%true_29698 = torch.constant.bool true
%none_29699 = torch.constant.none
%32201 = torch.aten.mean.dim %32190, %32200, %true_29698, %none_29699 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29700 = torch.constant.int -1
%32202 = torch.prim.ListConstruct %int-1_29700 : (!torch.int) -> !torch.list<int>
%true_29701 = torch.constant.bool true
%none_29702 = torch.constant.none
%32203 = torch.aten.mean.dim %32191, %32202, %true_29701, %none_29702 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29703 = torch.constant.int -1
%32204 = torch.prim.ListConstruct %int-1_29703 : (!torch.int) -> !torch.list<int>
%true_29704 = torch.constant.bool true
%none_29705 = torch.constant.none
%32205 = torch.aten.mean.dim %32192, %32204, %true_29704, %none_29705 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29706 = torch.constant.int -1
%32206 = torch.prim.ListConstruct %int-1_29706 : (!torch.int) -> !torch.list<int>
%true_29707 = torch.constant.bool true
%none_29708 = torch.constant.none
%32207 = torch.aten.mean.dim %32193, %32206, %true_29707, %none_29708 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29709 = torch.constant.int -1
%32208 = torch.prim.ListConstruct %int-1_29709 : (!torch.int) -> !torch.list<int>
%true_29710 = torch.constant.bool true
%none_29711 = torch.constant.none
%32209 = torch.aten.mean.dim %32194, %32208, %true_29710, %none_29711 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_29712 = torch.constant.int -1
%32210 = torch.prim.ListConstruct %int-1_29712 : (!torch.int) -> !torch.list<int>
%true_29713 = torch.constant.bool true
%none_29714 = torch.constant.none
%32211 = torch.aten.mean.dim %32195, %32210, %true_29713, %none_29714 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29715 = torch.constant.float 9.9999997473787516E-6
%int1_29716 = torch.constant.int 1
%32212 = torch.aten.add.Scalar %32197, %float9.999990e-06_29715, %int1_29716 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29717 = torch.constant.float 9.9999997473787516E-6
%int1_29718 = torch.constant.int 1
%32213 = torch.aten.add.Scalar %32199, %float9.999990e-06_29717, %int1_29718 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29719 = torch.constant.float 9.9999997473787516E-6
%int1_29720 = torch.constant.int 1
%32214 = torch.aten.add.Scalar %32201, %float9.999990e-06_29719, %int1_29720 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29721 = torch.constant.float 9.9999997473787516E-6
%int1_29722 = torch.constant.int 1
%32215 = torch.aten.add.Scalar %32203, %float9.999990e-06_29721, %int1_29722 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29723 = torch.constant.float 9.9999997473787516E-6
%int1_29724 = torch.constant.int 1
%32216 = torch.aten.add.Scalar %32205, %float9.999990e-06_29723, %int1_29724 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29725 = torch.constant.float 9.9999997473787516E-6
%int1_29726 = torch.constant.int 1
%32217 = torch.aten.add.Scalar %32207, %float9.999990e-06_29725, %int1_29726 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29727 = torch.constant.float 9.9999997473787516E-6
%int1_29728 = torch.constant.int 1
%32218 = torch.aten.add.Scalar %32209, %float9.999990e-06_29727, %int1_29728 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_29729 = torch.constant.float 9.9999997473787516E-6
%int1_29730 = torch.constant.int 1
%32219 = torch.aten.add.Scalar %32211, %float9.999990e-06_29729, %int1_29730 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32219, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32220 = torch.aten.rsqrt %32212 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32221 = torch.aten.rsqrt %32213 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32222 = torch.aten.rsqrt %32214 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32223 = torch.aten.rsqrt %32215 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32224 = torch.aten.rsqrt %32216 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32225 = torch.aten.rsqrt %32217 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32225, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32226 = torch.aten.rsqrt %32218 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32227 = torch.aten.rsqrt %32219 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %32227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%32228 = torch.aten.mul.Tensor %32180, %32220 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32229 = torch.aten.mul.Tensor %32181, %32221 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32230 = torch.aten.mul.Tensor %32182, %32222 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32231 = torch.aten.mul.Tensor %32183, %32223 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32232 = torch.aten.mul.Tensor %32184, %32224 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32233 = torch.aten.mul.Tensor %32185, %32225 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32234 = torch.aten.mul.Tensor %32186, %32226 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32235 = torch.aten.mul.Tensor %32187, %32227 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32236 = torch.aten.mul.Tensor %1160, %32228 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32236, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32237 = torch.aten.mul.Tensor %1161, %32229 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32238 = torch.aten.mul.Tensor %1162, %32230 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32239 = torch.aten.mul.Tensor %1163, %32231 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32240 = torch.aten.mul.Tensor %1164, %32232 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32241 = torch.aten.mul.Tensor %1165, %32233 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32242 = torch.aten.mul.Tensor %1166, %32234 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%32243 = torch.aten.mul.Tensor %1167, %32235 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %32243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_29731 = torch.constant.int 5
%32244 = torch.prims.convert_element_type %32236, %int5_29731 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29732 = torch.constant.int 5
%32245 = torch.prims.convert_element_type %32237, %int5_29732 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29733 = torch.constant.int 5
%32246 = torch.prims.convert_element_type %32238, %int5_29733 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29734 = torch.constant.int 5
%32247 = torch.prims.convert_element_type %32239, %int5_29734 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29735 = torch.constant.int 5
%32248 = torch.prims.convert_element_type %32240, %int5_29735 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29736 = torch.constant.int 5
%32249 = torch.prims.convert_element_type %32241, %int5_29736 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29737 = torch.constant.int 5
%32250 = torch.prims.convert_element_type %32242, %int5_29737 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_29738 = torch.constant.int 5
%32251 = torch.prims.convert_element_type %32243, %int5_29738 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %32251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_29739 = torch.constant.int 1
%int0_29740 = torch.constant.int 0
%32252 = torch.prim.ListConstruct %int1_29739, %int0_29740 : (!torch.int, !torch.int) -> !torch.list<int>
%32253 = torch.aten.permute %1168, %32252 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29741 = torch.constant.int 1
%int0_29742 = torch.constant.int 0
%32254 = torch.prim.ListConstruct %int1_29741, %int0_29742 : (!torch.int, !torch.int) -> !torch.list<int>
%32255 = torch.aten.permute %1169, %32254 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29743 = torch.constant.int 1
%int0_29744 = torch.constant.int 0
%32256 = torch.prim.ListConstruct %int1_29743, %int0_29744 : (!torch.int, !torch.int) -> !torch.list<int>
%32257 = torch.aten.permute %1170, %32256 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29745 = torch.constant.int 1
%int0_29746 = torch.constant.int 0
%32258 = torch.prim.ListConstruct %int1_29745, %int0_29746 : (!torch.int, !torch.int) -> !torch.list<int>
%32259 = torch.aten.permute %1171, %32258 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29747 = torch.constant.int 1
%int0_29748 = torch.constant.int 0
%32260 = torch.prim.ListConstruct %int1_29747, %int0_29748 : (!torch.int, !torch.int) -> !torch.list<int>
%32261 = torch.aten.permute %1172, %32260 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29749 = torch.constant.int 1
%int0_29750 = torch.constant.int 0
%32262 = torch.prim.ListConstruct %int1_29749, %int0_29750 : (!torch.int, !torch.int) -> !torch.list<int>
%32263 = torch.aten.permute %1173, %32262 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29751 = torch.constant.int 1
%int0_29752 = torch.constant.int 0
%32264 = torch.prim.ListConstruct %int1_29751, %int0_29752 : (!torch.int, !torch.int) -> !torch.list<int>
%32265 = torch.aten.permute %1174, %32264 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_29753 = torch.constant.int 1
%int0_29754 = torch.constant.int 0
%32266 = torch.prim.ListConstruct %int1_29753, %int0_29754 : (!torch.int, !torch.int) -> !torch.list<int>
%32267 = torch.aten.permute %1175, %32266 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_29755 = torch.constant.int 4
%32268 = torch.aten.mul.int %int4_29755, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29756 = torch.constant.int 4096
%32269 = torch.prim.ListConstruct %32268, %int4096_29756 : (!torch.int, !torch.int) -> !torch.list<int>
%32270 = torch.aten.view %32244, %32269 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32270, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32271 = torch.aten.mm %32270, %32253 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32271, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29757 = torch.constant.int 4
%int512_29758 = torch.constant.int 512
%32272 = torch.prim.ListConstruct %int4_29757, %2482, %int512_29758 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32273 = torch.aten.view %32271, %32272 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29759 = torch.constant.int 4
%32274 = torch.aten.mul.int %int4_29759, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29760 = torch.constant.int 4096
%32275 = torch.prim.ListConstruct %32274, %int4096_29760 : (!torch.int, !torch.int) -> !torch.list<int>
%32276 = torch.aten.view %32245, %32275 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32276, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32277 = torch.aten.mm %32276, %32255 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32277, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29761 = torch.constant.int 4
%int512_29762 = torch.constant.int 512
%32278 = torch.prim.ListConstruct %int4_29761, %2482, %int512_29762 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32279 = torch.aten.view %32277, %32278 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29763 = torch.constant.int 4
%32280 = torch.aten.mul.int %int4_29763, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29764 = torch.constant.int 4096
%32281 = torch.prim.ListConstruct %32280, %int4096_29764 : (!torch.int, !torch.int) -> !torch.list<int>
%32282 = torch.aten.view %32246, %32281 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32282, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32283 = torch.aten.mm %32282, %32257 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32283, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29765 = torch.constant.int 4
%int512_29766 = torch.constant.int 512
%32284 = torch.prim.ListConstruct %int4_29765, %2482, %int512_29766 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32285 = torch.aten.view %32283, %32284 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29767 = torch.constant.int 4
%32286 = torch.aten.mul.int %int4_29767, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29768 = torch.constant.int 4096
%32287 = torch.prim.ListConstruct %32286, %int4096_29768 : (!torch.int, !torch.int) -> !torch.list<int>
%32288 = torch.aten.view %32247, %32287 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32288, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32289 = torch.aten.mm %32288, %32259 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32289, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29769 = torch.constant.int 4
%int512_29770 = torch.constant.int 512
%32290 = torch.prim.ListConstruct %int4_29769, %2482, %int512_29770 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32291 = torch.aten.view %32289, %32290 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29771 = torch.constant.int 4
%32292 = torch.aten.mul.int %int4_29771, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29772 = torch.constant.int 4096
%32293 = torch.prim.ListConstruct %32292, %int4096_29772 : (!torch.int, !torch.int) -> !torch.list<int>
%32294 = torch.aten.view %32248, %32293 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32294, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32295 = torch.aten.mm %32294, %32261 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32295, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29773 = torch.constant.int 4
%int512_29774 = torch.constant.int 512
%32296 = torch.prim.ListConstruct %int4_29773, %2482, %int512_29774 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32297 = torch.aten.view %32295, %32296 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29775 = torch.constant.int 4
%32298 = torch.aten.mul.int %int4_29775, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29776 = torch.constant.int 4096
%32299 = torch.prim.ListConstruct %32298, %int4096_29776 : (!torch.int, !torch.int) -> !torch.list<int>
%32300 = torch.aten.view %32249, %32299 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32300, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32301 = torch.aten.mm %32300, %32263 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32301, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29777 = torch.constant.int 4
%int512_29778 = torch.constant.int 512
%32302 = torch.prim.ListConstruct %int4_29777, %2482, %int512_29778 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32303 = torch.aten.view %32301, %32302 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29779 = torch.constant.int 4
%32304 = torch.aten.mul.int %int4_29779, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29780 = torch.constant.int 4096
%32305 = torch.prim.ListConstruct %32304, %int4096_29780 : (!torch.int, !torch.int) -> !torch.list<int>
%32306 = torch.aten.view %32250, %32305 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32306, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32307 = torch.aten.mm %32306, %32265 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32307, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29781 = torch.constant.int 4
%int512_29782 = torch.constant.int 512
%32308 = torch.prim.ListConstruct %int4_29781, %2482, %int512_29782 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32309 = torch.aten.view %32307, %32308 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_29783 = torch.constant.int 4
%32310 = torch.aten.mul.int %int4_29783, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29784 = torch.constant.int 4096
%32311 = torch.prim.ListConstruct %32310, %int4096_29784 : (!torch.int, !torch.int) -> !torch.list<int>
%32312 = torch.aten.view %32251, %32311 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32312, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32313 = torch.aten.mm %32312, %32267 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %32313, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_29785 = torch.constant.int 4
%int512_29786 = torch.constant.int 512
%32314 = torch.prim.ListConstruct %int4_29785, %2482, %int512_29786 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32315 = torch.aten.view %32313, %32314 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %32315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_29787 = torch.constant.int 1
%int0_29788 = torch.constant.int 0
%32316 = torch.prim.ListConstruct %int1_29787, %int0_29788 : (!torch.int, !torch.int) -> !torch.list<int>
%32317 = torch.aten.permute %1176, %32316 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29789 = torch.constant.int 1
%int0_29790 = torch.constant.int 0
%32318 = torch.prim.ListConstruct %int1_29789, %int0_29790 : (!torch.int, !torch.int) -> !torch.list<int>
%32319 = torch.aten.permute %1177, %32318 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29791 = torch.constant.int 1
%int0_29792 = torch.constant.int 0
%32320 = torch.prim.ListConstruct %int1_29791, %int0_29792 : (!torch.int, !torch.int) -> !torch.list<int>
%32321 = torch.aten.permute %1178, %32320 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29793 = torch.constant.int 1
%int0_29794 = torch.constant.int 0
%32322 = torch.prim.ListConstruct %int1_29793, %int0_29794 : (!torch.int, !torch.int) -> !torch.list<int>
%32323 = torch.aten.permute %1179, %32322 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29795 = torch.constant.int 1
%int0_29796 = torch.constant.int 0
%32324 = torch.prim.ListConstruct %int1_29795, %int0_29796 : (!torch.int, !torch.int) -> !torch.list<int>
%32325 = torch.aten.permute %1180, %32324 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29797 = torch.constant.int 1
%int0_29798 = torch.constant.int 0
%32326 = torch.prim.ListConstruct %int1_29797, %int0_29798 : (!torch.int, !torch.int) -> !torch.list<int>
%32327 = torch.aten.permute %1181, %32326 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29799 = torch.constant.int 1
%int0_29800 = torch.constant.int 0
%32328 = torch.prim.ListConstruct %int1_29799, %int0_29800 : (!torch.int, !torch.int) -> !torch.list<int>
%32329 = torch.aten.permute %1182, %32328 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29801 = torch.constant.int 1
%int0_29802 = torch.constant.int 0
%32330 = torch.prim.ListConstruct %int1_29801, %int0_29802 : (!torch.int, !torch.int) -> !torch.list<int>
%32331 = torch.aten.permute %1183, %32330 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_29803 = torch.constant.int 4
%32332 = torch.aten.mul.int %int4_29803, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29804 = torch.constant.int 4096
%32333 = torch.prim.ListConstruct %32332, %int4096_29804 : (!torch.int, !torch.int) -> !torch.list<int>
%32334 = torch.aten.view %32244, %32333 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32334, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32335 = torch.aten.mm %32334, %32317 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32335, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29805 = torch.constant.int 4
%int128_29806 = torch.constant.int 128
%32336 = torch.prim.ListConstruct %int4_29805, %2482, %int128_29806 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32337 = torch.aten.view %32335, %32336 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29807 = torch.constant.int 4
%32338 = torch.aten.mul.int %int4_29807, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29808 = torch.constant.int 4096
%32339 = torch.prim.ListConstruct %32338, %int4096_29808 : (!torch.int, !torch.int) -> !torch.list<int>
%32340 = torch.aten.view %32245, %32339 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32340, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32341 = torch.aten.mm %32340, %32319 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32341, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29809 = torch.constant.int 4
%int128_29810 = torch.constant.int 128
%32342 = torch.prim.ListConstruct %int4_29809, %2482, %int128_29810 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32343 = torch.aten.view %32341, %32342 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29811 = torch.constant.int 4
%32344 = torch.aten.mul.int %int4_29811, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29812 = torch.constant.int 4096
%32345 = torch.prim.ListConstruct %32344, %int4096_29812 : (!torch.int, !torch.int) -> !torch.list<int>
%32346 = torch.aten.view %32246, %32345 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32346, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32347 = torch.aten.mm %32346, %32321 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32347, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29813 = torch.constant.int 4
%int128_29814 = torch.constant.int 128
%32348 = torch.prim.ListConstruct %int4_29813, %2482, %int128_29814 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32349 = torch.aten.view %32347, %32348 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29815 = torch.constant.int 4
%32350 = torch.aten.mul.int %int4_29815, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29816 = torch.constant.int 4096
%32351 = torch.prim.ListConstruct %32350, %int4096_29816 : (!torch.int, !torch.int) -> !torch.list<int>
%32352 = torch.aten.view %32247, %32351 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32352, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32353 = torch.aten.mm %32352, %32323 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32353, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29817 = torch.constant.int 4
%int128_29818 = torch.constant.int 128
%32354 = torch.prim.ListConstruct %int4_29817, %2482, %int128_29818 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32355 = torch.aten.view %32353, %32354 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29819 = torch.constant.int 4
%32356 = torch.aten.mul.int %int4_29819, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29820 = torch.constant.int 4096
%32357 = torch.prim.ListConstruct %32356, %int4096_29820 : (!torch.int, !torch.int) -> !torch.list<int>
%32358 = torch.aten.view %32248, %32357 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32358, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32359 = torch.aten.mm %32358, %32325 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32359, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29821 = torch.constant.int 4
%int128_29822 = torch.constant.int 128
%32360 = torch.prim.ListConstruct %int4_29821, %2482, %int128_29822 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32361 = torch.aten.view %32359, %32360 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32361, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29823 = torch.constant.int 4
%32362 = torch.aten.mul.int %int4_29823, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29824 = torch.constant.int 4096
%32363 = torch.prim.ListConstruct %32362, %int4096_29824 : (!torch.int, !torch.int) -> !torch.list<int>
%32364 = torch.aten.view %32249, %32363 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32364, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32365 = torch.aten.mm %32364, %32327 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32365, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29825 = torch.constant.int 4
%int128_29826 = torch.constant.int 128
%32366 = torch.prim.ListConstruct %int4_29825, %2482, %int128_29826 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32367 = torch.aten.view %32365, %32366 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29827 = torch.constant.int 4
%32368 = torch.aten.mul.int %int4_29827, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29828 = torch.constant.int 4096
%32369 = torch.prim.ListConstruct %32368, %int4096_29828 : (!torch.int, !torch.int) -> !torch.list<int>
%32370 = torch.aten.view %32250, %32369 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32370, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32371 = torch.aten.mm %32370, %32329 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32371, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29829 = torch.constant.int 4
%int128_29830 = torch.constant.int 128
%32372 = torch.prim.ListConstruct %int4_29829, %2482, %int128_29830 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32373 = torch.aten.view %32371, %32372 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29831 = torch.constant.int 4
%32374 = torch.aten.mul.int %int4_29831, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29832 = torch.constant.int 4096
%32375 = torch.prim.ListConstruct %32374, %int4096_29832 : (!torch.int, !torch.int) -> !torch.list<int>
%32376 = torch.aten.view %32251, %32375 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32376, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32377 = torch.aten.mm %32376, %32331 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32377, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29833 = torch.constant.int 4
%int128_29834 = torch.constant.int 128
%32378 = torch.prim.ListConstruct %int4_29833, %2482, %int128_29834 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32379 = torch.aten.view %32377, %32378 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_29835 = torch.constant.int 1
%int0_29836 = torch.constant.int 0
%32380 = torch.prim.ListConstruct %int1_29835, %int0_29836 : (!torch.int, !torch.int) -> !torch.list<int>
%32381 = torch.aten.permute %1184, %32380 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29837 = torch.constant.int 1
%int0_29838 = torch.constant.int 0
%32382 = torch.prim.ListConstruct %int1_29837, %int0_29838 : (!torch.int, !torch.int) -> !torch.list<int>
%32383 = torch.aten.permute %1185, %32382 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29839 = torch.constant.int 1
%int0_29840 = torch.constant.int 0
%32384 = torch.prim.ListConstruct %int1_29839, %int0_29840 : (!torch.int, !torch.int) -> !torch.list<int>
%32385 = torch.aten.permute %1186, %32384 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29841 = torch.constant.int 1
%int0_29842 = torch.constant.int 0
%32386 = torch.prim.ListConstruct %int1_29841, %int0_29842 : (!torch.int, !torch.int) -> !torch.list<int>
%32387 = torch.aten.permute %1187, %32386 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29843 = torch.constant.int 1
%int0_29844 = torch.constant.int 0
%32388 = torch.prim.ListConstruct %int1_29843, %int0_29844 : (!torch.int, !torch.int) -> !torch.list<int>
%32389 = torch.aten.permute %1188, %32388 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29845 = torch.constant.int 1
%int0_29846 = torch.constant.int 0
%32390 = torch.prim.ListConstruct %int1_29845, %int0_29846 : (!torch.int, !torch.int) -> !torch.list<int>
%32391 = torch.aten.permute %1189, %32390 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29847 = torch.constant.int 1
%int0_29848 = torch.constant.int 0
%32392 = torch.prim.ListConstruct %int1_29847, %int0_29848 : (!torch.int, !torch.int) -> !torch.list<int>
%32393 = torch.aten.permute %1190, %32392 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_29849 = torch.constant.int 1
%int0_29850 = torch.constant.int 0
%32394 = torch.prim.ListConstruct %int1_29849, %int0_29850 : (!torch.int, !torch.int) -> !torch.list<int>
%32395 = torch.aten.permute %1191, %32394 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_29851 = torch.constant.int 4
%32396 = torch.aten.mul.int %int4_29851, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29852 = torch.constant.int 4096
%32397 = torch.prim.ListConstruct %32396, %int4096_29852 : (!torch.int, !torch.int) -> !torch.list<int>
%32398 = torch.aten.view %32244, %32397 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32398, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32399 = torch.aten.mm %32398, %32381 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32399, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29853 = torch.constant.int 4
%int128_29854 = torch.constant.int 128
%32400 = torch.prim.ListConstruct %int4_29853, %2482, %int128_29854 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32401 = torch.aten.view %32399, %32400 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29855 = torch.constant.int 4
%32402 = torch.aten.mul.int %int4_29855, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29856 = torch.constant.int 4096
%32403 = torch.prim.ListConstruct %32402, %int4096_29856 : (!torch.int, !torch.int) -> !torch.list<int>
%32404 = torch.aten.view %32245, %32403 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32404, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32405 = torch.aten.mm %32404, %32383 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32405, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29857 = torch.constant.int 4
%int128_29858 = torch.constant.int 128
%32406 = torch.prim.ListConstruct %int4_29857, %2482, %int128_29858 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32407 = torch.aten.view %32405, %32406 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29859 = torch.constant.int 4
%32408 = torch.aten.mul.int %int4_29859, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29860 = torch.constant.int 4096
%32409 = torch.prim.ListConstruct %32408, %int4096_29860 : (!torch.int, !torch.int) -> !torch.list<int>
%32410 = torch.aten.view %32246, %32409 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32410, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32411 = torch.aten.mm %32410, %32385 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32411, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29861 = torch.constant.int 4
%int128_29862 = torch.constant.int 128
%32412 = torch.prim.ListConstruct %int4_29861, %2482, %int128_29862 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32413 = torch.aten.view %32411, %32412 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29863 = torch.constant.int 4
%32414 = torch.aten.mul.int %int4_29863, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29864 = torch.constant.int 4096
%32415 = torch.prim.ListConstruct %32414, %int4096_29864 : (!torch.int, !torch.int) -> !torch.list<int>
%32416 = torch.aten.view %32247, %32415 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32416, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32417 = torch.aten.mm %32416, %32387 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32417, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29865 = torch.constant.int 4
%int128_29866 = torch.constant.int 128
%32418 = torch.prim.ListConstruct %int4_29865, %2482, %int128_29866 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32419 = torch.aten.view %32417, %32418 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29867 = torch.constant.int 4
%32420 = torch.aten.mul.int %int4_29867, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29868 = torch.constant.int 4096
%32421 = torch.prim.ListConstruct %32420, %int4096_29868 : (!torch.int, !torch.int) -> !torch.list<int>
%32422 = torch.aten.view %32248, %32421 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32422, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32423 = torch.aten.mm %32422, %32389 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32423, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29869 = torch.constant.int 4
%int128_29870 = torch.constant.int 128
%32424 = torch.prim.ListConstruct %int4_29869, %2482, %int128_29870 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32425 = torch.aten.view %32423, %32424 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29871 = torch.constant.int 4
%32426 = torch.aten.mul.int %int4_29871, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29872 = torch.constant.int 4096
%32427 = torch.prim.ListConstruct %32426, %int4096_29872 : (!torch.int, !torch.int) -> !torch.list<int>
%32428 = torch.aten.view %32249, %32427 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32428, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32429 = torch.aten.mm %32428, %32391 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32429, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29873 = torch.constant.int 4
%int128_29874 = torch.constant.int 128
%32430 = torch.prim.ListConstruct %int4_29873, %2482, %int128_29874 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32431 = torch.aten.view %32429, %32430 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29875 = torch.constant.int 4
%32432 = torch.aten.mul.int %int4_29875, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29876 = torch.constant.int 4096
%32433 = torch.prim.ListConstruct %32432, %int4096_29876 : (!torch.int, !torch.int) -> !torch.list<int>
%32434 = torch.aten.view %32250, %32433 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32434, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32435 = torch.aten.mm %32434, %32393 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32435, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29877 = torch.constant.int 4
%int128_29878 = torch.constant.int 128
%32436 = torch.prim.ListConstruct %int4_29877, %2482, %int128_29878 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32437 = torch.aten.view %32435, %32436 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29879 = torch.constant.int 4
%32438 = torch.aten.mul.int %int4_29879, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_29880 = torch.constant.int 4096
%32439 = torch.prim.ListConstruct %32438, %int4096_29880 : (!torch.int, !torch.int) -> !torch.list<int>
%32440 = torch.aten.view %32251, %32439 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %32440, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%32441 = torch.aten.mm %32440, %32395 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %32441, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_29881 = torch.constant.int 4
%int128_29882 = torch.constant.int 128
%32442 = torch.prim.ListConstruct %int4_29881, %2482, %int128_29882 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32443 = torch.aten.view %32441, %32442 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %32443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_29883 = torch.constant.int 4
%int4_29884 = torch.constant.int 4
%int128_29885 = torch.constant.int 128
%32444 = torch.prim.ListConstruct %int4_29883, %2482, %int4_29884, %int128_29885 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32445 = torch.aten.view %32273, %32444 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29886 = torch.constant.int 4
%int4_29887 = torch.constant.int 4
%int128_29888 = torch.constant.int 128
%32446 = torch.prim.ListConstruct %int4_29886, %2482, %int4_29887, %int128_29888 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32447 = torch.aten.view %32279, %32446 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29889 = torch.constant.int 4
%int4_29890 = torch.constant.int 4
%int128_29891 = torch.constant.int 128
%32448 = torch.prim.ListConstruct %int4_29889, %2482, %int4_29890, %int128_29891 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32449 = torch.aten.view %32285, %32448 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29892 = torch.constant.int 4
%int4_29893 = torch.constant.int 4
%int128_29894 = torch.constant.int 128
%32450 = torch.prim.ListConstruct %int4_29892, %2482, %int4_29893, %int128_29894 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32451 = torch.aten.view %32291, %32450 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29895 = torch.constant.int 4
%int4_29896 = torch.constant.int 4
%int128_29897 = torch.constant.int 128
%32452 = torch.prim.ListConstruct %int4_29895, %2482, %int4_29896, %int128_29897 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32453 = torch.aten.view %32297, %32452 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29898 = torch.constant.int 4
%int4_29899 = torch.constant.int 4
%int128_29900 = torch.constant.int 128
%32454 = torch.prim.ListConstruct %int4_29898, %2482, %int4_29899, %int128_29900 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32455 = torch.aten.view %32303, %32454 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29901 = torch.constant.int 4
%int4_29902 = torch.constant.int 4
%int128_29903 = torch.constant.int 128
%32456 = torch.prim.ListConstruct %int4_29901, %2482, %int4_29902, %int128_29903 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32457 = torch.aten.view %32309, %32456 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29904 = torch.constant.int 4
%int4_29905 = torch.constant.int 4
%int128_29906 = torch.constant.int 128
%32458 = torch.prim.ListConstruct %int4_29904, %2482, %int4_29905, %int128_29906 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32459 = torch.aten.view %32315, %32458 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_29907 = torch.constant.int 4
%int1_29908 = torch.constant.int 1
%int128_29909 = torch.constant.int 128
%32460 = torch.prim.ListConstruct %int4_29907, %2482, %int1_29908, %int128_29909 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32461 = torch.aten.view %32337, %32460 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29910 = torch.constant.int 4
%int1_29911 = torch.constant.int 1
%int128_29912 = torch.constant.int 128
%32462 = torch.prim.ListConstruct %int4_29910, %2482, %int1_29911, %int128_29912 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32463 = torch.aten.view %32343, %32462 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32463, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29913 = torch.constant.int 4
%int1_29914 = torch.constant.int 1
%int128_29915 = torch.constant.int 128
%32464 = torch.prim.ListConstruct %int4_29913, %2482, %int1_29914, %int128_29915 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32465 = torch.aten.view %32349, %32464 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29916 = torch.constant.int 4
%int1_29917 = torch.constant.int 1
%int128_29918 = torch.constant.int 128
%32466 = torch.prim.ListConstruct %int4_29916, %2482, %int1_29917, %int128_29918 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32467 = torch.aten.view %32355, %32466 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32467, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29919 = torch.constant.int 4
%int1_29920 = torch.constant.int 1
%int128_29921 = torch.constant.int 128
%32468 = torch.prim.ListConstruct %int4_29919, %2482, %int1_29920, %int128_29921 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32469 = torch.aten.view %32361, %32468 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29922 = torch.constant.int 4
%int1_29923 = torch.constant.int 1
%int128_29924 = torch.constant.int 128
%32470 = torch.prim.ListConstruct %int4_29922, %2482, %int1_29923, %int128_29924 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32471 = torch.aten.view %32367, %32470 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29925 = torch.constant.int 4
%int1_29926 = torch.constant.int 1
%int128_29927 = torch.constant.int 128
%32472 = torch.prim.ListConstruct %int4_29925, %2482, %int1_29926, %int128_29927 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32473 = torch.aten.view %32373, %32472 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29928 = torch.constant.int 4
%int1_29929 = torch.constant.int 1
%int128_29930 = torch.constant.int 128
%32474 = torch.prim.ListConstruct %int4_29928, %2482, %int1_29929, %int128_29930 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32475 = torch.aten.view %32379, %32474 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29931 = torch.constant.int 4
%int1_29932 = torch.constant.int 1
%int128_29933 = torch.constant.int 128
%32476 = torch.prim.ListConstruct %int4_29931, %2482, %int1_29932, %int128_29933 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32477 = torch.aten.view %32401, %32476 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29934 = torch.constant.int 4
%int1_29935 = torch.constant.int 1
%int128_29936 = torch.constant.int 128
%32478 = torch.prim.ListConstruct %int4_29934, %2482, %int1_29935, %int128_29936 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32479 = torch.aten.view %32407, %32478 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29937 = torch.constant.int 4
%int1_29938 = torch.constant.int 1
%int128_29939 = torch.constant.int 128
%32480 = torch.prim.ListConstruct %int4_29937, %2482, %int1_29938, %int128_29939 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32481 = torch.aten.view %32413, %32480 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29940 = torch.constant.int 4
%int1_29941 = torch.constant.int 1
%int128_29942 = torch.constant.int 128
%32482 = torch.prim.ListConstruct %int4_29940, %2482, %int1_29941, %int128_29942 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32483 = torch.aten.view %32419, %32482 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29943 = torch.constant.int 4
%int1_29944 = torch.constant.int 1
%int128_29945 = torch.constant.int 128
%32484 = torch.prim.ListConstruct %int4_29943, %2482, %int1_29944, %int128_29945 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32485 = torch.aten.view %32425, %32484 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32485, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29946 = torch.constant.int 4
%int1_29947 = torch.constant.int 1
%int128_29948 = torch.constant.int 128
%32486 = torch.prim.ListConstruct %int4_29946, %2482, %int1_29947, %int128_29948 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32487 = torch.aten.view %32431, %32486 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29949 = torch.constant.int 4
%int1_29950 = torch.constant.int 1
%int128_29951 = torch.constant.int 128
%32488 = torch.prim.ListConstruct %int4_29949, %2482, %int1_29950, %int128_29951 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32489 = torch.aten.view %32437, %32488 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32489, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_29952 = torch.constant.int 4
%int1_29953 = torch.constant.int 1
%int128_29954 = torch.constant.int 128
%32490 = torch.prim.ListConstruct %int4_29952, %2482, %int1_29953, %int128_29954 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32491 = torch.aten.view %32443, %32490 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_29955 = torch.constant.int 131072
%none_29956 = torch.constant.none
%none_29957 = torch.constant.none
%cpu_29958 = torch.constant.device "cpu"
%false_29959 = torch.constant.bool false
%32492 = torch.aten.arange %int131072_29955, %none_29956, %none_29957, %cpu_29958, %false_29959 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_29960 = torch.constant.int 0
%int128_29961 = torch.constant.int 128
%int2_29962 = torch.constant.int 2
%none_29963 = torch.constant.none
%none_29964 = torch.constant.none
%cpu_29965 = torch.constant.device "cpu"
%false_29966 = torch.constant.bool false
%32493 = torch.aten.arange.start_step %int0_29960, %int128_29961, %int2_29962, %none_29963, %none_29964, %cpu_29965, %false_29966 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_29967 = torch.constant.int 0
%int0_29968 = torch.constant.int 0
%int64_29969 = torch.constant.int 64
%int1_29970 = torch.constant.int 1
%32494 = torch.aten.slice.Tensor %32493, %int0_29967, %int0_29968, %int64_29969, %int1_29970 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_29971 = torch.constant.int 6
%32495 = torch.prims.convert_element_type %32494, %int6_29971 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_29972 = torch.constant.int 128
%32496 = torch.aten.div.Scalar %32495, %int128_29972 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_29973 = torch.constant.float 5.000000e+05
%32497 = torch.aten.pow.Scalar %float5.000000e05_29973, %32496 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%32498 = torch.aten.reciprocal %32497 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_29974 = torch.constant.float 1.000000e+00
%32499 = torch.aten.mul.Scalar %32498, %float1.000000e00_29974 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_29975 = torch.constant.int 131072
%int1_29976 = torch.constant.int 1
%32500 = torch.prim.ListConstruct %int131072_29975, %int1_29976 : (!torch.int, !torch.int) -> !torch.list<int>
%32501 = torch.aten.view %32492, %32500 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%32502 = torch.aten.mul.Tensor %32501, %32499 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%32503 = torch.aten.cos %32502 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%32504 = torch.aten.sin %32502 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%32505 = torch.aten.complex %32503, %32504 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%32506 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32507 = flow.tensor.transfer %32506 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%32508 = torch_c.from_builtin_tensor %32507 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32509 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32510 = flow.tensor.transfer %32509 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%32511 = torch_c.from_builtin_tensor %32510 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32512 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32513 = flow.tensor.transfer %32512 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%32514 = torch_c.from_builtin_tensor %32513 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32515 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32516 = flow.tensor.transfer %32515 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%32517 = torch_c.from_builtin_tensor %32516 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32518 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32519 = flow.tensor.transfer %32518 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%32520 = torch_c.from_builtin_tensor %32519 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32521 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32522 = flow.tensor.transfer %32521 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%32523 = torch_c.from_builtin_tensor %32522 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32524 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32525 = flow.tensor.transfer %32524 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%32526 = torch_c.from_builtin_tensor %32525 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32527 = torch_c.to_builtin_tensor %32505 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32528 = flow.tensor.transfer %32527 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%32529 = torch_c.from_builtin_tensor %32528 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_29977 = torch.constant.int 1
%32530 = torch.aten.size.int %32273, %int1_29977 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_29978 = torch.constant.int 0
%32531 = torch.aten.add.int %int0_29978, %32530 : !torch.int, !torch.int -> !torch.int
%int0_29979 = torch.constant.int 0
%int0_29980 = torch.constant.int 0
%int1_29981 = torch.constant.int 1
%32532 = torch.aten.slice.Tensor %32508, %int0_29979, %int0_29980, %32531, %int1_29981 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32532, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_29982 = torch.constant.int 1
%int0_29983 = torch.constant.int 0
%int9223372036854775807_29984 = torch.constant.int 9223372036854775807
%int1_29985 = torch.constant.int 1
%32533 = torch.aten.slice.Tensor %32532, %int1_29982, %int0_29983, %int9223372036854775807_29984, %int1_29985 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32533, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_29986 = torch.constant.int 0
%32534 = torch.aten.unsqueeze %32533, %int0_29986 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32534, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_29987 = torch.constant.int 2
%32535 = torch.aten.unsqueeze %32534, %int2_29987 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32535, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_29988 = torch.constant.int 3
%int0_29989 = torch.constant.int 0
%int9223372036854775807_29990 = torch.constant.int 9223372036854775807
%int1_29991 = torch.constant.int 1
%32536 = torch.aten.slice.Tensor %32535, %int3_29988, %int0_29989, %int9223372036854775807_29990, %int1_29991 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32536, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32537 = torch_c.to_builtin_tensor %32445 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_29992 = arith.constant 1 : index
%dim_29993 = tensor.dim %32537, %c1_29992 : tensor<4x?x4x128xf16>
%32538 = flow.tensor.bitcast %32537 : tensor<4x?x4x128xf16>{%dim_29993} -> tensor<4x?x4x64xcomplex<f16>>{%dim_29993}
%32539 = torch_c.from_builtin_tensor %32538 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32540 = torch.aten.mul.Tensor %32539, %32536 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32541 = torch_c.to_builtin_tensor %32540 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_29994 = arith.constant 1 : index
%dim_29995 = tensor.dim %32541, %c1_29994 : tensor<4x?x4x64xcomplex<f32>>
%32542 = flow.tensor.bitcast %32541 : tensor<4x?x4x64xcomplex<f32>>{%dim_29995} -> tensor<4x?x4x128xf32>{%dim_29995}
%32543 = torch_c.from_builtin_tensor %32542 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_29996 = torch.constant.int 5
%32544 = torch.prims.convert_element_type %32543, %int5_29996 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_29997 = torch.constant.int 1
%32545 = torch.aten.size.int %32279, %int1_29997 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_29998 = torch.constant.int 0
%32546 = torch.aten.add.int %int0_29998, %32545 : !torch.int, !torch.int -> !torch.int
%int0_29999 = torch.constant.int 0
%int0_30000 = torch.constant.int 0
%int1_30001 = torch.constant.int 1
%32547 = torch.aten.slice.Tensor %32511, %int0_29999, %int0_30000, %32546, %int1_30001 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32547, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30002 = torch.constant.int 1
%int0_30003 = torch.constant.int 0
%int9223372036854775807_30004 = torch.constant.int 9223372036854775807
%int1_30005 = torch.constant.int 1
%32548 = torch.aten.slice.Tensor %32547, %int1_30002, %int0_30003, %int9223372036854775807_30004, %int1_30005 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32548, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30006 = torch.constant.int 0
%32549 = torch.aten.unsqueeze %32548, %int0_30006 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32549, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30007 = torch.constant.int 2
%32550 = torch.aten.unsqueeze %32549, %int2_30007 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32550, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30008 = torch.constant.int 3
%int0_30009 = torch.constant.int 0
%int9223372036854775807_30010 = torch.constant.int 9223372036854775807
%int1_30011 = torch.constant.int 1
%32551 = torch.aten.slice.Tensor %32550, %int3_30008, %int0_30009, %int9223372036854775807_30010, %int1_30011 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32551, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32552 = torch_c.to_builtin_tensor %32447 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30012 = arith.constant 1 : index
%dim_30013 = tensor.dim %32552, %c1_30012 : tensor<4x?x4x128xf16>
%32553 = flow.tensor.bitcast %32552 : tensor<4x?x4x128xf16>{%dim_30013} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30013}
%32554 = torch_c.from_builtin_tensor %32553 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32555 = torch.aten.mul.Tensor %32554, %32551 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32556 = torch_c.to_builtin_tensor %32555 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30014 = arith.constant 1 : index
%dim_30015 = tensor.dim %32556, %c1_30014 : tensor<4x?x4x64xcomplex<f32>>
%32557 = flow.tensor.bitcast %32556 : tensor<4x?x4x64xcomplex<f32>>{%dim_30015} -> tensor<4x?x4x128xf32>{%dim_30015}
%32558 = torch_c.from_builtin_tensor %32557 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30016 = torch.constant.int 5
%32559 = torch.prims.convert_element_type %32558, %int5_30016 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30017 = torch.constant.int 1
%32560 = torch.aten.size.int %32285, %int1_30017 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_30018 = torch.constant.int 0
%32561 = torch.aten.add.int %int0_30018, %32560 : !torch.int, !torch.int -> !torch.int
%int0_30019 = torch.constant.int 0
%int0_30020 = torch.constant.int 0
%int1_30021 = torch.constant.int 1
%32562 = torch.aten.slice.Tensor %32514, %int0_30019, %int0_30020, %32561, %int1_30021 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32562, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30022 = torch.constant.int 1
%int0_30023 = torch.constant.int 0
%int9223372036854775807_30024 = torch.constant.int 9223372036854775807
%int1_30025 = torch.constant.int 1
%32563 = torch.aten.slice.Tensor %32562, %int1_30022, %int0_30023, %int9223372036854775807_30024, %int1_30025 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32563, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30026 = torch.constant.int 0
%32564 = torch.aten.unsqueeze %32563, %int0_30026 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32564, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30027 = torch.constant.int 2
%32565 = torch.aten.unsqueeze %32564, %int2_30027 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32565, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30028 = torch.constant.int 3
%int0_30029 = torch.constant.int 0
%int9223372036854775807_30030 = torch.constant.int 9223372036854775807
%int1_30031 = torch.constant.int 1
%32566 = torch.aten.slice.Tensor %32565, %int3_30028, %int0_30029, %int9223372036854775807_30030, %int1_30031 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32566, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32567 = torch_c.to_builtin_tensor %32449 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30032 = arith.constant 1 : index
%dim_30033 = tensor.dim %32567, %c1_30032 : tensor<4x?x4x128xf16>
%32568 = flow.tensor.bitcast %32567 : tensor<4x?x4x128xf16>{%dim_30033} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30033}
%32569 = torch_c.from_builtin_tensor %32568 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32570 = torch.aten.mul.Tensor %32569, %32566 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32571 = torch_c.to_builtin_tensor %32570 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30034 = arith.constant 1 : index
%dim_30035 = tensor.dim %32571, %c1_30034 : tensor<4x?x4x64xcomplex<f32>>
%32572 = flow.tensor.bitcast %32571 : tensor<4x?x4x64xcomplex<f32>>{%dim_30035} -> tensor<4x?x4x128xf32>{%dim_30035}
%32573 = torch_c.from_builtin_tensor %32572 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30036 = torch.constant.int 5
%32574 = torch.prims.convert_element_type %32573, %int5_30036 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30037 = torch.constant.int 1
%32575 = torch.aten.size.int %32291, %int1_30037 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_30038 = torch.constant.int 0
%32576 = torch.aten.add.int %int0_30038, %32575 : !torch.int, !torch.int -> !torch.int
%int0_30039 = torch.constant.int 0
%int0_30040 = torch.constant.int 0
%int1_30041 = torch.constant.int 1
%32577 = torch.aten.slice.Tensor %32517, %int0_30039, %int0_30040, %32576, %int1_30041 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32577, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30042 = torch.constant.int 1
%int0_30043 = torch.constant.int 0
%int9223372036854775807_30044 = torch.constant.int 9223372036854775807
%int1_30045 = torch.constant.int 1
%32578 = torch.aten.slice.Tensor %32577, %int1_30042, %int0_30043, %int9223372036854775807_30044, %int1_30045 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32578, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30046 = torch.constant.int 0
%32579 = torch.aten.unsqueeze %32578, %int0_30046 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32579, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30047 = torch.constant.int 2
%32580 = torch.aten.unsqueeze %32579, %int2_30047 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32580, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30048 = torch.constant.int 3
%int0_30049 = torch.constant.int 0
%int9223372036854775807_30050 = torch.constant.int 9223372036854775807
%int1_30051 = torch.constant.int 1
%32581 = torch.aten.slice.Tensor %32580, %int3_30048, %int0_30049, %int9223372036854775807_30050, %int1_30051 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32581, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32582 = torch_c.to_builtin_tensor %32451 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30052 = arith.constant 1 : index
%dim_30053 = tensor.dim %32582, %c1_30052 : tensor<4x?x4x128xf16>
%32583 = flow.tensor.bitcast %32582 : tensor<4x?x4x128xf16>{%dim_30053} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30053}
%32584 = torch_c.from_builtin_tensor %32583 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32585 = torch.aten.mul.Tensor %32584, %32581 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32586 = torch_c.to_builtin_tensor %32585 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30054 = arith.constant 1 : index
%dim_30055 = tensor.dim %32586, %c1_30054 : tensor<4x?x4x64xcomplex<f32>>
%32587 = flow.tensor.bitcast %32586 : tensor<4x?x4x64xcomplex<f32>>{%dim_30055} -> tensor<4x?x4x128xf32>{%dim_30055}
%32588 = torch_c.from_builtin_tensor %32587 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30056 = torch.constant.int 5
%32589 = torch.prims.convert_element_type %32588, %int5_30056 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30057 = torch.constant.int 1
%32590 = torch.aten.size.int %32297, %int1_30057 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_30058 = torch.constant.int 0
%32591 = torch.aten.add.int %int0_30058, %32590 : !torch.int, !torch.int -> !torch.int
%int0_30059 = torch.constant.int 0
%int0_30060 = torch.constant.int 0
%int1_30061 = torch.constant.int 1
%32592 = torch.aten.slice.Tensor %32520, %int0_30059, %int0_30060, %32591, %int1_30061 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32592, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30062 = torch.constant.int 1
%int0_30063 = torch.constant.int 0
%int9223372036854775807_30064 = torch.constant.int 9223372036854775807
%int1_30065 = torch.constant.int 1
%32593 = torch.aten.slice.Tensor %32592, %int1_30062, %int0_30063, %int9223372036854775807_30064, %int1_30065 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32593, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30066 = torch.constant.int 0
%32594 = torch.aten.unsqueeze %32593, %int0_30066 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32594, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30067 = torch.constant.int 2
%32595 = torch.aten.unsqueeze %32594, %int2_30067 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32595, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30068 = torch.constant.int 3
%int0_30069 = torch.constant.int 0
%int9223372036854775807_30070 = torch.constant.int 9223372036854775807
%int1_30071 = torch.constant.int 1
%32596 = torch.aten.slice.Tensor %32595, %int3_30068, %int0_30069, %int9223372036854775807_30070, %int1_30071 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32596, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32597 = torch_c.to_builtin_tensor %32453 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30072 = arith.constant 1 : index
%dim_30073 = tensor.dim %32597, %c1_30072 : tensor<4x?x4x128xf16>
%32598 = flow.tensor.bitcast %32597 : tensor<4x?x4x128xf16>{%dim_30073} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30073}
%32599 = torch_c.from_builtin_tensor %32598 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32600 = torch.aten.mul.Tensor %32599, %32596 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32600, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32601 = torch_c.to_builtin_tensor %32600 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30074 = arith.constant 1 : index
%dim_30075 = tensor.dim %32601, %c1_30074 : tensor<4x?x4x64xcomplex<f32>>
%32602 = flow.tensor.bitcast %32601 : tensor<4x?x4x64xcomplex<f32>>{%dim_30075} -> tensor<4x?x4x128xf32>{%dim_30075}
%32603 = torch_c.from_builtin_tensor %32602 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30076 = torch.constant.int 5
%32604 = torch.prims.convert_element_type %32603, %int5_30076 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30077 = torch.constant.int 1
%32605 = torch.aten.size.int %32303, %int1_30077 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_30078 = torch.constant.int 0
%32606 = torch.aten.add.int %int0_30078, %32605 : !torch.int, !torch.int -> !torch.int
%int0_30079 = torch.constant.int 0
%int0_30080 = torch.constant.int 0
%int1_30081 = torch.constant.int 1
%32607 = torch.aten.slice.Tensor %32523, %int0_30079, %int0_30080, %32606, %int1_30081 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32607, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30082 = torch.constant.int 1
%int0_30083 = torch.constant.int 0
%int9223372036854775807_30084 = torch.constant.int 9223372036854775807
%int1_30085 = torch.constant.int 1
%32608 = torch.aten.slice.Tensor %32607, %int1_30082, %int0_30083, %int9223372036854775807_30084, %int1_30085 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32608, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30086 = torch.constant.int 0
%32609 = torch.aten.unsqueeze %32608, %int0_30086 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32609, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30087 = torch.constant.int 2
%32610 = torch.aten.unsqueeze %32609, %int2_30087 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32610, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30088 = torch.constant.int 3
%int0_30089 = torch.constant.int 0
%int9223372036854775807_30090 = torch.constant.int 9223372036854775807
%int1_30091 = torch.constant.int 1
%32611 = torch.aten.slice.Tensor %32610, %int3_30088, %int0_30089, %int9223372036854775807_30090, %int1_30091 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32611, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32612 = torch_c.to_builtin_tensor %32455 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30092 = arith.constant 1 : index
%dim_30093 = tensor.dim %32612, %c1_30092 : tensor<4x?x4x128xf16>
%32613 = flow.tensor.bitcast %32612 : tensor<4x?x4x128xf16>{%dim_30093} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30093}
%32614 = torch_c.from_builtin_tensor %32613 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32615 = torch.aten.mul.Tensor %32614, %32611 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32616 = torch_c.to_builtin_tensor %32615 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30094 = arith.constant 1 : index
%dim_30095 = tensor.dim %32616, %c1_30094 : tensor<4x?x4x64xcomplex<f32>>
%32617 = flow.tensor.bitcast %32616 : tensor<4x?x4x64xcomplex<f32>>{%dim_30095} -> tensor<4x?x4x128xf32>{%dim_30095}
%32618 = torch_c.from_builtin_tensor %32617 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30096 = torch.constant.int 5
%32619 = torch.prims.convert_element_type %32618, %int5_30096 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30097 = torch.constant.int 1
%32620 = torch.aten.size.int %32309, %int1_30097 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_30098 = torch.constant.int 0
%32621 = torch.aten.add.int %int0_30098, %32620 : !torch.int, !torch.int -> !torch.int
%int0_30099 = torch.constant.int 0
%int0_30100 = torch.constant.int 0
%int1_30101 = torch.constant.int 1
%32622 = torch.aten.slice.Tensor %32526, %int0_30099, %int0_30100, %32621, %int1_30101 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32622, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30102 = torch.constant.int 1
%int0_30103 = torch.constant.int 0
%int9223372036854775807_30104 = torch.constant.int 9223372036854775807
%int1_30105 = torch.constant.int 1
%32623 = torch.aten.slice.Tensor %32622, %int1_30102, %int0_30103, %int9223372036854775807_30104, %int1_30105 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32623, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30106 = torch.constant.int 0
%32624 = torch.aten.unsqueeze %32623, %int0_30106 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32624, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30107 = torch.constant.int 2
%32625 = torch.aten.unsqueeze %32624, %int2_30107 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32625, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30108 = torch.constant.int 3
%int0_30109 = torch.constant.int 0
%int9223372036854775807_30110 = torch.constant.int 9223372036854775807
%int1_30111 = torch.constant.int 1
%32626 = torch.aten.slice.Tensor %32625, %int3_30108, %int0_30109, %int9223372036854775807_30110, %int1_30111 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32626, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32627 = torch_c.to_builtin_tensor %32457 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30112 = arith.constant 1 : index
%dim_30113 = tensor.dim %32627, %c1_30112 : tensor<4x?x4x128xf16>
%32628 = flow.tensor.bitcast %32627 : tensor<4x?x4x128xf16>{%dim_30113} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30113}
%32629 = torch_c.from_builtin_tensor %32628 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32630 = torch.aten.mul.Tensor %32629, %32626 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32631 = torch_c.to_builtin_tensor %32630 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30114 = arith.constant 1 : index
%dim_30115 = tensor.dim %32631, %c1_30114 : tensor<4x?x4x64xcomplex<f32>>
%32632 = flow.tensor.bitcast %32631 : tensor<4x?x4x64xcomplex<f32>>{%dim_30115} -> tensor<4x?x4x128xf32>{%dim_30115}
%32633 = torch_c.from_builtin_tensor %32632 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30116 = torch.constant.int 5
%32634 = torch.prims.convert_element_type %32633, %int5_30116 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30117 = torch.constant.int 1
%32635 = torch.aten.size.int %32315, %int1_30117 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_30118 = torch.constant.int 0
%32636 = torch.aten.add.int %int0_30118, %32635 : !torch.int, !torch.int -> !torch.int
%int0_30119 = torch.constant.int 0
%int0_30120 = torch.constant.int 0
%int1_30121 = torch.constant.int 1
%32637 = torch.aten.slice.Tensor %32529, %int0_30119, %int0_30120, %32636, %int1_30121 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32637, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30122 = torch.constant.int 1
%int0_30123 = torch.constant.int 0
%int9223372036854775807_30124 = torch.constant.int 9223372036854775807
%int1_30125 = torch.constant.int 1
%32638 = torch.aten.slice.Tensor %32637, %int1_30122, %int0_30123, %int9223372036854775807_30124, %int1_30125 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32638, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30126 = torch.constant.int 0
%32639 = torch.aten.unsqueeze %32638, %int0_30126 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32639, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30127 = torch.constant.int 2
%32640 = torch.aten.unsqueeze %32639, %int2_30127 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32640, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30128 = torch.constant.int 3
%int0_30129 = torch.constant.int 0
%int9223372036854775807_30130 = torch.constant.int 9223372036854775807
%int1_30131 = torch.constant.int 1
%32641 = torch.aten.slice.Tensor %32640, %int3_30128, %int0_30129, %int9223372036854775807_30130, %int1_30131 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32641, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32642 = torch_c.to_builtin_tensor %32459 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_30132 = arith.constant 1 : index
%dim_30133 = tensor.dim %32642, %c1_30132 : tensor<4x?x4x128xf16>
%32643 = flow.tensor.bitcast %32642 : tensor<4x?x4x128xf16>{%dim_30133} -> tensor<4x?x4x64xcomplex<f16>>{%dim_30133}
%32644 = torch_c.from_builtin_tensor %32643 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %32644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%32645 = torch.aten.mul.Tensor %32644, %32641 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %32645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%32646 = torch_c.to_builtin_tensor %32645 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_30134 = arith.constant 1 : index
%dim_30135 = tensor.dim %32646, %c1_30134 : tensor<4x?x4x64xcomplex<f32>>
%32647 = flow.tensor.bitcast %32646 : tensor<4x?x4x64xcomplex<f32>>{%dim_30135} -> tensor<4x?x4x128xf32>{%dim_30135}
%32648 = torch_c.from_builtin_tensor %32647 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %32648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_30136 = torch.constant.int 5
%32649 = torch.prims.convert_element_type %32648, %int5_30136 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %32649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_30137 = torch.constant.int 131072
%none_30138 = torch.constant.none
%none_30139 = torch.constant.none
%cpu_30140 = torch.constant.device "cpu"
%false_30141 = torch.constant.bool false
%32650 = torch.aten.arange %int131072_30137, %none_30138, %none_30139, %cpu_30140, %false_30141 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_30142 = torch.constant.int 0
%int128_30143 = torch.constant.int 128
%int2_30144 = torch.constant.int 2
%none_30145 = torch.constant.none
%none_30146 = torch.constant.none
%cpu_30147 = torch.constant.device "cpu"
%false_30148 = torch.constant.bool false
%32651 = torch.aten.arange.start_step %int0_30142, %int128_30143, %int2_30144, %none_30145, %none_30146, %cpu_30147, %false_30148 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_30149 = torch.constant.int 0
%int0_30150 = torch.constant.int 0
%int64_30151 = torch.constant.int 64
%int1_30152 = torch.constant.int 1
%32652 = torch.aten.slice.Tensor %32651, %int0_30149, %int0_30150, %int64_30151, %int1_30152 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_30153 = torch.constant.int 6
%32653 = torch.prims.convert_element_type %32652, %int6_30153 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_30154 = torch.constant.int 128
%32654 = torch.aten.div.Scalar %32653, %int128_30154 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_30155 = torch.constant.float 5.000000e+05
%32655 = torch.aten.pow.Scalar %float5.000000e05_30155, %32654 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%32656 = torch.aten.reciprocal %32655 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_30156 = torch.constant.float 1.000000e+00
%32657 = torch.aten.mul.Scalar %32656, %float1.000000e00_30156 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_30157 = torch.constant.int 131072
%int1_30158 = torch.constant.int 1
%32658 = torch.prim.ListConstruct %int131072_30157, %int1_30158 : (!torch.int, !torch.int) -> !torch.list<int>
%32659 = torch.aten.view %32650, %32658 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%32660 = torch.aten.mul.Tensor %32659, %32657 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%32661 = torch.aten.cos %32660 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%32662 = torch.aten.sin %32660 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%32663 = torch.aten.complex %32661, %32662 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%32664 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32665 = flow.tensor.transfer %32664 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%32666 = torch_c.from_builtin_tensor %32665 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32667 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32668 = flow.tensor.transfer %32667 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%32669 = torch_c.from_builtin_tensor %32668 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32670 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32671 = flow.tensor.transfer %32670 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%32672 = torch_c.from_builtin_tensor %32671 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32673 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32674 = flow.tensor.transfer %32673 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%32675 = torch_c.from_builtin_tensor %32674 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32676 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32677 = flow.tensor.transfer %32676 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%32678 = torch_c.from_builtin_tensor %32677 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32679 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32680 = flow.tensor.transfer %32679 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%32681 = torch_c.from_builtin_tensor %32680 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32682 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32683 = flow.tensor.transfer %32682 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%32684 = torch_c.from_builtin_tensor %32683 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%32685 = torch_c.to_builtin_tensor %32663 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%32686 = flow.tensor.transfer %32685 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%32687 = torch_c.from_builtin_tensor %32686 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_30159 = torch.constant.int 1
%32688 = torch.aten.size.int %32337, %int1_30159 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30160 = torch.constant.int 0
%32689 = torch.aten.add.int %int0_30160, %32688 : !torch.int, !torch.int -> !torch.int
%int0_30161 = torch.constant.int 0
%int0_30162 = torch.constant.int 0
%int1_30163 = torch.constant.int 1
%32690 = torch.aten.slice.Tensor %32666, %int0_30161, %int0_30162, %32689, %int1_30163 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32690, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30164 = torch.constant.int 1
%int0_30165 = torch.constant.int 0
%int9223372036854775807_30166 = torch.constant.int 9223372036854775807
%int1_30167 = torch.constant.int 1
%32691 = torch.aten.slice.Tensor %32690, %int1_30164, %int0_30165, %int9223372036854775807_30166, %int1_30167 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32691, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30168 = torch.constant.int 0
%32692 = torch.aten.unsqueeze %32691, %int0_30168 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32692, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30169 = torch.constant.int 2
%32693 = torch.aten.unsqueeze %32692, %int2_30169 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32693, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30170 = torch.constant.int 3
%int0_30171 = torch.constant.int 0
%int9223372036854775807_30172 = torch.constant.int 9223372036854775807
%int1_30173 = torch.constant.int 1
%32694 = torch.aten.slice.Tensor %32693, %int3_30170, %int0_30171, %int9223372036854775807_30172, %int1_30173 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32694, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32695 = torch_c.to_builtin_tensor %32461 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30174 = arith.constant 1 : index
%dim_30175 = tensor.dim %32695, %c1_30174 : tensor<4x?x1x128xf16>
%32696 = flow.tensor.bitcast %32695 : tensor<4x?x1x128xf16>{%dim_30175} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30175}
%32697 = torch_c.from_builtin_tensor %32696 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32698 = torch.aten.mul.Tensor %32697, %32694 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32699 = torch_c.to_builtin_tensor %32698 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30176 = arith.constant 1 : index
%dim_30177 = tensor.dim %32699, %c1_30176 : tensor<4x?x1x64xcomplex<f32>>
%32700 = flow.tensor.bitcast %32699 : tensor<4x?x1x64xcomplex<f32>>{%dim_30177} -> tensor<4x?x1x128xf32>{%dim_30177}
%32701 = torch_c.from_builtin_tensor %32700 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30178 = torch.constant.int 5
%32702 = torch.prims.convert_element_type %32701, %int5_30178 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30179 = torch.constant.int 1
%32703 = torch.aten.size.int %32343, %int1_30179 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30180 = torch.constant.int 0
%32704 = torch.aten.add.int %int0_30180, %32703 : !torch.int, !torch.int -> !torch.int
%int0_30181 = torch.constant.int 0
%int0_30182 = torch.constant.int 0
%int1_30183 = torch.constant.int 1
%32705 = torch.aten.slice.Tensor %32669, %int0_30181, %int0_30182, %32704, %int1_30183 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32705, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30184 = torch.constant.int 1
%int0_30185 = torch.constant.int 0
%int9223372036854775807_30186 = torch.constant.int 9223372036854775807
%int1_30187 = torch.constant.int 1
%32706 = torch.aten.slice.Tensor %32705, %int1_30184, %int0_30185, %int9223372036854775807_30186, %int1_30187 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32706, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30188 = torch.constant.int 0
%32707 = torch.aten.unsqueeze %32706, %int0_30188 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32707, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30189 = torch.constant.int 2
%32708 = torch.aten.unsqueeze %32707, %int2_30189 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32708, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30190 = torch.constant.int 3
%int0_30191 = torch.constant.int 0
%int9223372036854775807_30192 = torch.constant.int 9223372036854775807
%int1_30193 = torch.constant.int 1
%32709 = torch.aten.slice.Tensor %32708, %int3_30190, %int0_30191, %int9223372036854775807_30192, %int1_30193 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32709, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32710 = torch_c.to_builtin_tensor %32463 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30194 = arith.constant 1 : index
%dim_30195 = tensor.dim %32710, %c1_30194 : tensor<4x?x1x128xf16>
%32711 = flow.tensor.bitcast %32710 : tensor<4x?x1x128xf16>{%dim_30195} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30195}
%32712 = torch_c.from_builtin_tensor %32711 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32713 = torch.aten.mul.Tensor %32712, %32709 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32714 = torch_c.to_builtin_tensor %32713 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30196 = arith.constant 1 : index
%dim_30197 = tensor.dim %32714, %c1_30196 : tensor<4x?x1x64xcomplex<f32>>
%32715 = flow.tensor.bitcast %32714 : tensor<4x?x1x64xcomplex<f32>>{%dim_30197} -> tensor<4x?x1x128xf32>{%dim_30197}
%32716 = torch_c.from_builtin_tensor %32715 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30198 = torch.constant.int 5
%32717 = torch.prims.convert_element_type %32716, %int5_30198 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30199 = torch.constant.int 1
%32718 = torch.aten.size.int %32349, %int1_30199 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30200 = torch.constant.int 0
%32719 = torch.aten.add.int %int0_30200, %32718 : !torch.int, !torch.int -> !torch.int
%int0_30201 = torch.constant.int 0
%int0_30202 = torch.constant.int 0
%int1_30203 = torch.constant.int 1
%32720 = torch.aten.slice.Tensor %32672, %int0_30201, %int0_30202, %32719, %int1_30203 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32720, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30204 = torch.constant.int 1
%int0_30205 = torch.constant.int 0
%int9223372036854775807_30206 = torch.constant.int 9223372036854775807
%int1_30207 = torch.constant.int 1
%32721 = torch.aten.slice.Tensor %32720, %int1_30204, %int0_30205, %int9223372036854775807_30206, %int1_30207 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32721, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30208 = torch.constant.int 0
%32722 = torch.aten.unsqueeze %32721, %int0_30208 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32722, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30209 = torch.constant.int 2
%32723 = torch.aten.unsqueeze %32722, %int2_30209 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32723, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30210 = torch.constant.int 3
%int0_30211 = torch.constant.int 0
%int9223372036854775807_30212 = torch.constant.int 9223372036854775807
%int1_30213 = torch.constant.int 1
%32724 = torch.aten.slice.Tensor %32723, %int3_30210, %int0_30211, %int9223372036854775807_30212, %int1_30213 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32724, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32725 = torch_c.to_builtin_tensor %32465 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30214 = arith.constant 1 : index
%dim_30215 = tensor.dim %32725, %c1_30214 : tensor<4x?x1x128xf16>
%32726 = flow.tensor.bitcast %32725 : tensor<4x?x1x128xf16>{%dim_30215} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30215}
%32727 = torch_c.from_builtin_tensor %32726 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32728 = torch.aten.mul.Tensor %32727, %32724 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32729 = torch_c.to_builtin_tensor %32728 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30216 = arith.constant 1 : index
%dim_30217 = tensor.dim %32729, %c1_30216 : tensor<4x?x1x64xcomplex<f32>>
%32730 = flow.tensor.bitcast %32729 : tensor<4x?x1x64xcomplex<f32>>{%dim_30217} -> tensor<4x?x1x128xf32>{%dim_30217}
%32731 = torch_c.from_builtin_tensor %32730 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30218 = torch.constant.int 5
%32732 = torch.prims.convert_element_type %32731, %int5_30218 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30219 = torch.constant.int 1
%32733 = torch.aten.size.int %32355, %int1_30219 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30220 = torch.constant.int 0
%32734 = torch.aten.add.int %int0_30220, %32733 : !torch.int, !torch.int -> !torch.int
%int0_30221 = torch.constant.int 0
%int0_30222 = torch.constant.int 0
%int1_30223 = torch.constant.int 1
%32735 = torch.aten.slice.Tensor %32675, %int0_30221, %int0_30222, %32734, %int1_30223 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32735, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30224 = torch.constant.int 1
%int0_30225 = torch.constant.int 0
%int9223372036854775807_30226 = torch.constant.int 9223372036854775807
%int1_30227 = torch.constant.int 1
%32736 = torch.aten.slice.Tensor %32735, %int1_30224, %int0_30225, %int9223372036854775807_30226, %int1_30227 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32736, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30228 = torch.constant.int 0
%32737 = torch.aten.unsqueeze %32736, %int0_30228 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32737, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30229 = torch.constant.int 2
%32738 = torch.aten.unsqueeze %32737, %int2_30229 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32738, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30230 = torch.constant.int 3
%int0_30231 = torch.constant.int 0
%int9223372036854775807_30232 = torch.constant.int 9223372036854775807
%int1_30233 = torch.constant.int 1
%32739 = torch.aten.slice.Tensor %32738, %int3_30230, %int0_30231, %int9223372036854775807_30232, %int1_30233 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32739, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32740 = torch_c.to_builtin_tensor %32467 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30234 = arith.constant 1 : index
%dim_30235 = tensor.dim %32740, %c1_30234 : tensor<4x?x1x128xf16>
%32741 = flow.tensor.bitcast %32740 : tensor<4x?x1x128xf16>{%dim_30235} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30235}
%32742 = torch_c.from_builtin_tensor %32741 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32742, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32743 = torch.aten.mul.Tensor %32742, %32739 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32744 = torch_c.to_builtin_tensor %32743 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30236 = arith.constant 1 : index
%dim_30237 = tensor.dim %32744, %c1_30236 : tensor<4x?x1x64xcomplex<f32>>
%32745 = flow.tensor.bitcast %32744 : tensor<4x?x1x64xcomplex<f32>>{%dim_30237} -> tensor<4x?x1x128xf32>{%dim_30237}
%32746 = torch_c.from_builtin_tensor %32745 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30238 = torch.constant.int 5
%32747 = torch.prims.convert_element_type %32746, %int5_30238 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30239 = torch.constant.int 1
%32748 = torch.aten.size.int %32361, %int1_30239 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30240 = torch.constant.int 0
%32749 = torch.aten.add.int %int0_30240, %32748 : !torch.int, !torch.int -> !torch.int
%int0_30241 = torch.constant.int 0
%int0_30242 = torch.constant.int 0
%int1_30243 = torch.constant.int 1
%32750 = torch.aten.slice.Tensor %32678, %int0_30241, %int0_30242, %32749, %int1_30243 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32750, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30244 = torch.constant.int 1
%int0_30245 = torch.constant.int 0
%int9223372036854775807_30246 = torch.constant.int 9223372036854775807
%int1_30247 = torch.constant.int 1
%32751 = torch.aten.slice.Tensor %32750, %int1_30244, %int0_30245, %int9223372036854775807_30246, %int1_30247 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32751, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30248 = torch.constant.int 0
%32752 = torch.aten.unsqueeze %32751, %int0_30248 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32752, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30249 = torch.constant.int 2
%32753 = torch.aten.unsqueeze %32752, %int2_30249 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32753, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30250 = torch.constant.int 3
%int0_30251 = torch.constant.int 0
%int9223372036854775807_30252 = torch.constant.int 9223372036854775807
%int1_30253 = torch.constant.int 1
%32754 = torch.aten.slice.Tensor %32753, %int3_30250, %int0_30251, %int9223372036854775807_30252, %int1_30253 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32754, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32755 = torch_c.to_builtin_tensor %32469 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30254 = arith.constant 1 : index
%dim_30255 = tensor.dim %32755, %c1_30254 : tensor<4x?x1x128xf16>
%32756 = flow.tensor.bitcast %32755 : tensor<4x?x1x128xf16>{%dim_30255} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30255}
%32757 = torch_c.from_builtin_tensor %32756 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32758 = torch.aten.mul.Tensor %32757, %32754 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32759 = torch_c.to_builtin_tensor %32758 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30256 = arith.constant 1 : index
%dim_30257 = tensor.dim %32759, %c1_30256 : tensor<4x?x1x64xcomplex<f32>>
%32760 = flow.tensor.bitcast %32759 : tensor<4x?x1x64xcomplex<f32>>{%dim_30257} -> tensor<4x?x1x128xf32>{%dim_30257}
%32761 = torch_c.from_builtin_tensor %32760 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30258 = torch.constant.int 5
%32762 = torch.prims.convert_element_type %32761, %int5_30258 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30259 = torch.constant.int 1
%32763 = torch.aten.size.int %32367, %int1_30259 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30260 = torch.constant.int 0
%32764 = torch.aten.add.int %int0_30260, %32763 : !torch.int, !torch.int -> !torch.int
%int0_30261 = torch.constant.int 0
%int0_30262 = torch.constant.int 0
%int1_30263 = torch.constant.int 1
%32765 = torch.aten.slice.Tensor %32681, %int0_30261, %int0_30262, %32764, %int1_30263 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32765, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30264 = torch.constant.int 1
%int0_30265 = torch.constant.int 0
%int9223372036854775807_30266 = torch.constant.int 9223372036854775807
%int1_30267 = torch.constant.int 1
%32766 = torch.aten.slice.Tensor %32765, %int1_30264, %int0_30265, %int9223372036854775807_30266, %int1_30267 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32766, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30268 = torch.constant.int 0
%32767 = torch.aten.unsqueeze %32766, %int0_30268 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32767, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30269 = torch.constant.int 2
%32768 = torch.aten.unsqueeze %32767, %int2_30269 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32768, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30270 = torch.constant.int 3
%int0_30271 = torch.constant.int 0
%int9223372036854775807_30272 = torch.constant.int 9223372036854775807
%int1_30273 = torch.constant.int 1
%32769 = torch.aten.slice.Tensor %32768, %int3_30270, %int0_30271, %int9223372036854775807_30272, %int1_30273 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32769, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32770 = torch_c.to_builtin_tensor %32471 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30274 = arith.constant 1 : index
%dim_30275 = tensor.dim %32770, %c1_30274 : tensor<4x?x1x128xf16>
%32771 = flow.tensor.bitcast %32770 : tensor<4x?x1x128xf16>{%dim_30275} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30275}
%32772 = torch_c.from_builtin_tensor %32771 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32773 = torch.aten.mul.Tensor %32772, %32769 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32774 = torch_c.to_builtin_tensor %32773 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30276 = arith.constant 1 : index
%dim_30277 = tensor.dim %32774, %c1_30276 : tensor<4x?x1x64xcomplex<f32>>
%32775 = flow.tensor.bitcast %32774 : tensor<4x?x1x64xcomplex<f32>>{%dim_30277} -> tensor<4x?x1x128xf32>{%dim_30277}
%32776 = torch_c.from_builtin_tensor %32775 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30278 = torch.constant.int 5
%32777 = torch.prims.convert_element_type %32776, %int5_30278 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30279 = torch.constant.int 1
%32778 = torch.aten.size.int %32373, %int1_30279 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30280 = torch.constant.int 0
%32779 = torch.aten.add.int %int0_30280, %32778 : !torch.int, !torch.int -> !torch.int
%int0_30281 = torch.constant.int 0
%int0_30282 = torch.constant.int 0
%int1_30283 = torch.constant.int 1
%32780 = torch.aten.slice.Tensor %32684, %int0_30281, %int0_30282, %32779, %int1_30283 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32780, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30284 = torch.constant.int 1
%int0_30285 = torch.constant.int 0
%int9223372036854775807_30286 = torch.constant.int 9223372036854775807
%int1_30287 = torch.constant.int 1
%32781 = torch.aten.slice.Tensor %32780, %int1_30284, %int0_30285, %int9223372036854775807_30286, %int1_30287 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32781, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30288 = torch.constant.int 0
%32782 = torch.aten.unsqueeze %32781, %int0_30288 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32782, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30289 = torch.constant.int 2
%32783 = torch.aten.unsqueeze %32782, %int2_30289 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32783, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30290 = torch.constant.int 3
%int0_30291 = torch.constant.int 0
%int9223372036854775807_30292 = torch.constant.int 9223372036854775807
%int1_30293 = torch.constant.int 1
%32784 = torch.aten.slice.Tensor %32783, %int3_30290, %int0_30291, %int9223372036854775807_30292, %int1_30293 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32784, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32785 = torch_c.to_builtin_tensor %32473 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30294 = arith.constant 1 : index
%dim_30295 = tensor.dim %32785, %c1_30294 : tensor<4x?x1x128xf16>
%32786 = flow.tensor.bitcast %32785 : tensor<4x?x1x128xf16>{%dim_30295} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30295}
%32787 = torch_c.from_builtin_tensor %32786 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32788 = torch.aten.mul.Tensor %32787, %32784 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32789 = torch_c.to_builtin_tensor %32788 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30296 = arith.constant 1 : index
%dim_30297 = tensor.dim %32789, %c1_30296 : tensor<4x?x1x64xcomplex<f32>>
%32790 = flow.tensor.bitcast %32789 : tensor<4x?x1x64xcomplex<f32>>{%dim_30297} -> tensor<4x?x1x128xf32>{%dim_30297}
%32791 = torch_c.from_builtin_tensor %32790 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30298 = torch.constant.int 5
%32792 = torch.prims.convert_element_type %32791, %int5_30298 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_30299 = torch.constant.int 1
%32793 = torch.aten.size.int %32379, %int1_30299 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_30300 = torch.constant.int 0
%32794 = torch.aten.add.int %int0_30300, %32793 : !torch.int, !torch.int -> !torch.int
%int0_30301 = torch.constant.int 0
%int0_30302 = torch.constant.int 0
%int1_30303 = torch.constant.int 1
%32795 = torch.aten.slice.Tensor %32687, %int0_30301, %int0_30302, %32794, %int1_30303 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32795, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_30304 = torch.constant.int 1
%int0_30305 = torch.constant.int 0
%int9223372036854775807_30306 = torch.constant.int 9223372036854775807
%int1_30307 = torch.constant.int 1
%32796 = torch.aten.slice.Tensor %32795, %int1_30304, %int0_30305, %int9223372036854775807_30306, %int1_30307 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %32796, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_30308 = torch.constant.int 0
%32797 = torch.aten.unsqueeze %32796, %int0_30308 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %32797, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_30309 = torch.constant.int 2
%32798 = torch.aten.unsqueeze %32797, %int2_30309 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32798, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_30310 = torch.constant.int 3
%int0_30311 = torch.constant.int 0
%int9223372036854775807_30312 = torch.constant.int 9223372036854775807
%int1_30313 = torch.constant.int 1
%32799 = torch.aten.slice.Tensor %32798, %int3_30310, %int0_30311, %int9223372036854775807_30312, %int1_30313 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32799, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%32800 = torch_c.to_builtin_tensor %32475 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_30314 = arith.constant 1 : index
%dim_30315 = tensor.dim %32800, %c1_30314 : tensor<4x?x1x128xf16>
%32801 = flow.tensor.bitcast %32800 : tensor<4x?x1x128xf16>{%dim_30315} -> tensor<4x?x1x64xcomplex<f16>>{%dim_30315}
%32802 = torch_c.from_builtin_tensor %32801 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %32802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%32803 = torch.aten.mul.Tensor %32802, %32799 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %32803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%32804 = torch_c.to_builtin_tensor %32803 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_30316 = arith.constant 1 : index
%dim_30317 = tensor.dim %32804, %c1_30316 : tensor<4x?x1x64xcomplex<f32>>
%32805 = flow.tensor.bitcast %32804 : tensor<4x?x1x64xcomplex<f32>>{%dim_30317} -> tensor<4x?x1x128xf32>{%dim_30317}
%32806 = torch_c.from_builtin_tensor %32805 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %32806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_30318 = torch.constant.int 5
%32807 = torch.prims.convert_element_type %32806, %int5_30318 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %32807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_30319 = torch.constant.int 64
%32808 = torch.aten.mul.Scalar %2364, %int64_30319 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32808, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30320 = torch.constant.int 64
%32809 = torch.aten.mul.Scalar %2367, %int64_30320 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32809, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30321 = torch.constant.int 64
%32810 = torch.aten.mul.Scalar %2370, %int64_30321 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32810, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30322 = torch.constant.int 64
%32811 = torch.aten.mul.Scalar %2373, %int64_30322 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32811, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30323 = torch.constant.int 64
%32812 = torch.aten.mul.Scalar %2376, %int64_30323 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32812, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30324 = torch.constant.int 64
%32813 = torch.aten.mul.Scalar %2379, %int64_30324 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32813, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30325 = torch.constant.int 64
%32814 = torch.aten.mul.Scalar %2382, %int64_30325 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32814, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_30326 = torch.constant.int 64
%32815 = torch.aten.mul.Scalar %2385, %int64_30326 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32815, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30327 = torch.constant.int 32
%int1_30328 = torch.constant.int 1
%32816 = torch.aten.add.Scalar %32808, %int32_30327, %int1_30328 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32816, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30329 = torch.constant.int 32
%int1_30330 = torch.constant.int 1
%32817 = torch.aten.add.Scalar %32809, %int32_30329, %int1_30330 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32817, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30331 = torch.constant.int 32
%int1_30332 = torch.constant.int 1
%32818 = torch.aten.add.Scalar %32810, %int32_30331, %int1_30332 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32818, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30333 = torch.constant.int 32
%int1_30334 = torch.constant.int 1
%32819 = torch.aten.add.Scalar %32811, %int32_30333, %int1_30334 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32819, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30335 = torch.constant.int 32
%int1_30336 = torch.constant.int 1
%32820 = torch.aten.add.Scalar %32812, %int32_30335, %int1_30336 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32820, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30337 = torch.constant.int 32
%int1_30338 = torch.constant.int 1
%32821 = torch.aten.add.Scalar %32813, %int32_30337, %int1_30338 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32821, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30339 = torch.constant.int 32
%int1_30340 = torch.constant.int 1
%32822 = torch.aten.add.Scalar %32814, %int32_30339, %int1_30340 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32822, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int32_30341 = torch.constant.int 32
%int1_30342 = torch.constant.int 1
%32823 = torch.aten.add.Scalar %32815, %int32_30341, %int1_30342 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32823, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_30343 = torch.constant.int 4
%int16_30344 = torch.constant.int 16
%int1_30345 = torch.constant.int 1
%int128_30346 = torch.constant.int 128
%32824 = torch.prim.ListConstruct %int4_30343, %3095, %int16_30344, %int1_30345, %int128_30346 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32825 = torch.aten.view %32702, %32824 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32825, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30347 = torch.constant.int 4
%int16_30348 = torch.constant.int 16
%int1_30349 = torch.constant.int 1
%int128_30350 = torch.constant.int 128
%32826 = torch.prim.ListConstruct %int4_30347, %3095, %int16_30348, %int1_30349, %int128_30350 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32827 = torch.aten.view %32717, %32826 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32827, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30351 = torch.constant.int 4
%int16_30352 = torch.constant.int 16
%int1_30353 = torch.constant.int 1
%int128_30354 = torch.constant.int 128
%32828 = torch.prim.ListConstruct %int4_30351, %3095, %int16_30352, %int1_30353, %int128_30354 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32829 = torch.aten.view %32732, %32828 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32829, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30355 = torch.constant.int 4
%int16_30356 = torch.constant.int 16
%int1_30357 = torch.constant.int 1
%int128_30358 = torch.constant.int 128
%32830 = torch.prim.ListConstruct %int4_30355, %3095, %int16_30356, %int1_30357, %int128_30358 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32831 = torch.aten.view %32747, %32830 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32831, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30359 = torch.constant.int 4
%int16_30360 = torch.constant.int 16
%int1_30361 = torch.constant.int 1
%int128_30362 = torch.constant.int 128
%32832 = torch.prim.ListConstruct %int4_30359, %3095, %int16_30360, %int1_30361, %int128_30362 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32833 = torch.aten.view %32762, %32832 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32833, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30363 = torch.constant.int 4
%int16_30364 = torch.constant.int 16
%int1_30365 = torch.constant.int 1
%int128_30366 = torch.constant.int 128
%32834 = torch.prim.ListConstruct %int4_30363, %3095, %int16_30364, %int1_30365, %int128_30366 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32835 = torch.aten.view %32777, %32834 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32835, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30367 = torch.constant.int 4
%int16_30368 = torch.constant.int 16
%int1_30369 = torch.constant.int 1
%int128_30370 = torch.constant.int 128
%32836 = torch.prim.ListConstruct %int4_30367, %3095, %int16_30368, %int1_30369, %int128_30370 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32837 = torch.aten.view %32792, %32836 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32837, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30371 = torch.constant.int 4
%int16_30372 = torch.constant.int 16
%int1_30373 = torch.constant.int 1
%int128_30374 = torch.constant.int 128
%32838 = torch.prim.ListConstruct %int4_30371, %3095, %int16_30372, %int1_30373, %int128_30374 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32839 = torch.aten.view %32807, %32838 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32839, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30375 = torch.constant.int 4
%32840 = torch.aten.mul.int %int4_30375, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30376 = torch.constant.int 16
%int1_30377 = torch.constant.int 1
%int128_30378 = torch.constant.int 128
%32841 = torch.prim.ListConstruct %32840, %int16_30376, %int1_30377, %int128_30378 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32842 = torch.aten.view %32825, %32841 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32842, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30379 = torch.constant.int 4
%32843 = torch.aten.mul.int %int4_30379, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30380 = torch.constant.int 16
%int1_30381 = torch.constant.int 1
%int128_30382 = torch.constant.int 128
%32844 = torch.prim.ListConstruct %32843, %int16_30380, %int1_30381, %int128_30382 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32845 = torch.aten.view %32827, %32844 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32845, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30383 = torch.constant.int 4
%32846 = torch.aten.mul.int %int4_30383, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30384 = torch.constant.int 16
%int1_30385 = torch.constant.int 1
%int128_30386 = torch.constant.int 128
%32847 = torch.prim.ListConstruct %32846, %int16_30384, %int1_30385, %int128_30386 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32848 = torch.aten.view %32829, %32847 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32848, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30387 = torch.constant.int 4
%32849 = torch.aten.mul.int %int4_30387, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30388 = torch.constant.int 16
%int1_30389 = torch.constant.int 1
%int128_30390 = torch.constant.int 128
%32850 = torch.prim.ListConstruct %32849, %int16_30388, %int1_30389, %int128_30390 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32851 = torch.aten.view %32831, %32850 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32851, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30391 = torch.constant.int 4
%32852 = torch.aten.mul.int %int4_30391, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30392 = torch.constant.int 16
%int1_30393 = torch.constant.int 1
%int128_30394 = torch.constant.int 128
%32853 = torch.prim.ListConstruct %32852, %int16_30392, %int1_30393, %int128_30394 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32854 = torch.aten.view %32833, %32853 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32854, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30395 = torch.constant.int 4
%32855 = torch.aten.mul.int %int4_30395, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30396 = torch.constant.int 16
%int1_30397 = torch.constant.int 1
%int128_30398 = torch.constant.int 128
%32856 = torch.prim.ListConstruct %32855, %int16_30396, %int1_30397, %int128_30398 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32857 = torch.aten.view %32835, %32856 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32857, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30399 = torch.constant.int 4
%32858 = torch.aten.mul.int %int4_30399, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30400 = torch.constant.int 16
%int1_30401 = torch.constant.int 1
%int128_30402 = torch.constant.int 128
%32859 = torch.prim.ListConstruct %32858, %int16_30400, %int1_30401, %int128_30402 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32860 = torch.aten.view %32837, %32859 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32860, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30403 = torch.constant.int 4
%32861 = torch.aten.mul.int %int4_30403, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30404 = torch.constant.int 16
%int1_30405 = torch.constant.int 1
%int128_30406 = torch.constant.int 128
%32862 = torch.prim.ListConstruct %32861, %int16_30404, %int1_30405, %int128_30406 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32863 = torch.aten.view %32839, %32862 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32863, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30407 = torch.constant.int 4
%32864 = torch.aten.mul.int %int4_30407, %3095 : !torch.int, !torch.int -> !torch.int
%32865 = torch.prim.ListConstruct %32864 : (!torch.int) -> !torch.list<int>
%32866 = torch.aten.view %32816, %32865 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32866, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30408 = torch.constant.int 4
%32867 = torch.aten.mul.int %int4_30408, %3095 : !torch.int, !torch.int -> !torch.int
%32868 = torch.prim.ListConstruct %32867 : (!torch.int) -> !torch.list<int>
%32869 = torch.aten.view %32817, %32868 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32869, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30409 = torch.constant.int 4
%32870 = torch.aten.mul.int %int4_30409, %3095 : !torch.int, !torch.int -> !torch.int
%32871 = torch.prim.ListConstruct %32870 : (!torch.int) -> !torch.list<int>
%32872 = torch.aten.view %32818, %32871 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32872, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30410 = torch.constant.int 4
%32873 = torch.aten.mul.int %int4_30410, %3095 : !torch.int, !torch.int -> !torch.int
%32874 = torch.prim.ListConstruct %32873 : (!torch.int) -> !torch.list<int>
%32875 = torch.aten.view %32819, %32874 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32875, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30411 = torch.constant.int 4
%32876 = torch.aten.mul.int %int4_30411, %3095 : !torch.int, !torch.int -> !torch.int
%32877 = torch.prim.ListConstruct %32876 : (!torch.int) -> !torch.list<int>
%32878 = torch.aten.view %32820, %32877 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32878, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30412 = torch.constant.int 4
%32879 = torch.aten.mul.int %int4_30412, %3095 : !torch.int, !torch.int -> !torch.int
%32880 = torch.prim.ListConstruct %32879 : (!torch.int) -> !torch.list<int>
%32881 = torch.aten.view %32821, %32880 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32881, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30413 = torch.constant.int 4
%32882 = torch.aten.mul.int %int4_30413, %3095 : !torch.int, !torch.int -> !torch.int
%32883 = torch.prim.ListConstruct %32882 : (!torch.int) -> !torch.list<int>
%32884 = torch.aten.view %32822, %32883 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32884, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30414 = torch.constant.int 4
%32885 = torch.aten.mul.int %int4_30414, %3095 : !torch.int, !torch.int -> !torch.int
%32886 = torch.prim.ListConstruct %32885 : (!torch.int) -> !torch.list<int>
%32887 = torch.aten.view %32823, %32886 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32887, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30415 = torch.constant.int 4
%int16_30416 = torch.constant.int 16
%int1_30417 = torch.constant.int 1
%int128_30418 = torch.constant.int 128
%32888 = torch.prim.ListConstruct %int4_30415, %3095, %int16_30416, %int1_30417, %int128_30418 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32889 = torch.aten.view %32477, %32888 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32889, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30419 = torch.constant.int 4
%int16_30420 = torch.constant.int 16
%int1_30421 = torch.constant.int 1
%int128_30422 = torch.constant.int 128
%32890 = torch.prim.ListConstruct %int4_30419, %3095, %int16_30420, %int1_30421, %int128_30422 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32891 = torch.aten.view %32479, %32890 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32891, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30423 = torch.constant.int 4
%int16_30424 = torch.constant.int 16
%int1_30425 = torch.constant.int 1
%int128_30426 = torch.constant.int 128
%32892 = torch.prim.ListConstruct %int4_30423, %3095, %int16_30424, %int1_30425, %int128_30426 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32893 = torch.aten.view %32481, %32892 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32893, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30427 = torch.constant.int 4
%int16_30428 = torch.constant.int 16
%int1_30429 = torch.constant.int 1
%int128_30430 = torch.constant.int 128
%32894 = torch.prim.ListConstruct %int4_30427, %3095, %int16_30428, %int1_30429, %int128_30430 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32895 = torch.aten.view %32483, %32894 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32895, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30431 = torch.constant.int 4
%int16_30432 = torch.constant.int 16
%int1_30433 = torch.constant.int 1
%int128_30434 = torch.constant.int 128
%32896 = torch.prim.ListConstruct %int4_30431, %3095, %int16_30432, %int1_30433, %int128_30434 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32897 = torch.aten.view %32485, %32896 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32897, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30435 = torch.constant.int 4
%int16_30436 = torch.constant.int 16
%int1_30437 = torch.constant.int 1
%int128_30438 = torch.constant.int 128
%32898 = torch.prim.ListConstruct %int4_30435, %3095, %int16_30436, %int1_30437, %int128_30438 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32899 = torch.aten.view %32487, %32898 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32899, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30439 = torch.constant.int 4
%int16_30440 = torch.constant.int 16
%int1_30441 = torch.constant.int 1
%int128_30442 = torch.constant.int 128
%32900 = torch.prim.ListConstruct %int4_30439, %3095, %int16_30440, %int1_30441, %int128_30442 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32901 = torch.aten.view %32489, %32900 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32901, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30443 = torch.constant.int 4
%int16_30444 = torch.constant.int 16
%int1_30445 = torch.constant.int 1
%int128_30446 = torch.constant.int 128
%32902 = torch.prim.ListConstruct %int4_30443, %3095, %int16_30444, %int1_30445, %int128_30446 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32903 = torch.aten.view %32491, %32902 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %32903, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_30447 = torch.constant.int 4
%32904 = torch.aten.mul.int %int4_30447, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30448 = torch.constant.int 16
%int1_30449 = torch.constant.int 1
%int128_30450 = torch.constant.int 128
%32905 = torch.prim.ListConstruct %32904, %int16_30448, %int1_30449, %int128_30450 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32906 = torch.aten.view %32889, %32905 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32906, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30451 = torch.constant.int 4
%32907 = torch.aten.mul.int %int4_30451, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30452 = torch.constant.int 16
%int1_30453 = torch.constant.int 1
%int128_30454 = torch.constant.int 128
%32908 = torch.prim.ListConstruct %32907, %int16_30452, %int1_30453, %int128_30454 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32909 = torch.aten.view %32891, %32908 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32909, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30455 = torch.constant.int 4
%32910 = torch.aten.mul.int %int4_30455, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30456 = torch.constant.int 16
%int1_30457 = torch.constant.int 1
%int128_30458 = torch.constant.int 128
%32911 = torch.prim.ListConstruct %32910, %int16_30456, %int1_30457, %int128_30458 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32912 = torch.aten.view %32893, %32911 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32912, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30459 = torch.constant.int 4
%32913 = torch.aten.mul.int %int4_30459, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30460 = torch.constant.int 16
%int1_30461 = torch.constant.int 1
%int128_30462 = torch.constant.int 128
%32914 = torch.prim.ListConstruct %32913, %int16_30460, %int1_30461, %int128_30462 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32915 = torch.aten.view %32895, %32914 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32915, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30463 = torch.constant.int 4
%32916 = torch.aten.mul.int %int4_30463, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30464 = torch.constant.int 16
%int1_30465 = torch.constant.int 1
%int128_30466 = torch.constant.int 128
%32917 = torch.prim.ListConstruct %32916, %int16_30464, %int1_30465, %int128_30466 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32918 = torch.aten.view %32897, %32917 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32918, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30467 = torch.constant.int 4
%32919 = torch.aten.mul.int %int4_30467, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30468 = torch.constant.int 16
%int1_30469 = torch.constant.int 1
%int128_30470 = torch.constant.int 128
%32920 = torch.prim.ListConstruct %32919, %int16_30468, %int1_30469, %int128_30470 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32921 = torch.aten.view %32899, %32920 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32921, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30471 = torch.constant.int 4
%32922 = torch.aten.mul.int %int4_30471, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30472 = torch.constant.int 16
%int1_30473 = torch.constant.int 1
%int128_30474 = torch.constant.int 128
%32923 = torch.prim.ListConstruct %32922, %int16_30472, %int1_30473, %int128_30474 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32924 = torch.aten.view %32901, %32923 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32924, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_30475 = torch.constant.int 4
%32925 = torch.aten.mul.int %int4_30475, %3095 : !torch.int, !torch.int -> !torch.int
%int16_30476 = torch.constant.int 16
%int1_30477 = torch.constant.int 1
%int128_30478 = torch.constant.int 128
%32926 = torch.prim.ListConstruct %32925, %int16_30476, %int1_30477, %int128_30478 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32927 = torch.aten.view %32903, %32926 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32927, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_30479 = torch.constant.int 1
%int1_30480 = torch.constant.int 1
%32928 = torch.aten.add.Scalar %32816, %int1_30479, %int1_30480 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32928, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30481 = torch.constant.int 1
%int1_30482 = torch.constant.int 1
%32929 = torch.aten.add.Scalar %32817, %int1_30481, %int1_30482 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32929, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30483 = torch.constant.int 1
%int1_30484 = torch.constant.int 1
%32930 = torch.aten.add.Scalar %32818, %int1_30483, %int1_30484 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32930, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30485 = torch.constant.int 1
%int1_30486 = torch.constant.int 1
%32931 = torch.aten.add.Scalar %32819, %int1_30485, %int1_30486 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32931, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30487 = torch.constant.int 1
%int1_30488 = torch.constant.int 1
%32932 = torch.aten.add.Scalar %32820, %int1_30487, %int1_30488 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32932, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30489 = torch.constant.int 1
%int1_30490 = torch.constant.int 1
%32933 = torch.aten.add.Scalar %32821, %int1_30489, %int1_30490 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32933, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30491 = torch.constant.int 1
%int1_30492 = torch.constant.int 1
%32934 = torch.aten.add.Scalar %32822, %int1_30491, %int1_30492 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32934, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_30493 = torch.constant.int 1
%int1_30494 = torch.constant.int 1
%32935 = torch.aten.add.Scalar %32823, %int1_30493, %int1_30494 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %32935, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_30495 = torch.constant.int 4
%32936 = torch.aten.mul.int %int4_30495, %3095 : !torch.int, !torch.int -> !torch.int
%32937 = torch.prim.ListConstruct %32936 : (!torch.int) -> !torch.list<int>
%32938 = torch.aten.view %32928, %32937 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32938, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30496 = torch.constant.int 4
%32939 = torch.aten.mul.int %int4_30496, %3095 : !torch.int, !torch.int -> !torch.int
%32940 = torch.prim.ListConstruct %32939 : (!torch.int) -> !torch.list<int>
%32941 = torch.aten.view %32929, %32940 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32941, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30497 = torch.constant.int 4
%32942 = torch.aten.mul.int %int4_30497, %3095 : !torch.int, !torch.int -> !torch.int
%32943 = torch.prim.ListConstruct %32942 : (!torch.int) -> !torch.list<int>
%32944 = torch.aten.view %32930, %32943 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32944, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30498 = torch.constant.int 4
%32945 = torch.aten.mul.int %int4_30498, %3095 : !torch.int, !torch.int -> !torch.int
%32946 = torch.prim.ListConstruct %32945 : (!torch.int) -> !torch.list<int>
%32947 = torch.aten.view %32931, %32946 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32947, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30499 = torch.constant.int 4
%32948 = torch.aten.mul.int %int4_30499, %3095 : !torch.int, !torch.int -> !torch.int
%32949 = torch.prim.ListConstruct %32948 : (!torch.int) -> !torch.list<int>
%32950 = torch.aten.view %32932, %32949 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32950, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30500 = torch.constant.int 4
%32951 = torch.aten.mul.int %int4_30500, %3095 : !torch.int, !torch.int -> !torch.int
%32952 = torch.prim.ListConstruct %32951 : (!torch.int) -> !torch.list<int>
%32953 = torch.aten.view %32933, %32952 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32953, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30501 = torch.constant.int 4
%32954 = torch.aten.mul.int %int4_30501, %3095 : !torch.int, !torch.int -> !torch.int
%32955 = torch.prim.ListConstruct %32954 : (!torch.int) -> !torch.list<int>
%32956 = torch.aten.view %32934, %32955 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32956, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_30502 = torch.constant.int 4
%32957 = torch.aten.mul.int %int4_30502, %3095 : !torch.int, !torch.int -> !torch.int
%32958 = torch.prim.ListConstruct %32957 : (!torch.int) -> !torch.list<int>
%32959 = torch.aten.view %32935, %32958 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32959, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%32960 = torch.prim.ListConstruct %32866, %32938 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30503 = torch.constant.int 0
%32961 = torch.aten.cat %32960, %int0_30503 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32961, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32962 = torch.prim.ListConstruct %32869, %32941 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30504 = torch.constant.int 0
%32963 = torch.aten.cat %32962, %int0_30504 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32963, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32964 = torch.prim.ListConstruct %32872, %32944 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30505 = torch.constant.int 0
%32965 = torch.aten.cat %32964, %int0_30505 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32965, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32966 = torch.prim.ListConstruct %32875, %32947 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30506 = torch.constant.int 0
%32967 = torch.aten.cat %32966, %int0_30506 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32967, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32968 = torch.prim.ListConstruct %32878, %32950 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30507 = torch.constant.int 0
%32969 = torch.aten.cat %32968, %int0_30507 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32969, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32970 = torch.prim.ListConstruct %32881, %32953 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30508 = torch.constant.int 0
%32971 = torch.aten.cat %32970, %int0_30508 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32971, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32972 = torch.prim.ListConstruct %32884, %32956 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30509 = torch.constant.int 0
%32973 = torch.aten.cat %32972, %int0_30509 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32973, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32974 = torch.prim.ListConstruct %32887, %32959 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_30510 = torch.constant.int 0
%32975 = torch.aten.cat %32974, %int0_30510 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %32975, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%32976 = torch.prim.ListConstruct %32842, %32906 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30511 = torch.constant.int 0
%32977 = torch.aten.cat %32976, %int0_30511 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32977, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32978 = torch.prim.ListConstruct %32845, %32909 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30512 = torch.constant.int 0
%32979 = torch.aten.cat %32978, %int0_30512 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32979, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32980 = torch.prim.ListConstruct %32848, %32912 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30513 = torch.constant.int 0
%32981 = torch.aten.cat %32980, %int0_30513 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32981, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32982 = torch.prim.ListConstruct %32851, %32915 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30514 = torch.constant.int 0
%32983 = torch.aten.cat %32982, %int0_30514 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32983, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32984 = torch.prim.ListConstruct %32854, %32918 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30515 = torch.constant.int 0
%32985 = torch.aten.cat %32984, %int0_30515 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32985, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32986 = torch.prim.ListConstruct %32857, %32921 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30516 = torch.constant.int 0
%32987 = torch.aten.cat %32986, %int0_30516 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32987, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32988 = torch.prim.ListConstruct %32860, %32924 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30517 = torch.constant.int 0
%32989 = torch.aten.cat %32988, %int0_30517 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32989, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32990 = torch.prim.ListConstruct %32863, %32927 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_30518 = torch.constant.int 0
%32991 = torch.aten.cat %32990, %int0_30518 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32991, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30519 = torch.constant.int 32
%int2_30520 = torch.constant.int 2
%int16_30521 = torch.constant.int 16
%int1_30522 = torch.constant.int 1
%int128_30523 = torch.constant.int 128
%32992 = torch.prim.ListConstruct %3023, %int32_30519, %int2_30520, %int16_30521, %int1_30522, %int128_30523 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32993 = torch.aten.view %31142, %32992 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %32993, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30524 = torch.constant.int 32
%32994 = torch.aten.mul.int %3023, %int32_30524 : !torch.int, !torch.int -> !torch.int
%int2_30525 = torch.constant.int 2
%32995 = torch.aten.mul.int %32994, %int2_30525 : !torch.int, !torch.int -> !torch.int
%int16_30526 = torch.constant.int 16
%int1_30527 = torch.constant.int 1
%int128_30528 = torch.constant.int 128
%32996 = torch.prim.ListConstruct %32995, %int16_30526, %int1_30527, %int128_30528 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%32997 = torch.aten.view %32993, %32996 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32997, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%32998 = torch.prim.ListConstruct %32961 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30529 = torch.constant.bool false
%32999 = torch.aten.index_put %32997, %32998, %32977, %false_30529 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %32999, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30530 = torch.constant.int 32
%int2_30531 = torch.constant.int 2
%int16_30532 = torch.constant.int 16
%int1_30533 = torch.constant.int 1
%int128_30534 = torch.constant.int 128
%33000 = torch.prim.ListConstruct %3023, %int32_30530, %int2_30531, %int16_30532, %int1_30533, %int128_30534 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33001 = torch.aten.view %32999, %33000 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33001, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30535 = torch.constant.int 131072
%33002 = torch.prim.ListConstruct %3023, %int131072_30535 : (!torch.int, !torch.int) -> !torch.list<int>
%33003 = torch.aten.view %33001, %33002 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33003, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30536 = torch.constant.int 32
%int2_30537 = torch.constant.int 2
%int16_30538 = torch.constant.int 16
%int1_30539 = torch.constant.int 1
%int128_30540 = torch.constant.int 128
%33004 = torch.prim.ListConstruct %3026, %int32_30536, %int2_30537, %int16_30538, %int1_30539, %int128_30540 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33005 = torch.aten.view %31154, %33004 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33005, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30541 = torch.constant.int 32
%33006 = torch.aten.mul.int %3026, %int32_30541 : !torch.int, !torch.int -> !torch.int
%int2_30542 = torch.constant.int 2
%33007 = torch.aten.mul.int %33006, %int2_30542 : !torch.int, !torch.int -> !torch.int
%int16_30543 = torch.constant.int 16
%int1_30544 = torch.constant.int 1
%int128_30545 = torch.constant.int 128
%33008 = torch.prim.ListConstruct %33007, %int16_30543, %int1_30544, %int128_30545 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33009 = torch.aten.view %33005, %33008 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33009, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33010 = torch.prim.ListConstruct %32963 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30546 = torch.constant.bool false
%33011 = torch.aten.index_put %33009, %33010, %32979, %false_30546 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33011, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30547 = torch.constant.int 32
%int2_30548 = torch.constant.int 2
%int16_30549 = torch.constant.int 16
%int1_30550 = torch.constant.int 1
%int128_30551 = torch.constant.int 128
%33012 = torch.prim.ListConstruct %3026, %int32_30547, %int2_30548, %int16_30549, %int1_30550, %int128_30551 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33013 = torch.aten.view %33011, %33012 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33013, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30552 = torch.constant.int 131072
%33014 = torch.prim.ListConstruct %3026, %int131072_30552 : (!torch.int, !torch.int) -> !torch.list<int>
%33015 = torch.aten.view %33013, %33014 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33015, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30553 = torch.constant.int 32
%int2_30554 = torch.constant.int 2
%int16_30555 = torch.constant.int 16
%int1_30556 = torch.constant.int 1
%int128_30557 = torch.constant.int 128
%33016 = torch.prim.ListConstruct %3029, %int32_30553, %int2_30554, %int16_30555, %int1_30556, %int128_30557 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33017 = torch.aten.view %31166, %33016 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33017, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30558 = torch.constant.int 32
%33018 = torch.aten.mul.int %3029, %int32_30558 : !torch.int, !torch.int -> !torch.int
%int2_30559 = torch.constant.int 2
%33019 = torch.aten.mul.int %33018, %int2_30559 : !torch.int, !torch.int -> !torch.int
%int16_30560 = torch.constant.int 16
%int1_30561 = torch.constant.int 1
%int128_30562 = torch.constant.int 128
%33020 = torch.prim.ListConstruct %33019, %int16_30560, %int1_30561, %int128_30562 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33021 = torch.aten.view %33017, %33020 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33021, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33022 = torch.prim.ListConstruct %32965 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30563 = torch.constant.bool false
%33023 = torch.aten.index_put %33021, %33022, %32981, %false_30563 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33023, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30564 = torch.constant.int 32
%int2_30565 = torch.constant.int 2
%int16_30566 = torch.constant.int 16
%int1_30567 = torch.constant.int 1
%int128_30568 = torch.constant.int 128
%33024 = torch.prim.ListConstruct %3029, %int32_30564, %int2_30565, %int16_30566, %int1_30567, %int128_30568 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33025 = torch.aten.view %33023, %33024 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33025, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30569 = torch.constant.int 131072
%33026 = torch.prim.ListConstruct %3029, %int131072_30569 : (!torch.int, !torch.int) -> !torch.list<int>
%33027 = torch.aten.view %33025, %33026 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33027, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30570 = torch.constant.int 32
%int2_30571 = torch.constant.int 2
%int16_30572 = torch.constant.int 16
%int1_30573 = torch.constant.int 1
%int128_30574 = torch.constant.int 128
%33028 = torch.prim.ListConstruct %3032, %int32_30570, %int2_30571, %int16_30572, %int1_30573, %int128_30574 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33029 = torch.aten.view %31178, %33028 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33029, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30575 = torch.constant.int 32
%33030 = torch.aten.mul.int %3032, %int32_30575 : !torch.int, !torch.int -> !torch.int
%int2_30576 = torch.constant.int 2
%33031 = torch.aten.mul.int %33030, %int2_30576 : !torch.int, !torch.int -> !torch.int
%int16_30577 = torch.constant.int 16
%int1_30578 = torch.constant.int 1
%int128_30579 = torch.constant.int 128
%33032 = torch.prim.ListConstruct %33031, %int16_30577, %int1_30578, %int128_30579 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33033 = torch.aten.view %33029, %33032 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33033, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33034 = torch.prim.ListConstruct %32967 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30580 = torch.constant.bool false
%33035 = torch.aten.index_put %33033, %33034, %32983, %false_30580 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33035, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30581 = torch.constant.int 32
%int2_30582 = torch.constant.int 2
%int16_30583 = torch.constant.int 16
%int1_30584 = torch.constant.int 1
%int128_30585 = torch.constant.int 128
%33036 = torch.prim.ListConstruct %3032, %int32_30581, %int2_30582, %int16_30583, %int1_30584, %int128_30585 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33037 = torch.aten.view %33035, %33036 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33037, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30586 = torch.constant.int 131072
%33038 = torch.prim.ListConstruct %3032, %int131072_30586 : (!torch.int, !torch.int) -> !torch.list<int>
%33039 = torch.aten.view %33037, %33038 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33039, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30587 = torch.constant.int 32
%int2_30588 = torch.constant.int 2
%int16_30589 = torch.constant.int 16
%int1_30590 = torch.constant.int 1
%int128_30591 = torch.constant.int 128
%33040 = torch.prim.ListConstruct %3035, %int32_30587, %int2_30588, %int16_30589, %int1_30590, %int128_30591 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33041 = torch.aten.view %31190, %33040 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33041, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30592 = torch.constant.int 32
%33042 = torch.aten.mul.int %3035, %int32_30592 : !torch.int, !torch.int -> !torch.int
%int2_30593 = torch.constant.int 2
%33043 = torch.aten.mul.int %33042, %int2_30593 : !torch.int, !torch.int -> !torch.int
%int16_30594 = torch.constant.int 16
%int1_30595 = torch.constant.int 1
%int128_30596 = torch.constant.int 128
%33044 = torch.prim.ListConstruct %33043, %int16_30594, %int1_30595, %int128_30596 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33045 = torch.aten.view %33041, %33044 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33045, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33046 = torch.prim.ListConstruct %32969 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30597 = torch.constant.bool false
%33047 = torch.aten.index_put %33045, %33046, %32985, %false_30597 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33047, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30598 = torch.constant.int 32
%int2_30599 = torch.constant.int 2
%int16_30600 = torch.constant.int 16
%int1_30601 = torch.constant.int 1
%int128_30602 = torch.constant.int 128
%33048 = torch.prim.ListConstruct %3035, %int32_30598, %int2_30599, %int16_30600, %int1_30601, %int128_30602 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33049 = torch.aten.view %33047, %33048 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33049, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30603 = torch.constant.int 131072
%33050 = torch.prim.ListConstruct %3035, %int131072_30603 : (!torch.int, !torch.int) -> !torch.list<int>
%33051 = torch.aten.view %33049, %33050 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33051, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30604 = torch.constant.int 32
%int2_30605 = torch.constant.int 2
%int16_30606 = torch.constant.int 16
%int1_30607 = torch.constant.int 1
%int128_30608 = torch.constant.int 128
%33052 = torch.prim.ListConstruct %3038, %int32_30604, %int2_30605, %int16_30606, %int1_30607, %int128_30608 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33053 = torch.aten.view %31202, %33052 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33053, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30609 = torch.constant.int 32
%33054 = torch.aten.mul.int %3038, %int32_30609 : !torch.int, !torch.int -> !torch.int
%int2_30610 = torch.constant.int 2
%33055 = torch.aten.mul.int %33054, %int2_30610 : !torch.int, !torch.int -> !torch.int
%int16_30611 = torch.constant.int 16
%int1_30612 = torch.constant.int 1
%int128_30613 = torch.constant.int 128
%33056 = torch.prim.ListConstruct %33055, %int16_30611, %int1_30612, %int128_30613 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33057 = torch.aten.view %33053, %33056 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33057, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33058 = torch.prim.ListConstruct %32971 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30614 = torch.constant.bool false
%33059 = torch.aten.index_put %33057, %33058, %32987, %false_30614 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33059, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30615 = torch.constant.int 32
%int2_30616 = torch.constant.int 2
%int16_30617 = torch.constant.int 16
%int1_30618 = torch.constant.int 1
%int128_30619 = torch.constant.int 128
%33060 = torch.prim.ListConstruct %3038, %int32_30615, %int2_30616, %int16_30617, %int1_30618, %int128_30619 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33061 = torch.aten.view %33059, %33060 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33061, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30620 = torch.constant.int 131072
%33062 = torch.prim.ListConstruct %3038, %int131072_30620 : (!torch.int, !torch.int) -> !torch.list<int>
%33063 = torch.aten.view %33061, %33062 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33063, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30621 = torch.constant.int 32
%int2_30622 = torch.constant.int 2
%int16_30623 = torch.constant.int 16
%int1_30624 = torch.constant.int 1
%int128_30625 = torch.constant.int 128
%33064 = torch.prim.ListConstruct %3041, %int32_30621, %int2_30622, %int16_30623, %int1_30624, %int128_30625 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33065 = torch.aten.view %31214, %33064 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33065, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30626 = torch.constant.int 32
%33066 = torch.aten.mul.int %3041, %int32_30626 : !torch.int, !torch.int -> !torch.int
%int2_30627 = torch.constant.int 2
%33067 = torch.aten.mul.int %33066, %int2_30627 : !torch.int, !torch.int -> !torch.int
%int16_30628 = torch.constant.int 16
%int1_30629 = torch.constant.int 1
%int128_30630 = torch.constant.int 128
%33068 = torch.prim.ListConstruct %33067, %int16_30628, %int1_30629, %int128_30630 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33069 = torch.aten.view %33065, %33068 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33069, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33070 = torch.prim.ListConstruct %32973 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30631 = torch.constant.bool false
%33071 = torch.aten.index_put %33069, %33070, %32989, %false_30631 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33071, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30632 = torch.constant.int 32
%int2_30633 = torch.constant.int 2
%int16_30634 = torch.constant.int 16
%int1_30635 = torch.constant.int 1
%int128_30636 = torch.constant.int 128
%33072 = torch.prim.ListConstruct %3041, %int32_30632, %int2_30633, %int16_30634, %int1_30635, %int128_30636 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33073 = torch.aten.view %33071, %33072 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33073, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30637 = torch.constant.int 131072
%33074 = torch.prim.ListConstruct %3041, %int131072_30637 : (!torch.int, !torch.int) -> !torch.list<int>
%33075 = torch.aten.view %33073, %33074 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33075, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_30638 = torch.constant.int 32
%int2_30639 = torch.constant.int 2
%int16_30640 = torch.constant.int 16
%int1_30641 = torch.constant.int 1
%int128_30642 = torch.constant.int 128
%33076 = torch.prim.ListConstruct %3044, %int32_30638, %int2_30639, %int16_30640, %int1_30641, %int128_30642 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33077 = torch.aten.view %31226, %33076 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33077, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_30643 = torch.constant.int 32
%33078 = torch.aten.mul.int %3044, %int32_30643 : !torch.int, !torch.int -> !torch.int
%int2_30644 = torch.constant.int 2
%33079 = torch.aten.mul.int %33078, %int2_30644 : !torch.int, !torch.int -> !torch.int
%int16_30645 = torch.constant.int 16
%int1_30646 = torch.constant.int 1
%int128_30647 = torch.constant.int 128
%33080 = torch.prim.ListConstruct %33079, %int16_30645, %int1_30646, %int128_30647 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33081 = torch.aten.view %33077, %33080 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33081, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%33082 = torch.prim.ListConstruct %32975 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_30648 = torch.constant.bool false
%33083 = torch.aten.index_put %33081, %33082, %32991, %false_30648 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %33083, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_30649 = torch.constant.int 32
%int2_30650 = torch.constant.int 2
%int16_30651 = torch.constant.int 16
%int1_30652 = torch.constant.int 1
%int128_30653 = torch.constant.int 128
%33084 = torch.prim.ListConstruct %3044, %int32_30649, %int2_30650, %int16_30651, %int1_30652, %int128_30653 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33085 = torch.aten.view %33083, %33084 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %33085, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_30654 = torch.constant.int 131072
%33086 = torch.prim.ListConstruct %3044, %int131072_30654 : (!torch.int, !torch.int) -> !torch.list<int>
%33087 = torch.aten.view %33085, %33086 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %33087, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_30655 = torch.constant.int -2
%33088 = torch.aten.unsqueeze %32702, %int-2_30655 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30656 = torch.constant.int -2
%33089 = torch.aten.unsqueeze %32717, %int-2_30656 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30657 = torch.constant.int -2
%33090 = torch.aten.unsqueeze %32732, %int-2_30657 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30658 = torch.constant.int -2
%33091 = torch.aten.unsqueeze %32747, %int-2_30658 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30659 = torch.constant.int -2
%33092 = torch.aten.unsqueeze %32762, %int-2_30659 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30660 = torch.constant.int -2
%33093 = torch.aten.unsqueeze %32777, %int-2_30660 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30661 = torch.constant.int -2
%33094 = torch.aten.unsqueeze %32792, %int-2_30661 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30662 = torch.constant.int -2
%33095 = torch.aten.unsqueeze %32807, %int-2_30662 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_30663 = torch.constant.int 4
%int1_30664 = torch.constant.int 1
%int4_30665 = torch.constant.int 4
%int128_30666 = torch.constant.int 128
%33096 = torch.prim.ListConstruct %int4_30663, %32688, %int1_30664, %int4_30665, %int128_30666 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30667 = torch.constant.bool false
%33097 = torch.aten.expand %33088, %33096, %false_30667 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30668 = torch.constant.int 4
%int1_30669 = torch.constant.int 1
%int4_30670 = torch.constant.int 4
%int128_30671 = torch.constant.int 128
%33098 = torch.prim.ListConstruct %int4_30668, %32688, %int1_30669, %int4_30670, %int128_30671 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30672 = torch.constant.bool false
%33099 = torch.aten.expand %33089, %33098, %false_30672 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30673 = torch.constant.int 4
%int1_30674 = torch.constant.int 1
%int4_30675 = torch.constant.int 4
%int128_30676 = torch.constant.int 128
%33100 = torch.prim.ListConstruct %int4_30673, %32688, %int1_30674, %int4_30675, %int128_30676 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30677 = torch.constant.bool false
%33101 = torch.aten.expand %33090, %33100, %false_30677 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30678 = torch.constant.int 4
%int1_30679 = torch.constant.int 1
%int4_30680 = torch.constant.int 4
%int128_30681 = torch.constant.int 128
%33102 = torch.prim.ListConstruct %int4_30678, %32688, %int1_30679, %int4_30680, %int128_30681 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30682 = torch.constant.bool false
%33103 = torch.aten.expand %33091, %33102, %false_30682 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30683 = torch.constant.int 4
%int1_30684 = torch.constant.int 1
%int4_30685 = torch.constant.int 4
%int128_30686 = torch.constant.int 128
%33104 = torch.prim.ListConstruct %int4_30683, %32688, %int1_30684, %int4_30685, %int128_30686 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30687 = torch.constant.bool false
%33105 = torch.aten.expand %33092, %33104, %false_30687 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30688 = torch.constant.int 4
%int1_30689 = torch.constant.int 1
%int4_30690 = torch.constant.int 4
%int128_30691 = torch.constant.int 128
%33106 = torch.prim.ListConstruct %int4_30688, %32688, %int1_30689, %int4_30690, %int128_30691 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30692 = torch.constant.bool false
%33107 = torch.aten.expand %33093, %33106, %false_30692 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30693 = torch.constant.int 4
%int1_30694 = torch.constant.int 1
%int4_30695 = torch.constant.int 4
%int128_30696 = torch.constant.int 128
%33108 = torch.prim.ListConstruct %int4_30693, %32688, %int1_30694, %int4_30695, %int128_30696 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30697 = torch.constant.bool false
%33109 = torch.aten.expand %33094, %33108, %false_30697 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30698 = torch.constant.int 4
%int1_30699 = torch.constant.int 1
%int4_30700 = torch.constant.int 4
%int128_30701 = torch.constant.int 128
%33110 = torch.prim.ListConstruct %int4_30698, %32688, %int1_30699, %int4_30700, %int128_30701 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30702 = torch.constant.bool false
%33111 = torch.aten.expand %33095, %33110, %false_30702 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30703 = torch.constant.int 4
%int4_30704 = torch.constant.int 4
%int128_30705 = torch.constant.int 128
%33112 = torch.prim.ListConstruct %int4_30703, %32688, %int4_30704, %int128_30705 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33113 = torch.aten.view %33097, %33112 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30706 = torch.constant.int 4
%int4_30707 = torch.constant.int 4
%int128_30708 = torch.constant.int 128
%33114 = torch.prim.ListConstruct %int4_30706, %32688, %int4_30707, %int128_30708 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33115 = torch.aten.view %33099, %33114 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30709 = torch.constant.int 4
%int4_30710 = torch.constant.int 4
%int128_30711 = torch.constant.int 128
%33116 = torch.prim.ListConstruct %int4_30709, %32688, %int4_30710, %int128_30711 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33117 = torch.aten.view %33101, %33116 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30712 = torch.constant.int 4
%int4_30713 = torch.constant.int 4
%int128_30714 = torch.constant.int 128
%33118 = torch.prim.ListConstruct %int4_30712, %32688, %int4_30713, %int128_30714 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33119 = torch.aten.view %33103, %33118 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30715 = torch.constant.int 4
%int4_30716 = torch.constant.int 4
%int128_30717 = torch.constant.int 128
%33120 = torch.prim.ListConstruct %int4_30715, %32688, %int4_30716, %int128_30717 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33121 = torch.aten.view %33105, %33120 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30718 = torch.constant.int 4
%int4_30719 = torch.constant.int 4
%int128_30720 = torch.constant.int 128
%33122 = torch.prim.ListConstruct %int4_30718, %32688, %int4_30719, %int128_30720 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33123 = torch.aten.view %33107, %33122 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30721 = torch.constant.int 4
%int4_30722 = torch.constant.int 4
%int128_30723 = torch.constant.int 128
%33124 = torch.prim.ListConstruct %int4_30721, %32688, %int4_30722, %int128_30723 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33125 = torch.aten.view %33109, %33124 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30724 = torch.constant.int 4
%int4_30725 = torch.constant.int 4
%int128_30726 = torch.constant.int 128
%33126 = torch.prim.ListConstruct %int4_30724, %32688, %int4_30725, %int128_30726 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33127 = torch.aten.view %33111, %33126 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_30727 = torch.constant.int -2
%33128 = torch.aten.unsqueeze %32477, %int-2_30727 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30728 = torch.constant.int -2
%33129 = torch.aten.unsqueeze %32479, %int-2_30728 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30729 = torch.constant.int -2
%33130 = torch.aten.unsqueeze %32481, %int-2_30729 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33130, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30730 = torch.constant.int -2
%33131 = torch.aten.unsqueeze %32483, %int-2_30730 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30731 = torch.constant.int -2
%33132 = torch.aten.unsqueeze %32485, %int-2_30731 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33132, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30732 = torch.constant.int -2
%33133 = torch.aten.unsqueeze %32487, %int-2_30732 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30733 = torch.constant.int -2
%33134 = torch.aten.unsqueeze %32489, %int-2_30733 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_30734 = torch.constant.int -2
%33135 = torch.aten.unsqueeze %32491, %int-2_30734 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %33135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_30735 = torch.constant.int 1
%33136 = torch.aten.size.int %32401, %int1_30735 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_30736 = torch.constant.int 4
%int1_30737 = torch.constant.int 1
%int4_30738 = torch.constant.int 4
%int128_30739 = torch.constant.int 128
%33137 = torch.prim.ListConstruct %int4_30736, %33136, %int1_30737, %int4_30738, %int128_30739 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30740 = torch.constant.bool false
%33138 = torch.aten.expand %33128, %33137, %false_30740 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30741 = torch.constant.int 4
%int1_30742 = torch.constant.int 1
%int4_30743 = torch.constant.int 4
%int128_30744 = torch.constant.int 128
%33139 = torch.prim.ListConstruct %int4_30741, %33136, %int1_30742, %int4_30743, %int128_30744 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30745 = torch.constant.bool false
%33140 = torch.aten.expand %33129, %33139, %false_30745 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30746 = torch.constant.int 4
%int1_30747 = torch.constant.int 1
%int4_30748 = torch.constant.int 4
%int128_30749 = torch.constant.int 128
%33141 = torch.prim.ListConstruct %int4_30746, %33136, %int1_30747, %int4_30748, %int128_30749 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30750 = torch.constant.bool false
%33142 = torch.aten.expand %33130, %33141, %false_30750 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30751 = torch.constant.int 4
%int1_30752 = torch.constant.int 1
%int4_30753 = torch.constant.int 4
%int128_30754 = torch.constant.int 128
%33143 = torch.prim.ListConstruct %int4_30751, %33136, %int1_30752, %int4_30753, %int128_30754 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30755 = torch.constant.bool false
%33144 = torch.aten.expand %33131, %33143, %false_30755 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30756 = torch.constant.int 4
%int1_30757 = torch.constant.int 1
%int4_30758 = torch.constant.int 4
%int128_30759 = torch.constant.int 128
%33145 = torch.prim.ListConstruct %int4_30756, %33136, %int1_30757, %int4_30758, %int128_30759 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30760 = torch.constant.bool false
%33146 = torch.aten.expand %33132, %33145, %false_30760 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30761 = torch.constant.int 4
%int1_30762 = torch.constant.int 1
%int4_30763 = torch.constant.int 4
%int128_30764 = torch.constant.int 128
%33147 = torch.prim.ListConstruct %int4_30761, %33136, %int1_30762, %int4_30763, %int128_30764 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30765 = torch.constant.bool false
%33148 = torch.aten.expand %33133, %33147, %false_30765 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30766 = torch.constant.int 4
%int1_30767 = torch.constant.int 1
%int4_30768 = torch.constant.int 4
%int128_30769 = torch.constant.int 128
%33149 = torch.prim.ListConstruct %int4_30766, %33136, %int1_30767, %int4_30768, %int128_30769 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30770 = torch.constant.bool false
%33150 = torch.aten.expand %33134, %33149, %false_30770 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30771 = torch.constant.int 4
%int1_30772 = torch.constant.int 1
%int4_30773 = torch.constant.int 4
%int128_30774 = torch.constant.int 128
%33151 = torch.prim.ListConstruct %int4_30771, %33136, %int1_30772, %int4_30773, %int128_30774 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_30775 = torch.constant.bool false
%33152 = torch.aten.expand %33135, %33151, %false_30775 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %33152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_30776 = torch.constant.int 4
%int4_30777 = torch.constant.int 4
%int128_30778 = torch.constant.int 128
%33153 = torch.prim.ListConstruct %int4_30776, %33136, %int4_30777, %int128_30778 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33154 = torch.aten.view %33138, %33153 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30779 = torch.constant.int 4
%int4_30780 = torch.constant.int 4
%int128_30781 = torch.constant.int 128
%33155 = torch.prim.ListConstruct %int4_30779, %33136, %int4_30780, %int128_30781 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33156 = torch.aten.view %33140, %33155 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30782 = torch.constant.int 4
%int4_30783 = torch.constant.int 4
%int128_30784 = torch.constant.int 128
%33157 = torch.prim.ListConstruct %int4_30782, %33136, %int4_30783, %int128_30784 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33158 = torch.aten.view %33142, %33157 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30785 = torch.constant.int 4
%int4_30786 = torch.constant.int 4
%int128_30787 = torch.constant.int 128
%33159 = torch.prim.ListConstruct %int4_30785, %33136, %int4_30786, %int128_30787 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33160 = torch.aten.view %33144, %33159 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30788 = torch.constant.int 4
%int4_30789 = torch.constant.int 4
%int128_30790 = torch.constant.int 128
%33161 = torch.prim.ListConstruct %int4_30788, %33136, %int4_30789, %int128_30790 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33162 = torch.aten.view %33146, %33161 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30791 = torch.constant.int 4
%int4_30792 = torch.constant.int 4
%int128_30793 = torch.constant.int 128
%33163 = torch.prim.ListConstruct %int4_30791, %33136, %int4_30792, %int128_30793 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33164 = torch.aten.view %33148, %33163 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30794 = torch.constant.int 4
%int4_30795 = torch.constant.int 4
%int128_30796 = torch.constant.int 128
%33165 = torch.prim.ListConstruct %int4_30794, %33136, %int4_30795, %int128_30796 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33166 = torch.aten.view %33150, %33165 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30797 = torch.constant.int 4
%int4_30798 = torch.constant.int 4
%int128_30799 = torch.constant.int 128
%33167 = torch.prim.ListConstruct %int4_30797, %33136, %int4_30798, %int128_30799 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33168 = torch.aten.view %33152, %33167 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30800 = torch.constant.int 1
%int2_30801 = torch.constant.int 2
%33169 = torch.aten.transpose.int %32544, %int1_30800, %int2_30801 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33169, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30802 = torch.constant.int 1
%int2_30803 = torch.constant.int 2
%33170 = torch.aten.transpose.int %32559, %int1_30802, %int2_30803 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33170, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30804 = torch.constant.int 1
%int2_30805 = torch.constant.int 2
%33171 = torch.aten.transpose.int %32574, %int1_30804, %int2_30805 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33171, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30806 = torch.constant.int 1
%int2_30807 = torch.constant.int 2
%33172 = torch.aten.transpose.int %32589, %int1_30806, %int2_30807 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33172, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30808 = torch.constant.int 1
%int2_30809 = torch.constant.int 2
%33173 = torch.aten.transpose.int %32604, %int1_30808, %int2_30809 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33173, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30810 = torch.constant.int 1
%int2_30811 = torch.constant.int 2
%33174 = torch.aten.transpose.int %32619, %int1_30810, %int2_30811 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33174, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30812 = torch.constant.int 1
%int2_30813 = torch.constant.int 2
%33175 = torch.aten.transpose.int %32634, %int1_30812, %int2_30813 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33175, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30814 = torch.constant.int 1
%int2_30815 = torch.constant.int 2
%33176 = torch.aten.transpose.int %32649, %int1_30814, %int2_30815 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33176, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30816 = torch.constant.int 1
%int2_30817 = torch.constant.int 2
%33177 = torch.aten.transpose.int %33113, %int1_30816, %int2_30817 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33177, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30818 = torch.constant.int 1
%int2_30819 = torch.constant.int 2
%33178 = torch.aten.transpose.int %33115, %int1_30818, %int2_30819 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33178, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30820 = torch.constant.int 1
%int2_30821 = torch.constant.int 2
%33179 = torch.aten.transpose.int %33117, %int1_30820, %int2_30821 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33179, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30822 = torch.constant.int 1
%int2_30823 = torch.constant.int 2
%33180 = torch.aten.transpose.int %33119, %int1_30822, %int2_30823 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33180, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30824 = torch.constant.int 1
%int2_30825 = torch.constant.int 2
%33181 = torch.aten.transpose.int %33121, %int1_30824, %int2_30825 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33181, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30826 = torch.constant.int 1
%int2_30827 = torch.constant.int 2
%33182 = torch.aten.transpose.int %33123, %int1_30826, %int2_30827 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33182, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30828 = torch.constant.int 1
%int2_30829 = torch.constant.int 2
%33183 = torch.aten.transpose.int %33125, %int1_30828, %int2_30829 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33183, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30830 = torch.constant.int 1
%int2_30831 = torch.constant.int 2
%33184 = torch.aten.transpose.int %33127, %int1_30830, %int2_30831 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33184, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30832 = torch.constant.int 1
%int2_30833 = torch.constant.int 2
%33185 = torch.aten.transpose.int %33154, %int1_30832, %int2_30833 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33185, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30834 = torch.constant.int 1
%int2_30835 = torch.constant.int 2
%33186 = torch.aten.transpose.int %33156, %int1_30834, %int2_30835 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33186, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30836 = torch.constant.int 1
%int2_30837 = torch.constant.int 2
%33187 = torch.aten.transpose.int %33158, %int1_30836, %int2_30837 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33187, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30838 = torch.constant.int 1
%int2_30839 = torch.constant.int 2
%33188 = torch.aten.transpose.int %33160, %int1_30838, %int2_30839 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33188, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30840 = torch.constant.int 1
%int2_30841 = torch.constant.int 2
%33189 = torch.aten.transpose.int %33162, %int1_30840, %int2_30841 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33189, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30842 = torch.constant.int 1
%int2_30843 = torch.constant.int 2
%33190 = torch.aten.transpose.int %33164, %int1_30842, %int2_30843 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33190, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30844 = torch.constant.int 1
%int2_30845 = torch.constant.int 2
%33191 = torch.aten.transpose.int %33166, %int1_30844, %int2_30845 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33191, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30846 = torch.constant.int 1
%int2_30847 = torch.constant.int 2
%33192 = torch.aten.transpose.int %33168, %int1_30846, %int2_30847 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %33192, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30848 = torch.constant.float 0.000000e+00
%true_30849 = torch.constant.bool true
%none_30850 = torch.constant.none
%none_30851 = torch.constant.none
%33193:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33169, %33177, %33185, %float0.000000e00_30848, %true_30849, %none_30850, %none_30851) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33193#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30852 = torch.constant.float 0.000000e+00
%true_30853 = torch.constant.bool true
%none_30854 = torch.constant.none
%none_30855 = torch.constant.none
%33194:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33170, %33178, %33186, %float0.000000e00_30852, %true_30853, %none_30854, %none_30855) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33194#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30856 = torch.constant.float 0.000000e+00
%true_30857 = torch.constant.bool true
%none_30858 = torch.constant.none
%none_30859 = torch.constant.none
%33195:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33171, %33179, %33187, %float0.000000e00_30856, %true_30857, %none_30858, %none_30859) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33195#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30860 = torch.constant.float 0.000000e+00
%true_30861 = torch.constant.bool true
%none_30862 = torch.constant.none
%none_30863 = torch.constant.none
%33196:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33172, %33180, %33188, %float0.000000e00_30860, %true_30861, %none_30862, %none_30863) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33196#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30864 = torch.constant.float 0.000000e+00
%true_30865 = torch.constant.bool true
%none_30866 = torch.constant.none
%none_30867 = torch.constant.none
%33197:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33173, %33181, %33189, %float0.000000e00_30864, %true_30865, %none_30866, %none_30867) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33197#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30868 = torch.constant.float 0.000000e+00
%true_30869 = torch.constant.bool true
%none_30870 = torch.constant.none
%none_30871 = torch.constant.none
%33198:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33174, %33182, %33190, %float0.000000e00_30868, %true_30869, %none_30870, %none_30871) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33198#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30872 = torch.constant.float 0.000000e+00
%true_30873 = torch.constant.bool true
%none_30874 = torch.constant.none
%none_30875 = torch.constant.none
%33199:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33175, %33183, %33191, %float0.000000e00_30872, %true_30873, %none_30874, %none_30875) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33199#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_30876 = torch.constant.float 0.000000e+00
%true_30877 = torch.constant.bool true
%none_30878 = torch.constant.none
%none_30879 = torch.constant.none
%33200:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%33176, %33184, %33192, %float0.000000e00_30876, %true_30877, %none_30878, %none_30879) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %33200#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_30880 = torch.constant.int 1
%int2_30881 = torch.constant.int 2
%33201 = torch.aten.transpose.int %33193#0, %int1_30880, %int2_30881 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30882 = torch.constant.int 1
%int2_30883 = torch.constant.int 2
%33202 = torch.aten.transpose.int %33194#0, %int1_30882, %int2_30883 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30884 = torch.constant.int 1
%int2_30885 = torch.constant.int 2
%33203 = torch.aten.transpose.int %33195#0, %int1_30884, %int2_30885 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30886 = torch.constant.int 1
%int2_30887 = torch.constant.int 2
%33204 = torch.aten.transpose.int %33196#0, %int1_30886, %int2_30887 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30888 = torch.constant.int 1
%int2_30889 = torch.constant.int 2
%33205 = torch.aten.transpose.int %33197#0, %int1_30888, %int2_30889 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30890 = torch.constant.int 1
%int2_30891 = torch.constant.int 2
%33206 = torch.aten.transpose.int %33198#0, %int1_30890, %int2_30891 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30892 = torch.constant.int 1
%int2_30893 = torch.constant.int 2
%33207 = torch.aten.transpose.int %33199#0, %int1_30892, %int2_30893 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_30894 = torch.constant.int 1
%int2_30895 = torch.constant.int 2
%33208 = torch.aten.transpose.int %33200#0, %int1_30894, %int2_30895 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %33208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_30896 = torch.constant.int 4
%int512_30897 = torch.constant.int 512
%33209 = torch.prim.ListConstruct %int4_30896, %32530, %int512_30897 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33210 = torch.aten.view %33201, %33209 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30898 = torch.constant.int 4
%int512_30899 = torch.constant.int 512
%33211 = torch.prim.ListConstruct %int4_30898, %32545, %int512_30899 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33212 = torch.aten.view %33202, %33211 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30900 = torch.constant.int 4
%int512_30901 = torch.constant.int 512
%33213 = torch.prim.ListConstruct %int4_30900, %32560, %int512_30901 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33214 = torch.aten.view %33203, %33213 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30902 = torch.constant.int 4
%int512_30903 = torch.constant.int 512
%33215 = torch.prim.ListConstruct %int4_30902, %32575, %int512_30903 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33216 = torch.aten.view %33204, %33215 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30904 = torch.constant.int 4
%int512_30905 = torch.constant.int 512
%33217 = torch.prim.ListConstruct %int4_30904, %32590, %int512_30905 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33218 = torch.aten.view %33205, %33217 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30906 = torch.constant.int 4
%int512_30907 = torch.constant.int 512
%33219 = torch.prim.ListConstruct %int4_30906, %32605, %int512_30907 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33220 = torch.aten.view %33206, %33219 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30908 = torch.constant.int 4
%int512_30909 = torch.constant.int 512
%33221 = torch.prim.ListConstruct %int4_30908, %32620, %int512_30909 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33222 = torch.aten.view %33207, %33221 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_30910 = torch.constant.int 4
%int512_30911 = torch.constant.int 512
%33223 = torch.prim.ListConstruct %int4_30910, %32635, %int512_30911 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33224 = torch.aten.view %33208, %33223 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %33224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_30912 = torch.constant.int 1
%int0_30913 = torch.constant.int 0
%33225 = torch.prim.ListConstruct %int1_30912, %int0_30913 : (!torch.int, !torch.int) -> !torch.list<int>
%33226 = torch.aten.permute %1192, %33225 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30914 = torch.constant.int 1
%int0_30915 = torch.constant.int 0
%33227 = torch.prim.ListConstruct %int1_30914, %int0_30915 : (!torch.int, !torch.int) -> !torch.list<int>
%33228 = torch.aten.permute %1193, %33227 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30916 = torch.constant.int 1
%int0_30917 = torch.constant.int 0
%33229 = torch.prim.ListConstruct %int1_30916, %int0_30917 : (!torch.int, !torch.int) -> !torch.list<int>
%33230 = torch.aten.permute %1194, %33229 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30918 = torch.constant.int 1
%int0_30919 = torch.constant.int 0
%33231 = torch.prim.ListConstruct %int1_30918, %int0_30919 : (!torch.int, !torch.int) -> !torch.list<int>
%33232 = torch.aten.permute %1195, %33231 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30920 = torch.constant.int 1
%int0_30921 = torch.constant.int 0
%33233 = torch.prim.ListConstruct %int1_30920, %int0_30921 : (!torch.int, !torch.int) -> !torch.list<int>
%33234 = torch.aten.permute %1196, %33233 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30922 = torch.constant.int 1
%int0_30923 = torch.constant.int 0
%33235 = torch.prim.ListConstruct %int1_30922, %int0_30923 : (!torch.int, !torch.int) -> !torch.list<int>
%33236 = torch.aten.permute %1197, %33235 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30924 = torch.constant.int 1
%int0_30925 = torch.constant.int 0
%33237 = torch.prim.ListConstruct %int1_30924, %int0_30925 : (!torch.int, !torch.int) -> !torch.list<int>
%33238 = torch.aten.permute %1198, %33237 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_30926 = torch.constant.int 1
%int0_30927 = torch.constant.int 0
%33239 = torch.prim.ListConstruct %int1_30926, %int0_30927 : (!torch.int, !torch.int) -> !torch.list<int>
%33240 = torch.aten.permute %1199, %33239 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_30928 = torch.constant.int 4
%33241 = torch.aten.mul.int %int4_30928, %32530 : !torch.int, !torch.int -> !torch.int
%int512_30929 = torch.constant.int 512
%33242 = torch.prim.ListConstruct %33241, %int512_30929 : (!torch.int, !torch.int) -> !torch.list<int>
%33243 = torch.aten.view %33210, %33242 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33243, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33244 = torch.aten.mm %33243, %33226 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33244, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30930 = torch.constant.int 4
%int4096_30931 = torch.constant.int 4096
%33245 = torch.prim.ListConstruct %int4_30930, %32530, %int4096_30931 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33246 = torch.aten.view %33244, %33245 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30932 = torch.constant.int 4
%33247 = torch.aten.mul.int %int4_30932, %32545 : !torch.int, !torch.int -> !torch.int
%int512_30933 = torch.constant.int 512
%33248 = torch.prim.ListConstruct %33247, %int512_30933 : (!torch.int, !torch.int) -> !torch.list<int>
%33249 = torch.aten.view %33212, %33248 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33249, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33250 = torch.aten.mm %33249, %33228 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33250, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30934 = torch.constant.int 4
%int4096_30935 = torch.constant.int 4096
%33251 = torch.prim.ListConstruct %int4_30934, %32545, %int4096_30935 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33252 = torch.aten.view %33250, %33251 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30936 = torch.constant.int 4
%33253 = torch.aten.mul.int %int4_30936, %32560 : !torch.int, !torch.int -> !torch.int
%int512_30937 = torch.constant.int 512
%33254 = torch.prim.ListConstruct %33253, %int512_30937 : (!torch.int, !torch.int) -> !torch.list<int>
%33255 = torch.aten.view %33214, %33254 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33255, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33256 = torch.aten.mm %33255, %33230 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33256, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30938 = torch.constant.int 4
%int4096_30939 = torch.constant.int 4096
%33257 = torch.prim.ListConstruct %int4_30938, %32560, %int4096_30939 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33258 = torch.aten.view %33256, %33257 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30940 = torch.constant.int 4
%33259 = torch.aten.mul.int %int4_30940, %32575 : !torch.int, !torch.int -> !torch.int
%int512_30941 = torch.constant.int 512
%33260 = torch.prim.ListConstruct %33259, %int512_30941 : (!torch.int, !torch.int) -> !torch.list<int>
%33261 = torch.aten.view %33216, %33260 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33261, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33262 = torch.aten.mm %33261, %33232 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33262, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30942 = torch.constant.int 4
%int4096_30943 = torch.constant.int 4096
%33263 = torch.prim.ListConstruct %int4_30942, %32575, %int4096_30943 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33264 = torch.aten.view %33262, %33263 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30944 = torch.constant.int 4
%33265 = torch.aten.mul.int %int4_30944, %32590 : !torch.int, !torch.int -> !torch.int
%int512_30945 = torch.constant.int 512
%33266 = torch.prim.ListConstruct %33265, %int512_30945 : (!torch.int, !torch.int) -> !torch.list<int>
%33267 = torch.aten.view %33218, %33266 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33267, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33268 = torch.aten.mm %33267, %33234 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33268, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30946 = torch.constant.int 4
%int4096_30947 = torch.constant.int 4096
%33269 = torch.prim.ListConstruct %int4_30946, %32590, %int4096_30947 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33270 = torch.aten.view %33268, %33269 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30948 = torch.constant.int 4
%33271 = torch.aten.mul.int %int4_30948, %32605 : !torch.int, !torch.int -> !torch.int
%int512_30949 = torch.constant.int 512
%33272 = torch.prim.ListConstruct %33271, %int512_30949 : (!torch.int, !torch.int) -> !torch.list<int>
%33273 = torch.aten.view %33220, %33272 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33273, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33274 = torch.aten.mm %33273, %33236 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33274, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30950 = torch.constant.int 4
%int4096_30951 = torch.constant.int 4096
%33275 = torch.prim.ListConstruct %int4_30950, %32605, %int4096_30951 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33276 = torch.aten.view %33274, %33275 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30952 = torch.constant.int 4
%33277 = torch.aten.mul.int %int4_30952, %32620 : !torch.int, !torch.int -> !torch.int
%int512_30953 = torch.constant.int 512
%33278 = torch.prim.ListConstruct %33277, %int512_30953 : (!torch.int, !torch.int) -> !torch.list<int>
%33279 = torch.aten.view %33222, %33278 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33279, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33280 = torch.aten.mm %33279, %33238 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33280, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30954 = torch.constant.int 4
%int4096_30955 = torch.constant.int 4096
%33281 = torch.prim.ListConstruct %int4_30954, %32620, %int4096_30955 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33282 = torch.aten.view %33280, %33281 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_30956 = torch.constant.int 4
%33283 = torch.aten.mul.int %int4_30956, %32635 : !torch.int, !torch.int -> !torch.int
%int512_30957 = torch.constant.int 512
%33284 = torch.prim.ListConstruct %33283, %int512_30957 : (!torch.int, !torch.int) -> !torch.list<int>
%33285 = torch.aten.view %33224, %33284 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %33285, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%33286 = torch.aten.mm %33285, %33240 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33286, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_30958 = torch.constant.int 4
%int4096_30959 = torch.constant.int 4096
%33287 = torch.prim.ListConstruct %int4_30958, %32635, %int4096_30959 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33288 = torch.aten.view %33286, %33287 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33289 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30960 = arith.constant 1 : index
%dim_30961 = tensor.dim %33289, %c1_30960 : tensor<4x?x4096xf16>
%33290 = flow.tensor.transfer %33289 : tensor<4x?x4096xf16>{%dim_30961} to #hal.device.promise<@__device_0>
%33291 = torch_c.from_builtin_tensor %33290 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33292 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30962 = arith.constant 1 : index
%dim_30963 = tensor.dim %33292, %c1_30962 : tensor<4x?x4096xf16>
%33293 = flow.tensor.transfer %33292 : tensor<4x?x4096xf16>{%dim_30963} to #hal.device.promise<@__device_0>
%33294 = torch_c.from_builtin_tensor %33293 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33295 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30964 = arith.constant 1 : index
%dim_30965 = tensor.dim %33295, %c1_30964 : tensor<4x?x4096xf16>
%33296 = flow.tensor.transfer %33295 : tensor<4x?x4096xf16>{%dim_30965} to #hal.device.promise<@__device_0>
%33297 = torch_c.from_builtin_tensor %33296 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33298 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30966 = arith.constant 1 : index
%dim_30967 = tensor.dim %33298, %c1_30966 : tensor<4x?x4096xf16>
%33299 = flow.tensor.transfer %33298 : tensor<4x?x4096xf16>{%dim_30967} to #hal.device.promise<@__device_0>
%33300 = torch_c.from_builtin_tensor %33299 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33301 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30968 = arith.constant 1 : index
%dim_30969 = tensor.dim %33301, %c1_30968 : tensor<4x?x4096xf16>
%33302 = flow.tensor.transfer %33301 : tensor<4x?x4096xf16>{%dim_30969} to #hal.device.promise<@__device_0>
%33303 = torch_c.from_builtin_tensor %33302 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33304 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30970 = arith.constant 1 : index
%dim_30971 = tensor.dim %33304, %c1_30970 : tensor<4x?x4096xf16>
%33305 = flow.tensor.transfer %33304 : tensor<4x?x4096xf16>{%dim_30971} to #hal.device.promise<@__device_0>
%33306 = torch_c.from_builtin_tensor %33305 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33307 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30972 = arith.constant 1 : index
%dim_30973 = tensor.dim %33307, %c1_30972 : tensor<4x?x4096xf16>
%33308 = flow.tensor.transfer %33307 : tensor<4x?x4096xf16>{%dim_30973} to #hal.device.promise<@__device_0>
%33309 = torch_c.from_builtin_tensor %33308 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30974 = torch.constant.int 1
%33310 = torch.aten.add.Tensor %33246, %33291, %int1_30974 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30975 = torch.constant.int 1
%33311 = torch.aten.add.Tensor %33310, %33294, %int1_30975 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30976 = torch.constant.int 1
%33312 = torch.aten.add.Tensor %33311, %33297, %int1_30976 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30977 = torch.constant.int 1
%33313 = torch.aten.add.Tensor %33312, %33300, %int1_30977 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30978 = torch.constant.int 1
%33314 = torch.aten.add.Tensor %33313, %33303, %int1_30978 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30979 = torch.constant.int 1
%33315 = torch.aten.add.Tensor %33314, %33306, %int1_30979 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30980 = torch.constant.int 1
%33316 = torch.aten.add.Tensor %33315, %33309, %int1_30980 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33317 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30981 = arith.constant 1 : index
%dim_30982 = tensor.dim %33317, %c1_30981 : tensor<4x?x4096xf16>
%33318 = flow.tensor.transfer %33317 : tensor<4x?x4096xf16>{%dim_30982} to #hal.device.promise<@__device_1>
%33319 = torch_c.from_builtin_tensor %33318 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33320 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30983 = arith.constant 1 : index
%dim_30984 = tensor.dim %33320, %c1_30983 : tensor<4x?x4096xf16>
%33321 = flow.tensor.transfer %33320 : tensor<4x?x4096xf16>{%dim_30984} to #hal.device.promise<@__device_1>
%33322 = torch_c.from_builtin_tensor %33321 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33323 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30985 = arith.constant 1 : index
%dim_30986 = tensor.dim %33323, %c1_30985 : tensor<4x?x4096xf16>
%33324 = flow.tensor.transfer %33323 : tensor<4x?x4096xf16>{%dim_30986} to #hal.device.promise<@__device_1>
%33325 = torch_c.from_builtin_tensor %33324 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33326 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30987 = arith.constant 1 : index
%dim_30988 = tensor.dim %33326, %c1_30987 : tensor<4x?x4096xf16>
%33327 = flow.tensor.transfer %33326 : tensor<4x?x4096xf16>{%dim_30988} to #hal.device.promise<@__device_1>
%33328 = torch_c.from_builtin_tensor %33327 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33329 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30989 = arith.constant 1 : index
%dim_30990 = tensor.dim %33329, %c1_30989 : tensor<4x?x4096xf16>
%33330 = flow.tensor.transfer %33329 : tensor<4x?x4096xf16>{%dim_30990} to #hal.device.promise<@__device_1>
%33331 = torch_c.from_builtin_tensor %33330 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33331, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33332 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30991 = arith.constant 1 : index
%dim_30992 = tensor.dim %33332, %c1_30991 : tensor<4x?x4096xf16>
%33333 = flow.tensor.transfer %33332 : tensor<4x?x4096xf16>{%dim_30992} to #hal.device.promise<@__device_1>
%33334 = torch_c.from_builtin_tensor %33333 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33335 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_30993 = arith.constant 1 : index
%dim_30994 = tensor.dim %33335, %c1_30993 : tensor<4x?x4096xf16>
%33336 = flow.tensor.transfer %33335 : tensor<4x?x4096xf16>{%dim_30994} to #hal.device.promise<@__device_1>
%33337 = torch_c.from_builtin_tensor %33336 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30995 = torch.constant.int 1
%33338 = torch.aten.add.Tensor %33319, %33252, %int1_30995 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30996 = torch.constant.int 1
%33339 = torch.aten.add.Tensor %33338, %33322, %int1_30996 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30997 = torch.constant.int 1
%33340 = torch.aten.add.Tensor %33339, %33325, %int1_30997 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30998 = torch.constant.int 1
%33341 = torch.aten.add.Tensor %33340, %33328, %int1_30998 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_30999 = torch.constant.int 1
%33342 = torch.aten.add.Tensor %33341, %33331, %int1_30999 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31000 = torch.constant.int 1
%33343 = torch.aten.add.Tensor %33342, %33334, %int1_31000 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31001 = torch.constant.int 1
%33344 = torch.aten.add.Tensor %33343, %33337, %int1_31001 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33345 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31002 = arith.constant 1 : index
%dim_31003 = tensor.dim %33345, %c1_31002 : tensor<4x?x4096xf16>
%33346 = flow.tensor.transfer %33345 : tensor<4x?x4096xf16>{%dim_31003} to #hal.device.promise<@__device_2>
%33347 = torch_c.from_builtin_tensor %33346 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33347, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33348 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31004 = arith.constant 1 : index
%dim_31005 = tensor.dim %33348, %c1_31004 : tensor<4x?x4096xf16>
%33349 = flow.tensor.transfer %33348 : tensor<4x?x4096xf16>{%dim_31005} to #hal.device.promise<@__device_2>
%33350 = torch_c.from_builtin_tensor %33349 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33351 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31006 = arith.constant 1 : index
%dim_31007 = tensor.dim %33351, %c1_31006 : tensor<4x?x4096xf16>
%33352 = flow.tensor.transfer %33351 : tensor<4x?x4096xf16>{%dim_31007} to #hal.device.promise<@__device_2>
%33353 = torch_c.from_builtin_tensor %33352 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33353, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33354 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31008 = arith.constant 1 : index
%dim_31009 = tensor.dim %33354, %c1_31008 : tensor<4x?x4096xf16>
%33355 = flow.tensor.transfer %33354 : tensor<4x?x4096xf16>{%dim_31009} to #hal.device.promise<@__device_2>
%33356 = torch_c.from_builtin_tensor %33355 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33357 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31010 = arith.constant 1 : index
%dim_31011 = tensor.dim %33357, %c1_31010 : tensor<4x?x4096xf16>
%33358 = flow.tensor.transfer %33357 : tensor<4x?x4096xf16>{%dim_31011} to #hal.device.promise<@__device_2>
%33359 = torch_c.from_builtin_tensor %33358 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33360 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31012 = arith.constant 1 : index
%dim_31013 = tensor.dim %33360, %c1_31012 : tensor<4x?x4096xf16>
%33361 = flow.tensor.transfer %33360 : tensor<4x?x4096xf16>{%dim_31013} to #hal.device.promise<@__device_2>
%33362 = torch_c.from_builtin_tensor %33361 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33362, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33363 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31014 = arith.constant 1 : index
%dim_31015 = tensor.dim %33363, %c1_31014 : tensor<4x?x4096xf16>
%33364 = flow.tensor.transfer %33363 : tensor<4x?x4096xf16>{%dim_31015} to #hal.device.promise<@__device_2>
%33365 = torch_c.from_builtin_tensor %33364 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33365, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31016 = torch.constant.int 1
%33366 = torch.aten.add.Tensor %33347, %33350, %int1_31016 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31017 = torch.constant.int 1
%33367 = torch.aten.add.Tensor %33366, %33258, %int1_31017 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31018 = torch.constant.int 1
%33368 = torch.aten.add.Tensor %33367, %33353, %int1_31018 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31019 = torch.constant.int 1
%33369 = torch.aten.add.Tensor %33368, %33356, %int1_31019 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31020 = torch.constant.int 1
%33370 = torch.aten.add.Tensor %33369, %33359, %int1_31020 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31021 = torch.constant.int 1
%33371 = torch.aten.add.Tensor %33370, %33362, %int1_31021 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31022 = torch.constant.int 1
%33372 = torch.aten.add.Tensor %33371, %33365, %int1_31022 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33373 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31023 = arith.constant 1 : index
%dim_31024 = tensor.dim %33373, %c1_31023 : tensor<4x?x4096xf16>
%33374 = flow.tensor.transfer %33373 : tensor<4x?x4096xf16>{%dim_31024} to #hal.device.promise<@__device_3>
%33375 = torch_c.from_builtin_tensor %33374 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33376 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31025 = arith.constant 1 : index
%dim_31026 = tensor.dim %33376, %c1_31025 : tensor<4x?x4096xf16>
%33377 = flow.tensor.transfer %33376 : tensor<4x?x4096xf16>{%dim_31026} to #hal.device.promise<@__device_3>
%33378 = torch_c.from_builtin_tensor %33377 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33379 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31027 = arith.constant 1 : index
%dim_31028 = tensor.dim %33379, %c1_31027 : tensor<4x?x4096xf16>
%33380 = flow.tensor.transfer %33379 : tensor<4x?x4096xf16>{%dim_31028} to #hal.device.promise<@__device_3>
%33381 = torch_c.from_builtin_tensor %33380 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33382 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31029 = arith.constant 1 : index
%dim_31030 = tensor.dim %33382, %c1_31029 : tensor<4x?x4096xf16>
%33383 = flow.tensor.transfer %33382 : tensor<4x?x4096xf16>{%dim_31030} to #hal.device.promise<@__device_3>
%33384 = torch_c.from_builtin_tensor %33383 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33385 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31031 = arith.constant 1 : index
%dim_31032 = tensor.dim %33385, %c1_31031 : tensor<4x?x4096xf16>
%33386 = flow.tensor.transfer %33385 : tensor<4x?x4096xf16>{%dim_31032} to #hal.device.promise<@__device_3>
%33387 = torch_c.from_builtin_tensor %33386 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33388 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31033 = arith.constant 1 : index
%dim_31034 = tensor.dim %33388, %c1_31033 : tensor<4x?x4096xf16>
%33389 = flow.tensor.transfer %33388 : tensor<4x?x4096xf16>{%dim_31034} to #hal.device.promise<@__device_3>
%33390 = torch_c.from_builtin_tensor %33389 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33391 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31035 = arith.constant 1 : index
%dim_31036 = tensor.dim %33391, %c1_31035 : tensor<4x?x4096xf16>
%33392 = flow.tensor.transfer %33391 : tensor<4x?x4096xf16>{%dim_31036} to #hal.device.promise<@__device_3>
%33393 = torch_c.from_builtin_tensor %33392 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31037 = torch.constant.int 1
%33394 = torch.aten.add.Tensor %33375, %33378, %int1_31037 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31038 = torch.constant.int 1
%33395 = torch.aten.add.Tensor %33394, %33381, %int1_31038 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31039 = torch.constant.int 1
%33396 = torch.aten.add.Tensor %33395, %33264, %int1_31039 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31040 = torch.constant.int 1
%33397 = torch.aten.add.Tensor %33396, %33384, %int1_31040 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31041 = torch.constant.int 1
%33398 = torch.aten.add.Tensor %33397, %33387, %int1_31041 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33398, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31042 = torch.constant.int 1
%33399 = torch.aten.add.Tensor %33398, %33390, %int1_31042 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31043 = torch.constant.int 1
%33400 = torch.aten.add.Tensor %33399, %33393, %int1_31043 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33401 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31044 = arith.constant 1 : index
%dim_31045 = tensor.dim %33401, %c1_31044 : tensor<4x?x4096xf16>
%33402 = flow.tensor.transfer %33401 : tensor<4x?x4096xf16>{%dim_31045} to #hal.device.promise<@__device_4>
%33403 = torch_c.from_builtin_tensor %33402 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33404 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31046 = arith.constant 1 : index
%dim_31047 = tensor.dim %33404, %c1_31046 : tensor<4x?x4096xf16>
%33405 = flow.tensor.transfer %33404 : tensor<4x?x4096xf16>{%dim_31047} to #hal.device.promise<@__device_4>
%33406 = torch_c.from_builtin_tensor %33405 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33407 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31048 = arith.constant 1 : index
%dim_31049 = tensor.dim %33407, %c1_31048 : tensor<4x?x4096xf16>
%33408 = flow.tensor.transfer %33407 : tensor<4x?x4096xf16>{%dim_31049} to #hal.device.promise<@__device_4>
%33409 = torch_c.from_builtin_tensor %33408 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33410 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31050 = arith.constant 1 : index
%dim_31051 = tensor.dim %33410, %c1_31050 : tensor<4x?x4096xf16>
%33411 = flow.tensor.transfer %33410 : tensor<4x?x4096xf16>{%dim_31051} to #hal.device.promise<@__device_4>
%33412 = torch_c.from_builtin_tensor %33411 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33413 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31052 = arith.constant 1 : index
%dim_31053 = tensor.dim %33413, %c1_31052 : tensor<4x?x4096xf16>
%33414 = flow.tensor.transfer %33413 : tensor<4x?x4096xf16>{%dim_31053} to #hal.device.promise<@__device_4>
%33415 = torch_c.from_builtin_tensor %33414 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33416 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31054 = arith.constant 1 : index
%dim_31055 = tensor.dim %33416, %c1_31054 : tensor<4x?x4096xf16>
%33417 = flow.tensor.transfer %33416 : tensor<4x?x4096xf16>{%dim_31055} to #hal.device.promise<@__device_4>
%33418 = torch_c.from_builtin_tensor %33417 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33419 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31056 = arith.constant 1 : index
%dim_31057 = tensor.dim %33419, %c1_31056 : tensor<4x?x4096xf16>
%33420 = flow.tensor.transfer %33419 : tensor<4x?x4096xf16>{%dim_31057} to #hal.device.promise<@__device_4>
%33421 = torch_c.from_builtin_tensor %33420 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31058 = torch.constant.int 1
%33422 = torch.aten.add.Tensor %33403, %33406, %int1_31058 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31059 = torch.constant.int 1
%33423 = torch.aten.add.Tensor %33422, %33409, %int1_31059 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31060 = torch.constant.int 1
%33424 = torch.aten.add.Tensor %33423, %33412, %int1_31060 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31061 = torch.constant.int 1
%33425 = torch.aten.add.Tensor %33424, %33270, %int1_31061 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31062 = torch.constant.int 1
%33426 = torch.aten.add.Tensor %33425, %33415, %int1_31062 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31063 = torch.constant.int 1
%33427 = torch.aten.add.Tensor %33426, %33418, %int1_31063 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31064 = torch.constant.int 1
%33428 = torch.aten.add.Tensor %33427, %33421, %int1_31064 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33429 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31065 = arith.constant 1 : index
%dim_31066 = tensor.dim %33429, %c1_31065 : tensor<4x?x4096xf16>
%33430 = flow.tensor.transfer %33429 : tensor<4x?x4096xf16>{%dim_31066} to #hal.device.promise<@__device_5>
%33431 = torch_c.from_builtin_tensor %33430 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33432 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31067 = arith.constant 1 : index
%dim_31068 = tensor.dim %33432, %c1_31067 : tensor<4x?x4096xf16>
%33433 = flow.tensor.transfer %33432 : tensor<4x?x4096xf16>{%dim_31068} to #hal.device.promise<@__device_5>
%33434 = torch_c.from_builtin_tensor %33433 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33435 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31069 = arith.constant 1 : index
%dim_31070 = tensor.dim %33435, %c1_31069 : tensor<4x?x4096xf16>
%33436 = flow.tensor.transfer %33435 : tensor<4x?x4096xf16>{%dim_31070} to #hal.device.promise<@__device_5>
%33437 = torch_c.from_builtin_tensor %33436 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33438 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31071 = arith.constant 1 : index
%dim_31072 = tensor.dim %33438, %c1_31071 : tensor<4x?x4096xf16>
%33439 = flow.tensor.transfer %33438 : tensor<4x?x4096xf16>{%dim_31072} to #hal.device.promise<@__device_5>
%33440 = torch_c.from_builtin_tensor %33439 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33441 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31073 = arith.constant 1 : index
%dim_31074 = tensor.dim %33441, %c1_31073 : tensor<4x?x4096xf16>
%33442 = flow.tensor.transfer %33441 : tensor<4x?x4096xf16>{%dim_31074} to #hal.device.promise<@__device_5>
%33443 = torch_c.from_builtin_tensor %33442 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33444 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31075 = arith.constant 1 : index
%dim_31076 = tensor.dim %33444, %c1_31075 : tensor<4x?x4096xf16>
%33445 = flow.tensor.transfer %33444 : tensor<4x?x4096xf16>{%dim_31076} to #hal.device.promise<@__device_5>
%33446 = torch_c.from_builtin_tensor %33445 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33447 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31077 = arith.constant 1 : index
%dim_31078 = tensor.dim %33447, %c1_31077 : tensor<4x?x4096xf16>
%33448 = flow.tensor.transfer %33447 : tensor<4x?x4096xf16>{%dim_31078} to #hal.device.promise<@__device_5>
%33449 = torch_c.from_builtin_tensor %33448 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31079 = torch.constant.int 1
%33450 = torch.aten.add.Tensor %33431, %33434, %int1_31079 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31080 = torch.constant.int 1
%33451 = torch.aten.add.Tensor %33450, %33437, %int1_31080 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31081 = torch.constant.int 1
%33452 = torch.aten.add.Tensor %33451, %33440, %int1_31081 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31082 = torch.constant.int 1
%33453 = torch.aten.add.Tensor %33452, %33443, %int1_31082 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31083 = torch.constant.int 1
%33454 = torch.aten.add.Tensor %33453, %33276, %int1_31083 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31084 = torch.constant.int 1
%33455 = torch.aten.add.Tensor %33454, %33446, %int1_31084 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31085 = torch.constant.int 1
%33456 = torch.aten.add.Tensor %33455, %33449, %int1_31085 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33457 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31086 = arith.constant 1 : index
%dim_31087 = tensor.dim %33457, %c1_31086 : tensor<4x?x4096xf16>
%33458 = flow.tensor.transfer %33457 : tensor<4x?x4096xf16>{%dim_31087} to #hal.device.promise<@__device_6>
%33459 = torch_c.from_builtin_tensor %33458 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33460 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31088 = arith.constant 1 : index
%dim_31089 = tensor.dim %33460, %c1_31088 : tensor<4x?x4096xf16>
%33461 = flow.tensor.transfer %33460 : tensor<4x?x4096xf16>{%dim_31089} to #hal.device.promise<@__device_6>
%33462 = torch_c.from_builtin_tensor %33461 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33463 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31090 = arith.constant 1 : index
%dim_31091 = tensor.dim %33463, %c1_31090 : tensor<4x?x4096xf16>
%33464 = flow.tensor.transfer %33463 : tensor<4x?x4096xf16>{%dim_31091} to #hal.device.promise<@__device_6>
%33465 = torch_c.from_builtin_tensor %33464 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33466 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31092 = arith.constant 1 : index
%dim_31093 = tensor.dim %33466, %c1_31092 : tensor<4x?x4096xf16>
%33467 = flow.tensor.transfer %33466 : tensor<4x?x4096xf16>{%dim_31093} to #hal.device.promise<@__device_6>
%33468 = torch_c.from_builtin_tensor %33467 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33469 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31094 = arith.constant 1 : index
%dim_31095 = tensor.dim %33469, %c1_31094 : tensor<4x?x4096xf16>
%33470 = flow.tensor.transfer %33469 : tensor<4x?x4096xf16>{%dim_31095} to #hal.device.promise<@__device_6>
%33471 = torch_c.from_builtin_tensor %33470 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33472 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31096 = arith.constant 1 : index
%dim_31097 = tensor.dim %33472, %c1_31096 : tensor<4x?x4096xf16>
%33473 = flow.tensor.transfer %33472 : tensor<4x?x4096xf16>{%dim_31097} to #hal.device.promise<@__device_6>
%33474 = torch_c.from_builtin_tensor %33473 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33475 = torch_c.to_builtin_tensor %33288 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31098 = arith.constant 1 : index
%dim_31099 = tensor.dim %33475, %c1_31098 : tensor<4x?x4096xf16>
%33476 = flow.tensor.transfer %33475 : tensor<4x?x4096xf16>{%dim_31099} to #hal.device.promise<@__device_6>
%33477 = torch_c.from_builtin_tensor %33476 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33477, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31100 = torch.constant.int 1
%33478 = torch.aten.add.Tensor %33459, %33462, %int1_31100 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31101 = torch.constant.int 1
%33479 = torch.aten.add.Tensor %33478, %33465, %int1_31101 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31102 = torch.constant.int 1
%33480 = torch.aten.add.Tensor %33479, %33468, %int1_31102 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31103 = torch.constant.int 1
%33481 = torch.aten.add.Tensor %33480, %33471, %int1_31103 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31104 = torch.constant.int 1
%33482 = torch.aten.add.Tensor %33481, %33474, %int1_31104 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33482, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31105 = torch.constant.int 1
%33483 = torch.aten.add.Tensor %33482, %33282, %int1_31105 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31106 = torch.constant.int 1
%33484 = torch.aten.add.Tensor %33483, %33477, %int1_31106 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33485 = torch_c.to_builtin_tensor %33246 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31107 = arith.constant 1 : index
%dim_31108 = tensor.dim %33485, %c1_31107 : tensor<4x?x4096xf16>
%33486 = flow.tensor.transfer %33485 : tensor<4x?x4096xf16>{%dim_31108} to #hal.device.promise<@__device_7>
%33487 = torch_c.from_builtin_tensor %33486 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33488 = torch_c.to_builtin_tensor %33252 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31109 = arith.constant 1 : index
%dim_31110 = tensor.dim %33488, %c1_31109 : tensor<4x?x4096xf16>
%33489 = flow.tensor.transfer %33488 : tensor<4x?x4096xf16>{%dim_31110} to #hal.device.promise<@__device_7>
%33490 = torch_c.from_builtin_tensor %33489 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33491 = torch_c.to_builtin_tensor %33258 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31111 = arith.constant 1 : index
%dim_31112 = tensor.dim %33491, %c1_31111 : tensor<4x?x4096xf16>
%33492 = flow.tensor.transfer %33491 : tensor<4x?x4096xf16>{%dim_31112} to #hal.device.promise<@__device_7>
%33493 = torch_c.from_builtin_tensor %33492 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33494 = torch_c.to_builtin_tensor %33264 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31113 = arith.constant 1 : index
%dim_31114 = tensor.dim %33494, %c1_31113 : tensor<4x?x4096xf16>
%33495 = flow.tensor.transfer %33494 : tensor<4x?x4096xf16>{%dim_31114} to #hal.device.promise<@__device_7>
%33496 = torch_c.from_builtin_tensor %33495 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33497 = torch_c.to_builtin_tensor %33270 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31115 = arith.constant 1 : index
%dim_31116 = tensor.dim %33497, %c1_31115 : tensor<4x?x4096xf16>
%33498 = flow.tensor.transfer %33497 : tensor<4x?x4096xf16>{%dim_31116} to #hal.device.promise<@__device_7>
%33499 = torch_c.from_builtin_tensor %33498 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33500 = torch_c.to_builtin_tensor %33276 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31117 = arith.constant 1 : index
%dim_31118 = tensor.dim %33500, %c1_31117 : tensor<4x?x4096xf16>
%33501 = flow.tensor.transfer %33500 : tensor<4x?x4096xf16>{%dim_31118} to #hal.device.promise<@__device_7>
%33502 = torch_c.from_builtin_tensor %33501 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33503 = torch_c.to_builtin_tensor %33282 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31119 = arith.constant 1 : index
%dim_31120 = tensor.dim %33503, %c1_31119 : tensor<4x?x4096xf16>
%33504 = flow.tensor.transfer %33503 : tensor<4x?x4096xf16>{%dim_31120} to #hal.device.promise<@__device_7>
%33505 = torch_c.from_builtin_tensor %33504 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31121 = torch.constant.int 1
%33506 = torch.aten.add.Tensor %33487, %33490, %int1_31121 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31122 = torch.constant.int 1
%33507 = torch.aten.add.Tensor %33506, %33493, %int1_31122 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33507, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31123 = torch.constant.int 1
%33508 = torch.aten.add.Tensor %33507, %33496, %int1_31123 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33508, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31124 = torch.constant.int 1
%33509 = torch.aten.add.Tensor %33508, %33499, %int1_31124 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31125 = torch.constant.int 1
%33510 = torch.aten.add.Tensor %33509, %33502, %int1_31125 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31126 = torch.constant.int 1
%33511 = torch.aten.add.Tensor %33510, %33505, %int1_31126 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31127 = torch.constant.int 1
%33512 = torch.aten.add.Tensor %33511, %33288, %int1_31127 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31128 = torch.constant.int 1
%33513 = torch.aten.add.Tensor %32172, %33316, %int1_31128 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31129 = torch.constant.int 1
%33514 = torch.aten.add.Tensor %32173, %33344, %int1_31129 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31130 = torch.constant.int 1
%33515 = torch.aten.add.Tensor %32174, %33372, %int1_31130 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31131 = torch.constant.int 1
%33516 = torch.aten.add.Tensor %32175, %33400, %int1_31131 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31132 = torch.constant.int 1
%33517 = torch.aten.add.Tensor %32176, %33428, %int1_31132 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31133 = torch.constant.int 1
%33518 = torch.aten.add.Tensor %32177, %33456, %int1_31133 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31134 = torch.constant.int 1
%33519 = torch.aten.add.Tensor %32178, %33484, %int1_31134 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31135 = torch.constant.int 1
%33520 = torch.aten.add.Tensor %32179, %33512, %int1_31135 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_31136 = torch.constant.int 6
%33521 = torch.prims.convert_element_type %33513, %int6_31136 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31137 = torch.constant.int 6
%33522 = torch.prims.convert_element_type %33514, %int6_31137 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31138 = torch.constant.int 6
%33523 = torch.prims.convert_element_type %33515, %int6_31138 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31139 = torch.constant.int 6
%33524 = torch.prims.convert_element_type %33516, %int6_31139 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31140 = torch.constant.int 6
%33525 = torch.prims.convert_element_type %33517, %int6_31140 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31141 = torch.constant.int 6
%33526 = torch.prims.convert_element_type %33518, %int6_31141 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31142 = torch.constant.int 6
%33527 = torch.prims.convert_element_type %33519, %int6_31142 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31143 = torch.constant.int 6
%33528 = torch.prims.convert_element_type %33520, %int6_31143 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31144 = torch.constant.int 2
%33529 = torch.aten.pow.Tensor_Scalar %33521, %int2_31144 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31145 = torch.constant.int 2
%33530 = torch.aten.pow.Tensor_Scalar %33522, %int2_31145 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31146 = torch.constant.int 2
%33531 = torch.aten.pow.Tensor_Scalar %33523, %int2_31146 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31147 = torch.constant.int 2
%33532 = torch.aten.pow.Tensor_Scalar %33524, %int2_31147 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33532, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31148 = torch.constant.int 2
%33533 = torch.aten.pow.Tensor_Scalar %33525, %int2_31148 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31149 = torch.constant.int 2
%33534 = torch.aten.pow.Tensor_Scalar %33526, %int2_31149 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31150 = torch.constant.int 2
%33535 = torch.aten.pow.Tensor_Scalar %33527, %int2_31150 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33535, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31151 = torch.constant.int 2
%33536 = torch.aten.pow.Tensor_Scalar %33528, %int2_31151 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_31152 = torch.constant.int -1
%33537 = torch.prim.ListConstruct %int-1_31152 : (!torch.int) -> !torch.list<int>
%true_31153 = torch.constant.bool true
%none_31154 = torch.constant.none
%33538 = torch.aten.mean.dim %33529, %33537, %true_31153, %none_31154 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33538, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31155 = torch.constant.int -1
%33539 = torch.prim.ListConstruct %int-1_31155 : (!torch.int) -> !torch.list<int>
%true_31156 = torch.constant.bool true
%none_31157 = torch.constant.none
%33540 = torch.aten.mean.dim %33530, %33539, %true_31156, %none_31157 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31158 = torch.constant.int -1
%33541 = torch.prim.ListConstruct %int-1_31158 : (!torch.int) -> !torch.list<int>
%true_31159 = torch.constant.bool true
%none_31160 = torch.constant.none
%33542 = torch.aten.mean.dim %33531, %33541, %true_31159, %none_31160 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31161 = torch.constant.int -1
%33543 = torch.prim.ListConstruct %int-1_31161 : (!torch.int) -> !torch.list<int>
%true_31162 = torch.constant.bool true
%none_31163 = torch.constant.none
%33544 = torch.aten.mean.dim %33532, %33543, %true_31162, %none_31163 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33544, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31164 = torch.constant.int -1
%33545 = torch.prim.ListConstruct %int-1_31164 : (!torch.int) -> !torch.list<int>
%true_31165 = torch.constant.bool true
%none_31166 = torch.constant.none
%33546 = torch.aten.mean.dim %33533, %33545, %true_31165, %none_31166 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31167 = torch.constant.int -1
%33547 = torch.prim.ListConstruct %int-1_31167 : (!torch.int) -> !torch.list<int>
%true_31168 = torch.constant.bool true
%none_31169 = torch.constant.none
%33548 = torch.aten.mean.dim %33534, %33547, %true_31168, %none_31169 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31170 = torch.constant.int -1
%33549 = torch.prim.ListConstruct %int-1_31170 : (!torch.int) -> !torch.list<int>
%true_31171 = torch.constant.bool true
%none_31172 = torch.constant.none
%33550 = torch.aten.mean.dim %33535, %33549, %true_31171, %none_31172 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33550, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31173 = torch.constant.int -1
%33551 = torch.prim.ListConstruct %int-1_31173 : (!torch.int) -> !torch.list<int>
%true_31174 = torch.constant.bool true
%none_31175 = torch.constant.none
%33552 = torch.aten.mean.dim %33536, %33551, %true_31174, %none_31175 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31176 = torch.constant.float 9.9999997473787516E-6
%int1_31177 = torch.constant.int 1
%33553 = torch.aten.add.Scalar %33538, %float9.999990e-06_31176, %int1_31177 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31178 = torch.constant.float 9.9999997473787516E-6
%int1_31179 = torch.constant.int 1
%33554 = torch.aten.add.Scalar %33540, %float9.999990e-06_31178, %int1_31179 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31180 = torch.constant.float 9.9999997473787516E-6
%int1_31181 = torch.constant.int 1
%33555 = torch.aten.add.Scalar %33542, %float9.999990e-06_31180, %int1_31181 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31182 = torch.constant.float 9.9999997473787516E-6
%int1_31183 = torch.constant.int 1
%33556 = torch.aten.add.Scalar %33544, %float9.999990e-06_31182, %int1_31183 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31184 = torch.constant.float 9.9999997473787516E-6
%int1_31185 = torch.constant.int 1
%33557 = torch.aten.add.Scalar %33546, %float9.999990e-06_31184, %int1_31185 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31186 = torch.constant.float 9.9999997473787516E-6
%int1_31187 = torch.constant.int 1
%33558 = torch.aten.add.Scalar %33548, %float9.999990e-06_31186, %int1_31187 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31188 = torch.constant.float 9.9999997473787516E-6
%int1_31189 = torch.constant.int 1
%33559 = torch.aten.add.Scalar %33550, %float9.999990e-06_31188, %int1_31189 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31190 = torch.constant.float 9.9999997473787516E-6
%int1_31191 = torch.constant.int 1
%33560 = torch.aten.add.Scalar %33552, %float9.999990e-06_31190, %int1_31191 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33560, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33561 = torch.aten.rsqrt %33553 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33562 = torch.aten.rsqrt %33554 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33563 = torch.aten.rsqrt %33555 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33564 = torch.aten.rsqrt %33556 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33565 = torch.aten.rsqrt %33557 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33566 = torch.aten.rsqrt %33558 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33566, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33567 = torch.aten.rsqrt %33559 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33568 = torch.aten.rsqrt %33560 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %33568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%33569 = torch.aten.mul.Tensor %33521, %33561 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33569, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33570 = torch.aten.mul.Tensor %33522, %33562 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33571 = torch.aten.mul.Tensor %33523, %33563 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33572 = torch.aten.mul.Tensor %33524, %33564 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33572, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33573 = torch.aten.mul.Tensor %33525, %33565 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33574 = torch.aten.mul.Tensor %33526, %33566 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33575 = torch.aten.mul.Tensor %33527, %33567 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33575, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33576 = torch.aten.mul.Tensor %33528, %33568 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33577 = torch.aten.mul.Tensor %1200, %33569 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33578 = torch.aten.mul.Tensor %1201, %33570 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33579 = torch.aten.mul.Tensor %1202, %33571 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33580 = torch.aten.mul.Tensor %1203, %33572 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33581 = torch.aten.mul.Tensor %1204, %33573 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33582 = torch.aten.mul.Tensor %1205, %33574 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33583 = torch.aten.mul.Tensor %1206, %33575 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%33584 = torch.aten.mul.Tensor %1207, %33576 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %33584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_31192 = torch.constant.int 5
%33585 = torch.prims.convert_element_type %33577, %int5_31192 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31193 = torch.constant.int 5
%33586 = torch.prims.convert_element_type %33578, %int5_31193 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31194 = torch.constant.int 5
%33587 = torch.prims.convert_element_type %33579, %int5_31194 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31195 = torch.constant.int 5
%33588 = torch.prims.convert_element_type %33580, %int5_31195 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31196 = torch.constant.int 5
%33589 = torch.prims.convert_element_type %33581, %int5_31196 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31197 = torch.constant.int 5
%33590 = torch.prims.convert_element_type %33582, %int5_31197 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31198 = torch.constant.int 5
%33591 = torch.prims.convert_element_type %33583, %int5_31198 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31199 = torch.constant.int 5
%33592 = torch.prims.convert_element_type %33584, %int5_31199 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31200 = torch.constant.int 1
%int0_31201 = torch.constant.int 0
%33593 = torch.prim.ListConstruct %int1_31200, %int0_31201 : (!torch.int, !torch.int) -> !torch.list<int>
%33594 = torch.aten.permute %1208, %33593 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31202 = torch.constant.int 1
%int0_31203 = torch.constant.int 0
%33595 = torch.prim.ListConstruct %int1_31202, %int0_31203 : (!torch.int, !torch.int) -> !torch.list<int>
%33596 = torch.aten.permute %1209, %33595 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31204 = torch.constant.int 1
%int0_31205 = torch.constant.int 0
%33597 = torch.prim.ListConstruct %int1_31204, %int0_31205 : (!torch.int, !torch.int) -> !torch.list<int>
%33598 = torch.aten.permute %1210, %33597 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31206 = torch.constant.int 1
%int0_31207 = torch.constant.int 0
%33599 = torch.prim.ListConstruct %int1_31206, %int0_31207 : (!torch.int, !torch.int) -> !torch.list<int>
%33600 = torch.aten.permute %1211, %33599 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31208 = torch.constant.int 1
%int0_31209 = torch.constant.int 0
%33601 = torch.prim.ListConstruct %int1_31208, %int0_31209 : (!torch.int, !torch.int) -> !torch.list<int>
%33602 = torch.aten.permute %1212, %33601 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31210 = torch.constant.int 1
%int0_31211 = torch.constant.int 0
%33603 = torch.prim.ListConstruct %int1_31210, %int0_31211 : (!torch.int, !torch.int) -> !torch.list<int>
%33604 = torch.aten.permute %1213, %33603 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31212 = torch.constant.int 1
%int0_31213 = torch.constant.int 0
%33605 = torch.prim.ListConstruct %int1_31212, %int0_31213 : (!torch.int, !torch.int) -> !torch.list<int>
%33606 = torch.aten.permute %1214, %33605 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31214 = torch.constant.int 1
%int0_31215 = torch.constant.int 0
%33607 = torch.prim.ListConstruct %int1_31214, %int0_31215 : (!torch.int, !torch.int) -> !torch.list<int>
%33608 = torch.aten.permute %1215, %33607 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_31216 = torch.constant.int 4
%33609 = torch.aten.mul.int %int4_31216, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31217 = torch.constant.int 4096
%33610 = torch.prim.ListConstruct %33609, %int4096_31217 : (!torch.int, !torch.int) -> !torch.list<int>
%33611 = torch.aten.view %33585, %33610 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33611, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33612 = torch.aten.mm %33611, %33594 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33612, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31218 = torch.constant.int 4
%int1792_31219 = torch.constant.int 1792
%33613 = torch.prim.ListConstruct %int4_31218, %2482, %int1792_31219 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33614 = torch.aten.view %33612, %33613 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31220 = torch.constant.int 4
%33615 = torch.aten.mul.int %int4_31220, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31221 = torch.constant.int 4096
%33616 = torch.prim.ListConstruct %33615, %int4096_31221 : (!torch.int, !torch.int) -> !torch.list<int>
%33617 = torch.aten.view %33586, %33616 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33617, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33618 = torch.aten.mm %33617, %33596 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33618, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31222 = torch.constant.int 4
%int1792_31223 = torch.constant.int 1792
%33619 = torch.prim.ListConstruct %int4_31222, %2482, %int1792_31223 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33620 = torch.aten.view %33618, %33619 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31224 = torch.constant.int 4
%33621 = torch.aten.mul.int %int4_31224, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31225 = torch.constant.int 4096
%33622 = torch.prim.ListConstruct %33621, %int4096_31225 : (!torch.int, !torch.int) -> !torch.list<int>
%33623 = torch.aten.view %33587, %33622 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33623, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33624 = torch.aten.mm %33623, %33598 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33624, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31226 = torch.constant.int 4
%int1792_31227 = torch.constant.int 1792
%33625 = torch.prim.ListConstruct %int4_31226, %2482, %int1792_31227 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33626 = torch.aten.view %33624, %33625 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31228 = torch.constant.int 4
%33627 = torch.aten.mul.int %int4_31228, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31229 = torch.constant.int 4096
%33628 = torch.prim.ListConstruct %33627, %int4096_31229 : (!torch.int, !torch.int) -> !torch.list<int>
%33629 = torch.aten.view %33588, %33628 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33629, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33630 = torch.aten.mm %33629, %33600 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33630, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31230 = torch.constant.int 4
%int1792_31231 = torch.constant.int 1792
%33631 = torch.prim.ListConstruct %int4_31230, %2482, %int1792_31231 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33632 = torch.aten.view %33630, %33631 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31232 = torch.constant.int 4
%33633 = torch.aten.mul.int %int4_31232, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31233 = torch.constant.int 4096
%33634 = torch.prim.ListConstruct %33633, %int4096_31233 : (!torch.int, !torch.int) -> !torch.list<int>
%33635 = torch.aten.view %33589, %33634 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33635, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33636 = torch.aten.mm %33635, %33602 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33636, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31234 = torch.constant.int 4
%int1792_31235 = torch.constant.int 1792
%33637 = torch.prim.ListConstruct %int4_31234, %2482, %int1792_31235 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33638 = torch.aten.view %33636, %33637 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31236 = torch.constant.int 4
%33639 = torch.aten.mul.int %int4_31236, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31237 = torch.constant.int 4096
%33640 = torch.prim.ListConstruct %33639, %int4096_31237 : (!torch.int, !torch.int) -> !torch.list<int>
%33641 = torch.aten.view %33590, %33640 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33641, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33642 = torch.aten.mm %33641, %33604 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33642, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31238 = torch.constant.int 4
%int1792_31239 = torch.constant.int 1792
%33643 = torch.prim.ListConstruct %int4_31238, %2482, %int1792_31239 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33644 = torch.aten.view %33642, %33643 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33644, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31240 = torch.constant.int 4
%33645 = torch.aten.mul.int %int4_31240, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31241 = torch.constant.int 4096
%33646 = torch.prim.ListConstruct %33645, %int4096_31241 : (!torch.int, !torch.int) -> !torch.list<int>
%33647 = torch.aten.view %33591, %33646 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33647, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33648 = torch.aten.mm %33647, %33606 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33648, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31242 = torch.constant.int 4
%int1792_31243 = torch.constant.int 1792
%33649 = torch.prim.ListConstruct %int4_31242, %2482, %int1792_31243 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33650 = torch.aten.view %33648, %33649 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33650, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31244 = torch.constant.int 4
%33651 = torch.aten.mul.int %int4_31244, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31245 = torch.constant.int 4096
%33652 = torch.prim.ListConstruct %33651, %int4096_31245 : (!torch.int, !torch.int) -> !torch.list<int>
%33653 = torch.aten.view %33592, %33652 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33653, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33654 = torch.aten.mm %33653, %33608 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33654, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31246 = torch.constant.int 4
%int1792_31247 = torch.constant.int 1792
%33655 = torch.prim.ListConstruct %int4_31246, %2482, %int1792_31247 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33656 = torch.aten.view %33654, %33655 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33657 = torch.aten.silu %33614 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33658 = torch.aten.silu %33620 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33659 = torch.aten.silu %33626 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33660 = torch.aten.silu %33632 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33661 = torch.aten.silu %33638 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33662 = torch.aten.silu %33644 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33663 = torch.aten.silu %33650 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33664 = torch.aten.silu %33656 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_31248 = torch.constant.int 1
%int0_31249 = torch.constant.int 0
%33665 = torch.prim.ListConstruct %int1_31248, %int0_31249 : (!torch.int, !torch.int) -> !torch.list<int>
%33666 = torch.aten.permute %1216, %33665 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31250 = torch.constant.int 1
%int0_31251 = torch.constant.int 0
%33667 = torch.prim.ListConstruct %int1_31250, %int0_31251 : (!torch.int, !torch.int) -> !torch.list<int>
%33668 = torch.aten.permute %1217, %33667 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31252 = torch.constant.int 1
%int0_31253 = torch.constant.int 0
%33669 = torch.prim.ListConstruct %int1_31252, %int0_31253 : (!torch.int, !torch.int) -> !torch.list<int>
%33670 = torch.aten.permute %1218, %33669 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31254 = torch.constant.int 1
%int0_31255 = torch.constant.int 0
%33671 = torch.prim.ListConstruct %int1_31254, %int0_31255 : (!torch.int, !torch.int) -> !torch.list<int>
%33672 = torch.aten.permute %1219, %33671 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31256 = torch.constant.int 1
%int0_31257 = torch.constant.int 0
%33673 = torch.prim.ListConstruct %int1_31256, %int0_31257 : (!torch.int, !torch.int) -> !torch.list<int>
%33674 = torch.aten.permute %1220, %33673 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31258 = torch.constant.int 1
%int0_31259 = torch.constant.int 0
%33675 = torch.prim.ListConstruct %int1_31258, %int0_31259 : (!torch.int, !torch.int) -> !torch.list<int>
%33676 = torch.aten.permute %1221, %33675 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31260 = torch.constant.int 1
%int0_31261 = torch.constant.int 0
%33677 = torch.prim.ListConstruct %int1_31260, %int0_31261 : (!torch.int, !torch.int) -> !torch.list<int>
%33678 = torch.aten.permute %1222, %33677 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_31262 = torch.constant.int 1
%int0_31263 = torch.constant.int 0
%33679 = torch.prim.ListConstruct %int1_31262, %int0_31263 : (!torch.int, !torch.int) -> !torch.list<int>
%33680 = torch.aten.permute %1223, %33679 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_31264 = torch.constant.int 4
%33681 = torch.aten.mul.int %int4_31264, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31265 = torch.constant.int 4096
%33682 = torch.prim.ListConstruct %33681, %int4096_31265 : (!torch.int, !torch.int) -> !torch.list<int>
%33683 = torch.aten.view %33585, %33682 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33683, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33684 = torch.aten.mm %33683, %33666 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33684, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31266 = torch.constant.int 4
%int1792_31267 = torch.constant.int 1792
%33685 = torch.prim.ListConstruct %int4_31266, %2482, %int1792_31267 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33686 = torch.aten.view %33684, %33685 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31268 = torch.constant.int 4
%33687 = torch.aten.mul.int %int4_31268, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31269 = torch.constant.int 4096
%33688 = torch.prim.ListConstruct %33687, %int4096_31269 : (!torch.int, !torch.int) -> !torch.list<int>
%33689 = torch.aten.view %33586, %33688 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33689, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33690 = torch.aten.mm %33689, %33668 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33690, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31270 = torch.constant.int 4
%int1792_31271 = torch.constant.int 1792
%33691 = torch.prim.ListConstruct %int4_31270, %2482, %int1792_31271 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33692 = torch.aten.view %33690, %33691 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31272 = torch.constant.int 4
%33693 = torch.aten.mul.int %int4_31272, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31273 = torch.constant.int 4096
%33694 = torch.prim.ListConstruct %33693, %int4096_31273 : (!torch.int, !torch.int) -> !torch.list<int>
%33695 = torch.aten.view %33587, %33694 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33695, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33696 = torch.aten.mm %33695, %33670 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33696, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31274 = torch.constant.int 4
%int1792_31275 = torch.constant.int 1792
%33697 = torch.prim.ListConstruct %int4_31274, %2482, %int1792_31275 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33698 = torch.aten.view %33696, %33697 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31276 = torch.constant.int 4
%33699 = torch.aten.mul.int %int4_31276, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31277 = torch.constant.int 4096
%33700 = torch.prim.ListConstruct %33699, %int4096_31277 : (!torch.int, !torch.int) -> !torch.list<int>
%33701 = torch.aten.view %33588, %33700 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33701, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33702 = torch.aten.mm %33701, %33672 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33702, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31278 = torch.constant.int 4
%int1792_31279 = torch.constant.int 1792
%33703 = torch.prim.ListConstruct %int4_31278, %2482, %int1792_31279 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33704 = torch.aten.view %33702, %33703 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31280 = torch.constant.int 4
%33705 = torch.aten.mul.int %int4_31280, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31281 = torch.constant.int 4096
%33706 = torch.prim.ListConstruct %33705, %int4096_31281 : (!torch.int, !torch.int) -> !torch.list<int>
%33707 = torch.aten.view %33589, %33706 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33707, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33708 = torch.aten.mm %33707, %33674 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33708, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31282 = torch.constant.int 4
%int1792_31283 = torch.constant.int 1792
%33709 = torch.prim.ListConstruct %int4_31282, %2482, %int1792_31283 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33710 = torch.aten.view %33708, %33709 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31284 = torch.constant.int 4
%33711 = torch.aten.mul.int %int4_31284, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31285 = torch.constant.int 4096
%33712 = torch.prim.ListConstruct %33711, %int4096_31285 : (!torch.int, !torch.int) -> !torch.list<int>
%33713 = torch.aten.view %33590, %33712 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33713, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33714 = torch.aten.mm %33713, %33676 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33714, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31286 = torch.constant.int 4
%int1792_31287 = torch.constant.int 1792
%33715 = torch.prim.ListConstruct %int4_31286, %2482, %int1792_31287 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33716 = torch.aten.view %33714, %33715 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31288 = torch.constant.int 4
%33717 = torch.aten.mul.int %int4_31288, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31289 = torch.constant.int 4096
%33718 = torch.prim.ListConstruct %33717, %int4096_31289 : (!torch.int, !torch.int) -> !torch.list<int>
%33719 = torch.aten.view %33591, %33718 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33719, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33720 = torch.aten.mm %33719, %33678 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33720, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31290 = torch.constant.int 4
%int1792_31291 = torch.constant.int 1792
%33721 = torch.prim.ListConstruct %int4_31290, %2482, %int1792_31291 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33722 = torch.aten.view %33720, %33721 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_31292 = torch.constant.int 4
%33723 = torch.aten.mul.int %int4_31292, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31293 = torch.constant.int 4096
%33724 = torch.prim.ListConstruct %33723, %int4096_31293 : (!torch.int, !torch.int) -> !torch.list<int>
%33725 = torch.aten.view %33592, %33724 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33725, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%33726 = torch.aten.mm %33725, %33680 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33726, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_31294 = torch.constant.int 4
%int1792_31295 = torch.constant.int 1792
%33727 = torch.prim.ListConstruct %int4_31294, %2482, %int1792_31295 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33728 = torch.aten.view %33726, %33727 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33729 = torch.aten.mul.Tensor %33657, %33686 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33730 = torch.aten.mul.Tensor %33658, %33692 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33730, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33731 = torch.aten.mul.Tensor %33659, %33698 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33732 = torch.aten.mul.Tensor %33660, %33704 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33733 = torch.aten.mul.Tensor %33661, %33710 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33734 = torch.aten.mul.Tensor %33662, %33716 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33735 = torch.aten.mul.Tensor %33663, %33722 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%33736 = torch.aten.mul.Tensor %33664, %33728 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %33736, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_31296 = torch.constant.int 1
%int0_31297 = torch.constant.int 0
%33737 = torch.prim.ListConstruct %int1_31296, %int0_31297 : (!torch.int, !torch.int) -> !torch.list<int>
%33738 = torch.aten.permute %1224, %33737 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31298 = torch.constant.int 1
%int0_31299 = torch.constant.int 0
%33739 = torch.prim.ListConstruct %int1_31298, %int0_31299 : (!torch.int, !torch.int) -> !torch.list<int>
%33740 = torch.aten.permute %1225, %33739 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31300 = torch.constant.int 1
%int0_31301 = torch.constant.int 0
%33741 = torch.prim.ListConstruct %int1_31300, %int0_31301 : (!torch.int, !torch.int) -> !torch.list<int>
%33742 = torch.aten.permute %1226, %33741 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31302 = torch.constant.int 1
%int0_31303 = torch.constant.int 0
%33743 = torch.prim.ListConstruct %int1_31302, %int0_31303 : (!torch.int, !torch.int) -> !torch.list<int>
%33744 = torch.aten.permute %1227, %33743 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31304 = torch.constant.int 1
%int0_31305 = torch.constant.int 0
%33745 = torch.prim.ListConstruct %int1_31304, %int0_31305 : (!torch.int, !torch.int) -> !torch.list<int>
%33746 = torch.aten.permute %1228, %33745 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31306 = torch.constant.int 1
%int0_31307 = torch.constant.int 0
%33747 = torch.prim.ListConstruct %int1_31306, %int0_31307 : (!torch.int, !torch.int) -> !torch.list<int>
%33748 = torch.aten.permute %1229, %33747 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31308 = torch.constant.int 1
%int0_31309 = torch.constant.int 0
%33749 = torch.prim.ListConstruct %int1_31308, %int0_31309 : (!torch.int, !torch.int) -> !torch.list<int>
%33750 = torch.aten.permute %1230, %33749 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31310 = torch.constant.int 1
%int0_31311 = torch.constant.int 0
%33751 = torch.prim.ListConstruct %int1_31310, %int0_31311 : (!torch.int, !torch.int) -> !torch.list<int>
%33752 = torch.aten.permute %1231, %33751 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_31312 = torch.constant.int 1
%33753 = torch.aten.size.int %33614, %int1_31312 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31313 = torch.constant.int 4
%33754 = torch.aten.mul.int %int4_31313, %33753 : !torch.int, !torch.int -> !torch.int
%int1792_31314 = torch.constant.int 1792
%33755 = torch.prim.ListConstruct %33754, %int1792_31314 : (!torch.int, !torch.int) -> !torch.list<int>
%33756 = torch.aten.view %33729, %33755 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33756, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33757 = torch.aten.mm %33756, %33738 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33757, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31315 = torch.constant.int 4
%int4096_31316 = torch.constant.int 4096
%33758 = torch.prim.ListConstruct %int4_31315, %33753, %int4096_31316 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33759 = torch.aten.view %33757, %33758 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31317 = torch.constant.int 1
%33760 = torch.aten.size.int %33620, %int1_31317 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31318 = torch.constant.int 4
%33761 = torch.aten.mul.int %int4_31318, %33760 : !torch.int, !torch.int -> !torch.int
%int1792_31319 = torch.constant.int 1792
%33762 = torch.prim.ListConstruct %33761, %int1792_31319 : (!torch.int, !torch.int) -> !torch.list<int>
%33763 = torch.aten.view %33730, %33762 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33763, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33764 = torch.aten.mm %33763, %33740 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33764, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31320 = torch.constant.int 4
%int4096_31321 = torch.constant.int 4096
%33765 = torch.prim.ListConstruct %int4_31320, %33760, %int4096_31321 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33766 = torch.aten.view %33764, %33765 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31322 = torch.constant.int 1
%33767 = torch.aten.size.int %33626, %int1_31322 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31323 = torch.constant.int 4
%33768 = torch.aten.mul.int %int4_31323, %33767 : !torch.int, !torch.int -> !torch.int
%int1792_31324 = torch.constant.int 1792
%33769 = torch.prim.ListConstruct %33768, %int1792_31324 : (!torch.int, !torch.int) -> !torch.list<int>
%33770 = torch.aten.view %33731, %33769 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33770, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33771 = torch.aten.mm %33770, %33742 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33771, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31325 = torch.constant.int 4
%int4096_31326 = torch.constant.int 4096
%33772 = torch.prim.ListConstruct %int4_31325, %33767, %int4096_31326 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33773 = torch.aten.view %33771, %33772 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31327 = torch.constant.int 1
%33774 = torch.aten.size.int %33632, %int1_31327 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31328 = torch.constant.int 4
%33775 = torch.aten.mul.int %int4_31328, %33774 : !torch.int, !torch.int -> !torch.int
%int1792_31329 = torch.constant.int 1792
%33776 = torch.prim.ListConstruct %33775, %int1792_31329 : (!torch.int, !torch.int) -> !torch.list<int>
%33777 = torch.aten.view %33732, %33776 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33777, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33778 = torch.aten.mm %33777, %33744 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33778, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31330 = torch.constant.int 4
%int4096_31331 = torch.constant.int 4096
%33779 = torch.prim.ListConstruct %int4_31330, %33774, %int4096_31331 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33780 = torch.aten.view %33778, %33779 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31332 = torch.constant.int 1
%33781 = torch.aten.size.int %33638, %int1_31332 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31333 = torch.constant.int 4
%33782 = torch.aten.mul.int %int4_31333, %33781 : !torch.int, !torch.int -> !torch.int
%int1792_31334 = torch.constant.int 1792
%33783 = torch.prim.ListConstruct %33782, %int1792_31334 : (!torch.int, !torch.int) -> !torch.list<int>
%33784 = torch.aten.view %33733, %33783 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33784, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33785 = torch.aten.mm %33784, %33746 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33785, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31335 = torch.constant.int 4
%int4096_31336 = torch.constant.int 4096
%33786 = torch.prim.ListConstruct %int4_31335, %33781, %int4096_31336 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33787 = torch.aten.view %33785, %33786 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31337 = torch.constant.int 1
%33788 = torch.aten.size.int %33644, %int1_31337 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31338 = torch.constant.int 4
%33789 = torch.aten.mul.int %int4_31338, %33788 : !torch.int, !torch.int -> !torch.int
%int1792_31339 = torch.constant.int 1792
%33790 = torch.prim.ListConstruct %33789, %int1792_31339 : (!torch.int, !torch.int) -> !torch.list<int>
%33791 = torch.aten.view %33734, %33790 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33791, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33792 = torch.aten.mm %33791, %33748 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33792, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31340 = torch.constant.int 4
%int4096_31341 = torch.constant.int 4096
%33793 = torch.prim.ListConstruct %int4_31340, %33788, %int4096_31341 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33794 = torch.aten.view %33792, %33793 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31342 = torch.constant.int 1
%33795 = torch.aten.size.int %33650, %int1_31342 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31343 = torch.constant.int 4
%33796 = torch.aten.mul.int %int4_31343, %33795 : !torch.int, !torch.int -> !torch.int
%int1792_31344 = torch.constant.int 1792
%33797 = torch.prim.ListConstruct %33796, %int1792_31344 : (!torch.int, !torch.int) -> !torch.list<int>
%33798 = torch.aten.view %33735, %33797 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33798, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33799 = torch.aten.mm %33798, %33750 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33799, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31345 = torch.constant.int 4
%int4096_31346 = torch.constant.int 4096
%33800 = torch.prim.ListConstruct %int4_31345, %33795, %int4096_31346 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33801 = torch.aten.view %33799, %33800 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31347 = torch.constant.int 1
%33802 = torch.aten.size.int %33656, %int1_31347 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_31348 = torch.constant.int 4
%33803 = torch.aten.mul.int %int4_31348, %33802 : !torch.int, !torch.int -> !torch.int
%int1792_31349 = torch.constant.int 1792
%33804 = torch.prim.ListConstruct %33803, %int1792_31349 : (!torch.int, !torch.int) -> !torch.list<int>
%33805 = torch.aten.view %33736, %33804 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %33805, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%33806 = torch.aten.mm %33805, %33752 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %33806, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_31350 = torch.constant.int 4
%int4096_31351 = torch.constant.int 4096
%33807 = torch.prim.ListConstruct %int4_31350, %33802, %int4096_31351 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%33808 = torch.aten.view %33806, %33807 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33809 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31352 = arith.constant 1 : index
%dim_31353 = tensor.dim %33809, %c1_31352 : tensor<4x?x4096xf16>
%33810 = flow.tensor.transfer %33809 : tensor<4x?x4096xf16>{%dim_31353} to #hal.device.promise<@__device_0>
%33811 = torch_c.from_builtin_tensor %33810 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33812 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31354 = arith.constant 1 : index
%dim_31355 = tensor.dim %33812, %c1_31354 : tensor<4x?x4096xf16>
%33813 = flow.tensor.transfer %33812 : tensor<4x?x4096xf16>{%dim_31355} to #hal.device.promise<@__device_0>
%33814 = torch_c.from_builtin_tensor %33813 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33815 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31356 = arith.constant 1 : index
%dim_31357 = tensor.dim %33815, %c1_31356 : tensor<4x?x4096xf16>
%33816 = flow.tensor.transfer %33815 : tensor<4x?x4096xf16>{%dim_31357} to #hal.device.promise<@__device_0>
%33817 = torch_c.from_builtin_tensor %33816 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33818 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31358 = arith.constant 1 : index
%dim_31359 = tensor.dim %33818, %c1_31358 : tensor<4x?x4096xf16>
%33819 = flow.tensor.transfer %33818 : tensor<4x?x4096xf16>{%dim_31359} to #hal.device.promise<@__device_0>
%33820 = torch_c.from_builtin_tensor %33819 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33821 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31360 = arith.constant 1 : index
%dim_31361 = tensor.dim %33821, %c1_31360 : tensor<4x?x4096xf16>
%33822 = flow.tensor.transfer %33821 : tensor<4x?x4096xf16>{%dim_31361} to #hal.device.promise<@__device_0>
%33823 = torch_c.from_builtin_tensor %33822 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33824 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31362 = arith.constant 1 : index
%dim_31363 = tensor.dim %33824, %c1_31362 : tensor<4x?x4096xf16>
%33825 = flow.tensor.transfer %33824 : tensor<4x?x4096xf16>{%dim_31363} to #hal.device.promise<@__device_0>
%33826 = torch_c.from_builtin_tensor %33825 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33827 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31364 = arith.constant 1 : index
%dim_31365 = tensor.dim %33827, %c1_31364 : tensor<4x?x4096xf16>
%33828 = flow.tensor.transfer %33827 : tensor<4x?x4096xf16>{%dim_31365} to #hal.device.promise<@__device_0>
%33829 = torch_c.from_builtin_tensor %33828 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31366 = torch.constant.int 1
%33830 = torch.aten.add.Tensor %33759, %33811, %int1_31366 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31367 = torch.constant.int 1
%33831 = torch.aten.add.Tensor %33830, %33814, %int1_31367 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31368 = torch.constant.int 1
%33832 = torch.aten.add.Tensor %33831, %33817, %int1_31368 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31369 = torch.constant.int 1
%33833 = torch.aten.add.Tensor %33832, %33820, %int1_31369 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31370 = torch.constant.int 1
%33834 = torch.aten.add.Tensor %33833, %33823, %int1_31370 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31371 = torch.constant.int 1
%33835 = torch.aten.add.Tensor %33834, %33826, %int1_31371 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31372 = torch.constant.int 1
%33836 = torch.aten.add.Tensor %33835, %33829, %int1_31372 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33837 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31373 = arith.constant 1 : index
%dim_31374 = tensor.dim %33837, %c1_31373 : tensor<4x?x4096xf16>
%33838 = flow.tensor.transfer %33837 : tensor<4x?x4096xf16>{%dim_31374} to #hal.device.promise<@__device_1>
%33839 = torch_c.from_builtin_tensor %33838 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33840 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31375 = arith.constant 1 : index
%dim_31376 = tensor.dim %33840, %c1_31375 : tensor<4x?x4096xf16>
%33841 = flow.tensor.transfer %33840 : tensor<4x?x4096xf16>{%dim_31376} to #hal.device.promise<@__device_1>
%33842 = torch_c.from_builtin_tensor %33841 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33842, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33843 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31377 = arith.constant 1 : index
%dim_31378 = tensor.dim %33843, %c1_31377 : tensor<4x?x4096xf16>
%33844 = flow.tensor.transfer %33843 : tensor<4x?x4096xf16>{%dim_31378} to #hal.device.promise<@__device_1>
%33845 = torch_c.from_builtin_tensor %33844 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33846 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31379 = arith.constant 1 : index
%dim_31380 = tensor.dim %33846, %c1_31379 : tensor<4x?x4096xf16>
%33847 = flow.tensor.transfer %33846 : tensor<4x?x4096xf16>{%dim_31380} to #hal.device.promise<@__device_1>
%33848 = torch_c.from_builtin_tensor %33847 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33848, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33849 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31381 = arith.constant 1 : index
%dim_31382 = tensor.dim %33849, %c1_31381 : tensor<4x?x4096xf16>
%33850 = flow.tensor.transfer %33849 : tensor<4x?x4096xf16>{%dim_31382} to #hal.device.promise<@__device_1>
%33851 = torch_c.from_builtin_tensor %33850 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33852 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31383 = arith.constant 1 : index
%dim_31384 = tensor.dim %33852, %c1_31383 : tensor<4x?x4096xf16>
%33853 = flow.tensor.transfer %33852 : tensor<4x?x4096xf16>{%dim_31384} to #hal.device.promise<@__device_1>
%33854 = torch_c.from_builtin_tensor %33853 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33855 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31385 = arith.constant 1 : index
%dim_31386 = tensor.dim %33855, %c1_31385 : tensor<4x?x4096xf16>
%33856 = flow.tensor.transfer %33855 : tensor<4x?x4096xf16>{%dim_31386} to #hal.device.promise<@__device_1>
%33857 = torch_c.from_builtin_tensor %33856 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31387 = torch.constant.int 1
%33858 = torch.aten.add.Tensor %33839, %33766, %int1_31387 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31388 = torch.constant.int 1
%33859 = torch.aten.add.Tensor %33858, %33842, %int1_31388 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31389 = torch.constant.int 1
%33860 = torch.aten.add.Tensor %33859, %33845, %int1_31389 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31390 = torch.constant.int 1
%33861 = torch.aten.add.Tensor %33860, %33848, %int1_31390 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31391 = torch.constant.int 1
%33862 = torch.aten.add.Tensor %33861, %33851, %int1_31391 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31392 = torch.constant.int 1
%33863 = torch.aten.add.Tensor %33862, %33854, %int1_31392 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31393 = torch.constant.int 1
%33864 = torch.aten.add.Tensor %33863, %33857, %int1_31393 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33865 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31394 = arith.constant 1 : index
%dim_31395 = tensor.dim %33865, %c1_31394 : tensor<4x?x4096xf16>
%33866 = flow.tensor.transfer %33865 : tensor<4x?x4096xf16>{%dim_31395} to #hal.device.promise<@__device_2>
%33867 = torch_c.from_builtin_tensor %33866 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33867, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33868 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31396 = arith.constant 1 : index
%dim_31397 = tensor.dim %33868, %c1_31396 : tensor<4x?x4096xf16>
%33869 = flow.tensor.transfer %33868 : tensor<4x?x4096xf16>{%dim_31397} to #hal.device.promise<@__device_2>
%33870 = torch_c.from_builtin_tensor %33869 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33871 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31398 = arith.constant 1 : index
%dim_31399 = tensor.dim %33871, %c1_31398 : tensor<4x?x4096xf16>
%33872 = flow.tensor.transfer %33871 : tensor<4x?x4096xf16>{%dim_31399} to #hal.device.promise<@__device_2>
%33873 = torch_c.from_builtin_tensor %33872 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33873, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33874 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31400 = arith.constant 1 : index
%dim_31401 = tensor.dim %33874, %c1_31400 : tensor<4x?x4096xf16>
%33875 = flow.tensor.transfer %33874 : tensor<4x?x4096xf16>{%dim_31401} to #hal.device.promise<@__device_2>
%33876 = torch_c.from_builtin_tensor %33875 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33877 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31402 = arith.constant 1 : index
%dim_31403 = tensor.dim %33877, %c1_31402 : tensor<4x?x4096xf16>
%33878 = flow.tensor.transfer %33877 : tensor<4x?x4096xf16>{%dim_31403} to #hal.device.promise<@__device_2>
%33879 = torch_c.from_builtin_tensor %33878 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33879, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33880 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31404 = arith.constant 1 : index
%dim_31405 = tensor.dim %33880, %c1_31404 : tensor<4x?x4096xf16>
%33881 = flow.tensor.transfer %33880 : tensor<4x?x4096xf16>{%dim_31405} to #hal.device.promise<@__device_2>
%33882 = torch_c.from_builtin_tensor %33881 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33883 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31406 = arith.constant 1 : index
%dim_31407 = tensor.dim %33883, %c1_31406 : tensor<4x?x4096xf16>
%33884 = flow.tensor.transfer %33883 : tensor<4x?x4096xf16>{%dim_31407} to #hal.device.promise<@__device_2>
%33885 = torch_c.from_builtin_tensor %33884 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33885, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31408 = torch.constant.int 1
%33886 = torch.aten.add.Tensor %33867, %33870, %int1_31408 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31409 = torch.constant.int 1
%33887 = torch.aten.add.Tensor %33886, %33773, %int1_31409 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31410 = torch.constant.int 1
%33888 = torch.aten.add.Tensor %33887, %33873, %int1_31410 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31411 = torch.constant.int 1
%33889 = torch.aten.add.Tensor %33888, %33876, %int1_31411 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31412 = torch.constant.int 1
%33890 = torch.aten.add.Tensor %33889, %33879, %int1_31412 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31413 = torch.constant.int 1
%33891 = torch.aten.add.Tensor %33890, %33882, %int1_31413 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31414 = torch.constant.int 1
%33892 = torch.aten.add.Tensor %33891, %33885, %int1_31414 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33893 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31415 = arith.constant 1 : index
%dim_31416 = tensor.dim %33893, %c1_31415 : tensor<4x?x4096xf16>
%33894 = flow.tensor.transfer %33893 : tensor<4x?x4096xf16>{%dim_31416} to #hal.device.promise<@__device_3>
%33895 = torch_c.from_builtin_tensor %33894 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33896 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31417 = arith.constant 1 : index
%dim_31418 = tensor.dim %33896, %c1_31417 : tensor<4x?x4096xf16>
%33897 = flow.tensor.transfer %33896 : tensor<4x?x4096xf16>{%dim_31418} to #hal.device.promise<@__device_3>
%33898 = torch_c.from_builtin_tensor %33897 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33899 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31419 = arith.constant 1 : index
%dim_31420 = tensor.dim %33899, %c1_31419 : tensor<4x?x4096xf16>
%33900 = flow.tensor.transfer %33899 : tensor<4x?x4096xf16>{%dim_31420} to #hal.device.promise<@__device_3>
%33901 = torch_c.from_builtin_tensor %33900 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33902 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31421 = arith.constant 1 : index
%dim_31422 = tensor.dim %33902, %c1_31421 : tensor<4x?x4096xf16>
%33903 = flow.tensor.transfer %33902 : tensor<4x?x4096xf16>{%dim_31422} to #hal.device.promise<@__device_3>
%33904 = torch_c.from_builtin_tensor %33903 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33905 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31423 = arith.constant 1 : index
%dim_31424 = tensor.dim %33905, %c1_31423 : tensor<4x?x4096xf16>
%33906 = flow.tensor.transfer %33905 : tensor<4x?x4096xf16>{%dim_31424} to #hal.device.promise<@__device_3>
%33907 = torch_c.from_builtin_tensor %33906 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33908 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31425 = arith.constant 1 : index
%dim_31426 = tensor.dim %33908, %c1_31425 : tensor<4x?x4096xf16>
%33909 = flow.tensor.transfer %33908 : tensor<4x?x4096xf16>{%dim_31426} to #hal.device.promise<@__device_3>
%33910 = torch_c.from_builtin_tensor %33909 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33911 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31427 = arith.constant 1 : index
%dim_31428 = tensor.dim %33911, %c1_31427 : tensor<4x?x4096xf16>
%33912 = flow.tensor.transfer %33911 : tensor<4x?x4096xf16>{%dim_31428} to #hal.device.promise<@__device_3>
%33913 = torch_c.from_builtin_tensor %33912 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31429 = torch.constant.int 1
%33914 = torch.aten.add.Tensor %33895, %33898, %int1_31429 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31430 = torch.constant.int 1
%33915 = torch.aten.add.Tensor %33914, %33901, %int1_31430 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31431 = torch.constant.int 1
%33916 = torch.aten.add.Tensor %33915, %33780, %int1_31431 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31432 = torch.constant.int 1
%33917 = torch.aten.add.Tensor %33916, %33904, %int1_31432 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31433 = torch.constant.int 1
%33918 = torch.aten.add.Tensor %33917, %33907, %int1_31433 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33918, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31434 = torch.constant.int 1
%33919 = torch.aten.add.Tensor %33918, %33910, %int1_31434 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31435 = torch.constant.int 1
%33920 = torch.aten.add.Tensor %33919, %33913, %int1_31435 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33921 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31436 = arith.constant 1 : index
%dim_31437 = tensor.dim %33921, %c1_31436 : tensor<4x?x4096xf16>
%33922 = flow.tensor.transfer %33921 : tensor<4x?x4096xf16>{%dim_31437} to #hal.device.promise<@__device_4>
%33923 = torch_c.from_builtin_tensor %33922 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33924 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31438 = arith.constant 1 : index
%dim_31439 = tensor.dim %33924, %c1_31438 : tensor<4x?x4096xf16>
%33925 = flow.tensor.transfer %33924 : tensor<4x?x4096xf16>{%dim_31439} to #hal.device.promise<@__device_4>
%33926 = torch_c.from_builtin_tensor %33925 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33927 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31440 = arith.constant 1 : index
%dim_31441 = tensor.dim %33927, %c1_31440 : tensor<4x?x4096xf16>
%33928 = flow.tensor.transfer %33927 : tensor<4x?x4096xf16>{%dim_31441} to #hal.device.promise<@__device_4>
%33929 = torch_c.from_builtin_tensor %33928 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33930 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31442 = arith.constant 1 : index
%dim_31443 = tensor.dim %33930, %c1_31442 : tensor<4x?x4096xf16>
%33931 = flow.tensor.transfer %33930 : tensor<4x?x4096xf16>{%dim_31443} to #hal.device.promise<@__device_4>
%33932 = torch_c.from_builtin_tensor %33931 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33933 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31444 = arith.constant 1 : index
%dim_31445 = tensor.dim %33933, %c1_31444 : tensor<4x?x4096xf16>
%33934 = flow.tensor.transfer %33933 : tensor<4x?x4096xf16>{%dim_31445} to #hal.device.promise<@__device_4>
%33935 = torch_c.from_builtin_tensor %33934 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33936 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31446 = arith.constant 1 : index
%dim_31447 = tensor.dim %33936, %c1_31446 : tensor<4x?x4096xf16>
%33937 = flow.tensor.transfer %33936 : tensor<4x?x4096xf16>{%dim_31447} to #hal.device.promise<@__device_4>
%33938 = torch_c.from_builtin_tensor %33937 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33939 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31448 = arith.constant 1 : index
%dim_31449 = tensor.dim %33939, %c1_31448 : tensor<4x?x4096xf16>
%33940 = flow.tensor.transfer %33939 : tensor<4x?x4096xf16>{%dim_31449} to #hal.device.promise<@__device_4>
%33941 = torch_c.from_builtin_tensor %33940 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31450 = torch.constant.int 1
%33942 = torch.aten.add.Tensor %33923, %33926, %int1_31450 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31451 = torch.constant.int 1
%33943 = torch.aten.add.Tensor %33942, %33929, %int1_31451 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31452 = torch.constant.int 1
%33944 = torch.aten.add.Tensor %33943, %33932, %int1_31452 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31453 = torch.constant.int 1
%33945 = torch.aten.add.Tensor %33944, %33787, %int1_31453 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31454 = torch.constant.int 1
%33946 = torch.aten.add.Tensor %33945, %33935, %int1_31454 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31455 = torch.constant.int 1
%33947 = torch.aten.add.Tensor %33946, %33938, %int1_31455 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31456 = torch.constant.int 1
%33948 = torch.aten.add.Tensor %33947, %33941, %int1_31456 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33949 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31457 = arith.constant 1 : index
%dim_31458 = tensor.dim %33949, %c1_31457 : tensor<4x?x4096xf16>
%33950 = flow.tensor.transfer %33949 : tensor<4x?x4096xf16>{%dim_31458} to #hal.device.promise<@__device_5>
%33951 = torch_c.from_builtin_tensor %33950 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33952 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31459 = arith.constant 1 : index
%dim_31460 = tensor.dim %33952, %c1_31459 : tensor<4x?x4096xf16>
%33953 = flow.tensor.transfer %33952 : tensor<4x?x4096xf16>{%dim_31460} to #hal.device.promise<@__device_5>
%33954 = torch_c.from_builtin_tensor %33953 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33955 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31461 = arith.constant 1 : index
%dim_31462 = tensor.dim %33955, %c1_31461 : tensor<4x?x4096xf16>
%33956 = flow.tensor.transfer %33955 : tensor<4x?x4096xf16>{%dim_31462} to #hal.device.promise<@__device_5>
%33957 = torch_c.from_builtin_tensor %33956 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33958 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31463 = arith.constant 1 : index
%dim_31464 = tensor.dim %33958, %c1_31463 : tensor<4x?x4096xf16>
%33959 = flow.tensor.transfer %33958 : tensor<4x?x4096xf16>{%dim_31464} to #hal.device.promise<@__device_5>
%33960 = torch_c.from_builtin_tensor %33959 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33961 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31465 = arith.constant 1 : index
%dim_31466 = tensor.dim %33961, %c1_31465 : tensor<4x?x4096xf16>
%33962 = flow.tensor.transfer %33961 : tensor<4x?x4096xf16>{%dim_31466} to #hal.device.promise<@__device_5>
%33963 = torch_c.from_builtin_tensor %33962 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33964 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31467 = arith.constant 1 : index
%dim_31468 = tensor.dim %33964, %c1_31467 : tensor<4x?x4096xf16>
%33965 = flow.tensor.transfer %33964 : tensor<4x?x4096xf16>{%dim_31468} to #hal.device.promise<@__device_5>
%33966 = torch_c.from_builtin_tensor %33965 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33967 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31469 = arith.constant 1 : index
%dim_31470 = tensor.dim %33967, %c1_31469 : tensor<4x?x4096xf16>
%33968 = flow.tensor.transfer %33967 : tensor<4x?x4096xf16>{%dim_31470} to #hal.device.promise<@__device_5>
%33969 = torch_c.from_builtin_tensor %33968 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31471 = torch.constant.int 1
%33970 = torch.aten.add.Tensor %33951, %33954, %int1_31471 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31472 = torch.constant.int 1
%33971 = torch.aten.add.Tensor %33970, %33957, %int1_31472 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31473 = torch.constant.int 1
%33972 = torch.aten.add.Tensor %33971, %33960, %int1_31473 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31474 = torch.constant.int 1
%33973 = torch.aten.add.Tensor %33972, %33963, %int1_31474 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31475 = torch.constant.int 1
%33974 = torch.aten.add.Tensor %33973, %33794, %int1_31475 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31476 = torch.constant.int 1
%33975 = torch.aten.add.Tensor %33974, %33966, %int1_31476 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33975, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31477 = torch.constant.int 1
%33976 = torch.aten.add.Tensor %33975, %33969, %int1_31477 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33977 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31478 = arith.constant 1 : index
%dim_31479 = tensor.dim %33977, %c1_31478 : tensor<4x?x4096xf16>
%33978 = flow.tensor.transfer %33977 : tensor<4x?x4096xf16>{%dim_31479} to #hal.device.promise<@__device_6>
%33979 = torch_c.from_builtin_tensor %33978 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33980 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31480 = arith.constant 1 : index
%dim_31481 = tensor.dim %33980, %c1_31480 : tensor<4x?x4096xf16>
%33981 = flow.tensor.transfer %33980 : tensor<4x?x4096xf16>{%dim_31481} to #hal.device.promise<@__device_6>
%33982 = torch_c.from_builtin_tensor %33981 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33983 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31482 = arith.constant 1 : index
%dim_31483 = tensor.dim %33983, %c1_31482 : tensor<4x?x4096xf16>
%33984 = flow.tensor.transfer %33983 : tensor<4x?x4096xf16>{%dim_31483} to #hal.device.promise<@__device_6>
%33985 = torch_c.from_builtin_tensor %33984 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33985, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33986 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31484 = arith.constant 1 : index
%dim_31485 = tensor.dim %33986, %c1_31484 : tensor<4x?x4096xf16>
%33987 = flow.tensor.transfer %33986 : tensor<4x?x4096xf16>{%dim_31485} to #hal.device.promise<@__device_6>
%33988 = torch_c.from_builtin_tensor %33987 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33989 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31486 = arith.constant 1 : index
%dim_31487 = tensor.dim %33989, %c1_31486 : tensor<4x?x4096xf16>
%33990 = flow.tensor.transfer %33989 : tensor<4x?x4096xf16>{%dim_31487} to #hal.device.promise<@__device_6>
%33991 = torch_c.from_builtin_tensor %33990 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33992 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31488 = arith.constant 1 : index
%dim_31489 = tensor.dim %33992, %c1_31488 : tensor<4x?x4096xf16>
%33993 = flow.tensor.transfer %33992 : tensor<4x?x4096xf16>{%dim_31489} to #hal.device.promise<@__device_6>
%33994 = torch_c.from_builtin_tensor %33993 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%33995 = torch_c.to_builtin_tensor %33808 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31490 = arith.constant 1 : index
%dim_31491 = tensor.dim %33995, %c1_31490 : tensor<4x?x4096xf16>
%33996 = flow.tensor.transfer %33995 : tensor<4x?x4096xf16>{%dim_31491} to #hal.device.promise<@__device_6>
%33997 = torch_c.from_builtin_tensor %33996 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33997, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31492 = torch.constant.int 1
%33998 = torch.aten.add.Tensor %33979, %33982, %int1_31492 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31493 = torch.constant.int 1
%33999 = torch.aten.add.Tensor %33998, %33985, %int1_31493 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %33999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31494 = torch.constant.int 1
%34000 = torch.aten.add.Tensor %33999, %33988, %int1_31494 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34000, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31495 = torch.constant.int 1
%34001 = torch.aten.add.Tensor %34000, %33991, %int1_31495 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31496 = torch.constant.int 1
%34002 = torch.aten.add.Tensor %34001, %33994, %int1_31496 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31497 = torch.constant.int 1
%34003 = torch.aten.add.Tensor %34002, %33801, %int1_31497 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31498 = torch.constant.int 1
%34004 = torch.aten.add.Tensor %34003, %33997, %int1_31498 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34005 = torch_c.to_builtin_tensor %33759 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31499 = arith.constant 1 : index
%dim_31500 = tensor.dim %34005, %c1_31499 : tensor<4x?x4096xf16>
%34006 = flow.tensor.transfer %34005 : tensor<4x?x4096xf16>{%dim_31500} to #hal.device.promise<@__device_7>
%34007 = torch_c.from_builtin_tensor %34006 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34008 = torch_c.to_builtin_tensor %33766 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31501 = arith.constant 1 : index
%dim_31502 = tensor.dim %34008, %c1_31501 : tensor<4x?x4096xf16>
%34009 = flow.tensor.transfer %34008 : tensor<4x?x4096xf16>{%dim_31502} to #hal.device.promise<@__device_7>
%34010 = torch_c.from_builtin_tensor %34009 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34011 = torch_c.to_builtin_tensor %33773 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31503 = arith.constant 1 : index
%dim_31504 = tensor.dim %34011, %c1_31503 : tensor<4x?x4096xf16>
%34012 = flow.tensor.transfer %34011 : tensor<4x?x4096xf16>{%dim_31504} to #hal.device.promise<@__device_7>
%34013 = torch_c.from_builtin_tensor %34012 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34014 = torch_c.to_builtin_tensor %33780 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31505 = arith.constant 1 : index
%dim_31506 = tensor.dim %34014, %c1_31505 : tensor<4x?x4096xf16>
%34015 = flow.tensor.transfer %34014 : tensor<4x?x4096xf16>{%dim_31506} to #hal.device.promise<@__device_7>
%34016 = torch_c.from_builtin_tensor %34015 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34017 = torch_c.to_builtin_tensor %33787 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31507 = arith.constant 1 : index
%dim_31508 = tensor.dim %34017, %c1_31507 : tensor<4x?x4096xf16>
%34018 = flow.tensor.transfer %34017 : tensor<4x?x4096xf16>{%dim_31508} to #hal.device.promise<@__device_7>
%34019 = torch_c.from_builtin_tensor %34018 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34020 = torch_c.to_builtin_tensor %33794 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31509 = arith.constant 1 : index
%dim_31510 = tensor.dim %34020, %c1_31509 : tensor<4x?x4096xf16>
%34021 = flow.tensor.transfer %34020 : tensor<4x?x4096xf16>{%dim_31510} to #hal.device.promise<@__device_7>
%34022 = torch_c.from_builtin_tensor %34021 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%34023 = torch_c.to_builtin_tensor %33801 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_31511 = arith.constant 1 : index
%dim_31512 = tensor.dim %34023, %c1_31511 : tensor<4x?x4096xf16>
%34024 = flow.tensor.transfer %34023 : tensor<4x?x4096xf16>{%dim_31512} to #hal.device.promise<@__device_7>
%34025 = torch_c.from_builtin_tensor %34024 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31513 = torch.constant.int 1
%34026 = torch.aten.add.Tensor %34007, %34010, %int1_31513 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31514 = torch.constant.int 1
%34027 = torch.aten.add.Tensor %34026, %34013, %int1_31514 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31515 = torch.constant.int 1
%34028 = torch.aten.add.Tensor %34027, %34016, %int1_31515 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31516 = torch.constant.int 1
%34029 = torch.aten.add.Tensor %34028, %34019, %int1_31516 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31517 = torch.constant.int 1
%34030 = torch.aten.add.Tensor %34029, %34022, %int1_31517 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31518 = torch.constant.int 1
%34031 = torch.aten.add.Tensor %34030, %34025, %int1_31518 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31519 = torch.constant.int 1
%34032 = torch.aten.add.Tensor %34031, %33808, %int1_31519 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31520 = torch.constant.int 1
%34033 = torch.aten.add.Tensor %33513, %33836, %int1_31520 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31521 = torch.constant.int 1
%34034 = torch.aten.add.Tensor %33514, %33864, %int1_31521 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31522 = torch.constant.int 1
%34035 = torch.aten.add.Tensor %33515, %33892, %int1_31522 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31523 = torch.constant.int 1
%34036 = torch.aten.add.Tensor %33516, %33920, %int1_31523 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31524 = torch.constant.int 1
%34037 = torch.aten.add.Tensor %33517, %33948, %int1_31524 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31525 = torch.constant.int 1
%34038 = torch.aten.add.Tensor %33518, %33976, %int1_31525 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31526 = torch.constant.int 1
%34039 = torch.aten.add.Tensor %33519, %34004, %int1_31526 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31527 = torch.constant.int 1
%34040 = torch.aten.add.Tensor %33520, %34032, %int1_31527 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_31528 = torch.constant.int 6
%34041 = torch.prims.convert_element_type %34033, %int6_31528 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31529 = torch.constant.int 6
%34042 = torch.prims.convert_element_type %34034, %int6_31529 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31530 = torch.constant.int 6
%34043 = torch.prims.convert_element_type %34035, %int6_31530 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31531 = torch.constant.int 6
%34044 = torch.prims.convert_element_type %34036, %int6_31531 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31532 = torch.constant.int 6
%34045 = torch.prims.convert_element_type %34037, %int6_31532 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31533 = torch.constant.int 6
%34046 = torch.prims.convert_element_type %34038, %int6_31533 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31534 = torch.constant.int 6
%34047 = torch.prims.convert_element_type %34039, %int6_31534 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_31535 = torch.constant.int 6
%34048 = torch.prims.convert_element_type %34040, %int6_31535 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31536 = torch.constant.int 2
%34049 = torch.aten.pow.Tensor_Scalar %34041, %int2_31536 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34049, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31537 = torch.constant.int 2
%34050 = torch.aten.pow.Tensor_Scalar %34042, %int2_31537 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31538 = torch.constant.int 2
%34051 = torch.aten.pow.Tensor_Scalar %34043, %int2_31538 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31539 = torch.constant.int 2
%34052 = torch.aten.pow.Tensor_Scalar %34044, %int2_31539 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31540 = torch.constant.int 2
%34053 = torch.aten.pow.Tensor_Scalar %34045, %int2_31540 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31541 = torch.constant.int 2
%34054 = torch.aten.pow.Tensor_Scalar %34046, %int2_31541 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31542 = torch.constant.int 2
%34055 = torch.aten.pow.Tensor_Scalar %34047, %int2_31542 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34055, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_31543 = torch.constant.int 2
%34056 = torch.aten.pow.Tensor_Scalar %34048, %int2_31543 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_31544 = torch.constant.int -1
%34057 = torch.prim.ListConstruct %int-1_31544 : (!torch.int) -> !torch.list<int>
%true_31545 = torch.constant.bool true
%none_31546 = torch.constant.none
%34058 = torch.aten.mean.dim %34049, %34057, %true_31545, %none_31546 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31547 = torch.constant.int -1
%34059 = torch.prim.ListConstruct %int-1_31547 : (!torch.int) -> !torch.list<int>
%true_31548 = torch.constant.bool true
%none_31549 = torch.constant.none
%34060 = torch.aten.mean.dim %34050, %34059, %true_31548, %none_31549 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31550 = torch.constant.int -1
%34061 = torch.prim.ListConstruct %int-1_31550 : (!torch.int) -> !torch.list<int>
%true_31551 = torch.constant.bool true
%none_31552 = torch.constant.none
%34062 = torch.aten.mean.dim %34051, %34061, %true_31551, %none_31552 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31553 = torch.constant.int -1
%34063 = torch.prim.ListConstruct %int-1_31553 : (!torch.int) -> !torch.list<int>
%true_31554 = torch.constant.bool true
%none_31555 = torch.constant.none
%34064 = torch.aten.mean.dim %34052, %34063, %true_31554, %none_31555 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31556 = torch.constant.int -1
%34065 = torch.prim.ListConstruct %int-1_31556 : (!torch.int) -> !torch.list<int>
%true_31557 = torch.constant.bool true
%none_31558 = torch.constant.none
%34066 = torch.aten.mean.dim %34053, %34065, %true_31557, %none_31558 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31559 = torch.constant.int -1
%34067 = torch.prim.ListConstruct %int-1_31559 : (!torch.int) -> !torch.list<int>
%true_31560 = torch.constant.bool true
%none_31561 = torch.constant.none
%34068 = torch.aten.mean.dim %34054, %34067, %true_31560, %none_31561 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31562 = torch.constant.int -1
%34069 = torch.prim.ListConstruct %int-1_31562 : (!torch.int) -> !torch.list<int>
%true_31563 = torch.constant.bool true
%none_31564 = torch.constant.none
%34070 = torch.aten.mean.dim %34055, %34069, %true_31563, %none_31564 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_31565 = torch.constant.int -1
%34071 = torch.prim.ListConstruct %int-1_31565 : (!torch.int) -> !torch.list<int>
%true_31566 = torch.constant.bool true
%none_31567 = torch.constant.none
%34072 = torch.aten.mean.dim %34056, %34071, %true_31566, %none_31567 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31568 = torch.constant.float 9.9999997473787516E-6
%int1_31569 = torch.constant.int 1
%34073 = torch.aten.add.Scalar %34058, %float9.999990e-06_31568, %int1_31569 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31570 = torch.constant.float 9.9999997473787516E-6
%int1_31571 = torch.constant.int 1
%34074 = torch.aten.add.Scalar %34060, %float9.999990e-06_31570, %int1_31571 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31572 = torch.constant.float 9.9999997473787516E-6
%int1_31573 = torch.constant.int 1
%34075 = torch.aten.add.Scalar %34062, %float9.999990e-06_31572, %int1_31573 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31574 = torch.constant.float 9.9999997473787516E-6
%int1_31575 = torch.constant.int 1
%34076 = torch.aten.add.Scalar %34064, %float9.999990e-06_31574, %int1_31575 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31576 = torch.constant.float 9.9999997473787516E-6
%int1_31577 = torch.constant.int 1
%34077 = torch.aten.add.Scalar %34066, %float9.999990e-06_31576, %int1_31577 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31578 = torch.constant.float 9.9999997473787516E-6
%int1_31579 = torch.constant.int 1
%34078 = torch.aten.add.Scalar %34068, %float9.999990e-06_31578, %int1_31579 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31580 = torch.constant.float 9.9999997473787516E-6
%int1_31581 = torch.constant.int 1
%34079 = torch.aten.add.Scalar %34070, %float9.999990e-06_31580, %int1_31581 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_31582 = torch.constant.float 9.9999997473787516E-6
%int1_31583 = torch.constant.int 1
%34080 = torch.aten.add.Scalar %34072, %float9.999990e-06_31582, %int1_31583 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34080, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34081 = torch.aten.rsqrt %34073 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34082 = torch.aten.rsqrt %34074 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34083 = torch.aten.rsqrt %34075 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34084 = torch.aten.rsqrt %34076 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34085 = torch.aten.rsqrt %34077 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34086 = torch.aten.rsqrt %34078 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34086, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34087 = torch.aten.rsqrt %34079 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34088 = torch.aten.rsqrt %34080 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %34088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%34089 = torch.aten.mul.Tensor %34041, %34081 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34090 = torch.aten.mul.Tensor %34042, %34082 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34091 = torch.aten.mul.Tensor %34043, %34083 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34092 = torch.aten.mul.Tensor %34044, %34084 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34093 = torch.aten.mul.Tensor %34045, %34085 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34094 = torch.aten.mul.Tensor %34046, %34086 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34095 = torch.aten.mul.Tensor %34047, %34087 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34096 = torch.aten.mul.Tensor %34048, %34088 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34097 = torch.aten.mul.Tensor %1232, %34089 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34098 = torch.aten.mul.Tensor %1233, %34090 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34099 = torch.aten.mul.Tensor %1234, %34091 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34100 = torch.aten.mul.Tensor %1235, %34092 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34101 = torch.aten.mul.Tensor %1236, %34093 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34102 = torch.aten.mul.Tensor %1237, %34094 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34103 = torch.aten.mul.Tensor %1238, %34095 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%34104 = torch.aten.mul.Tensor %1239, %34096 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %34104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_31584 = torch.constant.int 5
%34105 = torch.prims.convert_element_type %34097, %int5_31584 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31585 = torch.constant.int 5
%34106 = torch.prims.convert_element_type %34098, %int5_31585 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31586 = torch.constant.int 5
%34107 = torch.prims.convert_element_type %34099, %int5_31586 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31587 = torch.constant.int 5
%34108 = torch.prims.convert_element_type %34100, %int5_31587 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31588 = torch.constant.int 5
%34109 = torch.prims.convert_element_type %34101, %int5_31588 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31589 = torch.constant.int 5
%34110 = torch.prims.convert_element_type %34102, %int5_31589 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31590 = torch.constant.int 5
%34111 = torch.prims.convert_element_type %34103, %int5_31590 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_31591 = torch.constant.int 5
%34112 = torch.prims.convert_element_type %34104, %int5_31591 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %34112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_31592 = torch.constant.int 1
%int0_31593 = torch.constant.int 0
%34113 = torch.prim.ListConstruct %int1_31592, %int0_31593 : (!torch.int, !torch.int) -> !torch.list<int>
%34114 = torch.aten.permute %1240, %34113 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31594 = torch.constant.int 1
%int0_31595 = torch.constant.int 0
%34115 = torch.prim.ListConstruct %int1_31594, %int0_31595 : (!torch.int, !torch.int) -> !torch.list<int>
%34116 = torch.aten.permute %1241, %34115 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31596 = torch.constant.int 1
%int0_31597 = torch.constant.int 0
%34117 = torch.prim.ListConstruct %int1_31596, %int0_31597 : (!torch.int, !torch.int) -> !torch.list<int>
%34118 = torch.aten.permute %1242, %34117 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31598 = torch.constant.int 1
%int0_31599 = torch.constant.int 0
%34119 = torch.prim.ListConstruct %int1_31598, %int0_31599 : (!torch.int, !torch.int) -> !torch.list<int>
%34120 = torch.aten.permute %1243, %34119 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31600 = torch.constant.int 1
%int0_31601 = torch.constant.int 0
%34121 = torch.prim.ListConstruct %int1_31600, %int0_31601 : (!torch.int, !torch.int) -> !torch.list<int>
%34122 = torch.aten.permute %1244, %34121 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31602 = torch.constant.int 1
%int0_31603 = torch.constant.int 0
%34123 = torch.prim.ListConstruct %int1_31602, %int0_31603 : (!torch.int, !torch.int) -> !torch.list<int>
%34124 = torch.aten.permute %1245, %34123 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31604 = torch.constant.int 1
%int0_31605 = torch.constant.int 0
%34125 = torch.prim.ListConstruct %int1_31604, %int0_31605 : (!torch.int, !torch.int) -> !torch.list<int>
%34126 = torch.aten.permute %1246, %34125 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_31606 = torch.constant.int 1
%int0_31607 = torch.constant.int 0
%34127 = torch.prim.ListConstruct %int1_31606, %int0_31607 : (!torch.int, !torch.int) -> !torch.list<int>
%34128 = torch.aten.permute %1247, %34127 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_31608 = torch.constant.int 4
%34129 = torch.aten.mul.int %int4_31608, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31609 = torch.constant.int 4096
%34130 = torch.prim.ListConstruct %34129, %int4096_31609 : (!torch.int, !torch.int) -> !torch.list<int>
%34131 = torch.aten.view %34105, %34130 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34131, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34132 = torch.aten.mm %34131, %34114 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34132, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31610 = torch.constant.int 4
%int512_31611 = torch.constant.int 512
%34133 = torch.prim.ListConstruct %int4_31610, %2482, %int512_31611 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34134 = torch.aten.view %34132, %34133 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31612 = torch.constant.int 4
%34135 = torch.aten.mul.int %int4_31612, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31613 = torch.constant.int 4096
%34136 = torch.prim.ListConstruct %34135, %int4096_31613 : (!torch.int, !torch.int) -> !torch.list<int>
%34137 = torch.aten.view %34106, %34136 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34137, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34138 = torch.aten.mm %34137, %34116 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34138, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31614 = torch.constant.int 4
%int512_31615 = torch.constant.int 512
%34139 = torch.prim.ListConstruct %int4_31614, %2482, %int512_31615 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34140 = torch.aten.view %34138, %34139 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31616 = torch.constant.int 4
%34141 = torch.aten.mul.int %int4_31616, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31617 = torch.constant.int 4096
%34142 = torch.prim.ListConstruct %34141, %int4096_31617 : (!torch.int, !torch.int) -> !torch.list<int>
%34143 = torch.aten.view %34107, %34142 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34143, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34144 = torch.aten.mm %34143, %34118 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34144, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31618 = torch.constant.int 4
%int512_31619 = torch.constant.int 512
%34145 = torch.prim.ListConstruct %int4_31618, %2482, %int512_31619 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34146 = torch.aten.view %34144, %34145 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31620 = torch.constant.int 4
%34147 = torch.aten.mul.int %int4_31620, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31621 = torch.constant.int 4096
%34148 = torch.prim.ListConstruct %34147, %int4096_31621 : (!torch.int, !torch.int) -> !torch.list<int>
%34149 = torch.aten.view %34108, %34148 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34149, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34150 = torch.aten.mm %34149, %34120 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34150, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31622 = torch.constant.int 4
%int512_31623 = torch.constant.int 512
%34151 = torch.prim.ListConstruct %int4_31622, %2482, %int512_31623 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34152 = torch.aten.view %34150, %34151 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31624 = torch.constant.int 4
%34153 = torch.aten.mul.int %int4_31624, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31625 = torch.constant.int 4096
%34154 = torch.prim.ListConstruct %34153, %int4096_31625 : (!torch.int, !torch.int) -> !torch.list<int>
%34155 = torch.aten.view %34109, %34154 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34155, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34156 = torch.aten.mm %34155, %34122 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34156, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31626 = torch.constant.int 4
%int512_31627 = torch.constant.int 512
%34157 = torch.prim.ListConstruct %int4_31626, %2482, %int512_31627 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34158 = torch.aten.view %34156, %34157 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31628 = torch.constant.int 4
%34159 = torch.aten.mul.int %int4_31628, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31629 = torch.constant.int 4096
%34160 = torch.prim.ListConstruct %34159, %int4096_31629 : (!torch.int, !torch.int) -> !torch.list<int>
%34161 = torch.aten.view %34110, %34160 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34161, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34162 = torch.aten.mm %34161, %34124 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34162, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31630 = torch.constant.int 4
%int512_31631 = torch.constant.int 512
%34163 = torch.prim.ListConstruct %int4_31630, %2482, %int512_31631 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34164 = torch.aten.view %34162, %34163 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31632 = torch.constant.int 4
%34165 = torch.aten.mul.int %int4_31632, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31633 = torch.constant.int 4096
%34166 = torch.prim.ListConstruct %34165, %int4096_31633 : (!torch.int, !torch.int) -> !torch.list<int>
%34167 = torch.aten.view %34111, %34166 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34167, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34168 = torch.aten.mm %34167, %34126 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34168, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31634 = torch.constant.int 4
%int512_31635 = torch.constant.int 512
%34169 = torch.prim.ListConstruct %int4_31634, %2482, %int512_31635 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34170 = torch.aten.view %34168, %34169 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_31636 = torch.constant.int 4
%34171 = torch.aten.mul.int %int4_31636, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31637 = torch.constant.int 4096
%34172 = torch.prim.ListConstruct %34171, %int4096_31637 : (!torch.int, !torch.int) -> !torch.list<int>
%34173 = torch.aten.view %34112, %34172 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34173, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34174 = torch.aten.mm %34173, %34128 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %34174, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_31638 = torch.constant.int 4
%int512_31639 = torch.constant.int 512
%34175 = torch.prim.ListConstruct %int4_31638, %2482, %int512_31639 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34176 = torch.aten.view %34174, %34175 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %34176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_31640 = torch.constant.int 1
%int0_31641 = torch.constant.int 0
%34177 = torch.prim.ListConstruct %int1_31640, %int0_31641 : (!torch.int, !torch.int) -> !torch.list<int>
%34178 = torch.aten.permute %1248, %34177 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31642 = torch.constant.int 1
%int0_31643 = torch.constant.int 0
%34179 = torch.prim.ListConstruct %int1_31642, %int0_31643 : (!torch.int, !torch.int) -> !torch.list<int>
%34180 = torch.aten.permute %1249, %34179 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31644 = torch.constant.int 1
%int0_31645 = torch.constant.int 0
%34181 = torch.prim.ListConstruct %int1_31644, %int0_31645 : (!torch.int, !torch.int) -> !torch.list<int>
%34182 = torch.aten.permute %1250, %34181 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31646 = torch.constant.int 1
%int0_31647 = torch.constant.int 0
%34183 = torch.prim.ListConstruct %int1_31646, %int0_31647 : (!torch.int, !torch.int) -> !torch.list<int>
%34184 = torch.aten.permute %1251, %34183 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31648 = torch.constant.int 1
%int0_31649 = torch.constant.int 0
%34185 = torch.prim.ListConstruct %int1_31648, %int0_31649 : (!torch.int, !torch.int) -> !torch.list<int>
%34186 = torch.aten.permute %1252, %34185 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31650 = torch.constant.int 1
%int0_31651 = torch.constant.int 0
%34187 = torch.prim.ListConstruct %int1_31650, %int0_31651 : (!torch.int, !torch.int) -> !torch.list<int>
%34188 = torch.aten.permute %1253, %34187 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31652 = torch.constant.int 1
%int0_31653 = torch.constant.int 0
%34189 = torch.prim.ListConstruct %int1_31652, %int0_31653 : (!torch.int, !torch.int) -> !torch.list<int>
%34190 = torch.aten.permute %1254, %34189 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31654 = torch.constant.int 1
%int0_31655 = torch.constant.int 0
%34191 = torch.prim.ListConstruct %int1_31654, %int0_31655 : (!torch.int, !torch.int) -> !torch.list<int>
%34192 = torch.aten.permute %1255, %34191 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_31656 = torch.constant.int 4
%34193 = torch.aten.mul.int %int4_31656, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31657 = torch.constant.int 4096
%34194 = torch.prim.ListConstruct %34193, %int4096_31657 : (!torch.int, !torch.int) -> !torch.list<int>
%34195 = torch.aten.view %34105, %34194 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34195, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34196 = torch.aten.mm %34195, %34178 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34196, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31658 = torch.constant.int 4
%int128_31659 = torch.constant.int 128
%34197 = torch.prim.ListConstruct %int4_31658, %2482, %int128_31659 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34198 = torch.aten.view %34196, %34197 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31660 = torch.constant.int 4
%34199 = torch.aten.mul.int %int4_31660, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31661 = torch.constant.int 4096
%34200 = torch.prim.ListConstruct %34199, %int4096_31661 : (!torch.int, !torch.int) -> !torch.list<int>
%34201 = torch.aten.view %34106, %34200 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34201, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34202 = torch.aten.mm %34201, %34180 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34202, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31662 = torch.constant.int 4
%int128_31663 = torch.constant.int 128
%34203 = torch.prim.ListConstruct %int4_31662, %2482, %int128_31663 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34204 = torch.aten.view %34202, %34203 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31664 = torch.constant.int 4
%34205 = torch.aten.mul.int %int4_31664, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31665 = torch.constant.int 4096
%34206 = torch.prim.ListConstruct %34205, %int4096_31665 : (!torch.int, !torch.int) -> !torch.list<int>
%34207 = torch.aten.view %34107, %34206 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34207, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34208 = torch.aten.mm %34207, %34182 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34208, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31666 = torch.constant.int 4
%int128_31667 = torch.constant.int 128
%34209 = torch.prim.ListConstruct %int4_31666, %2482, %int128_31667 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34210 = torch.aten.view %34208, %34209 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34210, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31668 = torch.constant.int 4
%34211 = torch.aten.mul.int %int4_31668, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31669 = torch.constant.int 4096
%34212 = torch.prim.ListConstruct %34211, %int4096_31669 : (!torch.int, !torch.int) -> !torch.list<int>
%34213 = torch.aten.view %34108, %34212 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34213, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34214 = torch.aten.mm %34213, %34184 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34214, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31670 = torch.constant.int 4
%int128_31671 = torch.constant.int 128
%34215 = torch.prim.ListConstruct %int4_31670, %2482, %int128_31671 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34216 = torch.aten.view %34214, %34215 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31672 = torch.constant.int 4
%34217 = torch.aten.mul.int %int4_31672, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31673 = torch.constant.int 4096
%34218 = torch.prim.ListConstruct %34217, %int4096_31673 : (!torch.int, !torch.int) -> !torch.list<int>
%34219 = torch.aten.view %34109, %34218 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34219, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34220 = torch.aten.mm %34219, %34186 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34220, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31674 = torch.constant.int 4
%int128_31675 = torch.constant.int 128
%34221 = torch.prim.ListConstruct %int4_31674, %2482, %int128_31675 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34222 = torch.aten.view %34220, %34221 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34222, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31676 = torch.constant.int 4
%34223 = torch.aten.mul.int %int4_31676, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31677 = torch.constant.int 4096
%34224 = torch.prim.ListConstruct %34223, %int4096_31677 : (!torch.int, !torch.int) -> !torch.list<int>
%34225 = torch.aten.view %34110, %34224 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34225, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34226 = torch.aten.mm %34225, %34188 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34226, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31678 = torch.constant.int 4
%int128_31679 = torch.constant.int 128
%34227 = torch.prim.ListConstruct %int4_31678, %2482, %int128_31679 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34228 = torch.aten.view %34226, %34227 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31680 = torch.constant.int 4
%34229 = torch.aten.mul.int %int4_31680, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31681 = torch.constant.int 4096
%34230 = torch.prim.ListConstruct %34229, %int4096_31681 : (!torch.int, !torch.int) -> !torch.list<int>
%34231 = torch.aten.view %34111, %34230 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34231, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34232 = torch.aten.mm %34231, %34190 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34232, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31682 = torch.constant.int 4
%int128_31683 = torch.constant.int 128
%34233 = torch.prim.ListConstruct %int4_31682, %2482, %int128_31683 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34234 = torch.aten.view %34232, %34233 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31684 = torch.constant.int 4
%34235 = torch.aten.mul.int %int4_31684, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31685 = torch.constant.int 4096
%34236 = torch.prim.ListConstruct %34235, %int4096_31685 : (!torch.int, !torch.int) -> !torch.list<int>
%34237 = torch.aten.view %34112, %34236 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34237, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34238 = torch.aten.mm %34237, %34192 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34238, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31686 = torch.constant.int 4
%int128_31687 = torch.constant.int 128
%34239 = torch.prim.ListConstruct %int4_31686, %2482, %int128_31687 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34240 = torch.aten.view %34238, %34239 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_31688 = torch.constant.int 1
%int0_31689 = torch.constant.int 0
%34241 = torch.prim.ListConstruct %int1_31688, %int0_31689 : (!torch.int, !torch.int) -> !torch.list<int>
%34242 = torch.aten.permute %1256, %34241 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31690 = torch.constant.int 1
%int0_31691 = torch.constant.int 0
%34243 = torch.prim.ListConstruct %int1_31690, %int0_31691 : (!torch.int, !torch.int) -> !torch.list<int>
%34244 = torch.aten.permute %1257, %34243 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31692 = torch.constant.int 1
%int0_31693 = torch.constant.int 0
%34245 = torch.prim.ListConstruct %int1_31692, %int0_31693 : (!torch.int, !torch.int) -> !torch.list<int>
%34246 = torch.aten.permute %1258, %34245 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31694 = torch.constant.int 1
%int0_31695 = torch.constant.int 0
%34247 = torch.prim.ListConstruct %int1_31694, %int0_31695 : (!torch.int, !torch.int) -> !torch.list<int>
%34248 = torch.aten.permute %1259, %34247 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31696 = torch.constant.int 1
%int0_31697 = torch.constant.int 0
%34249 = torch.prim.ListConstruct %int1_31696, %int0_31697 : (!torch.int, !torch.int) -> !torch.list<int>
%34250 = torch.aten.permute %1260, %34249 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31698 = torch.constant.int 1
%int0_31699 = torch.constant.int 0
%34251 = torch.prim.ListConstruct %int1_31698, %int0_31699 : (!torch.int, !torch.int) -> !torch.list<int>
%34252 = torch.aten.permute %1261, %34251 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31700 = torch.constant.int 1
%int0_31701 = torch.constant.int 0
%34253 = torch.prim.ListConstruct %int1_31700, %int0_31701 : (!torch.int, !torch.int) -> !torch.list<int>
%34254 = torch.aten.permute %1262, %34253 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_31702 = torch.constant.int 1
%int0_31703 = torch.constant.int 0
%34255 = torch.prim.ListConstruct %int1_31702, %int0_31703 : (!torch.int, !torch.int) -> !torch.list<int>
%34256 = torch.aten.permute %1263, %34255 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_31704 = torch.constant.int 4
%34257 = torch.aten.mul.int %int4_31704, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31705 = torch.constant.int 4096
%34258 = torch.prim.ListConstruct %34257, %int4096_31705 : (!torch.int, !torch.int) -> !torch.list<int>
%34259 = torch.aten.view %34105, %34258 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34259, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34260 = torch.aten.mm %34259, %34242 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34260, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31706 = torch.constant.int 4
%int128_31707 = torch.constant.int 128
%34261 = torch.prim.ListConstruct %int4_31706, %2482, %int128_31707 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34262 = torch.aten.view %34260, %34261 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31708 = torch.constant.int 4
%34263 = torch.aten.mul.int %int4_31708, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31709 = torch.constant.int 4096
%34264 = torch.prim.ListConstruct %34263, %int4096_31709 : (!torch.int, !torch.int) -> !torch.list<int>
%34265 = torch.aten.view %34106, %34264 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34265, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34266 = torch.aten.mm %34265, %34244 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34266, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31710 = torch.constant.int 4
%int128_31711 = torch.constant.int 128
%34267 = torch.prim.ListConstruct %int4_31710, %2482, %int128_31711 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34268 = torch.aten.view %34266, %34267 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31712 = torch.constant.int 4
%34269 = torch.aten.mul.int %int4_31712, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31713 = torch.constant.int 4096
%34270 = torch.prim.ListConstruct %34269, %int4096_31713 : (!torch.int, !torch.int) -> !torch.list<int>
%34271 = torch.aten.view %34107, %34270 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34271, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34272 = torch.aten.mm %34271, %34246 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34272, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31714 = torch.constant.int 4
%int128_31715 = torch.constant.int 128
%34273 = torch.prim.ListConstruct %int4_31714, %2482, %int128_31715 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34274 = torch.aten.view %34272, %34273 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31716 = torch.constant.int 4
%34275 = torch.aten.mul.int %int4_31716, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31717 = torch.constant.int 4096
%34276 = torch.prim.ListConstruct %34275, %int4096_31717 : (!torch.int, !torch.int) -> !torch.list<int>
%34277 = torch.aten.view %34108, %34276 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34277, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34278 = torch.aten.mm %34277, %34248 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34278, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31718 = torch.constant.int 4
%int128_31719 = torch.constant.int 128
%34279 = torch.prim.ListConstruct %int4_31718, %2482, %int128_31719 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34280 = torch.aten.view %34278, %34279 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31720 = torch.constant.int 4
%34281 = torch.aten.mul.int %int4_31720, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31721 = torch.constant.int 4096
%34282 = torch.prim.ListConstruct %34281, %int4096_31721 : (!torch.int, !torch.int) -> !torch.list<int>
%34283 = torch.aten.view %34109, %34282 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34283, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34284 = torch.aten.mm %34283, %34250 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34284, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31722 = torch.constant.int 4
%int128_31723 = torch.constant.int 128
%34285 = torch.prim.ListConstruct %int4_31722, %2482, %int128_31723 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34286 = torch.aten.view %34284, %34285 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31724 = torch.constant.int 4
%34287 = torch.aten.mul.int %int4_31724, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31725 = torch.constant.int 4096
%34288 = torch.prim.ListConstruct %34287, %int4096_31725 : (!torch.int, !torch.int) -> !torch.list<int>
%34289 = torch.aten.view %34110, %34288 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34289, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34290 = torch.aten.mm %34289, %34252 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34290, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31726 = torch.constant.int 4
%int128_31727 = torch.constant.int 128
%34291 = torch.prim.ListConstruct %int4_31726, %2482, %int128_31727 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34292 = torch.aten.view %34290, %34291 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31728 = torch.constant.int 4
%34293 = torch.aten.mul.int %int4_31728, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31729 = torch.constant.int 4096
%34294 = torch.prim.ListConstruct %34293, %int4096_31729 : (!torch.int, !torch.int) -> !torch.list<int>
%34295 = torch.aten.view %34111, %34294 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34295, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34296 = torch.aten.mm %34295, %34254 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34296, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31730 = torch.constant.int 4
%int128_31731 = torch.constant.int 128
%34297 = torch.prim.ListConstruct %int4_31730, %2482, %int128_31731 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34298 = torch.aten.view %34296, %34297 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31732 = torch.constant.int 4
%34299 = torch.aten.mul.int %int4_31732, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_31733 = torch.constant.int 4096
%34300 = torch.prim.ListConstruct %34299, %int4096_31733 : (!torch.int, !torch.int) -> !torch.list<int>
%34301 = torch.aten.view %34112, %34300 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %34301, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%34302 = torch.aten.mm %34301, %34256 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %34302, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_31734 = torch.constant.int 4
%int128_31735 = torch.constant.int 128
%34303 = torch.prim.ListConstruct %int4_31734, %2482, %int128_31735 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34304 = torch.aten.view %34302, %34303 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %34304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_31736 = torch.constant.int 4
%int4_31737 = torch.constant.int 4
%int128_31738 = torch.constant.int 128
%34305 = torch.prim.ListConstruct %int4_31736, %2482, %int4_31737, %int128_31738 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34306 = torch.aten.view %34134, %34305 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31739 = torch.constant.int 4
%int4_31740 = torch.constant.int 4
%int128_31741 = torch.constant.int 128
%34307 = torch.prim.ListConstruct %int4_31739, %2482, %int4_31740, %int128_31741 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34308 = torch.aten.view %34140, %34307 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31742 = torch.constant.int 4
%int4_31743 = torch.constant.int 4
%int128_31744 = torch.constant.int 128
%34309 = torch.prim.ListConstruct %int4_31742, %2482, %int4_31743, %int128_31744 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34310 = torch.aten.view %34146, %34309 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31745 = torch.constant.int 4
%int4_31746 = torch.constant.int 4
%int128_31747 = torch.constant.int 128
%34311 = torch.prim.ListConstruct %int4_31745, %2482, %int4_31746, %int128_31747 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34312 = torch.aten.view %34152, %34311 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31748 = torch.constant.int 4
%int4_31749 = torch.constant.int 4
%int128_31750 = torch.constant.int 128
%34313 = torch.prim.ListConstruct %int4_31748, %2482, %int4_31749, %int128_31750 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34314 = torch.aten.view %34158, %34313 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31751 = torch.constant.int 4
%int4_31752 = torch.constant.int 4
%int128_31753 = torch.constant.int 128
%34315 = torch.prim.ListConstruct %int4_31751, %2482, %int4_31752, %int128_31753 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34316 = torch.aten.view %34164, %34315 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31754 = torch.constant.int 4
%int4_31755 = torch.constant.int 4
%int128_31756 = torch.constant.int 128
%34317 = torch.prim.ListConstruct %int4_31754, %2482, %int4_31755, %int128_31756 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34318 = torch.aten.view %34170, %34317 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31757 = torch.constant.int 4
%int4_31758 = torch.constant.int 4
%int128_31759 = torch.constant.int 128
%34319 = torch.prim.ListConstruct %int4_31757, %2482, %int4_31758, %int128_31759 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34320 = torch.aten.view %34176, %34319 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_31760 = torch.constant.int 4
%int1_31761 = torch.constant.int 1
%int128_31762 = torch.constant.int 128
%34321 = torch.prim.ListConstruct %int4_31760, %2482, %int1_31761, %int128_31762 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34322 = torch.aten.view %34198, %34321 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31763 = torch.constant.int 4
%int1_31764 = torch.constant.int 1
%int128_31765 = torch.constant.int 128
%34323 = torch.prim.ListConstruct %int4_31763, %2482, %int1_31764, %int128_31765 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34324 = torch.aten.view %34204, %34323 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34324, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31766 = torch.constant.int 4
%int1_31767 = torch.constant.int 1
%int128_31768 = torch.constant.int 128
%34325 = torch.prim.ListConstruct %int4_31766, %2482, %int1_31767, %int128_31768 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34326 = torch.aten.view %34210, %34325 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31769 = torch.constant.int 4
%int1_31770 = torch.constant.int 1
%int128_31771 = torch.constant.int 128
%34327 = torch.prim.ListConstruct %int4_31769, %2482, %int1_31770, %int128_31771 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34328 = torch.aten.view %34216, %34327 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34328, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31772 = torch.constant.int 4
%int1_31773 = torch.constant.int 1
%int128_31774 = torch.constant.int 128
%34329 = torch.prim.ListConstruct %int4_31772, %2482, %int1_31773, %int128_31774 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34330 = torch.aten.view %34222, %34329 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31775 = torch.constant.int 4
%int1_31776 = torch.constant.int 1
%int128_31777 = torch.constant.int 128
%34331 = torch.prim.ListConstruct %int4_31775, %2482, %int1_31776, %int128_31777 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34332 = torch.aten.view %34228, %34331 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31778 = torch.constant.int 4
%int1_31779 = torch.constant.int 1
%int128_31780 = torch.constant.int 128
%34333 = torch.prim.ListConstruct %int4_31778, %2482, %int1_31779, %int128_31780 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34334 = torch.aten.view %34234, %34333 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34334, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31781 = torch.constant.int 4
%int1_31782 = torch.constant.int 1
%int128_31783 = torch.constant.int 128
%34335 = torch.prim.ListConstruct %int4_31781, %2482, %int1_31782, %int128_31783 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34336 = torch.aten.view %34240, %34335 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31784 = torch.constant.int 4
%int1_31785 = torch.constant.int 1
%int128_31786 = torch.constant.int 128
%34337 = torch.prim.ListConstruct %int4_31784, %2482, %int1_31785, %int128_31786 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34338 = torch.aten.view %34262, %34337 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31787 = torch.constant.int 4
%int1_31788 = torch.constant.int 1
%int128_31789 = torch.constant.int 128
%34339 = torch.prim.ListConstruct %int4_31787, %2482, %int1_31788, %int128_31789 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34340 = torch.aten.view %34268, %34339 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31790 = torch.constant.int 4
%int1_31791 = torch.constant.int 1
%int128_31792 = torch.constant.int 128
%34341 = torch.prim.ListConstruct %int4_31790, %2482, %int1_31791, %int128_31792 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34342 = torch.aten.view %34274, %34341 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31793 = torch.constant.int 4
%int1_31794 = torch.constant.int 1
%int128_31795 = torch.constant.int 128
%34343 = torch.prim.ListConstruct %int4_31793, %2482, %int1_31794, %int128_31795 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34344 = torch.aten.view %34280, %34343 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31796 = torch.constant.int 4
%int1_31797 = torch.constant.int 1
%int128_31798 = torch.constant.int 128
%34345 = torch.prim.ListConstruct %int4_31796, %2482, %int1_31797, %int128_31798 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34346 = torch.aten.view %34286, %34345 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34346, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31799 = torch.constant.int 4
%int1_31800 = torch.constant.int 1
%int128_31801 = torch.constant.int 128
%34347 = torch.prim.ListConstruct %int4_31799, %2482, %int1_31800, %int128_31801 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34348 = torch.aten.view %34292, %34347 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31802 = torch.constant.int 4
%int1_31803 = torch.constant.int 1
%int128_31804 = torch.constant.int 128
%34349 = torch.prim.ListConstruct %int4_31802, %2482, %int1_31803, %int128_31804 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34350 = torch.aten.view %34298, %34349 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34350, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_31805 = torch.constant.int 4
%int1_31806 = torch.constant.int 1
%int128_31807 = torch.constant.int 128
%34351 = torch.prim.ListConstruct %int4_31805, %2482, %int1_31806, %int128_31807 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34352 = torch.aten.view %34304, %34351 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_31808 = torch.constant.int 131072
%none_31809 = torch.constant.none
%none_31810 = torch.constant.none
%cpu_31811 = torch.constant.device "cpu"
%false_31812 = torch.constant.bool false
%34353 = torch.aten.arange %int131072_31808, %none_31809, %none_31810, %cpu_31811, %false_31812 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_31813 = torch.constant.int 0
%int128_31814 = torch.constant.int 128
%int2_31815 = torch.constant.int 2
%none_31816 = torch.constant.none
%none_31817 = torch.constant.none
%cpu_31818 = torch.constant.device "cpu"
%false_31819 = torch.constant.bool false
%34354 = torch.aten.arange.start_step %int0_31813, %int128_31814, %int2_31815, %none_31816, %none_31817, %cpu_31818, %false_31819 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_31820 = torch.constant.int 0
%int0_31821 = torch.constant.int 0
%int64_31822 = torch.constant.int 64
%int1_31823 = torch.constant.int 1
%34355 = torch.aten.slice.Tensor %34354, %int0_31820, %int0_31821, %int64_31822, %int1_31823 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_31824 = torch.constant.int 6
%34356 = torch.prims.convert_element_type %34355, %int6_31824 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_31825 = torch.constant.int 128
%34357 = torch.aten.div.Scalar %34356, %int128_31825 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_31826 = torch.constant.float 5.000000e+05
%34358 = torch.aten.pow.Scalar %float5.000000e05_31826, %34357 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%34359 = torch.aten.reciprocal %34358 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_31827 = torch.constant.float 1.000000e+00
%34360 = torch.aten.mul.Scalar %34359, %float1.000000e00_31827 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_31828 = torch.constant.int 131072
%int1_31829 = torch.constant.int 1
%34361 = torch.prim.ListConstruct %int131072_31828, %int1_31829 : (!torch.int, !torch.int) -> !torch.list<int>
%34362 = torch.aten.view %34353, %34361 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%34363 = torch.aten.mul.Tensor %34362, %34360 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%34364 = torch.aten.cos %34363 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%34365 = torch.aten.sin %34363 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%34366 = torch.aten.complex %34364, %34365 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%34367 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34368 = flow.tensor.transfer %34367 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%34369 = torch_c.from_builtin_tensor %34368 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34370 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34371 = flow.tensor.transfer %34370 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%34372 = torch_c.from_builtin_tensor %34371 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34373 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34374 = flow.tensor.transfer %34373 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%34375 = torch_c.from_builtin_tensor %34374 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34376 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34377 = flow.tensor.transfer %34376 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%34378 = torch_c.from_builtin_tensor %34377 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34379 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34380 = flow.tensor.transfer %34379 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%34381 = torch_c.from_builtin_tensor %34380 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34382 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34383 = flow.tensor.transfer %34382 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%34384 = torch_c.from_builtin_tensor %34383 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34385 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34386 = flow.tensor.transfer %34385 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%34387 = torch_c.from_builtin_tensor %34386 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34388 = torch_c.to_builtin_tensor %34366 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34389 = flow.tensor.transfer %34388 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%34390 = torch_c.from_builtin_tensor %34389 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_31830 = torch.constant.int 1
%34391 = torch.aten.size.int %34134, %int1_31830 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31831 = torch.constant.int 0
%34392 = torch.aten.add.int %int0_31831, %34391 : !torch.int, !torch.int -> !torch.int
%int0_31832 = torch.constant.int 0
%int0_31833 = torch.constant.int 0
%int1_31834 = torch.constant.int 1
%34393 = torch.aten.slice.Tensor %34369, %int0_31832, %int0_31833, %34392, %int1_31834 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34393, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31835 = torch.constant.int 1
%int0_31836 = torch.constant.int 0
%int9223372036854775807_31837 = torch.constant.int 9223372036854775807
%int1_31838 = torch.constant.int 1
%34394 = torch.aten.slice.Tensor %34393, %int1_31835, %int0_31836, %int9223372036854775807_31837, %int1_31838 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34394, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31839 = torch.constant.int 0
%34395 = torch.aten.unsqueeze %34394, %int0_31839 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34395, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31840 = torch.constant.int 2
%34396 = torch.aten.unsqueeze %34395, %int2_31840 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34396, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31841 = torch.constant.int 3
%int0_31842 = torch.constant.int 0
%int9223372036854775807_31843 = torch.constant.int 9223372036854775807
%int1_31844 = torch.constant.int 1
%34397 = torch.aten.slice.Tensor %34396, %int3_31841, %int0_31842, %int9223372036854775807_31843, %int1_31844 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34397, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34398 = torch_c.to_builtin_tensor %34306 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31845 = arith.constant 1 : index
%dim_31846 = tensor.dim %34398, %c1_31845 : tensor<4x?x4x128xf16>
%34399 = flow.tensor.bitcast %34398 : tensor<4x?x4x128xf16>{%dim_31846} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31846}
%34400 = torch_c.from_builtin_tensor %34399 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34401 = torch.aten.mul.Tensor %34400, %34397 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34402 = torch_c.to_builtin_tensor %34401 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31847 = arith.constant 1 : index
%dim_31848 = tensor.dim %34402, %c1_31847 : tensor<4x?x4x64xcomplex<f32>>
%34403 = flow.tensor.bitcast %34402 : tensor<4x?x4x64xcomplex<f32>>{%dim_31848} -> tensor<4x?x4x128xf32>{%dim_31848}
%34404 = torch_c.from_builtin_tensor %34403 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34404, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31849 = torch.constant.int 5
%34405 = torch.prims.convert_element_type %34404, %int5_31849 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31850 = torch.constant.int 1
%34406 = torch.aten.size.int %34140, %int1_31850 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31851 = torch.constant.int 0
%34407 = torch.aten.add.int %int0_31851, %34406 : !torch.int, !torch.int -> !torch.int
%int0_31852 = torch.constant.int 0
%int0_31853 = torch.constant.int 0
%int1_31854 = torch.constant.int 1
%34408 = torch.aten.slice.Tensor %34372, %int0_31852, %int0_31853, %34407, %int1_31854 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34408, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31855 = torch.constant.int 1
%int0_31856 = torch.constant.int 0
%int9223372036854775807_31857 = torch.constant.int 9223372036854775807
%int1_31858 = torch.constant.int 1
%34409 = torch.aten.slice.Tensor %34408, %int1_31855, %int0_31856, %int9223372036854775807_31857, %int1_31858 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34409, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31859 = torch.constant.int 0
%34410 = torch.aten.unsqueeze %34409, %int0_31859 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34410, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31860 = torch.constant.int 2
%34411 = torch.aten.unsqueeze %34410, %int2_31860 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34411, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31861 = torch.constant.int 3
%int0_31862 = torch.constant.int 0
%int9223372036854775807_31863 = torch.constant.int 9223372036854775807
%int1_31864 = torch.constant.int 1
%34412 = torch.aten.slice.Tensor %34411, %int3_31861, %int0_31862, %int9223372036854775807_31863, %int1_31864 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34412, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34413 = torch_c.to_builtin_tensor %34308 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31865 = arith.constant 1 : index
%dim_31866 = tensor.dim %34413, %c1_31865 : tensor<4x?x4x128xf16>
%34414 = flow.tensor.bitcast %34413 : tensor<4x?x4x128xf16>{%dim_31866} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31866}
%34415 = torch_c.from_builtin_tensor %34414 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34416 = torch.aten.mul.Tensor %34415, %34412 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34417 = torch_c.to_builtin_tensor %34416 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31867 = arith.constant 1 : index
%dim_31868 = tensor.dim %34417, %c1_31867 : tensor<4x?x4x64xcomplex<f32>>
%34418 = flow.tensor.bitcast %34417 : tensor<4x?x4x64xcomplex<f32>>{%dim_31868} -> tensor<4x?x4x128xf32>{%dim_31868}
%34419 = torch_c.from_builtin_tensor %34418 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31869 = torch.constant.int 5
%34420 = torch.prims.convert_element_type %34419, %int5_31869 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31870 = torch.constant.int 1
%34421 = torch.aten.size.int %34146, %int1_31870 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31871 = torch.constant.int 0
%34422 = torch.aten.add.int %int0_31871, %34421 : !torch.int, !torch.int -> !torch.int
%int0_31872 = torch.constant.int 0
%int0_31873 = torch.constant.int 0
%int1_31874 = torch.constant.int 1
%34423 = torch.aten.slice.Tensor %34375, %int0_31872, %int0_31873, %34422, %int1_31874 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34423, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31875 = torch.constant.int 1
%int0_31876 = torch.constant.int 0
%int9223372036854775807_31877 = torch.constant.int 9223372036854775807
%int1_31878 = torch.constant.int 1
%34424 = torch.aten.slice.Tensor %34423, %int1_31875, %int0_31876, %int9223372036854775807_31877, %int1_31878 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34424, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31879 = torch.constant.int 0
%34425 = torch.aten.unsqueeze %34424, %int0_31879 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34425, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31880 = torch.constant.int 2
%34426 = torch.aten.unsqueeze %34425, %int2_31880 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34426, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31881 = torch.constant.int 3
%int0_31882 = torch.constant.int 0
%int9223372036854775807_31883 = torch.constant.int 9223372036854775807
%int1_31884 = torch.constant.int 1
%34427 = torch.aten.slice.Tensor %34426, %int3_31881, %int0_31882, %int9223372036854775807_31883, %int1_31884 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34427, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34428 = torch_c.to_builtin_tensor %34310 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31885 = arith.constant 1 : index
%dim_31886 = tensor.dim %34428, %c1_31885 : tensor<4x?x4x128xf16>
%34429 = flow.tensor.bitcast %34428 : tensor<4x?x4x128xf16>{%dim_31886} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31886}
%34430 = torch_c.from_builtin_tensor %34429 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34431 = torch.aten.mul.Tensor %34430, %34427 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34432 = torch_c.to_builtin_tensor %34431 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31887 = arith.constant 1 : index
%dim_31888 = tensor.dim %34432, %c1_31887 : tensor<4x?x4x64xcomplex<f32>>
%34433 = flow.tensor.bitcast %34432 : tensor<4x?x4x64xcomplex<f32>>{%dim_31888} -> tensor<4x?x4x128xf32>{%dim_31888}
%34434 = torch_c.from_builtin_tensor %34433 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31889 = torch.constant.int 5
%34435 = torch.prims.convert_element_type %34434, %int5_31889 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31890 = torch.constant.int 1
%34436 = torch.aten.size.int %34152, %int1_31890 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31891 = torch.constant.int 0
%34437 = torch.aten.add.int %int0_31891, %34436 : !torch.int, !torch.int -> !torch.int
%int0_31892 = torch.constant.int 0
%int0_31893 = torch.constant.int 0
%int1_31894 = torch.constant.int 1
%34438 = torch.aten.slice.Tensor %34378, %int0_31892, %int0_31893, %34437, %int1_31894 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34438, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31895 = torch.constant.int 1
%int0_31896 = torch.constant.int 0
%int9223372036854775807_31897 = torch.constant.int 9223372036854775807
%int1_31898 = torch.constant.int 1
%34439 = torch.aten.slice.Tensor %34438, %int1_31895, %int0_31896, %int9223372036854775807_31897, %int1_31898 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34439, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31899 = torch.constant.int 0
%34440 = torch.aten.unsqueeze %34439, %int0_31899 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34440, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31900 = torch.constant.int 2
%34441 = torch.aten.unsqueeze %34440, %int2_31900 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34441, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31901 = torch.constant.int 3
%int0_31902 = torch.constant.int 0
%int9223372036854775807_31903 = torch.constant.int 9223372036854775807
%int1_31904 = torch.constant.int 1
%34442 = torch.aten.slice.Tensor %34441, %int3_31901, %int0_31902, %int9223372036854775807_31903, %int1_31904 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34442, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34443 = torch_c.to_builtin_tensor %34312 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31905 = arith.constant 1 : index
%dim_31906 = tensor.dim %34443, %c1_31905 : tensor<4x?x4x128xf16>
%34444 = flow.tensor.bitcast %34443 : tensor<4x?x4x128xf16>{%dim_31906} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31906}
%34445 = torch_c.from_builtin_tensor %34444 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34446 = torch.aten.mul.Tensor %34445, %34442 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34447 = torch_c.to_builtin_tensor %34446 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31907 = arith.constant 1 : index
%dim_31908 = tensor.dim %34447, %c1_31907 : tensor<4x?x4x64xcomplex<f32>>
%34448 = flow.tensor.bitcast %34447 : tensor<4x?x4x64xcomplex<f32>>{%dim_31908} -> tensor<4x?x4x128xf32>{%dim_31908}
%34449 = torch_c.from_builtin_tensor %34448 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31909 = torch.constant.int 5
%34450 = torch.prims.convert_element_type %34449, %int5_31909 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31910 = torch.constant.int 1
%34451 = torch.aten.size.int %34158, %int1_31910 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31911 = torch.constant.int 0
%34452 = torch.aten.add.int %int0_31911, %34451 : !torch.int, !torch.int -> !torch.int
%int0_31912 = torch.constant.int 0
%int0_31913 = torch.constant.int 0
%int1_31914 = torch.constant.int 1
%34453 = torch.aten.slice.Tensor %34381, %int0_31912, %int0_31913, %34452, %int1_31914 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34453, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31915 = torch.constant.int 1
%int0_31916 = torch.constant.int 0
%int9223372036854775807_31917 = torch.constant.int 9223372036854775807
%int1_31918 = torch.constant.int 1
%34454 = torch.aten.slice.Tensor %34453, %int1_31915, %int0_31916, %int9223372036854775807_31917, %int1_31918 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34454, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31919 = torch.constant.int 0
%34455 = torch.aten.unsqueeze %34454, %int0_31919 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34455, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31920 = torch.constant.int 2
%34456 = torch.aten.unsqueeze %34455, %int2_31920 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34456, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31921 = torch.constant.int 3
%int0_31922 = torch.constant.int 0
%int9223372036854775807_31923 = torch.constant.int 9223372036854775807
%int1_31924 = torch.constant.int 1
%34457 = torch.aten.slice.Tensor %34456, %int3_31921, %int0_31922, %int9223372036854775807_31923, %int1_31924 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34457, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34458 = torch_c.to_builtin_tensor %34314 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31925 = arith.constant 1 : index
%dim_31926 = tensor.dim %34458, %c1_31925 : tensor<4x?x4x128xf16>
%34459 = flow.tensor.bitcast %34458 : tensor<4x?x4x128xf16>{%dim_31926} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31926}
%34460 = torch_c.from_builtin_tensor %34459 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34460, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34461 = torch.aten.mul.Tensor %34460, %34457 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34461, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34462 = torch_c.to_builtin_tensor %34461 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31927 = arith.constant 1 : index
%dim_31928 = tensor.dim %34462, %c1_31927 : tensor<4x?x4x64xcomplex<f32>>
%34463 = flow.tensor.bitcast %34462 : tensor<4x?x4x64xcomplex<f32>>{%dim_31928} -> tensor<4x?x4x128xf32>{%dim_31928}
%34464 = torch_c.from_builtin_tensor %34463 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31929 = torch.constant.int 5
%34465 = torch.prims.convert_element_type %34464, %int5_31929 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31930 = torch.constant.int 1
%34466 = torch.aten.size.int %34164, %int1_31930 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31931 = torch.constant.int 0
%34467 = torch.aten.add.int %int0_31931, %34466 : !torch.int, !torch.int -> !torch.int
%int0_31932 = torch.constant.int 0
%int0_31933 = torch.constant.int 0
%int1_31934 = torch.constant.int 1
%34468 = torch.aten.slice.Tensor %34384, %int0_31932, %int0_31933, %34467, %int1_31934 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34468, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31935 = torch.constant.int 1
%int0_31936 = torch.constant.int 0
%int9223372036854775807_31937 = torch.constant.int 9223372036854775807
%int1_31938 = torch.constant.int 1
%34469 = torch.aten.slice.Tensor %34468, %int1_31935, %int0_31936, %int9223372036854775807_31937, %int1_31938 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34469, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31939 = torch.constant.int 0
%34470 = torch.aten.unsqueeze %34469, %int0_31939 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34470, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31940 = torch.constant.int 2
%34471 = torch.aten.unsqueeze %34470, %int2_31940 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34471, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31941 = torch.constant.int 3
%int0_31942 = torch.constant.int 0
%int9223372036854775807_31943 = torch.constant.int 9223372036854775807
%int1_31944 = torch.constant.int 1
%34472 = torch.aten.slice.Tensor %34471, %int3_31941, %int0_31942, %int9223372036854775807_31943, %int1_31944 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34472, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34473 = torch_c.to_builtin_tensor %34316 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31945 = arith.constant 1 : index
%dim_31946 = tensor.dim %34473, %c1_31945 : tensor<4x?x4x128xf16>
%34474 = flow.tensor.bitcast %34473 : tensor<4x?x4x128xf16>{%dim_31946} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31946}
%34475 = torch_c.from_builtin_tensor %34474 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34476 = torch.aten.mul.Tensor %34475, %34472 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34476, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34477 = torch_c.to_builtin_tensor %34476 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31947 = arith.constant 1 : index
%dim_31948 = tensor.dim %34477, %c1_31947 : tensor<4x?x4x64xcomplex<f32>>
%34478 = flow.tensor.bitcast %34477 : tensor<4x?x4x64xcomplex<f32>>{%dim_31948} -> tensor<4x?x4x128xf32>{%dim_31948}
%34479 = torch_c.from_builtin_tensor %34478 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31949 = torch.constant.int 5
%34480 = torch.prims.convert_element_type %34479, %int5_31949 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31950 = torch.constant.int 1
%34481 = torch.aten.size.int %34170, %int1_31950 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31951 = torch.constant.int 0
%34482 = torch.aten.add.int %int0_31951, %34481 : !torch.int, !torch.int -> !torch.int
%int0_31952 = torch.constant.int 0
%int0_31953 = torch.constant.int 0
%int1_31954 = torch.constant.int 1
%34483 = torch.aten.slice.Tensor %34387, %int0_31952, %int0_31953, %34482, %int1_31954 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34483, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31955 = torch.constant.int 1
%int0_31956 = torch.constant.int 0
%int9223372036854775807_31957 = torch.constant.int 9223372036854775807
%int1_31958 = torch.constant.int 1
%34484 = torch.aten.slice.Tensor %34483, %int1_31955, %int0_31956, %int9223372036854775807_31957, %int1_31958 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34484, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31959 = torch.constant.int 0
%34485 = torch.aten.unsqueeze %34484, %int0_31959 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34485, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31960 = torch.constant.int 2
%34486 = torch.aten.unsqueeze %34485, %int2_31960 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34486, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31961 = torch.constant.int 3
%int0_31962 = torch.constant.int 0
%int9223372036854775807_31963 = torch.constant.int 9223372036854775807
%int1_31964 = torch.constant.int 1
%34487 = torch.aten.slice.Tensor %34486, %int3_31961, %int0_31962, %int9223372036854775807_31963, %int1_31964 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34487, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34488 = torch_c.to_builtin_tensor %34318 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31965 = arith.constant 1 : index
%dim_31966 = tensor.dim %34488, %c1_31965 : tensor<4x?x4x128xf16>
%34489 = flow.tensor.bitcast %34488 : tensor<4x?x4x128xf16>{%dim_31966} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31966}
%34490 = torch_c.from_builtin_tensor %34489 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34491 = torch.aten.mul.Tensor %34490, %34487 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34491, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34492 = torch_c.to_builtin_tensor %34491 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31967 = arith.constant 1 : index
%dim_31968 = tensor.dim %34492, %c1_31967 : tensor<4x?x4x64xcomplex<f32>>
%34493 = flow.tensor.bitcast %34492 : tensor<4x?x4x64xcomplex<f32>>{%dim_31968} -> tensor<4x?x4x128xf32>{%dim_31968}
%34494 = torch_c.from_builtin_tensor %34493 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31969 = torch.constant.int 5
%34495 = torch.prims.convert_element_type %34494, %int5_31969 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_31970 = torch.constant.int 1
%34496 = torch.aten.size.int %34176, %int1_31970 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_31971 = torch.constant.int 0
%34497 = torch.aten.add.int %int0_31971, %34496 : !torch.int, !torch.int -> !torch.int
%int0_31972 = torch.constant.int 0
%int0_31973 = torch.constant.int 0
%int1_31974 = torch.constant.int 1
%34498 = torch.aten.slice.Tensor %34390, %int0_31972, %int0_31973, %34497, %int1_31974 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34498, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_31975 = torch.constant.int 1
%int0_31976 = torch.constant.int 0
%int9223372036854775807_31977 = torch.constant.int 9223372036854775807
%int1_31978 = torch.constant.int 1
%34499 = torch.aten.slice.Tensor %34498, %int1_31975, %int0_31976, %int9223372036854775807_31977, %int1_31978 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34499, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_31979 = torch.constant.int 0
%34500 = torch.aten.unsqueeze %34499, %int0_31979 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34500, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_31980 = torch.constant.int 2
%34501 = torch.aten.unsqueeze %34500, %int2_31980 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34501, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_31981 = torch.constant.int 3
%int0_31982 = torch.constant.int 0
%int9223372036854775807_31983 = torch.constant.int 9223372036854775807
%int1_31984 = torch.constant.int 1
%34502 = torch.aten.slice.Tensor %34501, %int3_31981, %int0_31982, %int9223372036854775807_31983, %int1_31984 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34502, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34503 = torch_c.to_builtin_tensor %34320 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_31985 = arith.constant 1 : index
%dim_31986 = tensor.dim %34503, %c1_31985 : tensor<4x?x4x128xf16>
%34504 = flow.tensor.bitcast %34503 : tensor<4x?x4x128xf16>{%dim_31986} -> tensor<4x?x4x64xcomplex<f16>>{%dim_31986}
%34505 = torch_c.from_builtin_tensor %34504 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %34505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%34506 = torch.aten.mul.Tensor %34505, %34502 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %34506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%34507 = torch_c.to_builtin_tensor %34506 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_31987 = arith.constant 1 : index
%dim_31988 = tensor.dim %34507, %c1_31987 : tensor<4x?x4x64xcomplex<f32>>
%34508 = flow.tensor.bitcast %34507 : tensor<4x?x4x64xcomplex<f32>>{%dim_31988} -> tensor<4x?x4x128xf32>{%dim_31988}
%34509 = torch_c.from_builtin_tensor %34508 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %34509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_31989 = torch.constant.int 5
%34510 = torch.prims.convert_element_type %34509, %int5_31989 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_31990 = torch.constant.int 131072
%none_31991 = torch.constant.none
%none_31992 = torch.constant.none
%cpu_31993 = torch.constant.device "cpu"
%false_31994 = torch.constant.bool false
%34511 = torch.aten.arange %int131072_31990, %none_31991, %none_31992, %cpu_31993, %false_31994 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_31995 = torch.constant.int 0
%int128_31996 = torch.constant.int 128
%int2_31997 = torch.constant.int 2
%none_31998 = torch.constant.none
%none_31999 = torch.constant.none
%cpu_32000 = torch.constant.device "cpu"
%false_32001 = torch.constant.bool false
%34512 = torch.aten.arange.start_step %int0_31995, %int128_31996, %int2_31997, %none_31998, %none_31999, %cpu_32000, %false_32001 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_32002 = torch.constant.int 0
%int0_32003 = torch.constant.int 0
%int64_32004 = torch.constant.int 64
%int1_32005 = torch.constant.int 1
%34513 = torch.aten.slice.Tensor %34512, %int0_32002, %int0_32003, %int64_32004, %int1_32005 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_32006 = torch.constant.int 6
%34514 = torch.prims.convert_element_type %34513, %int6_32006 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_32007 = torch.constant.int 128
%34515 = torch.aten.div.Scalar %34514, %int128_32007 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_32008 = torch.constant.float 5.000000e+05
%34516 = torch.aten.pow.Scalar %float5.000000e05_32008, %34515 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%34517 = torch.aten.reciprocal %34516 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_32009 = torch.constant.float 1.000000e+00
%34518 = torch.aten.mul.Scalar %34517, %float1.000000e00_32009 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_32010 = torch.constant.int 131072
%int1_32011 = torch.constant.int 1
%34519 = torch.prim.ListConstruct %int131072_32010, %int1_32011 : (!torch.int, !torch.int) -> !torch.list<int>
%34520 = torch.aten.view %34511, %34519 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%34521 = torch.aten.mul.Tensor %34520, %34518 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%34522 = torch.aten.cos %34521 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%34523 = torch.aten.sin %34521 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%34524 = torch.aten.complex %34522, %34523 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%34525 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34526 = flow.tensor.transfer %34525 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%34527 = torch_c.from_builtin_tensor %34526 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34528 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34529 = flow.tensor.transfer %34528 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%34530 = torch_c.from_builtin_tensor %34529 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34531 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34532 = flow.tensor.transfer %34531 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%34533 = torch_c.from_builtin_tensor %34532 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34534 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34535 = flow.tensor.transfer %34534 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%34536 = torch_c.from_builtin_tensor %34535 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34537 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34538 = flow.tensor.transfer %34537 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%34539 = torch_c.from_builtin_tensor %34538 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34540 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34541 = flow.tensor.transfer %34540 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%34542 = torch_c.from_builtin_tensor %34541 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34543 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34544 = flow.tensor.transfer %34543 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%34545 = torch_c.from_builtin_tensor %34544 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%34546 = torch_c.to_builtin_tensor %34524 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%34547 = flow.tensor.transfer %34546 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%34548 = torch_c.from_builtin_tensor %34547 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_32012 = torch.constant.int 1
%34549 = torch.aten.size.int %34198, %int1_32012 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32013 = torch.constant.int 0
%34550 = torch.aten.add.int %int0_32013, %34549 : !torch.int, !torch.int -> !torch.int
%int0_32014 = torch.constant.int 0
%int0_32015 = torch.constant.int 0
%int1_32016 = torch.constant.int 1
%34551 = torch.aten.slice.Tensor %34527, %int0_32014, %int0_32015, %34550, %int1_32016 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34551, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32017 = torch.constant.int 1
%int0_32018 = torch.constant.int 0
%int9223372036854775807_32019 = torch.constant.int 9223372036854775807
%int1_32020 = torch.constant.int 1
%34552 = torch.aten.slice.Tensor %34551, %int1_32017, %int0_32018, %int9223372036854775807_32019, %int1_32020 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34552, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32021 = torch.constant.int 0
%34553 = torch.aten.unsqueeze %34552, %int0_32021 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34553, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32022 = torch.constant.int 2
%34554 = torch.aten.unsqueeze %34553, %int2_32022 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34554, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32023 = torch.constant.int 3
%int0_32024 = torch.constant.int 0
%int9223372036854775807_32025 = torch.constant.int 9223372036854775807
%int1_32026 = torch.constant.int 1
%34555 = torch.aten.slice.Tensor %34554, %int3_32023, %int0_32024, %int9223372036854775807_32025, %int1_32026 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34555, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34556 = torch_c.to_builtin_tensor %34322 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32027 = arith.constant 1 : index
%dim_32028 = tensor.dim %34556, %c1_32027 : tensor<4x?x1x128xf16>
%34557 = flow.tensor.bitcast %34556 : tensor<4x?x1x128xf16>{%dim_32028} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32028}
%34558 = torch_c.from_builtin_tensor %34557 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34559 = torch.aten.mul.Tensor %34558, %34555 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34560 = torch_c.to_builtin_tensor %34559 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32029 = arith.constant 1 : index
%dim_32030 = tensor.dim %34560, %c1_32029 : tensor<4x?x1x64xcomplex<f32>>
%34561 = flow.tensor.bitcast %34560 : tensor<4x?x1x64xcomplex<f32>>{%dim_32030} -> tensor<4x?x1x128xf32>{%dim_32030}
%34562 = torch_c.from_builtin_tensor %34561 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32031 = torch.constant.int 5
%34563 = torch.prims.convert_element_type %34562, %int5_32031 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34563, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32032 = torch.constant.int 1
%34564 = torch.aten.size.int %34204, %int1_32032 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32033 = torch.constant.int 0
%34565 = torch.aten.add.int %int0_32033, %34564 : !torch.int, !torch.int -> !torch.int
%int0_32034 = torch.constant.int 0
%int0_32035 = torch.constant.int 0
%int1_32036 = torch.constant.int 1
%34566 = torch.aten.slice.Tensor %34530, %int0_32034, %int0_32035, %34565, %int1_32036 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34566, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32037 = torch.constant.int 1
%int0_32038 = torch.constant.int 0
%int9223372036854775807_32039 = torch.constant.int 9223372036854775807
%int1_32040 = torch.constant.int 1
%34567 = torch.aten.slice.Tensor %34566, %int1_32037, %int0_32038, %int9223372036854775807_32039, %int1_32040 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34567, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32041 = torch.constant.int 0
%34568 = torch.aten.unsqueeze %34567, %int0_32041 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34568, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32042 = torch.constant.int 2
%34569 = torch.aten.unsqueeze %34568, %int2_32042 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34569, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32043 = torch.constant.int 3
%int0_32044 = torch.constant.int 0
%int9223372036854775807_32045 = torch.constant.int 9223372036854775807
%int1_32046 = torch.constant.int 1
%34570 = torch.aten.slice.Tensor %34569, %int3_32043, %int0_32044, %int9223372036854775807_32045, %int1_32046 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34570, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34571 = torch_c.to_builtin_tensor %34324 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32047 = arith.constant 1 : index
%dim_32048 = tensor.dim %34571, %c1_32047 : tensor<4x?x1x128xf16>
%34572 = flow.tensor.bitcast %34571 : tensor<4x?x1x128xf16>{%dim_32048} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32048}
%34573 = torch_c.from_builtin_tensor %34572 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34574 = torch.aten.mul.Tensor %34573, %34570 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34575 = torch_c.to_builtin_tensor %34574 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32049 = arith.constant 1 : index
%dim_32050 = tensor.dim %34575, %c1_32049 : tensor<4x?x1x64xcomplex<f32>>
%34576 = flow.tensor.bitcast %34575 : tensor<4x?x1x64xcomplex<f32>>{%dim_32050} -> tensor<4x?x1x128xf32>{%dim_32050}
%34577 = torch_c.from_builtin_tensor %34576 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32051 = torch.constant.int 5
%34578 = torch.prims.convert_element_type %34577, %int5_32051 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34578, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32052 = torch.constant.int 1
%34579 = torch.aten.size.int %34210, %int1_32052 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32053 = torch.constant.int 0
%34580 = torch.aten.add.int %int0_32053, %34579 : !torch.int, !torch.int -> !torch.int
%int0_32054 = torch.constant.int 0
%int0_32055 = torch.constant.int 0
%int1_32056 = torch.constant.int 1
%34581 = torch.aten.slice.Tensor %34533, %int0_32054, %int0_32055, %34580, %int1_32056 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34581, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32057 = torch.constant.int 1
%int0_32058 = torch.constant.int 0
%int9223372036854775807_32059 = torch.constant.int 9223372036854775807
%int1_32060 = torch.constant.int 1
%34582 = torch.aten.slice.Tensor %34581, %int1_32057, %int0_32058, %int9223372036854775807_32059, %int1_32060 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34582, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32061 = torch.constant.int 0
%34583 = torch.aten.unsqueeze %34582, %int0_32061 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34583, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32062 = torch.constant.int 2
%34584 = torch.aten.unsqueeze %34583, %int2_32062 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34584, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32063 = torch.constant.int 3
%int0_32064 = torch.constant.int 0
%int9223372036854775807_32065 = torch.constant.int 9223372036854775807
%int1_32066 = torch.constant.int 1
%34585 = torch.aten.slice.Tensor %34584, %int3_32063, %int0_32064, %int9223372036854775807_32065, %int1_32066 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34585, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34586 = torch_c.to_builtin_tensor %34326 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32067 = arith.constant 1 : index
%dim_32068 = tensor.dim %34586, %c1_32067 : tensor<4x?x1x128xf16>
%34587 = flow.tensor.bitcast %34586 : tensor<4x?x1x128xf16>{%dim_32068} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32068}
%34588 = torch_c.from_builtin_tensor %34587 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34588, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34589 = torch.aten.mul.Tensor %34588, %34585 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34590 = torch_c.to_builtin_tensor %34589 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32069 = arith.constant 1 : index
%dim_32070 = tensor.dim %34590, %c1_32069 : tensor<4x?x1x64xcomplex<f32>>
%34591 = flow.tensor.bitcast %34590 : tensor<4x?x1x64xcomplex<f32>>{%dim_32070} -> tensor<4x?x1x128xf32>{%dim_32070}
%34592 = torch_c.from_builtin_tensor %34591 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32071 = torch.constant.int 5
%34593 = torch.prims.convert_element_type %34592, %int5_32071 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32072 = torch.constant.int 1
%34594 = torch.aten.size.int %34216, %int1_32072 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32073 = torch.constant.int 0
%34595 = torch.aten.add.int %int0_32073, %34594 : !torch.int, !torch.int -> !torch.int
%int0_32074 = torch.constant.int 0
%int0_32075 = torch.constant.int 0
%int1_32076 = torch.constant.int 1
%34596 = torch.aten.slice.Tensor %34536, %int0_32074, %int0_32075, %34595, %int1_32076 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34596, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32077 = torch.constant.int 1
%int0_32078 = torch.constant.int 0
%int9223372036854775807_32079 = torch.constant.int 9223372036854775807
%int1_32080 = torch.constant.int 1
%34597 = torch.aten.slice.Tensor %34596, %int1_32077, %int0_32078, %int9223372036854775807_32079, %int1_32080 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34597, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32081 = torch.constant.int 0
%34598 = torch.aten.unsqueeze %34597, %int0_32081 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34598, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32082 = torch.constant.int 2
%34599 = torch.aten.unsqueeze %34598, %int2_32082 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34599, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32083 = torch.constant.int 3
%int0_32084 = torch.constant.int 0
%int9223372036854775807_32085 = torch.constant.int 9223372036854775807
%int1_32086 = torch.constant.int 1
%34600 = torch.aten.slice.Tensor %34599, %int3_32083, %int0_32084, %int9223372036854775807_32085, %int1_32086 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34600, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34601 = torch_c.to_builtin_tensor %34328 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32087 = arith.constant 1 : index
%dim_32088 = tensor.dim %34601, %c1_32087 : tensor<4x?x1x128xf16>
%34602 = flow.tensor.bitcast %34601 : tensor<4x?x1x128xf16>{%dim_32088} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32088}
%34603 = torch_c.from_builtin_tensor %34602 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34603, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34604 = torch.aten.mul.Tensor %34603, %34600 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34605 = torch_c.to_builtin_tensor %34604 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32089 = arith.constant 1 : index
%dim_32090 = tensor.dim %34605, %c1_32089 : tensor<4x?x1x64xcomplex<f32>>
%34606 = flow.tensor.bitcast %34605 : tensor<4x?x1x64xcomplex<f32>>{%dim_32090} -> tensor<4x?x1x128xf32>{%dim_32090}
%34607 = torch_c.from_builtin_tensor %34606 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32091 = torch.constant.int 5
%34608 = torch.prims.convert_element_type %34607, %int5_32091 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32092 = torch.constant.int 1
%34609 = torch.aten.size.int %34222, %int1_32092 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32093 = torch.constant.int 0
%34610 = torch.aten.add.int %int0_32093, %34609 : !torch.int, !torch.int -> !torch.int
%int0_32094 = torch.constant.int 0
%int0_32095 = torch.constant.int 0
%int1_32096 = torch.constant.int 1
%34611 = torch.aten.slice.Tensor %34539, %int0_32094, %int0_32095, %34610, %int1_32096 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34611, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32097 = torch.constant.int 1
%int0_32098 = torch.constant.int 0
%int9223372036854775807_32099 = torch.constant.int 9223372036854775807
%int1_32100 = torch.constant.int 1
%34612 = torch.aten.slice.Tensor %34611, %int1_32097, %int0_32098, %int9223372036854775807_32099, %int1_32100 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34612, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32101 = torch.constant.int 0
%34613 = torch.aten.unsqueeze %34612, %int0_32101 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34613, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32102 = torch.constant.int 2
%34614 = torch.aten.unsqueeze %34613, %int2_32102 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34614, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32103 = torch.constant.int 3
%int0_32104 = torch.constant.int 0
%int9223372036854775807_32105 = torch.constant.int 9223372036854775807
%int1_32106 = torch.constant.int 1
%34615 = torch.aten.slice.Tensor %34614, %int3_32103, %int0_32104, %int9223372036854775807_32105, %int1_32106 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34615, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34616 = torch_c.to_builtin_tensor %34330 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32107 = arith.constant 1 : index
%dim_32108 = tensor.dim %34616, %c1_32107 : tensor<4x?x1x128xf16>
%34617 = flow.tensor.bitcast %34616 : tensor<4x?x1x128xf16>{%dim_32108} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32108}
%34618 = torch_c.from_builtin_tensor %34617 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34619 = torch.aten.mul.Tensor %34618, %34615 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34620 = torch_c.to_builtin_tensor %34619 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32109 = arith.constant 1 : index
%dim_32110 = tensor.dim %34620, %c1_32109 : tensor<4x?x1x64xcomplex<f32>>
%34621 = flow.tensor.bitcast %34620 : tensor<4x?x1x64xcomplex<f32>>{%dim_32110} -> tensor<4x?x1x128xf32>{%dim_32110}
%34622 = torch_c.from_builtin_tensor %34621 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32111 = torch.constant.int 5
%34623 = torch.prims.convert_element_type %34622, %int5_32111 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32112 = torch.constant.int 1
%34624 = torch.aten.size.int %34228, %int1_32112 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32113 = torch.constant.int 0
%34625 = torch.aten.add.int %int0_32113, %34624 : !torch.int, !torch.int -> !torch.int
%int0_32114 = torch.constant.int 0
%int0_32115 = torch.constant.int 0
%int1_32116 = torch.constant.int 1
%34626 = torch.aten.slice.Tensor %34542, %int0_32114, %int0_32115, %34625, %int1_32116 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34626, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32117 = torch.constant.int 1
%int0_32118 = torch.constant.int 0
%int9223372036854775807_32119 = torch.constant.int 9223372036854775807
%int1_32120 = torch.constant.int 1
%34627 = torch.aten.slice.Tensor %34626, %int1_32117, %int0_32118, %int9223372036854775807_32119, %int1_32120 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34627, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32121 = torch.constant.int 0
%34628 = torch.aten.unsqueeze %34627, %int0_32121 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34628, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32122 = torch.constant.int 2
%34629 = torch.aten.unsqueeze %34628, %int2_32122 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34629, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32123 = torch.constant.int 3
%int0_32124 = torch.constant.int 0
%int9223372036854775807_32125 = torch.constant.int 9223372036854775807
%int1_32126 = torch.constant.int 1
%34630 = torch.aten.slice.Tensor %34629, %int3_32123, %int0_32124, %int9223372036854775807_32125, %int1_32126 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34630, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34631 = torch_c.to_builtin_tensor %34332 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32127 = arith.constant 1 : index
%dim_32128 = tensor.dim %34631, %c1_32127 : tensor<4x?x1x128xf16>
%34632 = flow.tensor.bitcast %34631 : tensor<4x?x1x128xf16>{%dim_32128} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32128}
%34633 = torch_c.from_builtin_tensor %34632 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34634 = torch.aten.mul.Tensor %34633, %34630 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34635 = torch_c.to_builtin_tensor %34634 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32129 = arith.constant 1 : index
%dim_32130 = tensor.dim %34635, %c1_32129 : tensor<4x?x1x64xcomplex<f32>>
%34636 = flow.tensor.bitcast %34635 : tensor<4x?x1x64xcomplex<f32>>{%dim_32130} -> tensor<4x?x1x128xf32>{%dim_32130}
%34637 = torch_c.from_builtin_tensor %34636 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32131 = torch.constant.int 5
%34638 = torch.prims.convert_element_type %34637, %int5_32131 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32132 = torch.constant.int 1
%34639 = torch.aten.size.int %34234, %int1_32132 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32133 = torch.constant.int 0
%34640 = torch.aten.add.int %int0_32133, %34639 : !torch.int, !torch.int -> !torch.int
%int0_32134 = torch.constant.int 0
%int0_32135 = torch.constant.int 0
%int1_32136 = torch.constant.int 1
%34641 = torch.aten.slice.Tensor %34545, %int0_32134, %int0_32135, %34640, %int1_32136 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34641, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32137 = torch.constant.int 1
%int0_32138 = torch.constant.int 0
%int9223372036854775807_32139 = torch.constant.int 9223372036854775807
%int1_32140 = torch.constant.int 1
%34642 = torch.aten.slice.Tensor %34641, %int1_32137, %int0_32138, %int9223372036854775807_32139, %int1_32140 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34642, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32141 = torch.constant.int 0
%34643 = torch.aten.unsqueeze %34642, %int0_32141 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34643, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32142 = torch.constant.int 2
%34644 = torch.aten.unsqueeze %34643, %int2_32142 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34644, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32143 = torch.constant.int 3
%int0_32144 = torch.constant.int 0
%int9223372036854775807_32145 = torch.constant.int 9223372036854775807
%int1_32146 = torch.constant.int 1
%34645 = torch.aten.slice.Tensor %34644, %int3_32143, %int0_32144, %int9223372036854775807_32145, %int1_32146 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34645, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34646 = torch_c.to_builtin_tensor %34334 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32147 = arith.constant 1 : index
%dim_32148 = tensor.dim %34646, %c1_32147 : tensor<4x?x1x128xf16>
%34647 = flow.tensor.bitcast %34646 : tensor<4x?x1x128xf16>{%dim_32148} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32148}
%34648 = torch_c.from_builtin_tensor %34647 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34649 = torch.aten.mul.Tensor %34648, %34645 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34650 = torch_c.to_builtin_tensor %34649 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32149 = arith.constant 1 : index
%dim_32150 = tensor.dim %34650, %c1_32149 : tensor<4x?x1x64xcomplex<f32>>
%34651 = flow.tensor.bitcast %34650 : tensor<4x?x1x64xcomplex<f32>>{%dim_32150} -> tensor<4x?x1x128xf32>{%dim_32150}
%34652 = torch_c.from_builtin_tensor %34651 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34652, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32151 = torch.constant.int 5
%34653 = torch.prims.convert_element_type %34652, %int5_32151 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_32152 = torch.constant.int 1
%34654 = torch.aten.size.int %34240, %int1_32152 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_32153 = torch.constant.int 0
%34655 = torch.aten.add.int %int0_32153, %34654 : !torch.int, !torch.int -> !torch.int
%int0_32154 = torch.constant.int 0
%int0_32155 = torch.constant.int 0
%int1_32156 = torch.constant.int 1
%34656 = torch.aten.slice.Tensor %34548, %int0_32154, %int0_32155, %34655, %int1_32156 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34656, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_32157 = torch.constant.int 1
%int0_32158 = torch.constant.int 0
%int9223372036854775807_32159 = torch.constant.int 9223372036854775807
%int1_32160 = torch.constant.int 1
%34657 = torch.aten.slice.Tensor %34656, %int1_32157, %int0_32158, %int9223372036854775807_32159, %int1_32160 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %34657, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_32161 = torch.constant.int 0
%34658 = torch.aten.unsqueeze %34657, %int0_32161 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %34658, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_32162 = torch.constant.int 2
%34659 = torch.aten.unsqueeze %34658, %int2_32162 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34659, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_32163 = torch.constant.int 3
%int0_32164 = torch.constant.int 0
%int9223372036854775807_32165 = torch.constant.int 9223372036854775807
%int1_32166 = torch.constant.int 1
%34660 = torch.aten.slice.Tensor %34659, %int3_32163, %int0_32164, %int9223372036854775807_32165, %int1_32166 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34660, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%34661 = torch_c.to_builtin_tensor %34336 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_32167 = arith.constant 1 : index
%dim_32168 = tensor.dim %34661, %c1_32167 : tensor<4x?x1x128xf16>
%34662 = flow.tensor.bitcast %34661 : tensor<4x?x1x128xf16>{%dim_32168} -> tensor<4x?x1x64xcomplex<f16>>{%dim_32168}
%34663 = torch_c.from_builtin_tensor %34662 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %34663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%34664 = torch.aten.mul.Tensor %34663, %34660 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %34664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%34665 = torch_c.to_builtin_tensor %34664 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_32169 = arith.constant 1 : index
%dim_32170 = tensor.dim %34665, %c1_32169 : tensor<4x?x1x64xcomplex<f32>>
%34666 = flow.tensor.bitcast %34665 : tensor<4x?x1x64xcomplex<f32>>{%dim_32170} -> tensor<4x?x1x128xf32>{%dim_32170}
%34667 = torch_c.from_builtin_tensor %34666 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %34667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_32171 = torch.constant.int 5
%34668 = torch.prims.convert_element_type %34667, %int5_32171 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %34668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_32172 = torch.constant.int 64
%34669 = torch.aten.mul.Scalar %2364, %int64_32172 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34669, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32173 = torch.constant.int 64
%34670 = torch.aten.mul.Scalar %2367, %int64_32173 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34670, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32174 = torch.constant.int 64
%34671 = torch.aten.mul.Scalar %2370, %int64_32174 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34671, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32175 = torch.constant.int 64
%34672 = torch.aten.mul.Scalar %2373, %int64_32175 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34672, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32176 = torch.constant.int 64
%34673 = torch.aten.mul.Scalar %2376, %int64_32176 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34673, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32177 = torch.constant.int 64
%34674 = torch.aten.mul.Scalar %2379, %int64_32177 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34674, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32178 = torch.constant.int 64
%34675 = torch.aten.mul.Scalar %2382, %int64_32178 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34675, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_32179 = torch.constant.int 64
%34676 = torch.aten.mul.Scalar %2385, %int64_32179 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34676, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34 = torch.constant.int 34
%int1_32180 = torch.constant.int 1
%34677 = torch.aten.add.Scalar %34669, %int34, %int1_32180 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34677, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32181 = torch.constant.int 34
%int1_32182 = torch.constant.int 1
%34678 = torch.aten.add.Scalar %34670, %int34_32181, %int1_32182 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34678, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32183 = torch.constant.int 34
%int1_32184 = torch.constant.int 1
%34679 = torch.aten.add.Scalar %34671, %int34_32183, %int1_32184 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34679, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32185 = torch.constant.int 34
%int1_32186 = torch.constant.int 1
%34680 = torch.aten.add.Scalar %34672, %int34_32185, %int1_32186 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34680, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32187 = torch.constant.int 34
%int1_32188 = torch.constant.int 1
%34681 = torch.aten.add.Scalar %34673, %int34_32187, %int1_32188 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34681, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32189 = torch.constant.int 34
%int1_32190 = torch.constant.int 1
%34682 = torch.aten.add.Scalar %34674, %int34_32189, %int1_32190 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34682, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32191 = torch.constant.int 34
%int1_32192 = torch.constant.int 1
%34683 = torch.aten.add.Scalar %34675, %int34_32191, %int1_32192 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34683, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int34_32193 = torch.constant.int 34
%int1_32194 = torch.constant.int 1
%34684 = torch.aten.add.Scalar %34676, %int34_32193, %int1_32194 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34684, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_32195 = torch.constant.int 4
%int16_32196 = torch.constant.int 16
%int1_32197 = torch.constant.int 1
%int128_32198 = torch.constant.int 128
%34685 = torch.prim.ListConstruct %int4_32195, %3095, %int16_32196, %int1_32197, %int128_32198 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34686 = torch.aten.view %34563, %34685 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34686, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32199 = torch.constant.int 4
%int16_32200 = torch.constant.int 16
%int1_32201 = torch.constant.int 1
%int128_32202 = torch.constant.int 128
%34687 = torch.prim.ListConstruct %int4_32199, %3095, %int16_32200, %int1_32201, %int128_32202 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34688 = torch.aten.view %34578, %34687 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34688, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32203 = torch.constant.int 4
%int16_32204 = torch.constant.int 16
%int1_32205 = torch.constant.int 1
%int128_32206 = torch.constant.int 128
%34689 = torch.prim.ListConstruct %int4_32203, %3095, %int16_32204, %int1_32205, %int128_32206 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34690 = torch.aten.view %34593, %34689 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34690, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32207 = torch.constant.int 4
%int16_32208 = torch.constant.int 16
%int1_32209 = torch.constant.int 1
%int128_32210 = torch.constant.int 128
%34691 = torch.prim.ListConstruct %int4_32207, %3095, %int16_32208, %int1_32209, %int128_32210 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34692 = torch.aten.view %34608, %34691 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34692, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32211 = torch.constant.int 4
%int16_32212 = torch.constant.int 16
%int1_32213 = torch.constant.int 1
%int128_32214 = torch.constant.int 128
%34693 = torch.prim.ListConstruct %int4_32211, %3095, %int16_32212, %int1_32213, %int128_32214 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34694 = torch.aten.view %34623, %34693 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34694, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32215 = torch.constant.int 4
%int16_32216 = torch.constant.int 16
%int1_32217 = torch.constant.int 1
%int128_32218 = torch.constant.int 128
%34695 = torch.prim.ListConstruct %int4_32215, %3095, %int16_32216, %int1_32217, %int128_32218 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34696 = torch.aten.view %34638, %34695 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34696, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32219 = torch.constant.int 4
%int16_32220 = torch.constant.int 16
%int1_32221 = torch.constant.int 1
%int128_32222 = torch.constant.int 128
%34697 = torch.prim.ListConstruct %int4_32219, %3095, %int16_32220, %int1_32221, %int128_32222 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34698 = torch.aten.view %34653, %34697 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34698, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32223 = torch.constant.int 4
%int16_32224 = torch.constant.int 16
%int1_32225 = torch.constant.int 1
%int128_32226 = torch.constant.int 128
%34699 = torch.prim.ListConstruct %int4_32223, %3095, %int16_32224, %int1_32225, %int128_32226 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34700 = torch.aten.view %34668, %34699 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34700, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32227 = torch.constant.int 4
%34701 = torch.aten.mul.int %int4_32227, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32228 = torch.constant.int 16
%int1_32229 = torch.constant.int 1
%int128_32230 = torch.constant.int 128
%34702 = torch.prim.ListConstruct %34701, %int16_32228, %int1_32229, %int128_32230 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34703 = torch.aten.view %34686, %34702 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34703, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32231 = torch.constant.int 4
%34704 = torch.aten.mul.int %int4_32231, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32232 = torch.constant.int 16
%int1_32233 = torch.constant.int 1
%int128_32234 = torch.constant.int 128
%34705 = torch.prim.ListConstruct %34704, %int16_32232, %int1_32233, %int128_32234 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34706 = torch.aten.view %34688, %34705 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34706, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32235 = torch.constant.int 4
%34707 = torch.aten.mul.int %int4_32235, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32236 = torch.constant.int 16
%int1_32237 = torch.constant.int 1
%int128_32238 = torch.constant.int 128
%34708 = torch.prim.ListConstruct %34707, %int16_32236, %int1_32237, %int128_32238 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34709 = torch.aten.view %34690, %34708 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34709, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32239 = torch.constant.int 4
%34710 = torch.aten.mul.int %int4_32239, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32240 = torch.constant.int 16
%int1_32241 = torch.constant.int 1
%int128_32242 = torch.constant.int 128
%34711 = torch.prim.ListConstruct %34710, %int16_32240, %int1_32241, %int128_32242 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34712 = torch.aten.view %34692, %34711 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34712, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32243 = torch.constant.int 4
%34713 = torch.aten.mul.int %int4_32243, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32244 = torch.constant.int 16
%int1_32245 = torch.constant.int 1
%int128_32246 = torch.constant.int 128
%34714 = torch.prim.ListConstruct %34713, %int16_32244, %int1_32245, %int128_32246 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34715 = torch.aten.view %34694, %34714 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34715, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32247 = torch.constant.int 4
%34716 = torch.aten.mul.int %int4_32247, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32248 = torch.constant.int 16
%int1_32249 = torch.constant.int 1
%int128_32250 = torch.constant.int 128
%34717 = torch.prim.ListConstruct %34716, %int16_32248, %int1_32249, %int128_32250 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34718 = torch.aten.view %34696, %34717 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34718, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32251 = torch.constant.int 4
%34719 = torch.aten.mul.int %int4_32251, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32252 = torch.constant.int 16
%int1_32253 = torch.constant.int 1
%int128_32254 = torch.constant.int 128
%34720 = torch.prim.ListConstruct %34719, %int16_32252, %int1_32253, %int128_32254 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34721 = torch.aten.view %34698, %34720 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34721, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32255 = torch.constant.int 4
%34722 = torch.aten.mul.int %int4_32255, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32256 = torch.constant.int 16
%int1_32257 = torch.constant.int 1
%int128_32258 = torch.constant.int 128
%34723 = torch.prim.ListConstruct %34722, %int16_32256, %int1_32257, %int128_32258 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34724 = torch.aten.view %34700, %34723 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34724, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32259 = torch.constant.int 4
%34725 = torch.aten.mul.int %int4_32259, %3095 : !torch.int, !torch.int -> !torch.int
%34726 = torch.prim.ListConstruct %34725 : (!torch.int) -> !torch.list<int>
%34727 = torch.aten.view %34677, %34726 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34727, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32260 = torch.constant.int 4
%34728 = torch.aten.mul.int %int4_32260, %3095 : !torch.int, !torch.int -> !torch.int
%34729 = torch.prim.ListConstruct %34728 : (!torch.int) -> !torch.list<int>
%34730 = torch.aten.view %34678, %34729 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34730, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32261 = torch.constant.int 4
%34731 = torch.aten.mul.int %int4_32261, %3095 : !torch.int, !torch.int -> !torch.int
%34732 = torch.prim.ListConstruct %34731 : (!torch.int) -> !torch.list<int>
%34733 = torch.aten.view %34679, %34732 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34733, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32262 = torch.constant.int 4
%34734 = torch.aten.mul.int %int4_32262, %3095 : !torch.int, !torch.int -> !torch.int
%34735 = torch.prim.ListConstruct %34734 : (!torch.int) -> !torch.list<int>
%34736 = torch.aten.view %34680, %34735 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34736, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32263 = torch.constant.int 4
%34737 = torch.aten.mul.int %int4_32263, %3095 : !torch.int, !torch.int -> !torch.int
%34738 = torch.prim.ListConstruct %34737 : (!torch.int) -> !torch.list<int>
%34739 = torch.aten.view %34681, %34738 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34739, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32264 = torch.constant.int 4
%34740 = torch.aten.mul.int %int4_32264, %3095 : !torch.int, !torch.int -> !torch.int
%34741 = torch.prim.ListConstruct %34740 : (!torch.int) -> !torch.list<int>
%34742 = torch.aten.view %34682, %34741 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34742, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32265 = torch.constant.int 4
%34743 = torch.aten.mul.int %int4_32265, %3095 : !torch.int, !torch.int -> !torch.int
%34744 = torch.prim.ListConstruct %34743 : (!torch.int) -> !torch.list<int>
%34745 = torch.aten.view %34683, %34744 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34745, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32266 = torch.constant.int 4
%34746 = torch.aten.mul.int %int4_32266, %3095 : !torch.int, !torch.int -> !torch.int
%34747 = torch.prim.ListConstruct %34746 : (!torch.int) -> !torch.list<int>
%34748 = torch.aten.view %34684, %34747 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34748, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32267 = torch.constant.int 4
%int16_32268 = torch.constant.int 16
%int1_32269 = torch.constant.int 1
%int128_32270 = torch.constant.int 128
%34749 = torch.prim.ListConstruct %int4_32267, %3095, %int16_32268, %int1_32269, %int128_32270 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34750 = torch.aten.view %34338, %34749 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34750, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32271 = torch.constant.int 4
%int16_32272 = torch.constant.int 16
%int1_32273 = torch.constant.int 1
%int128_32274 = torch.constant.int 128
%34751 = torch.prim.ListConstruct %int4_32271, %3095, %int16_32272, %int1_32273, %int128_32274 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34752 = torch.aten.view %34340, %34751 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34752, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32275 = torch.constant.int 4
%int16_32276 = torch.constant.int 16
%int1_32277 = torch.constant.int 1
%int128_32278 = torch.constant.int 128
%34753 = torch.prim.ListConstruct %int4_32275, %3095, %int16_32276, %int1_32277, %int128_32278 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34754 = torch.aten.view %34342, %34753 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34754, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32279 = torch.constant.int 4
%int16_32280 = torch.constant.int 16
%int1_32281 = torch.constant.int 1
%int128_32282 = torch.constant.int 128
%34755 = torch.prim.ListConstruct %int4_32279, %3095, %int16_32280, %int1_32281, %int128_32282 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34756 = torch.aten.view %34344, %34755 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34756, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32283 = torch.constant.int 4
%int16_32284 = torch.constant.int 16
%int1_32285 = torch.constant.int 1
%int128_32286 = torch.constant.int 128
%34757 = torch.prim.ListConstruct %int4_32283, %3095, %int16_32284, %int1_32285, %int128_32286 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34758 = torch.aten.view %34346, %34757 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34758, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32287 = torch.constant.int 4
%int16_32288 = torch.constant.int 16
%int1_32289 = torch.constant.int 1
%int128_32290 = torch.constant.int 128
%34759 = torch.prim.ListConstruct %int4_32287, %3095, %int16_32288, %int1_32289, %int128_32290 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34760 = torch.aten.view %34348, %34759 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34760, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32291 = torch.constant.int 4
%int16_32292 = torch.constant.int 16
%int1_32293 = torch.constant.int 1
%int128_32294 = torch.constant.int 128
%34761 = torch.prim.ListConstruct %int4_32291, %3095, %int16_32292, %int1_32293, %int128_32294 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34762 = torch.aten.view %34350, %34761 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34762, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32295 = torch.constant.int 4
%int16_32296 = torch.constant.int 16
%int1_32297 = torch.constant.int 1
%int128_32298 = torch.constant.int 128
%34763 = torch.prim.ListConstruct %int4_32295, %3095, %int16_32296, %int1_32297, %int128_32298 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34764 = torch.aten.view %34352, %34763 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %34764, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_32299 = torch.constant.int 4
%34765 = torch.aten.mul.int %int4_32299, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32300 = torch.constant.int 16
%int1_32301 = torch.constant.int 1
%int128_32302 = torch.constant.int 128
%34766 = torch.prim.ListConstruct %34765, %int16_32300, %int1_32301, %int128_32302 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34767 = torch.aten.view %34750, %34766 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34767, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32303 = torch.constant.int 4
%34768 = torch.aten.mul.int %int4_32303, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32304 = torch.constant.int 16
%int1_32305 = torch.constant.int 1
%int128_32306 = torch.constant.int 128
%34769 = torch.prim.ListConstruct %34768, %int16_32304, %int1_32305, %int128_32306 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34770 = torch.aten.view %34752, %34769 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34770, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32307 = torch.constant.int 4
%34771 = torch.aten.mul.int %int4_32307, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32308 = torch.constant.int 16
%int1_32309 = torch.constant.int 1
%int128_32310 = torch.constant.int 128
%34772 = torch.prim.ListConstruct %34771, %int16_32308, %int1_32309, %int128_32310 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34773 = torch.aten.view %34754, %34772 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34773, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32311 = torch.constant.int 4
%34774 = torch.aten.mul.int %int4_32311, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32312 = torch.constant.int 16
%int1_32313 = torch.constant.int 1
%int128_32314 = torch.constant.int 128
%34775 = torch.prim.ListConstruct %34774, %int16_32312, %int1_32313, %int128_32314 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34776 = torch.aten.view %34756, %34775 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34776, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32315 = torch.constant.int 4
%34777 = torch.aten.mul.int %int4_32315, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32316 = torch.constant.int 16
%int1_32317 = torch.constant.int 1
%int128_32318 = torch.constant.int 128
%34778 = torch.prim.ListConstruct %34777, %int16_32316, %int1_32317, %int128_32318 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34779 = torch.aten.view %34758, %34778 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34779, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32319 = torch.constant.int 4
%34780 = torch.aten.mul.int %int4_32319, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32320 = torch.constant.int 16
%int1_32321 = torch.constant.int 1
%int128_32322 = torch.constant.int 128
%34781 = torch.prim.ListConstruct %34780, %int16_32320, %int1_32321, %int128_32322 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34782 = torch.aten.view %34760, %34781 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34782, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32323 = torch.constant.int 4
%34783 = torch.aten.mul.int %int4_32323, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32324 = torch.constant.int 16
%int1_32325 = torch.constant.int 1
%int128_32326 = torch.constant.int 128
%34784 = torch.prim.ListConstruct %34783, %int16_32324, %int1_32325, %int128_32326 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34785 = torch.aten.view %34762, %34784 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34785, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_32327 = torch.constant.int 4
%34786 = torch.aten.mul.int %int4_32327, %3095 : !torch.int, !torch.int -> !torch.int
%int16_32328 = torch.constant.int 16
%int1_32329 = torch.constant.int 1
%int128_32330 = torch.constant.int 128
%34787 = torch.prim.ListConstruct %34786, %int16_32328, %int1_32329, %int128_32330 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34788 = torch.aten.view %34764, %34787 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34788, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_32331 = torch.constant.int 1
%int1_32332 = torch.constant.int 1
%34789 = torch.aten.add.Scalar %34677, %int1_32331, %int1_32332 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34789, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32333 = torch.constant.int 1
%int1_32334 = torch.constant.int 1
%34790 = torch.aten.add.Scalar %34678, %int1_32333, %int1_32334 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34790, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32335 = torch.constant.int 1
%int1_32336 = torch.constant.int 1
%34791 = torch.aten.add.Scalar %34679, %int1_32335, %int1_32336 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34791, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32337 = torch.constant.int 1
%int1_32338 = torch.constant.int 1
%34792 = torch.aten.add.Scalar %34680, %int1_32337, %int1_32338 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34792, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32339 = torch.constant.int 1
%int1_32340 = torch.constant.int 1
%34793 = torch.aten.add.Scalar %34681, %int1_32339, %int1_32340 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34793, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32341 = torch.constant.int 1
%int1_32342 = torch.constant.int 1
%34794 = torch.aten.add.Scalar %34682, %int1_32341, %int1_32342 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34794, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32343 = torch.constant.int 1
%int1_32344 = torch.constant.int 1
%34795 = torch.aten.add.Scalar %34683, %int1_32343, %int1_32344 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34795, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_32345 = torch.constant.int 1
%int1_32346 = torch.constant.int 1
%34796 = torch.aten.add.Scalar %34684, %int1_32345, %int1_32346 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %34796, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_32347 = torch.constant.int 4
%34797 = torch.aten.mul.int %int4_32347, %3095 : !torch.int, !torch.int -> !torch.int
%34798 = torch.prim.ListConstruct %34797 : (!torch.int) -> !torch.list<int>
%34799 = torch.aten.view %34789, %34798 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34799, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32348 = torch.constant.int 4
%34800 = torch.aten.mul.int %int4_32348, %3095 : !torch.int, !torch.int -> !torch.int
%34801 = torch.prim.ListConstruct %34800 : (!torch.int) -> !torch.list<int>
%34802 = torch.aten.view %34790, %34801 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34802, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32349 = torch.constant.int 4
%34803 = torch.aten.mul.int %int4_32349, %3095 : !torch.int, !torch.int -> !torch.int
%34804 = torch.prim.ListConstruct %34803 : (!torch.int) -> !torch.list<int>
%34805 = torch.aten.view %34791, %34804 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34805, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32350 = torch.constant.int 4
%34806 = torch.aten.mul.int %int4_32350, %3095 : !torch.int, !torch.int -> !torch.int
%34807 = torch.prim.ListConstruct %34806 : (!torch.int) -> !torch.list<int>
%34808 = torch.aten.view %34792, %34807 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34808, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32351 = torch.constant.int 4
%34809 = torch.aten.mul.int %int4_32351, %3095 : !torch.int, !torch.int -> !torch.int
%34810 = torch.prim.ListConstruct %34809 : (!torch.int) -> !torch.list<int>
%34811 = torch.aten.view %34793, %34810 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34811, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32352 = torch.constant.int 4
%34812 = torch.aten.mul.int %int4_32352, %3095 : !torch.int, !torch.int -> !torch.int
%34813 = torch.prim.ListConstruct %34812 : (!torch.int) -> !torch.list<int>
%34814 = torch.aten.view %34794, %34813 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34814, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32353 = torch.constant.int 4
%34815 = torch.aten.mul.int %int4_32353, %3095 : !torch.int, !torch.int -> !torch.int
%34816 = torch.prim.ListConstruct %34815 : (!torch.int) -> !torch.list<int>
%34817 = torch.aten.view %34795, %34816 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34817, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_32354 = torch.constant.int 4
%34818 = torch.aten.mul.int %int4_32354, %3095 : !torch.int, !torch.int -> !torch.int
%34819 = torch.prim.ListConstruct %34818 : (!torch.int) -> !torch.list<int>
%34820 = torch.aten.view %34796, %34819 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34820, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%34821 = torch.prim.ListConstruct %34727, %34799 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32355 = torch.constant.int 0
%34822 = torch.aten.cat %34821, %int0_32355 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34822, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34823 = torch.prim.ListConstruct %34730, %34802 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32356 = torch.constant.int 0
%34824 = torch.aten.cat %34823, %int0_32356 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34824, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34825 = torch.prim.ListConstruct %34733, %34805 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32357 = torch.constant.int 0
%34826 = torch.aten.cat %34825, %int0_32357 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34826, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34827 = torch.prim.ListConstruct %34736, %34808 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32358 = torch.constant.int 0
%34828 = torch.aten.cat %34827, %int0_32358 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34828, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34829 = torch.prim.ListConstruct %34739, %34811 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32359 = torch.constant.int 0
%34830 = torch.aten.cat %34829, %int0_32359 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34830, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34831 = torch.prim.ListConstruct %34742, %34814 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32360 = torch.constant.int 0
%34832 = torch.aten.cat %34831, %int0_32360 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34832, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34833 = torch.prim.ListConstruct %34745, %34817 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32361 = torch.constant.int 0
%34834 = torch.aten.cat %34833, %int0_32361 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34834, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34835 = torch.prim.ListConstruct %34748, %34820 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_32362 = torch.constant.int 0
%34836 = torch.aten.cat %34835, %int0_32362 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %34836, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%34837 = torch.prim.ListConstruct %34703, %34767 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32363 = torch.constant.int 0
%34838 = torch.aten.cat %34837, %int0_32363 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34838, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34839 = torch.prim.ListConstruct %34706, %34770 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32364 = torch.constant.int 0
%34840 = torch.aten.cat %34839, %int0_32364 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34840, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34841 = torch.prim.ListConstruct %34709, %34773 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32365 = torch.constant.int 0
%34842 = torch.aten.cat %34841, %int0_32365 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34842, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34843 = torch.prim.ListConstruct %34712, %34776 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32366 = torch.constant.int 0
%34844 = torch.aten.cat %34843, %int0_32366 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34844, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34845 = torch.prim.ListConstruct %34715, %34779 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32367 = torch.constant.int 0
%34846 = torch.aten.cat %34845, %int0_32367 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34846, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34847 = torch.prim.ListConstruct %34718, %34782 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32368 = torch.constant.int 0
%34848 = torch.aten.cat %34847, %int0_32368 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34848, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34849 = torch.prim.ListConstruct %34721, %34785 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32369 = torch.constant.int 0
%34850 = torch.aten.cat %34849, %int0_32369 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34850, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34851 = torch.prim.ListConstruct %34724, %34788 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_32370 = torch.constant.int 0
%34852 = torch.aten.cat %34851, %int0_32370 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34852, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32371 = torch.constant.int 32
%int2_32372 = torch.constant.int 2
%int16_32373 = torch.constant.int 16
%int1_32374 = torch.constant.int 1
%int128_32375 = torch.constant.int 128
%34853 = torch.prim.ListConstruct %3023, %int32_32371, %int2_32372, %int16_32373, %int1_32374, %int128_32375 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34854 = torch.aten.view %33003, %34853 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34854, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32376 = torch.constant.int 32
%34855 = torch.aten.mul.int %3023, %int32_32376 : !torch.int, !torch.int -> !torch.int
%int2_32377 = torch.constant.int 2
%34856 = torch.aten.mul.int %34855, %int2_32377 : !torch.int, !torch.int -> !torch.int
%int16_32378 = torch.constant.int 16
%int1_32379 = torch.constant.int 1
%int128_32380 = torch.constant.int 128
%34857 = torch.prim.ListConstruct %34856, %int16_32378, %int1_32379, %int128_32380 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34858 = torch.aten.view %34854, %34857 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34858, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34859 = torch.prim.ListConstruct %34822 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32381 = torch.constant.bool false
%34860 = torch.aten.index_put %34858, %34859, %34838, %false_32381 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34860, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32382 = torch.constant.int 32
%int2_32383 = torch.constant.int 2
%int16_32384 = torch.constant.int 16
%int1_32385 = torch.constant.int 1
%int128_32386 = torch.constant.int 128
%34861 = torch.prim.ListConstruct %3023, %int32_32382, %int2_32383, %int16_32384, %int1_32385, %int128_32386 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34862 = torch.aten.view %34860, %34861 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34862, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32387 = torch.constant.int 131072
%34863 = torch.prim.ListConstruct %3023, %int131072_32387 : (!torch.int, !torch.int) -> !torch.list<int>
%34864 = torch.aten.view %34862, %34863 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34864, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32388 = torch.constant.int 32
%int2_32389 = torch.constant.int 2
%int16_32390 = torch.constant.int 16
%int1_32391 = torch.constant.int 1
%int128_32392 = torch.constant.int 128
%34865 = torch.prim.ListConstruct %3026, %int32_32388, %int2_32389, %int16_32390, %int1_32391, %int128_32392 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34866 = torch.aten.view %33015, %34865 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34866, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32393 = torch.constant.int 32
%34867 = torch.aten.mul.int %3026, %int32_32393 : !torch.int, !torch.int -> !torch.int
%int2_32394 = torch.constant.int 2
%34868 = torch.aten.mul.int %34867, %int2_32394 : !torch.int, !torch.int -> !torch.int
%int16_32395 = torch.constant.int 16
%int1_32396 = torch.constant.int 1
%int128_32397 = torch.constant.int 128
%34869 = torch.prim.ListConstruct %34868, %int16_32395, %int1_32396, %int128_32397 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34870 = torch.aten.view %34866, %34869 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34870, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34871 = torch.prim.ListConstruct %34824 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32398 = torch.constant.bool false
%34872 = torch.aten.index_put %34870, %34871, %34840, %false_32398 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34872, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32399 = torch.constant.int 32
%int2_32400 = torch.constant.int 2
%int16_32401 = torch.constant.int 16
%int1_32402 = torch.constant.int 1
%int128_32403 = torch.constant.int 128
%34873 = torch.prim.ListConstruct %3026, %int32_32399, %int2_32400, %int16_32401, %int1_32402, %int128_32403 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34874 = torch.aten.view %34872, %34873 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34874, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32404 = torch.constant.int 131072
%34875 = torch.prim.ListConstruct %3026, %int131072_32404 : (!torch.int, !torch.int) -> !torch.list<int>
%34876 = torch.aten.view %34874, %34875 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34876, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32405 = torch.constant.int 32
%int2_32406 = torch.constant.int 2
%int16_32407 = torch.constant.int 16
%int1_32408 = torch.constant.int 1
%int128_32409 = torch.constant.int 128
%34877 = torch.prim.ListConstruct %3029, %int32_32405, %int2_32406, %int16_32407, %int1_32408, %int128_32409 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34878 = torch.aten.view %33027, %34877 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34878, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32410 = torch.constant.int 32
%34879 = torch.aten.mul.int %3029, %int32_32410 : !torch.int, !torch.int -> !torch.int
%int2_32411 = torch.constant.int 2
%34880 = torch.aten.mul.int %34879, %int2_32411 : !torch.int, !torch.int -> !torch.int
%int16_32412 = torch.constant.int 16
%int1_32413 = torch.constant.int 1
%int128_32414 = torch.constant.int 128
%34881 = torch.prim.ListConstruct %34880, %int16_32412, %int1_32413, %int128_32414 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34882 = torch.aten.view %34878, %34881 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34882, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34883 = torch.prim.ListConstruct %34826 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32415 = torch.constant.bool false
%34884 = torch.aten.index_put %34882, %34883, %34842, %false_32415 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34884, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32416 = torch.constant.int 32
%int2_32417 = torch.constant.int 2
%int16_32418 = torch.constant.int 16
%int1_32419 = torch.constant.int 1
%int128_32420 = torch.constant.int 128
%34885 = torch.prim.ListConstruct %3029, %int32_32416, %int2_32417, %int16_32418, %int1_32419, %int128_32420 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34886 = torch.aten.view %34884, %34885 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34886, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32421 = torch.constant.int 131072
%34887 = torch.prim.ListConstruct %3029, %int131072_32421 : (!torch.int, !torch.int) -> !torch.list<int>
%34888 = torch.aten.view %34886, %34887 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34888, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32422 = torch.constant.int 32
%int2_32423 = torch.constant.int 2
%int16_32424 = torch.constant.int 16
%int1_32425 = torch.constant.int 1
%int128_32426 = torch.constant.int 128
%34889 = torch.prim.ListConstruct %3032, %int32_32422, %int2_32423, %int16_32424, %int1_32425, %int128_32426 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34890 = torch.aten.view %33039, %34889 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34890, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32427 = torch.constant.int 32
%34891 = torch.aten.mul.int %3032, %int32_32427 : !torch.int, !torch.int -> !torch.int
%int2_32428 = torch.constant.int 2
%34892 = torch.aten.mul.int %34891, %int2_32428 : !torch.int, !torch.int -> !torch.int
%int16_32429 = torch.constant.int 16
%int1_32430 = torch.constant.int 1
%int128_32431 = torch.constant.int 128
%34893 = torch.prim.ListConstruct %34892, %int16_32429, %int1_32430, %int128_32431 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34894 = torch.aten.view %34890, %34893 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34894, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34895 = torch.prim.ListConstruct %34828 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32432 = torch.constant.bool false
%34896 = torch.aten.index_put %34894, %34895, %34844, %false_32432 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34896, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32433 = torch.constant.int 32
%int2_32434 = torch.constant.int 2
%int16_32435 = torch.constant.int 16
%int1_32436 = torch.constant.int 1
%int128_32437 = torch.constant.int 128
%34897 = torch.prim.ListConstruct %3032, %int32_32433, %int2_32434, %int16_32435, %int1_32436, %int128_32437 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34898 = torch.aten.view %34896, %34897 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34898, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32438 = torch.constant.int 131072
%34899 = torch.prim.ListConstruct %3032, %int131072_32438 : (!torch.int, !torch.int) -> !torch.list<int>
%34900 = torch.aten.view %34898, %34899 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34900, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32439 = torch.constant.int 32
%int2_32440 = torch.constant.int 2
%int16_32441 = torch.constant.int 16
%int1_32442 = torch.constant.int 1
%int128_32443 = torch.constant.int 128
%34901 = torch.prim.ListConstruct %3035, %int32_32439, %int2_32440, %int16_32441, %int1_32442, %int128_32443 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34902 = torch.aten.view %33051, %34901 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34902, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32444 = torch.constant.int 32
%34903 = torch.aten.mul.int %3035, %int32_32444 : !torch.int, !torch.int -> !torch.int
%int2_32445 = torch.constant.int 2
%34904 = torch.aten.mul.int %34903, %int2_32445 : !torch.int, !torch.int -> !torch.int
%int16_32446 = torch.constant.int 16
%int1_32447 = torch.constant.int 1
%int128_32448 = torch.constant.int 128
%34905 = torch.prim.ListConstruct %34904, %int16_32446, %int1_32447, %int128_32448 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34906 = torch.aten.view %34902, %34905 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34906, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34907 = torch.prim.ListConstruct %34830 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32449 = torch.constant.bool false
%34908 = torch.aten.index_put %34906, %34907, %34846, %false_32449 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34908, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32450 = torch.constant.int 32
%int2_32451 = torch.constant.int 2
%int16_32452 = torch.constant.int 16
%int1_32453 = torch.constant.int 1
%int128_32454 = torch.constant.int 128
%34909 = torch.prim.ListConstruct %3035, %int32_32450, %int2_32451, %int16_32452, %int1_32453, %int128_32454 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34910 = torch.aten.view %34908, %34909 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34910, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32455 = torch.constant.int 131072
%34911 = torch.prim.ListConstruct %3035, %int131072_32455 : (!torch.int, !torch.int) -> !torch.list<int>
%34912 = torch.aten.view %34910, %34911 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34912, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32456 = torch.constant.int 32
%int2_32457 = torch.constant.int 2
%int16_32458 = torch.constant.int 16
%int1_32459 = torch.constant.int 1
%int128_32460 = torch.constant.int 128
%34913 = torch.prim.ListConstruct %3038, %int32_32456, %int2_32457, %int16_32458, %int1_32459, %int128_32460 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34914 = torch.aten.view %33063, %34913 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34914, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32461 = torch.constant.int 32
%34915 = torch.aten.mul.int %3038, %int32_32461 : !torch.int, !torch.int -> !torch.int
%int2_32462 = torch.constant.int 2
%34916 = torch.aten.mul.int %34915, %int2_32462 : !torch.int, !torch.int -> !torch.int
%int16_32463 = torch.constant.int 16
%int1_32464 = torch.constant.int 1
%int128_32465 = torch.constant.int 128
%34917 = torch.prim.ListConstruct %34916, %int16_32463, %int1_32464, %int128_32465 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34918 = torch.aten.view %34914, %34917 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34918, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34919 = torch.prim.ListConstruct %34832 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32466 = torch.constant.bool false
%34920 = torch.aten.index_put %34918, %34919, %34848, %false_32466 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34920, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32467 = torch.constant.int 32
%int2_32468 = torch.constant.int 2
%int16_32469 = torch.constant.int 16
%int1_32470 = torch.constant.int 1
%int128_32471 = torch.constant.int 128
%34921 = torch.prim.ListConstruct %3038, %int32_32467, %int2_32468, %int16_32469, %int1_32470, %int128_32471 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34922 = torch.aten.view %34920, %34921 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34922, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32472 = torch.constant.int 131072
%34923 = torch.prim.ListConstruct %3038, %int131072_32472 : (!torch.int, !torch.int) -> !torch.list<int>
%34924 = torch.aten.view %34922, %34923 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34924, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32473 = torch.constant.int 32
%int2_32474 = torch.constant.int 2
%int16_32475 = torch.constant.int 16
%int1_32476 = torch.constant.int 1
%int128_32477 = torch.constant.int 128
%34925 = torch.prim.ListConstruct %3041, %int32_32473, %int2_32474, %int16_32475, %int1_32476, %int128_32477 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34926 = torch.aten.view %33075, %34925 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34926, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32478 = torch.constant.int 32
%34927 = torch.aten.mul.int %3041, %int32_32478 : !torch.int, !torch.int -> !torch.int
%int2_32479 = torch.constant.int 2
%34928 = torch.aten.mul.int %34927, %int2_32479 : !torch.int, !torch.int -> !torch.int
%int16_32480 = torch.constant.int 16
%int1_32481 = torch.constant.int 1
%int128_32482 = torch.constant.int 128
%34929 = torch.prim.ListConstruct %34928, %int16_32480, %int1_32481, %int128_32482 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34930 = torch.aten.view %34926, %34929 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34930, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34931 = torch.prim.ListConstruct %34834 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32483 = torch.constant.bool false
%34932 = torch.aten.index_put %34930, %34931, %34850, %false_32483 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34932, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32484 = torch.constant.int 32
%int2_32485 = torch.constant.int 2
%int16_32486 = torch.constant.int 16
%int1_32487 = torch.constant.int 1
%int128_32488 = torch.constant.int 128
%34933 = torch.prim.ListConstruct %3041, %int32_32484, %int2_32485, %int16_32486, %int1_32487, %int128_32488 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34934 = torch.aten.view %34932, %34933 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34934, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32489 = torch.constant.int 131072
%34935 = torch.prim.ListConstruct %3041, %int131072_32489 : (!torch.int, !torch.int) -> !torch.list<int>
%34936 = torch.aten.view %34934, %34935 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34936, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_32490 = torch.constant.int 32
%int2_32491 = torch.constant.int 2
%int16_32492 = torch.constant.int 16
%int1_32493 = torch.constant.int 1
%int128_32494 = torch.constant.int 128
%34937 = torch.prim.ListConstruct %3044, %int32_32490, %int2_32491, %int16_32492, %int1_32493, %int128_32494 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34938 = torch.aten.view %33087, %34937 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34938, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_32495 = torch.constant.int 32
%34939 = torch.aten.mul.int %3044, %int32_32495 : !torch.int, !torch.int -> !torch.int
%int2_32496 = torch.constant.int 2
%34940 = torch.aten.mul.int %34939, %int2_32496 : !torch.int, !torch.int -> !torch.int
%int16_32497 = torch.constant.int 16
%int1_32498 = torch.constant.int 1
%int128_32499 = torch.constant.int 128
%34941 = torch.prim.ListConstruct %34940, %int16_32497, %int1_32498, %int128_32499 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34942 = torch.aten.view %34938, %34941 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34942, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%34943 = torch.prim.ListConstruct %34836 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_32500 = torch.constant.bool false
%34944 = torch.aten.index_put %34942, %34943, %34852, %false_32500 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %34944, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_32501 = torch.constant.int 32
%int2_32502 = torch.constant.int 2
%int16_32503 = torch.constant.int 16
%int1_32504 = torch.constant.int 1
%int128_32505 = torch.constant.int 128
%34945 = torch.prim.ListConstruct %3044, %int32_32501, %int2_32502, %int16_32503, %int1_32504, %int128_32505 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34946 = torch.aten.view %34944, %34945 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %34946, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_32506 = torch.constant.int 131072
%34947 = torch.prim.ListConstruct %3044, %int131072_32506 : (!torch.int, !torch.int) -> !torch.list<int>
%34948 = torch.aten.view %34946, %34947 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %34948, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_32507 = torch.constant.int -2
%34949 = torch.aten.unsqueeze %34563, %int-2_32507 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32508 = torch.constant.int -2
%34950 = torch.aten.unsqueeze %34578, %int-2_32508 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32509 = torch.constant.int -2
%34951 = torch.aten.unsqueeze %34593, %int-2_32509 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32510 = torch.constant.int -2
%34952 = torch.aten.unsqueeze %34608, %int-2_32510 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32511 = torch.constant.int -2
%34953 = torch.aten.unsqueeze %34623, %int-2_32511 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32512 = torch.constant.int -2
%34954 = torch.aten.unsqueeze %34638, %int-2_32512 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32513 = torch.constant.int -2
%34955 = torch.aten.unsqueeze %34653, %int-2_32513 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32514 = torch.constant.int -2
%34956 = torch.aten.unsqueeze %34668, %int-2_32514 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_32515 = torch.constant.int 4
%int1_32516 = torch.constant.int 1
%int4_32517 = torch.constant.int 4
%int128_32518 = torch.constant.int 128
%34957 = torch.prim.ListConstruct %int4_32515, %34549, %int1_32516, %int4_32517, %int128_32518 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32519 = torch.constant.bool false
%34958 = torch.aten.expand %34949, %34957, %false_32519 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32520 = torch.constant.int 4
%int1_32521 = torch.constant.int 1
%int4_32522 = torch.constant.int 4
%int128_32523 = torch.constant.int 128
%34959 = torch.prim.ListConstruct %int4_32520, %34549, %int1_32521, %int4_32522, %int128_32523 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32524 = torch.constant.bool false
%34960 = torch.aten.expand %34950, %34959, %false_32524 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32525 = torch.constant.int 4
%int1_32526 = torch.constant.int 1
%int4_32527 = torch.constant.int 4
%int128_32528 = torch.constant.int 128
%34961 = torch.prim.ListConstruct %int4_32525, %34549, %int1_32526, %int4_32527, %int128_32528 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32529 = torch.constant.bool false
%34962 = torch.aten.expand %34951, %34961, %false_32529 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32530 = torch.constant.int 4
%int1_32531 = torch.constant.int 1
%int4_32532 = torch.constant.int 4
%int128_32533 = torch.constant.int 128
%34963 = torch.prim.ListConstruct %int4_32530, %34549, %int1_32531, %int4_32532, %int128_32533 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32534 = torch.constant.bool false
%34964 = torch.aten.expand %34952, %34963, %false_32534 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32535 = torch.constant.int 4
%int1_32536 = torch.constant.int 1
%int4_32537 = torch.constant.int 4
%int128_32538 = torch.constant.int 128
%34965 = torch.prim.ListConstruct %int4_32535, %34549, %int1_32536, %int4_32537, %int128_32538 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32539 = torch.constant.bool false
%34966 = torch.aten.expand %34953, %34965, %false_32539 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32540 = torch.constant.int 4
%int1_32541 = torch.constant.int 1
%int4_32542 = torch.constant.int 4
%int128_32543 = torch.constant.int 128
%34967 = torch.prim.ListConstruct %int4_32540, %34549, %int1_32541, %int4_32542, %int128_32543 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32544 = torch.constant.bool false
%34968 = torch.aten.expand %34954, %34967, %false_32544 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32545 = torch.constant.int 4
%int1_32546 = torch.constant.int 1
%int4_32547 = torch.constant.int 4
%int128_32548 = torch.constant.int 128
%34969 = torch.prim.ListConstruct %int4_32545, %34549, %int1_32546, %int4_32547, %int128_32548 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32549 = torch.constant.bool false
%34970 = torch.aten.expand %34955, %34969, %false_32549 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32550 = torch.constant.int 4
%int1_32551 = torch.constant.int 1
%int4_32552 = torch.constant.int 4
%int128_32553 = torch.constant.int 128
%34971 = torch.prim.ListConstruct %int4_32550, %34549, %int1_32551, %int4_32552, %int128_32553 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32554 = torch.constant.bool false
%34972 = torch.aten.expand %34956, %34971, %false_32554 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32555 = torch.constant.int 4
%int4_32556 = torch.constant.int 4
%int128_32557 = torch.constant.int 128
%34973 = torch.prim.ListConstruct %int4_32555, %34549, %int4_32556, %int128_32557 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34974 = torch.aten.view %34958, %34973 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32558 = torch.constant.int 4
%int4_32559 = torch.constant.int 4
%int128_32560 = torch.constant.int 128
%34975 = torch.prim.ListConstruct %int4_32558, %34549, %int4_32559, %int128_32560 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34976 = torch.aten.view %34960, %34975 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32561 = torch.constant.int 4
%int4_32562 = torch.constant.int 4
%int128_32563 = torch.constant.int 128
%34977 = torch.prim.ListConstruct %int4_32561, %34549, %int4_32562, %int128_32563 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34978 = torch.aten.view %34962, %34977 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32564 = torch.constant.int 4
%int4_32565 = torch.constant.int 4
%int128_32566 = torch.constant.int 128
%34979 = torch.prim.ListConstruct %int4_32564, %34549, %int4_32565, %int128_32566 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34980 = torch.aten.view %34964, %34979 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32567 = torch.constant.int 4
%int4_32568 = torch.constant.int 4
%int128_32569 = torch.constant.int 128
%34981 = torch.prim.ListConstruct %int4_32567, %34549, %int4_32568, %int128_32569 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34982 = torch.aten.view %34966, %34981 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32570 = torch.constant.int 4
%int4_32571 = torch.constant.int 4
%int128_32572 = torch.constant.int 128
%34983 = torch.prim.ListConstruct %int4_32570, %34549, %int4_32571, %int128_32572 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34984 = torch.aten.view %34968, %34983 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32573 = torch.constant.int 4
%int4_32574 = torch.constant.int 4
%int128_32575 = torch.constant.int 128
%34985 = torch.prim.ListConstruct %int4_32573, %34549, %int4_32574, %int128_32575 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34986 = torch.aten.view %34970, %34985 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32576 = torch.constant.int 4
%int4_32577 = torch.constant.int 4
%int128_32578 = torch.constant.int 128
%34987 = torch.prim.ListConstruct %int4_32576, %34549, %int4_32577, %int128_32578 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%34988 = torch.aten.view %34972, %34987 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %34988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_32579 = torch.constant.int -2
%34989 = torch.aten.unsqueeze %34338, %int-2_32579 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32580 = torch.constant.int -2
%34990 = torch.aten.unsqueeze %34340, %int-2_32580 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32581 = torch.constant.int -2
%34991 = torch.aten.unsqueeze %34342, %int-2_32581 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34991, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32582 = torch.constant.int -2
%34992 = torch.aten.unsqueeze %34344, %int-2_32582 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32583 = torch.constant.int -2
%34993 = torch.aten.unsqueeze %34346, %int-2_32583 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34993, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32584 = torch.constant.int -2
%34994 = torch.aten.unsqueeze %34348, %int-2_32584 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34994, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32585 = torch.constant.int -2
%34995 = torch.aten.unsqueeze %34350, %int-2_32585 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_32586 = torch.constant.int -2
%34996 = torch.aten.unsqueeze %34352, %int-2_32586 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %34996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_32587 = torch.constant.int 1
%34997 = torch.aten.size.int %34262, %int1_32587 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_32588 = torch.constant.int 4
%int1_32589 = torch.constant.int 1
%int4_32590 = torch.constant.int 4
%int128_32591 = torch.constant.int 128
%34998 = torch.prim.ListConstruct %int4_32588, %34997, %int1_32589, %int4_32590, %int128_32591 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32592 = torch.constant.bool false
%34999 = torch.aten.expand %34989, %34998, %false_32592 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %34999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32593 = torch.constant.int 4
%int1_32594 = torch.constant.int 1
%int4_32595 = torch.constant.int 4
%int128_32596 = torch.constant.int 128
%35000 = torch.prim.ListConstruct %int4_32593, %34997, %int1_32594, %int4_32595, %int128_32596 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32597 = torch.constant.bool false
%35001 = torch.aten.expand %34990, %35000, %false_32597 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32598 = torch.constant.int 4
%int1_32599 = torch.constant.int 1
%int4_32600 = torch.constant.int 4
%int128_32601 = torch.constant.int 128
%35002 = torch.prim.ListConstruct %int4_32598, %34997, %int1_32599, %int4_32600, %int128_32601 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32602 = torch.constant.bool false
%35003 = torch.aten.expand %34991, %35002, %false_32602 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32603 = torch.constant.int 4
%int1_32604 = torch.constant.int 1
%int4_32605 = torch.constant.int 4
%int128_32606 = torch.constant.int 128
%35004 = torch.prim.ListConstruct %int4_32603, %34997, %int1_32604, %int4_32605, %int128_32606 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32607 = torch.constant.bool false
%35005 = torch.aten.expand %34992, %35004, %false_32607 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32608 = torch.constant.int 4
%int1_32609 = torch.constant.int 1
%int4_32610 = torch.constant.int 4
%int128_32611 = torch.constant.int 128
%35006 = torch.prim.ListConstruct %int4_32608, %34997, %int1_32609, %int4_32610, %int128_32611 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32612 = torch.constant.bool false
%35007 = torch.aten.expand %34993, %35006, %false_32612 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32613 = torch.constant.int 4
%int1_32614 = torch.constant.int 1
%int4_32615 = torch.constant.int 4
%int128_32616 = torch.constant.int 128
%35008 = torch.prim.ListConstruct %int4_32613, %34997, %int1_32614, %int4_32615, %int128_32616 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32617 = torch.constant.bool false
%35009 = torch.aten.expand %34994, %35008, %false_32617 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32618 = torch.constant.int 4
%int1_32619 = torch.constant.int 1
%int4_32620 = torch.constant.int 4
%int128_32621 = torch.constant.int 128
%35010 = torch.prim.ListConstruct %int4_32618, %34997, %int1_32619, %int4_32620, %int128_32621 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32622 = torch.constant.bool false
%35011 = torch.aten.expand %34995, %35010, %false_32622 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32623 = torch.constant.int 4
%int1_32624 = torch.constant.int 1
%int4_32625 = torch.constant.int 4
%int128_32626 = torch.constant.int 128
%35012 = torch.prim.ListConstruct %int4_32623, %34997, %int1_32624, %int4_32625, %int128_32626 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_32627 = torch.constant.bool false
%35013 = torch.aten.expand %34996, %35012, %false_32627 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %35013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_32628 = torch.constant.int 4
%int4_32629 = torch.constant.int 4
%int128_32630 = torch.constant.int 128
%35014 = torch.prim.ListConstruct %int4_32628, %34997, %int4_32629, %int128_32630 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35015 = torch.aten.view %34999, %35014 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35015, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32631 = torch.constant.int 4
%int4_32632 = torch.constant.int 4
%int128_32633 = torch.constant.int 128
%35016 = torch.prim.ListConstruct %int4_32631, %34997, %int4_32632, %int128_32633 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35017 = torch.aten.view %35001, %35016 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32634 = torch.constant.int 4
%int4_32635 = torch.constant.int 4
%int128_32636 = torch.constant.int 128
%35018 = torch.prim.ListConstruct %int4_32634, %34997, %int4_32635, %int128_32636 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35019 = torch.aten.view %35003, %35018 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32637 = torch.constant.int 4
%int4_32638 = torch.constant.int 4
%int128_32639 = torch.constant.int 128
%35020 = torch.prim.ListConstruct %int4_32637, %34997, %int4_32638, %int128_32639 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35021 = torch.aten.view %35005, %35020 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35021, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32640 = torch.constant.int 4
%int4_32641 = torch.constant.int 4
%int128_32642 = torch.constant.int 128
%35022 = torch.prim.ListConstruct %int4_32640, %34997, %int4_32641, %int128_32642 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35023 = torch.aten.view %35007, %35022 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32643 = torch.constant.int 4
%int4_32644 = torch.constant.int 4
%int128_32645 = torch.constant.int 128
%35024 = torch.prim.ListConstruct %int4_32643, %34997, %int4_32644, %int128_32645 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35025 = torch.aten.view %35009, %35024 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32646 = torch.constant.int 4
%int4_32647 = torch.constant.int 4
%int128_32648 = torch.constant.int 128
%35026 = torch.prim.ListConstruct %int4_32646, %34997, %int4_32647, %int128_32648 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35027 = torch.aten.view %35011, %35026 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35027, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32649 = torch.constant.int 4
%int4_32650 = torch.constant.int 4
%int128_32651 = torch.constant.int 128
%35028 = torch.prim.ListConstruct %int4_32649, %34997, %int4_32650, %int128_32651 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35029 = torch.aten.view %35013, %35028 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32652 = torch.constant.int 1
%int2_32653 = torch.constant.int 2
%35030 = torch.aten.transpose.int %34405, %int1_32652, %int2_32653 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35030, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32654 = torch.constant.int 1
%int2_32655 = torch.constant.int 2
%35031 = torch.aten.transpose.int %34420, %int1_32654, %int2_32655 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35031, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32656 = torch.constant.int 1
%int2_32657 = torch.constant.int 2
%35032 = torch.aten.transpose.int %34435, %int1_32656, %int2_32657 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35032, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32658 = torch.constant.int 1
%int2_32659 = torch.constant.int 2
%35033 = torch.aten.transpose.int %34450, %int1_32658, %int2_32659 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35033, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32660 = torch.constant.int 1
%int2_32661 = torch.constant.int 2
%35034 = torch.aten.transpose.int %34465, %int1_32660, %int2_32661 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35034, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32662 = torch.constant.int 1
%int2_32663 = torch.constant.int 2
%35035 = torch.aten.transpose.int %34480, %int1_32662, %int2_32663 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35035, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32664 = torch.constant.int 1
%int2_32665 = torch.constant.int 2
%35036 = torch.aten.transpose.int %34495, %int1_32664, %int2_32665 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35036, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32666 = torch.constant.int 1
%int2_32667 = torch.constant.int 2
%35037 = torch.aten.transpose.int %34510, %int1_32666, %int2_32667 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35037, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32668 = torch.constant.int 1
%int2_32669 = torch.constant.int 2
%35038 = torch.aten.transpose.int %34974, %int1_32668, %int2_32669 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35038, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32670 = torch.constant.int 1
%int2_32671 = torch.constant.int 2
%35039 = torch.aten.transpose.int %34976, %int1_32670, %int2_32671 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35039, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32672 = torch.constant.int 1
%int2_32673 = torch.constant.int 2
%35040 = torch.aten.transpose.int %34978, %int1_32672, %int2_32673 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35040, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32674 = torch.constant.int 1
%int2_32675 = torch.constant.int 2
%35041 = torch.aten.transpose.int %34980, %int1_32674, %int2_32675 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35041, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32676 = torch.constant.int 1
%int2_32677 = torch.constant.int 2
%35042 = torch.aten.transpose.int %34982, %int1_32676, %int2_32677 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35042, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32678 = torch.constant.int 1
%int2_32679 = torch.constant.int 2
%35043 = torch.aten.transpose.int %34984, %int1_32678, %int2_32679 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35043, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32680 = torch.constant.int 1
%int2_32681 = torch.constant.int 2
%35044 = torch.aten.transpose.int %34986, %int1_32680, %int2_32681 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35044, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32682 = torch.constant.int 1
%int2_32683 = torch.constant.int 2
%35045 = torch.aten.transpose.int %34988, %int1_32682, %int2_32683 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35045, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32684 = torch.constant.int 1
%int2_32685 = torch.constant.int 2
%35046 = torch.aten.transpose.int %35015, %int1_32684, %int2_32685 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35046, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32686 = torch.constant.int 1
%int2_32687 = torch.constant.int 2
%35047 = torch.aten.transpose.int %35017, %int1_32686, %int2_32687 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35047, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32688 = torch.constant.int 1
%int2_32689 = torch.constant.int 2
%35048 = torch.aten.transpose.int %35019, %int1_32688, %int2_32689 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35048, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32690 = torch.constant.int 1
%int2_32691 = torch.constant.int 2
%35049 = torch.aten.transpose.int %35021, %int1_32690, %int2_32691 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35049, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32692 = torch.constant.int 1
%int2_32693 = torch.constant.int 2
%35050 = torch.aten.transpose.int %35023, %int1_32692, %int2_32693 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35050, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32694 = torch.constant.int 1
%int2_32695 = torch.constant.int 2
%35051 = torch.aten.transpose.int %35025, %int1_32694, %int2_32695 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35051, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32696 = torch.constant.int 1
%int2_32697 = torch.constant.int 2
%35052 = torch.aten.transpose.int %35027, %int1_32696, %int2_32697 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35052, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32698 = torch.constant.int 1
%int2_32699 = torch.constant.int 2
%35053 = torch.aten.transpose.int %35029, %int1_32698, %int2_32699 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %35053, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32700 = torch.constant.float 0.000000e+00
%true_32701 = torch.constant.bool true
%none_32702 = torch.constant.none
%none_32703 = torch.constant.none
%35054:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35030, %35038, %35046, %float0.000000e00_32700, %true_32701, %none_32702, %none_32703) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35054#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32704 = torch.constant.float 0.000000e+00
%true_32705 = torch.constant.bool true
%none_32706 = torch.constant.none
%none_32707 = torch.constant.none
%35055:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35031, %35039, %35047, %float0.000000e00_32704, %true_32705, %none_32706, %none_32707) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35055#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32708 = torch.constant.float 0.000000e+00
%true_32709 = torch.constant.bool true
%none_32710 = torch.constant.none
%none_32711 = torch.constant.none
%35056:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35032, %35040, %35048, %float0.000000e00_32708, %true_32709, %none_32710, %none_32711) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35056#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32712 = torch.constant.float 0.000000e+00
%true_32713 = torch.constant.bool true
%none_32714 = torch.constant.none
%none_32715 = torch.constant.none
%35057:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35033, %35041, %35049, %float0.000000e00_32712, %true_32713, %none_32714, %none_32715) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35057#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32716 = torch.constant.float 0.000000e+00
%true_32717 = torch.constant.bool true
%none_32718 = torch.constant.none
%none_32719 = torch.constant.none
%35058:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35034, %35042, %35050, %float0.000000e00_32716, %true_32717, %none_32718, %none_32719) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35058#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32720 = torch.constant.float 0.000000e+00
%true_32721 = torch.constant.bool true
%none_32722 = torch.constant.none
%none_32723 = torch.constant.none
%35059:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35035, %35043, %35051, %float0.000000e00_32720, %true_32721, %none_32722, %none_32723) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35059#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32724 = torch.constant.float 0.000000e+00
%true_32725 = torch.constant.bool true
%none_32726 = torch.constant.none
%none_32727 = torch.constant.none
%35060:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35036, %35044, %35052, %float0.000000e00_32724, %true_32725, %none_32726, %none_32727) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35060#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_32728 = torch.constant.float 0.000000e+00
%true_32729 = torch.constant.bool true
%none_32730 = torch.constant.none
%none_32731 = torch.constant.none
%35061:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%35037, %35045, %35053, %float0.000000e00_32728, %true_32729, %none_32730, %none_32731) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %35061#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_32732 = torch.constant.int 1
%int2_32733 = torch.constant.int 2
%35062 = torch.aten.transpose.int %35054#0, %int1_32732, %int2_32733 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32734 = torch.constant.int 1
%int2_32735 = torch.constant.int 2
%35063 = torch.aten.transpose.int %35055#0, %int1_32734, %int2_32735 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32736 = torch.constant.int 1
%int2_32737 = torch.constant.int 2
%35064 = torch.aten.transpose.int %35056#0, %int1_32736, %int2_32737 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32738 = torch.constant.int 1
%int2_32739 = torch.constant.int 2
%35065 = torch.aten.transpose.int %35057#0, %int1_32738, %int2_32739 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32740 = torch.constant.int 1
%int2_32741 = torch.constant.int 2
%35066 = torch.aten.transpose.int %35058#0, %int1_32740, %int2_32741 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32742 = torch.constant.int 1
%int2_32743 = torch.constant.int 2
%35067 = torch.aten.transpose.int %35059#0, %int1_32742, %int2_32743 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32744 = torch.constant.int 1
%int2_32745 = torch.constant.int 2
%35068 = torch.aten.transpose.int %35060#0, %int1_32744, %int2_32745 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_32746 = torch.constant.int 1
%int2_32747 = torch.constant.int 2
%35069 = torch.aten.transpose.int %35061#0, %int1_32746, %int2_32747 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %35069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_32748 = torch.constant.int 4
%int512_32749 = torch.constant.int 512
%35070 = torch.prim.ListConstruct %int4_32748, %34391, %int512_32749 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35071 = torch.aten.view %35062, %35070 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32750 = torch.constant.int 4
%int512_32751 = torch.constant.int 512
%35072 = torch.prim.ListConstruct %int4_32750, %34406, %int512_32751 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35073 = torch.aten.view %35063, %35072 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32752 = torch.constant.int 4
%int512_32753 = torch.constant.int 512
%35074 = torch.prim.ListConstruct %int4_32752, %34421, %int512_32753 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35075 = torch.aten.view %35064, %35074 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32754 = torch.constant.int 4
%int512_32755 = torch.constant.int 512
%35076 = torch.prim.ListConstruct %int4_32754, %34436, %int512_32755 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35077 = torch.aten.view %35065, %35076 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32756 = torch.constant.int 4
%int512_32757 = torch.constant.int 512
%35078 = torch.prim.ListConstruct %int4_32756, %34451, %int512_32757 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35079 = torch.aten.view %35066, %35078 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32758 = torch.constant.int 4
%int512_32759 = torch.constant.int 512
%35080 = torch.prim.ListConstruct %int4_32758, %34466, %int512_32759 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35081 = torch.aten.view %35067, %35080 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32760 = torch.constant.int 4
%int512_32761 = torch.constant.int 512
%35082 = torch.prim.ListConstruct %int4_32760, %34481, %int512_32761 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35083 = torch.aten.view %35068, %35082 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_32762 = torch.constant.int 4
%int512_32763 = torch.constant.int 512
%35084 = torch.prim.ListConstruct %int4_32762, %34496, %int512_32763 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35085 = torch.aten.view %35069, %35084 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_32764 = torch.constant.int 1
%int0_32765 = torch.constant.int 0
%35086 = torch.prim.ListConstruct %int1_32764, %int0_32765 : (!torch.int, !torch.int) -> !torch.list<int>
%35087 = torch.aten.permute %1264, %35086 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32766 = torch.constant.int 1
%int0_32767 = torch.constant.int 0
%35088 = torch.prim.ListConstruct %int1_32766, %int0_32767 : (!torch.int, !torch.int) -> !torch.list<int>
%35089 = torch.aten.permute %1265, %35088 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32768 = torch.constant.int 1
%int0_32769 = torch.constant.int 0
%35090 = torch.prim.ListConstruct %int1_32768, %int0_32769 : (!torch.int, !torch.int) -> !torch.list<int>
%35091 = torch.aten.permute %1266, %35090 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32770 = torch.constant.int 1
%int0_32771 = torch.constant.int 0
%35092 = torch.prim.ListConstruct %int1_32770, %int0_32771 : (!torch.int, !torch.int) -> !torch.list<int>
%35093 = torch.aten.permute %1267, %35092 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32772 = torch.constant.int 1
%int0_32773 = torch.constant.int 0
%35094 = torch.prim.ListConstruct %int1_32772, %int0_32773 : (!torch.int, !torch.int) -> !torch.list<int>
%35095 = torch.aten.permute %1268, %35094 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32774 = torch.constant.int 1
%int0_32775 = torch.constant.int 0
%35096 = torch.prim.ListConstruct %int1_32774, %int0_32775 : (!torch.int, !torch.int) -> !torch.list<int>
%35097 = torch.aten.permute %1269, %35096 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32776 = torch.constant.int 1
%int0_32777 = torch.constant.int 0
%35098 = torch.prim.ListConstruct %int1_32776, %int0_32777 : (!torch.int, !torch.int) -> !torch.list<int>
%35099 = torch.aten.permute %1270, %35098 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_32778 = torch.constant.int 1
%int0_32779 = torch.constant.int 0
%35100 = torch.prim.ListConstruct %int1_32778, %int0_32779 : (!torch.int, !torch.int) -> !torch.list<int>
%35101 = torch.aten.permute %1271, %35100 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_32780 = torch.constant.int 4
%35102 = torch.aten.mul.int %int4_32780, %34391 : !torch.int, !torch.int -> !torch.int
%int512_32781 = torch.constant.int 512
%35103 = torch.prim.ListConstruct %35102, %int512_32781 : (!torch.int, !torch.int) -> !torch.list<int>
%35104 = torch.aten.view %35071, %35103 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35104, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35105 = torch.aten.mm %35104, %35087 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35105, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32782 = torch.constant.int 4
%int4096_32783 = torch.constant.int 4096
%35106 = torch.prim.ListConstruct %int4_32782, %34391, %int4096_32783 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35107 = torch.aten.view %35105, %35106 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32784 = torch.constant.int 4
%35108 = torch.aten.mul.int %int4_32784, %34406 : !torch.int, !torch.int -> !torch.int
%int512_32785 = torch.constant.int 512
%35109 = torch.prim.ListConstruct %35108, %int512_32785 : (!torch.int, !torch.int) -> !torch.list<int>
%35110 = torch.aten.view %35073, %35109 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35110, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35111 = torch.aten.mm %35110, %35089 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35111, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32786 = torch.constant.int 4
%int4096_32787 = torch.constant.int 4096
%35112 = torch.prim.ListConstruct %int4_32786, %34406, %int4096_32787 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35113 = torch.aten.view %35111, %35112 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32788 = torch.constant.int 4
%35114 = torch.aten.mul.int %int4_32788, %34421 : !torch.int, !torch.int -> !torch.int
%int512_32789 = torch.constant.int 512
%35115 = torch.prim.ListConstruct %35114, %int512_32789 : (!torch.int, !torch.int) -> !torch.list<int>
%35116 = torch.aten.view %35075, %35115 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35116, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35117 = torch.aten.mm %35116, %35091 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35117, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32790 = torch.constant.int 4
%int4096_32791 = torch.constant.int 4096
%35118 = torch.prim.ListConstruct %int4_32790, %34421, %int4096_32791 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35119 = torch.aten.view %35117, %35118 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32792 = torch.constant.int 4
%35120 = torch.aten.mul.int %int4_32792, %34436 : !torch.int, !torch.int -> !torch.int
%int512_32793 = torch.constant.int 512
%35121 = torch.prim.ListConstruct %35120, %int512_32793 : (!torch.int, !torch.int) -> !torch.list<int>
%35122 = torch.aten.view %35077, %35121 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35122, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35123 = torch.aten.mm %35122, %35093 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35123, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32794 = torch.constant.int 4
%int4096_32795 = torch.constant.int 4096
%35124 = torch.prim.ListConstruct %int4_32794, %34436, %int4096_32795 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35125 = torch.aten.view %35123, %35124 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32796 = torch.constant.int 4
%35126 = torch.aten.mul.int %int4_32796, %34451 : !torch.int, !torch.int -> !torch.int
%int512_32797 = torch.constant.int 512
%35127 = torch.prim.ListConstruct %35126, %int512_32797 : (!torch.int, !torch.int) -> !torch.list<int>
%35128 = torch.aten.view %35079, %35127 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35128, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35129 = torch.aten.mm %35128, %35095 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35129, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32798 = torch.constant.int 4
%int4096_32799 = torch.constant.int 4096
%35130 = torch.prim.ListConstruct %int4_32798, %34451, %int4096_32799 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35131 = torch.aten.view %35129, %35130 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32800 = torch.constant.int 4
%35132 = torch.aten.mul.int %int4_32800, %34466 : !torch.int, !torch.int -> !torch.int
%int512_32801 = torch.constant.int 512
%35133 = torch.prim.ListConstruct %35132, %int512_32801 : (!torch.int, !torch.int) -> !torch.list<int>
%35134 = torch.aten.view %35081, %35133 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35134, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35135 = torch.aten.mm %35134, %35097 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35135, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32802 = torch.constant.int 4
%int4096_32803 = torch.constant.int 4096
%35136 = torch.prim.ListConstruct %int4_32802, %34466, %int4096_32803 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35137 = torch.aten.view %35135, %35136 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32804 = torch.constant.int 4
%35138 = torch.aten.mul.int %int4_32804, %34481 : !torch.int, !torch.int -> !torch.int
%int512_32805 = torch.constant.int 512
%35139 = torch.prim.ListConstruct %35138, %int512_32805 : (!torch.int, !torch.int) -> !torch.list<int>
%35140 = torch.aten.view %35083, %35139 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35140, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35141 = torch.aten.mm %35140, %35099 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35141, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32806 = torch.constant.int 4
%int4096_32807 = torch.constant.int 4096
%35142 = torch.prim.ListConstruct %int4_32806, %34481, %int4096_32807 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35143 = torch.aten.view %35141, %35142 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_32808 = torch.constant.int 4
%35144 = torch.aten.mul.int %int4_32808, %34496 : !torch.int, !torch.int -> !torch.int
%int512_32809 = torch.constant.int 512
%35145 = torch.prim.ListConstruct %35144, %int512_32809 : (!torch.int, !torch.int) -> !torch.list<int>
%35146 = torch.aten.view %35085, %35145 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35146, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%35147 = torch.aten.mm %35146, %35101 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35147, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_32810 = torch.constant.int 4
%int4096_32811 = torch.constant.int 4096
%35148 = torch.prim.ListConstruct %int4_32810, %34496, %int4096_32811 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35149 = torch.aten.view %35147, %35148 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35150 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32812 = arith.constant 1 : index
%dim_32813 = tensor.dim %35150, %c1_32812 : tensor<4x?x4096xf16>
%35151 = flow.tensor.transfer %35150 : tensor<4x?x4096xf16>{%dim_32813} to #hal.device.promise<@__device_0>
%35152 = torch_c.from_builtin_tensor %35151 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35153 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32814 = arith.constant 1 : index
%dim_32815 = tensor.dim %35153, %c1_32814 : tensor<4x?x4096xf16>
%35154 = flow.tensor.transfer %35153 : tensor<4x?x4096xf16>{%dim_32815} to #hal.device.promise<@__device_0>
%35155 = torch_c.from_builtin_tensor %35154 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35156 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32816 = arith.constant 1 : index
%dim_32817 = tensor.dim %35156, %c1_32816 : tensor<4x?x4096xf16>
%35157 = flow.tensor.transfer %35156 : tensor<4x?x4096xf16>{%dim_32817} to #hal.device.promise<@__device_0>
%35158 = torch_c.from_builtin_tensor %35157 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35159 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32818 = arith.constant 1 : index
%dim_32819 = tensor.dim %35159, %c1_32818 : tensor<4x?x4096xf16>
%35160 = flow.tensor.transfer %35159 : tensor<4x?x4096xf16>{%dim_32819} to #hal.device.promise<@__device_0>
%35161 = torch_c.from_builtin_tensor %35160 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35162 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32820 = arith.constant 1 : index
%dim_32821 = tensor.dim %35162, %c1_32820 : tensor<4x?x4096xf16>
%35163 = flow.tensor.transfer %35162 : tensor<4x?x4096xf16>{%dim_32821} to #hal.device.promise<@__device_0>
%35164 = torch_c.from_builtin_tensor %35163 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35165 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32822 = arith.constant 1 : index
%dim_32823 = tensor.dim %35165, %c1_32822 : tensor<4x?x4096xf16>
%35166 = flow.tensor.transfer %35165 : tensor<4x?x4096xf16>{%dim_32823} to #hal.device.promise<@__device_0>
%35167 = torch_c.from_builtin_tensor %35166 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35168 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32824 = arith.constant 1 : index
%dim_32825 = tensor.dim %35168, %c1_32824 : tensor<4x?x4096xf16>
%35169 = flow.tensor.transfer %35168 : tensor<4x?x4096xf16>{%dim_32825} to #hal.device.promise<@__device_0>
%35170 = torch_c.from_builtin_tensor %35169 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32826 = torch.constant.int 1
%35171 = torch.aten.add.Tensor %35107, %35152, %int1_32826 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32827 = torch.constant.int 1
%35172 = torch.aten.add.Tensor %35171, %35155, %int1_32827 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32828 = torch.constant.int 1
%35173 = torch.aten.add.Tensor %35172, %35158, %int1_32828 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32829 = torch.constant.int 1
%35174 = torch.aten.add.Tensor %35173, %35161, %int1_32829 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32830 = torch.constant.int 1
%35175 = torch.aten.add.Tensor %35174, %35164, %int1_32830 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32831 = torch.constant.int 1
%35176 = torch.aten.add.Tensor %35175, %35167, %int1_32831 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32832 = torch.constant.int 1
%35177 = torch.aten.add.Tensor %35176, %35170, %int1_32832 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35178 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32833 = arith.constant 1 : index
%dim_32834 = tensor.dim %35178, %c1_32833 : tensor<4x?x4096xf16>
%35179 = flow.tensor.transfer %35178 : tensor<4x?x4096xf16>{%dim_32834} to #hal.device.promise<@__device_1>
%35180 = torch_c.from_builtin_tensor %35179 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35180, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35181 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32835 = arith.constant 1 : index
%dim_32836 = tensor.dim %35181, %c1_32835 : tensor<4x?x4096xf16>
%35182 = flow.tensor.transfer %35181 : tensor<4x?x4096xf16>{%dim_32836} to #hal.device.promise<@__device_1>
%35183 = torch_c.from_builtin_tensor %35182 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35184 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32837 = arith.constant 1 : index
%dim_32838 = tensor.dim %35184, %c1_32837 : tensor<4x?x4096xf16>
%35185 = flow.tensor.transfer %35184 : tensor<4x?x4096xf16>{%dim_32838} to #hal.device.promise<@__device_1>
%35186 = torch_c.from_builtin_tensor %35185 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35187 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32839 = arith.constant 1 : index
%dim_32840 = tensor.dim %35187, %c1_32839 : tensor<4x?x4096xf16>
%35188 = flow.tensor.transfer %35187 : tensor<4x?x4096xf16>{%dim_32840} to #hal.device.promise<@__device_1>
%35189 = torch_c.from_builtin_tensor %35188 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35190 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32841 = arith.constant 1 : index
%dim_32842 = tensor.dim %35190, %c1_32841 : tensor<4x?x4096xf16>
%35191 = flow.tensor.transfer %35190 : tensor<4x?x4096xf16>{%dim_32842} to #hal.device.promise<@__device_1>
%35192 = torch_c.from_builtin_tensor %35191 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35192, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35193 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32843 = arith.constant 1 : index
%dim_32844 = tensor.dim %35193, %c1_32843 : tensor<4x?x4096xf16>
%35194 = flow.tensor.transfer %35193 : tensor<4x?x4096xf16>{%dim_32844} to #hal.device.promise<@__device_1>
%35195 = torch_c.from_builtin_tensor %35194 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35196 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32845 = arith.constant 1 : index
%dim_32846 = tensor.dim %35196, %c1_32845 : tensor<4x?x4096xf16>
%35197 = flow.tensor.transfer %35196 : tensor<4x?x4096xf16>{%dim_32846} to #hal.device.promise<@__device_1>
%35198 = torch_c.from_builtin_tensor %35197 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32847 = torch.constant.int 1
%35199 = torch.aten.add.Tensor %35180, %35113, %int1_32847 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32848 = torch.constant.int 1
%35200 = torch.aten.add.Tensor %35199, %35183, %int1_32848 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32849 = torch.constant.int 1
%35201 = torch.aten.add.Tensor %35200, %35186, %int1_32849 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32850 = torch.constant.int 1
%35202 = torch.aten.add.Tensor %35201, %35189, %int1_32850 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32851 = torch.constant.int 1
%35203 = torch.aten.add.Tensor %35202, %35192, %int1_32851 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32852 = torch.constant.int 1
%35204 = torch.aten.add.Tensor %35203, %35195, %int1_32852 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32853 = torch.constant.int 1
%35205 = torch.aten.add.Tensor %35204, %35198, %int1_32853 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35206 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32854 = arith.constant 1 : index
%dim_32855 = tensor.dim %35206, %c1_32854 : tensor<4x?x4096xf16>
%35207 = flow.tensor.transfer %35206 : tensor<4x?x4096xf16>{%dim_32855} to #hal.device.promise<@__device_2>
%35208 = torch_c.from_builtin_tensor %35207 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35208, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35209 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32856 = arith.constant 1 : index
%dim_32857 = tensor.dim %35209, %c1_32856 : tensor<4x?x4096xf16>
%35210 = flow.tensor.transfer %35209 : tensor<4x?x4096xf16>{%dim_32857} to #hal.device.promise<@__device_2>
%35211 = torch_c.from_builtin_tensor %35210 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35212 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32858 = arith.constant 1 : index
%dim_32859 = tensor.dim %35212, %c1_32858 : tensor<4x?x4096xf16>
%35213 = flow.tensor.transfer %35212 : tensor<4x?x4096xf16>{%dim_32859} to #hal.device.promise<@__device_2>
%35214 = torch_c.from_builtin_tensor %35213 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35214, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35215 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32860 = arith.constant 1 : index
%dim_32861 = tensor.dim %35215, %c1_32860 : tensor<4x?x4096xf16>
%35216 = flow.tensor.transfer %35215 : tensor<4x?x4096xf16>{%dim_32861} to #hal.device.promise<@__device_2>
%35217 = torch_c.from_builtin_tensor %35216 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35218 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32862 = arith.constant 1 : index
%dim_32863 = tensor.dim %35218, %c1_32862 : tensor<4x?x4096xf16>
%35219 = flow.tensor.transfer %35218 : tensor<4x?x4096xf16>{%dim_32863} to #hal.device.promise<@__device_2>
%35220 = torch_c.from_builtin_tensor %35219 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35220, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35221 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32864 = arith.constant 1 : index
%dim_32865 = tensor.dim %35221, %c1_32864 : tensor<4x?x4096xf16>
%35222 = flow.tensor.transfer %35221 : tensor<4x?x4096xf16>{%dim_32865} to #hal.device.promise<@__device_2>
%35223 = torch_c.from_builtin_tensor %35222 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35223, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35224 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32866 = arith.constant 1 : index
%dim_32867 = tensor.dim %35224, %c1_32866 : tensor<4x?x4096xf16>
%35225 = flow.tensor.transfer %35224 : tensor<4x?x4096xf16>{%dim_32867} to #hal.device.promise<@__device_2>
%35226 = torch_c.from_builtin_tensor %35225 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35226, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32868 = torch.constant.int 1
%35227 = torch.aten.add.Tensor %35208, %35211, %int1_32868 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32869 = torch.constant.int 1
%35228 = torch.aten.add.Tensor %35227, %35119, %int1_32869 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32870 = torch.constant.int 1
%35229 = torch.aten.add.Tensor %35228, %35214, %int1_32870 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32871 = torch.constant.int 1
%35230 = torch.aten.add.Tensor %35229, %35217, %int1_32871 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32872 = torch.constant.int 1
%35231 = torch.aten.add.Tensor %35230, %35220, %int1_32872 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32873 = torch.constant.int 1
%35232 = torch.aten.add.Tensor %35231, %35223, %int1_32873 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32874 = torch.constant.int 1
%35233 = torch.aten.add.Tensor %35232, %35226, %int1_32874 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35234 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32875 = arith.constant 1 : index
%dim_32876 = tensor.dim %35234, %c1_32875 : tensor<4x?x4096xf16>
%35235 = flow.tensor.transfer %35234 : tensor<4x?x4096xf16>{%dim_32876} to #hal.device.promise<@__device_3>
%35236 = torch_c.from_builtin_tensor %35235 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35236, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35237 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32877 = arith.constant 1 : index
%dim_32878 = tensor.dim %35237, %c1_32877 : tensor<4x?x4096xf16>
%35238 = flow.tensor.transfer %35237 : tensor<4x?x4096xf16>{%dim_32878} to #hal.device.promise<@__device_3>
%35239 = torch_c.from_builtin_tensor %35238 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35240 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32879 = arith.constant 1 : index
%dim_32880 = tensor.dim %35240, %c1_32879 : tensor<4x?x4096xf16>
%35241 = flow.tensor.transfer %35240 : tensor<4x?x4096xf16>{%dim_32880} to #hal.device.promise<@__device_3>
%35242 = torch_c.from_builtin_tensor %35241 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35243 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32881 = arith.constant 1 : index
%dim_32882 = tensor.dim %35243, %c1_32881 : tensor<4x?x4096xf16>
%35244 = flow.tensor.transfer %35243 : tensor<4x?x4096xf16>{%dim_32882} to #hal.device.promise<@__device_3>
%35245 = torch_c.from_builtin_tensor %35244 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35246 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32883 = arith.constant 1 : index
%dim_32884 = tensor.dim %35246, %c1_32883 : tensor<4x?x4096xf16>
%35247 = flow.tensor.transfer %35246 : tensor<4x?x4096xf16>{%dim_32884} to #hal.device.promise<@__device_3>
%35248 = torch_c.from_builtin_tensor %35247 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35249 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32885 = arith.constant 1 : index
%dim_32886 = tensor.dim %35249, %c1_32885 : tensor<4x?x4096xf16>
%35250 = flow.tensor.transfer %35249 : tensor<4x?x4096xf16>{%dim_32886} to #hal.device.promise<@__device_3>
%35251 = torch_c.from_builtin_tensor %35250 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35252 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32887 = arith.constant 1 : index
%dim_32888 = tensor.dim %35252, %c1_32887 : tensor<4x?x4096xf16>
%35253 = flow.tensor.transfer %35252 : tensor<4x?x4096xf16>{%dim_32888} to #hal.device.promise<@__device_3>
%35254 = torch_c.from_builtin_tensor %35253 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32889 = torch.constant.int 1
%35255 = torch.aten.add.Tensor %35236, %35239, %int1_32889 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32890 = torch.constant.int 1
%35256 = torch.aten.add.Tensor %35255, %35242, %int1_32890 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32891 = torch.constant.int 1
%35257 = torch.aten.add.Tensor %35256, %35125, %int1_32891 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32892 = torch.constant.int 1
%35258 = torch.aten.add.Tensor %35257, %35245, %int1_32892 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32893 = torch.constant.int 1
%35259 = torch.aten.add.Tensor %35258, %35248, %int1_32893 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35259, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32894 = torch.constant.int 1
%35260 = torch.aten.add.Tensor %35259, %35251, %int1_32894 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32895 = torch.constant.int 1
%35261 = torch.aten.add.Tensor %35260, %35254, %int1_32895 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35262 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32896 = arith.constant 1 : index
%dim_32897 = tensor.dim %35262, %c1_32896 : tensor<4x?x4096xf16>
%35263 = flow.tensor.transfer %35262 : tensor<4x?x4096xf16>{%dim_32897} to #hal.device.promise<@__device_4>
%35264 = torch_c.from_builtin_tensor %35263 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35265 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32898 = arith.constant 1 : index
%dim_32899 = tensor.dim %35265, %c1_32898 : tensor<4x?x4096xf16>
%35266 = flow.tensor.transfer %35265 : tensor<4x?x4096xf16>{%dim_32899} to #hal.device.promise<@__device_4>
%35267 = torch_c.from_builtin_tensor %35266 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35267, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35268 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32900 = arith.constant 1 : index
%dim_32901 = tensor.dim %35268, %c1_32900 : tensor<4x?x4096xf16>
%35269 = flow.tensor.transfer %35268 : tensor<4x?x4096xf16>{%dim_32901} to #hal.device.promise<@__device_4>
%35270 = torch_c.from_builtin_tensor %35269 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35271 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32902 = arith.constant 1 : index
%dim_32903 = tensor.dim %35271, %c1_32902 : tensor<4x?x4096xf16>
%35272 = flow.tensor.transfer %35271 : tensor<4x?x4096xf16>{%dim_32903} to #hal.device.promise<@__device_4>
%35273 = torch_c.from_builtin_tensor %35272 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35273, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35274 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32904 = arith.constant 1 : index
%dim_32905 = tensor.dim %35274, %c1_32904 : tensor<4x?x4096xf16>
%35275 = flow.tensor.transfer %35274 : tensor<4x?x4096xf16>{%dim_32905} to #hal.device.promise<@__device_4>
%35276 = torch_c.from_builtin_tensor %35275 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35277 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32906 = arith.constant 1 : index
%dim_32907 = tensor.dim %35277, %c1_32906 : tensor<4x?x4096xf16>
%35278 = flow.tensor.transfer %35277 : tensor<4x?x4096xf16>{%dim_32907} to #hal.device.promise<@__device_4>
%35279 = torch_c.from_builtin_tensor %35278 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35280 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32908 = arith.constant 1 : index
%dim_32909 = tensor.dim %35280, %c1_32908 : tensor<4x?x4096xf16>
%35281 = flow.tensor.transfer %35280 : tensor<4x?x4096xf16>{%dim_32909} to #hal.device.promise<@__device_4>
%35282 = torch_c.from_builtin_tensor %35281 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32910 = torch.constant.int 1
%35283 = torch.aten.add.Tensor %35264, %35267, %int1_32910 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32911 = torch.constant.int 1
%35284 = torch.aten.add.Tensor %35283, %35270, %int1_32911 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32912 = torch.constant.int 1
%35285 = torch.aten.add.Tensor %35284, %35273, %int1_32912 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32913 = torch.constant.int 1
%35286 = torch.aten.add.Tensor %35285, %35131, %int1_32913 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32914 = torch.constant.int 1
%35287 = torch.aten.add.Tensor %35286, %35276, %int1_32914 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32915 = torch.constant.int 1
%35288 = torch.aten.add.Tensor %35287, %35279, %int1_32915 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32916 = torch.constant.int 1
%35289 = torch.aten.add.Tensor %35288, %35282, %int1_32916 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35290 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32917 = arith.constant 1 : index
%dim_32918 = tensor.dim %35290, %c1_32917 : tensor<4x?x4096xf16>
%35291 = flow.tensor.transfer %35290 : tensor<4x?x4096xf16>{%dim_32918} to #hal.device.promise<@__device_5>
%35292 = torch_c.from_builtin_tensor %35291 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35293 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32919 = arith.constant 1 : index
%dim_32920 = tensor.dim %35293, %c1_32919 : tensor<4x?x4096xf16>
%35294 = flow.tensor.transfer %35293 : tensor<4x?x4096xf16>{%dim_32920} to #hal.device.promise<@__device_5>
%35295 = torch_c.from_builtin_tensor %35294 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35296 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32921 = arith.constant 1 : index
%dim_32922 = tensor.dim %35296, %c1_32921 : tensor<4x?x4096xf16>
%35297 = flow.tensor.transfer %35296 : tensor<4x?x4096xf16>{%dim_32922} to #hal.device.promise<@__device_5>
%35298 = torch_c.from_builtin_tensor %35297 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35299 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32923 = arith.constant 1 : index
%dim_32924 = tensor.dim %35299, %c1_32923 : tensor<4x?x4096xf16>
%35300 = flow.tensor.transfer %35299 : tensor<4x?x4096xf16>{%dim_32924} to #hal.device.promise<@__device_5>
%35301 = torch_c.from_builtin_tensor %35300 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35302 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32925 = arith.constant 1 : index
%dim_32926 = tensor.dim %35302, %c1_32925 : tensor<4x?x4096xf16>
%35303 = flow.tensor.transfer %35302 : tensor<4x?x4096xf16>{%dim_32926} to #hal.device.promise<@__device_5>
%35304 = torch_c.from_builtin_tensor %35303 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35305 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32927 = arith.constant 1 : index
%dim_32928 = tensor.dim %35305, %c1_32927 : tensor<4x?x4096xf16>
%35306 = flow.tensor.transfer %35305 : tensor<4x?x4096xf16>{%dim_32928} to #hal.device.promise<@__device_5>
%35307 = torch_c.from_builtin_tensor %35306 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35308 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32929 = arith.constant 1 : index
%dim_32930 = tensor.dim %35308, %c1_32929 : tensor<4x?x4096xf16>
%35309 = flow.tensor.transfer %35308 : tensor<4x?x4096xf16>{%dim_32930} to #hal.device.promise<@__device_5>
%35310 = torch_c.from_builtin_tensor %35309 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32931 = torch.constant.int 1
%35311 = torch.aten.add.Tensor %35292, %35295, %int1_32931 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32932 = torch.constant.int 1
%35312 = torch.aten.add.Tensor %35311, %35298, %int1_32932 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32933 = torch.constant.int 1
%35313 = torch.aten.add.Tensor %35312, %35301, %int1_32933 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32934 = torch.constant.int 1
%35314 = torch.aten.add.Tensor %35313, %35304, %int1_32934 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32935 = torch.constant.int 1
%35315 = torch.aten.add.Tensor %35314, %35137, %int1_32935 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32936 = torch.constant.int 1
%35316 = torch.aten.add.Tensor %35315, %35307, %int1_32936 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32937 = torch.constant.int 1
%35317 = torch.aten.add.Tensor %35316, %35310, %int1_32937 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35318 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32938 = arith.constant 1 : index
%dim_32939 = tensor.dim %35318, %c1_32938 : tensor<4x?x4096xf16>
%35319 = flow.tensor.transfer %35318 : tensor<4x?x4096xf16>{%dim_32939} to #hal.device.promise<@__device_6>
%35320 = torch_c.from_builtin_tensor %35319 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35320, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35321 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32940 = arith.constant 1 : index
%dim_32941 = tensor.dim %35321, %c1_32940 : tensor<4x?x4096xf16>
%35322 = flow.tensor.transfer %35321 : tensor<4x?x4096xf16>{%dim_32941} to #hal.device.promise<@__device_6>
%35323 = torch_c.from_builtin_tensor %35322 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35323, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35324 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32942 = arith.constant 1 : index
%dim_32943 = tensor.dim %35324, %c1_32942 : tensor<4x?x4096xf16>
%35325 = flow.tensor.transfer %35324 : tensor<4x?x4096xf16>{%dim_32943} to #hal.device.promise<@__device_6>
%35326 = torch_c.from_builtin_tensor %35325 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35327 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32944 = arith.constant 1 : index
%dim_32945 = tensor.dim %35327, %c1_32944 : tensor<4x?x4096xf16>
%35328 = flow.tensor.transfer %35327 : tensor<4x?x4096xf16>{%dim_32945} to #hal.device.promise<@__device_6>
%35329 = torch_c.from_builtin_tensor %35328 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35330 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32946 = arith.constant 1 : index
%dim_32947 = tensor.dim %35330, %c1_32946 : tensor<4x?x4096xf16>
%35331 = flow.tensor.transfer %35330 : tensor<4x?x4096xf16>{%dim_32947} to #hal.device.promise<@__device_6>
%35332 = torch_c.from_builtin_tensor %35331 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35332, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35333 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32948 = arith.constant 1 : index
%dim_32949 = tensor.dim %35333, %c1_32948 : tensor<4x?x4096xf16>
%35334 = flow.tensor.transfer %35333 : tensor<4x?x4096xf16>{%dim_32949} to #hal.device.promise<@__device_6>
%35335 = torch_c.from_builtin_tensor %35334 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35335, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35336 = torch_c.to_builtin_tensor %35149 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32950 = arith.constant 1 : index
%dim_32951 = tensor.dim %35336, %c1_32950 : tensor<4x?x4096xf16>
%35337 = flow.tensor.transfer %35336 : tensor<4x?x4096xf16>{%dim_32951} to #hal.device.promise<@__device_6>
%35338 = torch_c.from_builtin_tensor %35337 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35338, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32952 = torch.constant.int 1
%35339 = torch.aten.add.Tensor %35320, %35323, %int1_32952 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35339, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32953 = torch.constant.int 1
%35340 = torch.aten.add.Tensor %35339, %35326, %int1_32953 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32954 = torch.constant.int 1
%35341 = torch.aten.add.Tensor %35340, %35329, %int1_32954 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32955 = torch.constant.int 1
%35342 = torch.aten.add.Tensor %35341, %35332, %int1_32955 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32956 = torch.constant.int 1
%35343 = torch.aten.add.Tensor %35342, %35335, %int1_32956 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35343, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32957 = torch.constant.int 1
%35344 = torch.aten.add.Tensor %35343, %35143, %int1_32957 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32958 = torch.constant.int 1
%35345 = torch.aten.add.Tensor %35344, %35338, %int1_32958 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35346 = torch_c.to_builtin_tensor %35107 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32959 = arith.constant 1 : index
%dim_32960 = tensor.dim %35346, %c1_32959 : tensor<4x?x4096xf16>
%35347 = flow.tensor.transfer %35346 : tensor<4x?x4096xf16>{%dim_32960} to #hal.device.promise<@__device_7>
%35348 = torch_c.from_builtin_tensor %35347 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35349 = torch_c.to_builtin_tensor %35113 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32961 = arith.constant 1 : index
%dim_32962 = tensor.dim %35349, %c1_32961 : tensor<4x?x4096xf16>
%35350 = flow.tensor.transfer %35349 : tensor<4x?x4096xf16>{%dim_32962} to #hal.device.promise<@__device_7>
%35351 = torch_c.from_builtin_tensor %35350 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35352 = torch_c.to_builtin_tensor %35119 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32963 = arith.constant 1 : index
%dim_32964 = tensor.dim %35352, %c1_32963 : tensor<4x?x4096xf16>
%35353 = flow.tensor.transfer %35352 : tensor<4x?x4096xf16>{%dim_32964} to #hal.device.promise<@__device_7>
%35354 = torch_c.from_builtin_tensor %35353 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35355 = torch_c.to_builtin_tensor %35125 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32965 = arith.constant 1 : index
%dim_32966 = tensor.dim %35355, %c1_32965 : tensor<4x?x4096xf16>
%35356 = flow.tensor.transfer %35355 : tensor<4x?x4096xf16>{%dim_32966} to #hal.device.promise<@__device_7>
%35357 = torch_c.from_builtin_tensor %35356 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35357, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35358 = torch_c.to_builtin_tensor %35131 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32967 = arith.constant 1 : index
%dim_32968 = tensor.dim %35358, %c1_32967 : tensor<4x?x4096xf16>
%35359 = flow.tensor.transfer %35358 : tensor<4x?x4096xf16>{%dim_32968} to #hal.device.promise<@__device_7>
%35360 = torch_c.from_builtin_tensor %35359 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35361 = torch_c.to_builtin_tensor %35137 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32969 = arith.constant 1 : index
%dim_32970 = tensor.dim %35361, %c1_32969 : tensor<4x?x4096xf16>
%35362 = flow.tensor.transfer %35361 : tensor<4x?x4096xf16>{%dim_32970} to #hal.device.promise<@__device_7>
%35363 = torch_c.from_builtin_tensor %35362 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35364 = torch_c.to_builtin_tensor %35143 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_32971 = arith.constant 1 : index
%dim_32972 = tensor.dim %35364, %c1_32971 : tensor<4x?x4096xf16>
%35365 = flow.tensor.transfer %35364 : tensor<4x?x4096xf16>{%dim_32972} to #hal.device.promise<@__device_7>
%35366 = torch_c.from_builtin_tensor %35365 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32973 = torch.constant.int 1
%35367 = torch.aten.add.Tensor %35348, %35351, %int1_32973 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32974 = torch.constant.int 1
%35368 = torch.aten.add.Tensor %35367, %35354, %int1_32974 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35368, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32975 = torch.constant.int 1
%35369 = torch.aten.add.Tensor %35368, %35357, %int1_32975 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35369, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32976 = torch.constant.int 1
%35370 = torch.aten.add.Tensor %35369, %35360, %int1_32976 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32977 = torch.constant.int 1
%35371 = torch.aten.add.Tensor %35370, %35363, %int1_32977 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32978 = torch.constant.int 1
%35372 = torch.aten.add.Tensor %35371, %35366, %int1_32978 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32979 = torch.constant.int 1
%35373 = torch.aten.add.Tensor %35372, %35149, %int1_32979 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35373, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32980 = torch.constant.int 1
%35374 = torch.aten.add.Tensor %34033, %35177, %int1_32980 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32981 = torch.constant.int 1
%35375 = torch.aten.add.Tensor %34034, %35205, %int1_32981 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32982 = torch.constant.int 1
%35376 = torch.aten.add.Tensor %34035, %35233, %int1_32982 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35376, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32983 = torch.constant.int 1
%35377 = torch.aten.add.Tensor %34036, %35261, %int1_32983 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32984 = torch.constant.int 1
%35378 = torch.aten.add.Tensor %34037, %35289, %int1_32984 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32985 = torch.constant.int 1
%35379 = torch.aten.add.Tensor %34038, %35317, %int1_32985 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32986 = torch.constant.int 1
%35380 = torch.aten.add.Tensor %34039, %35345, %int1_32986 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_32987 = torch.constant.int 1
%35381 = torch.aten.add.Tensor %34040, %35373, %int1_32987 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_32988 = torch.constant.int 6
%35382 = torch.prims.convert_element_type %35374, %int6_32988 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32989 = torch.constant.int 6
%35383 = torch.prims.convert_element_type %35375, %int6_32989 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32990 = torch.constant.int 6
%35384 = torch.prims.convert_element_type %35376, %int6_32990 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32991 = torch.constant.int 6
%35385 = torch.prims.convert_element_type %35377, %int6_32991 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32992 = torch.constant.int 6
%35386 = torch.prims.convert_element_type %35378, %int6_32992 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32993 = torch.constant.int 6
%35387 = torch.prims.convert_element_type %35379, %int6_32993 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35387, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32994 = torch.constant.int 6
%35388 = torch.prims.convert_element_type %35380, %int6_32994 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35388, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_32995 = torch.constant.int 6
%35389 = torch.prims.convert_element_type %35381, %int6_32995 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_32996 = torch.constant.int 2
%35390 = torch.aten.pow.Tensor_Scalar %35382, %int2_32996 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_32997 = torch.constant.int 2
%35391 = torch.aten.pow.Tensor_Scalar %35383, %int2_32997 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_32998 = torch.constant.int 2
%35392 = torch.aten.pow.Tensor_Scalar %35384, %int2_32998 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35392, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_32999 = torch.constant.int 2
%35393 = torch.aten.pow.Tensor_Scalar %35385, %int2_32999 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35393, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33000 = torch.constant.int 2
%35394 = torch.aten.pow.Tensor_Scalar %35386, %int2_33000 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33001 = torch.constant.int 2
%35395 = torch.aten.pow.Tensor_Scalar %35387, %int2_33001 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35395, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33002 = torch.constant.int 2
%35396 = torch.aten.pow.Tensor_Scalar %35388, %int2_33002 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35396, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33003 = torch.constant.int 2
%35397 = torch.aten.pow.Tensor_Scalar %35389, %int2_33003 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_33004 = torch.constant.int -1
%35398 = torch.prim.ListConstruct %int-1_33004 : (!torch.int) -> !torch.list<int>
%true_33005 = torch.constant.bool true
%none_33006 = torch.constant.none
%35399 = torch.aten.mean.dim %35390, %35398, %true_33005, %none_33006 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35399, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33007 = torch.constant.int -1
%35400 = torch.prim.ListConstruct %int-1_33007 : (!torch.int) -> !torch.list<int>
%true_33008 = torch.constant.bool true
%none_33009 = torch.constant.none
%35401 = torch.aten.mean.dim %35391, %35400, %true_33008, %none_33009 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35401, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33010 = torch.constant.int -1
%35402 = torch.prim.ListConstruct %int-1_33010 : (!torch.int) -> !torch.list<int>
%true_33011 = torch.constant.bool true
%none_33012 = torch.constant.none
%35403 = torch.aten.mean.dim %35392, %35402, %true_33011, %none_33012 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33013 = torch.constant.int -1
%35404 = torch.prim.ListConstruct %int-1_33013 : (!torch.int) -> !torch.list<int>
%true_33014 = torch.constant.bool true
%none_33015 = torch.constant.none
%35405 = torch.aten.mean.dim %35393, %35404, %true_33014, %none_33015 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35405, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33016 = torch.constant.int -1
%35406 = torch.prim.ListConstruct %int-1_33016 : (!torch.int) -> !torch.list<int>
%true_33017 = torch.constant.bool true
%none_33018 = torch.constant.none
%35407 = torch.aten.mean.dim %35394, %35406, %true_33017, %none_33018 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35407, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33019 = torch.constant.int -1
%35408 = torch.prim.ListConstruct %int-1_33019 : (!torch.int) -> !torch.list<int>
%true_33020 = torch.constant.bool true
%none_33021 = torch.constant.none
%35409 = torch.aten.mean.dim %35395, %35408, %true_33020, %none_33021 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33022 = torch.constant.int -1
%35410 = torch.prim.ListConstruct %int-1_33022 : (!torch.int) -> !torch.list<int>
%true_33023 = torch.constant.bool true
%none_33024 = torch.constant.none
%35411 = torch.aten.mean.dim %35396, %35410, %true_33023, %none_33024 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35411, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33025 = torch.constant.int -1
%35412 = torch.prim.ListConstruct %int-1_33025 : (!torch.int) -> !torch.list<int>
%true_33026 = torch.constant.bool true
%none_33027 = torch.constant.none
%35413 = torch.aten.mean.dim %35397, %35412, %true_33026, %none_33027 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33028 = torch.constant.float 9.9999997473787516E-6
%int1_33029 = torch.constant.int 1
%35414 = torch.aten.add.Scalar %35399, %float9.999990e-06_33028, %int1_33029 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33030 = torch.constant.float 9.9999997473787516E-6
%int1_33031 = torch.constant.int 1
%35415 = torch.aten.add.Scalar %35401, %float9.999990e-06_33030, %int1_33031 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33032 = torch.constant.float 9.9999997473787516E-6
%int1_33033 = torch.constant.int 1
%35416 = torch.aten.add.Scalar %35403, %float9.999990e-06_33032, %int1_33033 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33034 = torch.constant.float 9.9999997473787516E-6
%int1_33035 = torch.constant.int 1
%35417 = torch.aten.add.Scalar %35405, %float9.999990e-06_33034, %int1_33035 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33036 = torch.constant.float 9.9999997473787516E-6
%int1_33037 = torch.constant.int 1
%35418 = torch.aten.add.Scalar %35407, %float9.999990e-06_33036, %int1_33037 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33038 = torch.constant.float 9.9999997473787516E-6
%int1_33039 = torch.constant.int 1
%35419 = torch.aten.add.Scalar %35409, %float9.999990e-06_33038, %int1_33039 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33040 = torch.constant.float 9.9999997473787516E-6
%int1_33041 = torch.constant.int 1
%35420 = torch.aten.add.Scalar %35411, %float9.999990e-06_33040, %int1_33041 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33042 = torch.constant.float 9.9999997473787516E-6
%int1_33043 = torch.constant.int 1
%35421 = torch.aten.add.Scalar %35413, %float9.999990e-06_33042, %int1_33043 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35421, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35422 = torch.aten.rsqrt %35414 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35423 = torch.aten.rsqrt %35415 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35424 = torch.aten.rsqrt %35416 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35425 = torch.aten.rsqrt %35417 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35426 = torch.aten.rsqrt %35418 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35427 = torch.aten.rsqrt %35419 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35427, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35428 = torch.aten.rsqrt %35420 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35429 = torch.aten.rsqrt %35421 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35429, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35430 = torch.aten.mul.Tensor %35382, %35422 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35430, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35431 = torch.aten.mul.Tensor %35383, %35423 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35432 = torch.aten.mul.Tensor %35384, %35424 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35433 = torch.aten.mul.Tensor %35385, %35425 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35433, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35434 = torch.aten.mul.Tensor %35386, %35426 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35435 = torch.aten.mul.Tensor %35387, %35427 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35436 = torch.aten.mul.Tensor %35388, %35428 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35436, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35437 = torch.aten.mul.Tensor %35389, %35429 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35438 = torch.aten.mul.Tensor %1272, %35430 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35439 = torch.aten.mul.Tensor %1273, %35431 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35440 = torch.aten.mul.Tensor %1274, %35432 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35441 = torch.aten.mul.Tensor %1275, %35433 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35442 = torch.aten.mul.Tensor %1276, %35434 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35443 = torch.aten.mul.Tensor %1277, %35435 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35444 = torch.aten.mul.Tensor %1278, %35436 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35445 = torch.aten.mul.Tensor %1279, %35437 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_33044 = torch.constant.int 5
%35446 = torch.prims.convert_element_type %35438, %int5_33044 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33045 = torch.constant.int 5
%35447 = torch.prims.convert_element_type %35439, %int5_33045 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33046 = torch.constant.int 5
%35448 = torch.prims.convert_element_type %35440, %int5_33046 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35448, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33047 = torch.constant.int 5
%35449 = torch.prims.convert_element_type %35441, %int5_33047 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33048 = torch.constant.int 5
%35450 = torch.prims.convert_element_type %35442, %int5_33048 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33049 = torch.constant.int 5
%35451 = torch.prims.convert_element_type %35443, %int5_33049 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33050 = torch.constant.int 5
%35452 = torch.prims.convert_element_type %35444, %int5_33050 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33051 = torch.constant.int 5
%35453 = torch.prims.convert_element_type %35445, %int5_33051 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33052 = torch.constant.int 1
%int0_33053 = torch.constant.int 0
%35454 = torch.prim.ListConstruct %int1_33052, %int0_33053 : (!torch.int, !torch.int) -> !torch.list<int>
%35455 = torch.aten.permute %1280, %35454 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33054 = torch.constant.int 1
%int0_33055 = torch.constant.int 0
%35456 = torch.prim.ListConstruct %int1_33054, %int0_33055 : (!torch.int, !torch.int) -> !torch.list<int>
%35457 = torch.aten.permute %1281, %35456 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33056 = torch.constant.int 1
%int0_33057 = torch.constant.int 0
%35458 = torch.prim.ListConstruct %int1_33056, %int0_33057 : (!torch.int, !torch.int) -> !torch.list<int>
%35459 = torch.aten.permute %1282, %35458 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33058 = torch.constant.int 1
%int0_33059 = torch.constant.int 0
%35460 = torch.prim.ListConstruct %int1_33058, %int0_33059 : (!torch.int, !torch.int) -> !torch.list<int>
%35461 = torch.aten.permute %1283, %35460 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33060 = torch.constant.int 1
%int0_33061 = torch.constant.int 0
%35462 = torch.prim.ListConstruct %int1_33060, %int0_33061 : (!torch.int, !torch.int) -> !torch.list<int>
%35463 = torch.aten.permute %1284, %35462 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33062 = torch.constant.int 1
%int0_33063 = torch.constant.int 0
%35464 = torch.prim.ListConstruct %int1_33062, %int0_33063 : (!torch.int, !torch.int) -> !torch.list<int>
%35465 = torch.aten.permute %1285, %35464 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33064 = torch.constant.int 1
%int0_33065 = torch.constant.int 0
%35466 = torch.prim.ListConstruct %int1_33064, %int0_33065 : (!torch.int, !torch.int) -> !torch.list<int>
%35467 = torch.aten.permute %1286, %35466 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33066 = torch.constant.int 1
%int0_33067 = torch.constant.int 0
%35468 = torch.prim.ListConstruct %int1_33066, %int0_33067 : (!torch.int, !torch.int) -> !torch.list<int>
%35469 = torch.aten.permute %1287, %35468 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_33068 = torch.constant.int 4
%35470 = torch.aten.mul.int %int4_33068, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33069 = torch.constant.int 4096
%35471 = torch.prim.ListConstruct %35470, %int4096_33069 : (!torch.int, !torch.int) -> !torch.list<int>
%35472 = torch.aten.view %35446, %35471 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35472, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35473 = torch.aten.mm %35472, %35455 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35473, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33070 = torch.constant.int 4
%int1792_33071 = torch.constant.int 1792
%35474 = torch.prim.ListConstruct %int4_33070, %2482, %int1792_33071 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35475 = torch.aten.view %35473, %35474 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33072 = torch.constant.int 4
%35476 = torch.aten.mul.int %int4_33072, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33073 = torch.constant.int 4096
%35477 = torch.prim.ListConstruct %35476, %int4096_33073 : (!torch.int, !torch.int) -> !torch.list<int>
%35478 = torch.aten.view %35447, %35477 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35478, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35479 = torch.aten.mm %35478, %35457 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35479, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33074 = torch.constant.int 4
%int1792_33075 = torch.constant.int 1792
%35480 = torch.prim.ListConstruct %int4_33074, %2482, %int1792_33075 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35481 = torch.aten.view %35479, %35480 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33076 = torch.constant.int 4
%35482 = torch.aten.mul.int %int4_33076, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33077 = torch.constant.int 4096
%35483 = torch.prim.ListConstruct %35482, %int4096_33077 : (!torch.int, !torch.int) -> !torch.list<int>
%35484 = torch.aten.view %35448, %35483 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35484, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35485 = torch.aten.mm %35484, %35459 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35485, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33078 = torch.constant.int 4
%int1792_33079 = torch.constant.int 1792
%35486 = torch.prim.ListConstruct %int4_33078, %2482, %int1792_33079 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35487 = torch.aten.view %35485, %35486 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33080 = torch.constant.int 4
%35488 = torch.aten.mul.int %int4_33080, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33081 = torch.constant.int 4096
%35489 = torch.prim.ListConstruct %35488, %int4096_33081 : (!torch.int, !torch.int) -> !torch.list<int>
%35490 = torch.aten.view %35449, %35489 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35490, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35491 = torch.aten.mm %35490, %35461 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35491, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33082 = torch.constant.int 4
%int1792_33083 = torch.constant.int 1792
%35492 = torch.prim.ListConstruct %int4_33082, %2482, %int1792_33083 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35493 = torch.aten.view %35491, %35492 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33084 = torch.constant.int 4
%35494 = torch.aten.mul.int %int4_33084, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33085 = torch.constant.int 4096
%35495 = torch.prim.ListConstruct %35494, %int4096_33085 : (!torch.int, !torch.int) -> !torch.list<int>
%35496 = torch.aten.view %35450, %35495 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35496, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35497 = torch.aten.mm %35496, %35463 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35497, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33086 = torch.constant.int 4
%int1792_33087 = torch.constant.int 1792
%35498 = torch.prim.ListConstruct %int4_33086, %2482, %int1792_33087 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35499 = torch.aten.view %35497, %35498 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33088 = torch.constant.int 4
%35500 = torch.aten.mul.int %int4_33088, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33089 = torch.constant.int 4096
%35501 = torch.prim.ListConstruct %35500, %int4096_33089 : (!torch.int, !torch.int) -> !torch.list<int>
%35502 = torch.aten.view %35451, %35501 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35502, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35503 = torch.aten.mm %35502, %35465 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35503, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33090 = torch.constant.int 4
%int1792_33091 = torch.constant.int 1792
%35504 = torch.prim.ListConstruct %int4_33090, %2482, %int1792_33091 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35505 = torch.aten.view %35503, %35504 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35505, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33092 = torch.constant.int 4
%35506 = torch.aten.mul.int %int4_33092, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33093 = torch.constant.int 4096
%35507 = torch.prim.ListConstruct %35506, %int4096_33093 : (!torch.int, !torch.int) -> !torch.list<int>
%35508 = torch.aten.view %35452, %35507 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35508, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35509 = torch.aten.mm %35508, %35467 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35509, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33094 = torch.constant.int 4
%int1792_33095 = torch.constant.int 1792
%35510 = torch.prim.ListConstruct %int4_33094, %2482, %int1792_33095 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35511 = torch.aten.view %35509, %35510 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35511, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33096 = torch.constant.int 4
%35512 = torch.aten.mul.int %int4_33096, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33097 = torch.constant.int 4096
%35513 = torch.prim.ListConstruct %35512, %int4096_33097 : (!torch.int, !torch.int) -> !torch.list<int>
%35514 = torch.aten.view %35453, %35513 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35514, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35515 = torch.aten.mm %35514, %35469 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35515, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33098 = torch.constant.int 4
%int1792_33099 = torch.constant.int 1792
%35516 = torch.prim.ListConstruct %int4_33098, %2482, %int1792_33099 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35517 = torch.aten.view %35515, %35516 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35517, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35518 = torch.aten.silu %35475 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35519 = torch.aten.silu %35481 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35519, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35520 = torch.aten.silu %35487 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35520, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35521 = torch.aten.silu %35493 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35522 = torch.aten.silu %35499 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35522, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35523 = torch.aten.silu %35505 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35524 = torch.aten.silu %35511 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35525 = torch.aten.silu %35517 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_33100 = torch.constant.int 1
%int0_33101 = torch.constant.int 0
%35526 = torch.prim.ListConstruct %int1_33100, %int0_33101 : (!torch.int, !torch.int) -> !torch.list<int>
%35527 = torch.aten.permute %1288, %35526 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33102 = torch.constant.int 1
%int0_33103 = torch.constant.int 0
%35528 = torch.prim.ListConstruct %int1_33102, %int0_33103 : (!torch.int, !torch.int) -> !torch.list<int>
%35529 = torch.aten.permute %1289, %35528 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33104 = torch.constant.int 1
%int0_33105 = torch.constant.int 0
%35530 = torch.prim.ListConstruct %int1_33104, %int0_33105 : (!torch.int, !torch.int) -> !torch.list<int>
%35531 = torch.aten.permute %1290, %35530 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33106 = torch.constant.int 1
%int0_33107 = torch.constant.int 0
%35532 = torch.prim.ListConstruct %int1_33106, %int0_33107 : (!torch.int, !torch.int) -> !torch.list<int>
%35533 = torch.aten.permute %1291, %35532 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33108 = torch.constant.int 1
%int0_33109 = torch.constant.int 0
%35534 = torch.prim.ListConstruct %int1_33108, %int0_33109 : (!torch.int, !torch.int) -> !torch.list<int>
%35535 = torch.aten.permute %1292, %35534 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33110 = torch.constant.int 1
%int0_33111 = torch.constant.int 0
%35536 = torch.prim.ListConstruct %int1_33110, %int0_33111 : (!torch.int, !torch.int) -> !torch.list<int>
%35537 = torch.aten.permute %1293, %35536 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33112 = torch.constant.int 1
%int0_33113 = torch.constant.int 0
%35538 = torch.prim.ListConstruct %int1_33112, %int0_33113 : (!torch.int, !torch.int) -> !torch.list<int>
%35539 = torch.aten.permute %1294, %35538 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_33114 = torch.constant.int 1
%int0_33115 = torch.constant.int 0
%35540 = torch.prim.ListConstruct %int1_33114, %int0_33115 : (!torch.int, !torch.int) -> !torch.list<int>
%35541 = torch.aten.permute %1295, %35540 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_33116 = torch.constant.int 4
%35542 = torch.aten.mul.int %int4_33116, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33117 = torch.constant.int 4096
%35543 = torch.prim.ListConstruct %35542, %int4096_33117 : (!torch.int, !torch.int) -> !torch.list<int>
%35544 = torch.aten.view %35446, %35543 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35544, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35545 = torch.aten.mm %35544, %35527 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35545, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33118 = torch.constant.int 4
%int1792_33119 = torch.constant.int 1792
%35546 = torch.prim.ListConstruct %int4_33118, %2482, %int1792_33119 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35547 = torch.aten.view %35545, %35546 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35547, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33120 = torch.constant.int 4
%35548 = torch.aten.mul.int %int4_33120, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33121 = torch.constant.int 4096
%35549 = torch.prim.ListConstruct %35548, %int4096_33121 : (!torch.int, !torch.int) -> !torch.list<int>
%35550 = torch.aten.view %35447, %35549 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35550, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35551 = torch.aten.mm %35550, %35529 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35551, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33122 = torch.constant.int 4
%int1792_33123 = torch.constant.int 1792
%35552 = torch.prim.ListConstruct %int4_33122, %2482, %int1792_33123 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35553 = torch.aten.view %35551, %35552 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33124 = torch.constant.int 4
%35554 = torch.aten.mul.int %int4_33124, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33125 = torch.constant.int 4096
%35555 = torch.prim.ListConstruct %35554, %int4096_33125 : (!torch.int, !torch.int) -> !torch.list<int>
%35556 = torch.aten.view %35448, %35555 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35556, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35557 = torch.aten.mm %35556, %35531 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35557, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33126 = torch.constant.int 4
%int1792_33127 = torch.constant.int 1792
%35558 = torch.prim.ListConstruct %int4_33126, %2482, %int1792_33127 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35559 = torch.aten.view %35557, %35558 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33128 = torch.constant.int 4
%35560 = torch.aten.mul.int %int4_33128, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33129 = torch.constant.int 4096
%35561 = torch.prim.ListConstruct %35560, %int4096_33129 : (!torch.int, !torch.int) -> !torch.list<int>
%35562 = torch.aten.view %35449, %35561 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35562, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35563 = torch.aten.mm %35562, %35533 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35563, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33130 = torch.constant.int 4
%int1792_33131 = torch.constant.int 1792
%35564 = torch.prim.ListConstruct %int4_33130, %2482, %int1792_33131 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35565 = torch.aten.view %35563, %35564 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33132 = torch.constant.int 4
%35566 = torch.aten.mul.int %int4_33132, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33133 = torch.constant.int 4096
%35567 = torch.prim.ListConstruct %35566, %int4096_33133 : (!torch.int, !torch.int) -> !torch.list<int>
%35568 = torch.aten.view %35450, %35567 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35568, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35569 = torch.aten.mm %35568, %35535 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35569, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33134 = torch.constant.int 4
%int1792_33135 = torch.constant.int 1792
%35570 = torch.prim.ListConstruct %int4_33134, %2482, %int1792_33135 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35571 = torch.aten.view %35569, %35570 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33136 = torch.constant.int 4
%35572 = torch.aten.mul.int %int4_33136, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33137 = torch.constant.int 4096
%35573 = torch.prim.ListConstruct %35572, %int4096_33137 : (!torch.int, !torch.int) -> !torch.list<int>
%35574 = torch.aten.view %35451, %35573 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35574, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35575 = torch.aten.mm %35574, %35537 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35575, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33138 = torch.constant.int 4
%int1792_33139 = torch.constant.int 1792
%35576 = torch.prim.ListConstruct %int4_33138, %2482, %int1792_33139 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35577 = torch.aten.view %35575, %35576 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33140 = torch.constant.int 4
%35578 = torch.aten.mul.int %int4_33140, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33141 = torch.constant.int 4096
%35579 = torch.prim.ListConstruct %35578, %int4096_33141 : (!torch.int, !torch.int) -> !torch.list<int>
%35580 = torch.aten.view %35452, %35579 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35580, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35581 = torch.aten.mm %35580, %35539 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35581, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33142 = torch.constant.int 4
%int1792_33143 = torch.constant.int 1792
%35582 = torch.prim.ListConstruct %int4_33142, %2482, %int1792_33143 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35583 = torch.aten.view %35581, %35582 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_33144 = torch.constant.int 4
%35584 = torch.aten.mul.int %int4_33144, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33145 = torch.constant.int 4096
%35585 = torch.prim.ListConstruct %35584, %int4096_33145 : (!torch.int, !torch.int) -> !torch.list<int>
%35586 = torch.aten.view %35453, %35585 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35586, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35587 = torch.aten.mm %35586, %35541 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35587, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_33146 = torch.constant.int 4
%int1792_33147 = torch.constant.int 1792
%35588 = torch.prim.ListConstruct %int4_33146, %2482, %int1792_33147 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35589 = torch.aten.view %35587, %35588 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35590 = torch.aten.mul.Tensor %35518, %35547 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35591 = torch.aten.mul.Tensor %35519, %35553 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35591, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35592 = torch.aten.mul.Tensor %35520, %35559 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35593 = torch.aten.mul.Tensor %35521, %35565 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35594 = torch.aten.mul.Tensor %35522, %35571 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35594, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35595 = torch.aten.mul.Tensor %35523, %35577 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35596 = torch.aten.mul.Tensor %35524, %35583 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%35597 = torch.aten.mul.Tensor %35525, %35589 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %35597, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_33148 = torch.constant.int 1
%int0_33149 = torch.constant.int 0
%35598 = torch.prim.ListConstruct %int1_33148, %int0_33149 : (!torch.int, !torch.int) -> !torch.list<int>
%35599 = torch.aten.permute %1296, %35598 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33150 = torch.constant.int 1
%int0_33151 = torch.constant.int 0
%35600 = torch.prim.ListConstruct %int1_33150, %int0_33151 : (!torch.int, !torch.int) -> !torch.list<int>
%35601 = torch.aten.permute %1297, %35600 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33152 = torch.constant.int 1
%int0_33153 = torch.constant.int 0
%35602 = torch.prim.ListConstruct %int1_33152, %int0_33153 : (!torch.int, !torch.int) -> !torch.list<int>
%35603 = torch.aten.permute %1298, %35602 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33154 = torch.constant.int 1
%int0_33155 = torch.constant.int 0
%35604 = torch.prim.ListConstruct %int1_33154, %int0_33155 : (!torch.int, !torch.int) -> !torch.list<int>
%35605 = torch.aten.permute %1299, %35604 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33156 = torch.constant.int 1
%int0_33157 = torch.constant.int 0
%35606 = torch.prim.ListConstruct %int1_33156, %int0_33157 : (!torch.int, !torch.int) -> !torch.list<int>
%35607 = torch.aten.permute %1300, %35606 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33158 = torch.constant.int 1
%int0_33159 = torch.constant.int 0
%35608 = torch.prim.ListConstruct %int1_33158, %int0_33159 : (!torch.int, !torch.int) -> !torch.list<int>
%35609 = torch.aten.permute %1301, %35608 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33160 = torch.constant.int 1
%int0_33161 = torch.constant.int 0
%35610 = torch.prim.ListConstruct %int1_33160, %int0_33161 : (!torch.int, !torch.int) -> !torch.list<int>
%35611 = torch.aten.permute %1302, %35610 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33162 = torch.constant.int 1
%int0_33163 = torch.constant.int 0
%35612 = torch.prim.ListConstruct %int1_33162, %int0_33163 : (!torch.int, !torch.int) -> !torch.list<int>
%35613 = torch.aten.permute %1303, %35612 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_33164 = torch.constant.int 1
%35614 = torch.aten.size.int %35475, %int1_33164 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33165 = torch.constant.int 4
%35615 = torch.aten.mul.int %int4_33165, %35614 : !torch.int, !torch.int -> !torch.int
%int1792_33166 = torch.constant.int 1792
%35616 = torch.prim.ListConstruct %35615, %int1792_33166 : (!torch.int, !torch.int) -> !torch.list<int>
%35617 = torch.aten.view %35590, %35616 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35617, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35618 = torch.aten.mm %35617, %35599 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35618, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33167 = torch.constant.int 4
%int4096_33168 = torch.constant.int 4096
%35619 = torch.prim.ListConstruct %int4_33167, %35614, %int4096_33168 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35620 = torch.aten.view %35618, %35619 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33169 = torch.constant.int 1
%35621 = torch.aten.size.int %35481, %int1_33169 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33170 = torch.constant.int 4
%35622 = torch.aten.mul.int %int4_33170, %35621 : !torch.int, !torch.int -> !torch.int
%int1792_33171 = torch.constant.int 1792
%35623 = torch.prim.ListConstruct %35622, %int1792_33171 : (!torch.int, !torch.int) -> !torch.list<int>
%35624 = torch.aten.view %35591, %35623 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35624, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35625 = torch.aten.mm %35624, %35601 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35625, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33172 = torch.constant.int 4
%int4096_33173 = torch.constant.int 4096
%35626 = torch.prim.ListConstruct %int4_33172, %35621, %int4096_33173 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35627 = torch.aten.view %35625, %35626 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33174 = torch.constant.int 1
%35628 = torch.aten.size.int %35487, %int1_33174 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33175 = torch.constant.int 4
%35629 = torch.aten.mul.int %int4_33175, %35628 : !torch.int, !torch.int -> !torch.int
%int1792_33176 = torch.constant.int 1792
%35630 = torch.prim.ListConstruct %35629, %int1792_33176 : (!torch.int, !torch.int) -> !torch.list<int>
%35631 = torch.aten.view %35592, %35630 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35631, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35632 = torch.aten.mm %35631, %35603 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35632, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33177 = torch.constant.int 4
%int4096_33178 = torch.constant.int 4096
%35633 = torch.prim.ListConstruct %int4_33177, %35628, %int4096_33178 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35634 = torch.aten.view %35632, %35633 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33179 = torch.constant.int 1
%35635 = torch.aten.size.int %35493, %int1_33179 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33180 = torch.constant.int 4
%35636 = torch.aten.mul.int %int4_33180, %35635 : !torch.int, !torch.int -> !torch.int
%int1792_33181 = torch.constant.int 1792
%35637 = torch.prim.ListConstruct %35636, %int1792_33181 : (!torch.int, !torch.int) -> !torch.list<int>
%35638 = torch.aten.view %35593, %35637 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35638, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35639 = torch.aten.mm %35638, %35605 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35639, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33182 = torch.constant.int 4
%int4096_33183 = torch.constant.int 4096
%35640 = torch.prim.ListConstruct %int4_33182, %35635, %int4096_33183 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35641 = torch.aten.view %35639, %35640 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33184 = torch.constant.int 1
%35642 = torch.aten.size.int %35499, %int1_33184 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33185 = torch.constant.int 4
%35643 = torch.aten.mul.int %int4_33185, %35642 : !torch.int, !torch.int -> !torch.int
%int1792_33186 = torch.constant.int 1792
%35644 = torch.prim.ListConstruct %35643, %int1792_33186 : (!torch.int, !torch.int) -> !torch.list<int>
%35645 = torch.aten.view %35594, %35644 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35645, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35646 = torch.aten.mm %35645, %35607 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35646, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33187 = torch.constant.int 4
%int4096_33188 = torch.constant.int 4096
%35647 = torch.prim.ListConstruct %int4_33187, %35642, %int4096_33188 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35648 = torch.aten.view %35646, %35647 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33189 = torch.constant.int 1
%35649 = torch.aten.size.int %35505, %int1_33189 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33190 = torch.constant.int 4
%35650 = torch.aten.mul.int %int4_33190, %35649 : !torch.int, !torch.int -> !torch.int
%int1792_33191 = torch.constant.int 1792
%35651 = torch.prim.ListConstruct %35650, %int1792_33191 : (!torch.int, !torch.int) -> !torch.list<int>
%35652 = torch.aten.view %35595, %35651 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35652, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35653 = torch.aten.mm %35652, %35609 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35653, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33192 = torch.constant.int 4
%int4096_33193 = torch.constant.int 4096
%35654 = torch.prim.ListConstruct %int4_33192, %35649, %int4096_33193 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35655 = torch.aten.view %35653, %35654 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33194 = torch.constant.int 1
%35656 = torch.aten.size.int %35511, %int1_33194 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33195 = torch.constant.int 4
%35657 = torch.aten.mul.int %int4_33195, %35656 : !torch.int, !torch.int -> !torch.int
%int1792_33196 = torch.constant.int 1792
%35658 = torch.prim.ListConstruct %35657, %int1792_33196 : (!torch.int, !torch.int) -> !torch.list<int>
%35659 = torch.aten.view %35596, %35658 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35659, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35660 = torch.aten.mm %35659, %35611 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35660, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33197 = torch.constant.int 4
%int4096_33198 = torch.constant.int 4096
%35661 = torch.prim.ListConstruct %int4_33197, %35656, %int4096_33198 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35662 = torch.aten.view %35660, %35661 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33199 = torch.constant.int 1
%35663 = torch.aten.size.int %35517, %int1_33199 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_33200 = torch.constant.int 4
%35664 = torch.aten.mul.int %int4_33200, %35663 : !torch.int, !torch.int -> !torch.int
%int1792_33201 = torch.constant.int 1792
%35665 = torch.prim.ListConstruct %35664, %int1792_33201 : (!torch.int, !torch.int) -> !torch.list<int>
%35666 = torch.aten.view %35597, %35665 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %35666, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%35667 = torch.aten.mm %35666, %35613 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35667, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_33202 = torch.constant.int 4
%int4096_33203 = torch.constant.int 4096
%35668 = torch.prim.ListConstruct %int4_33202, %35663, %int4096_33203 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35669 = torch.aten.view %35667, %35668 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35670 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33204 = arith.constant 1 : index
%dim_33205 = tensor.dim %35670, %c1_33204 : tensor<4x?x4096xf16>
%35671 = flow.tensor.transfer %35670 : tensor<4x?x4096xf16>{%dim_33205} to #hal.device.promise<@__device_0>
%35672 = torch_c.from_builtin_tensor %35671 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35673 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33206 = arith.constant 1 : index
%dim_33207 = tensor.dim %35673, %c1_33206 : tensor<4x?x4096xf16>
%35674 = flow.tensor.transfer %35673 : tensor<4x?x4096xf16>{%dim_33207} to #hal.device.promise<@__device_0>
%35675 = torch_c.from_builtin_tensor %35674 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35676 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33208 = arith.constant 1 : index
%dim_33209 = tensor.dim %35676, %c1_33208 : tensor<4x?x4096xf16>
%35677 = flow.tensor.transfer %35676 : tensor<4x?x4096xf16>{%dim_33209} to #hal.device.promise<@__device_0>
%35678 = torch_c.from_builtin_tensor %35677 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35679 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33210 = arith.constant 1 : index
%dim_33211 = tensor.dim %35679, %c1_33210 : tensor<4x?x4096xf16>
%35680 = flow.tensor.transfer %35679 : tensor<4x?x4096xf16>{%dim_33211} to #hal.device.promise<@__device_0>
%35681 = torch_c.from_builtin_tensor %35680 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35682 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33212 = arith.constant 1 : index
%dim_33213 = tensor.dim %35682, %c1_33212 : tensor<4x?x4096xf16>
%35683 = flow.tensor.transfer %35682 : tensor<4x?x4096xf16>{%dim_33213} to #hal.device.promise<@__device_0>
%35684 = torch_c.from_builtin_tensor %35683 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35685 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33214 = arith.constant 1 : index
%dim_33215 = tensor.dim %35685, %c1_33214 : tensor<4x?x4096xf16>
%35686 = flow.tensor.transfer %35685 : tensor<4x?x4096xf16>{%dim_33215} to #hal.device.promise<@__device_0>
%35687 = torch_c.from_builtin_tensor %35686 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35688 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33216 = arith.constant 1 : index
%dim_33217 = tensor.dim %35688, %c1_33216 : tensor<4x?x4096xf16>
%35689 = flow.tensor.transfer %35688 : tensor<4x?x4096xf16>{%dim_33217} to #hal.device.promise<@__device_0>
%35690 = torch_c.from_builtin_tensor %35689 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33218 = torch.constant.int 1
%35691 = torch.aten.add.Tensor %35620, %35672, %int1_33218 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33219 = torch.constant.int 1
%35692 = torch.aten.add.Tensor %35691, %35675, %int1_33219 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33220 = torch.constant.int 1
%35693 = torch.aten.add.Tensor %35692, %35678, %int1_33220 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33221 = torch.constant.int 1
%35694 = torch.aten.add.Tensor %35693, %35681, %int1_33221 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33222 = torch.constant.int 1
%35695 = torch.aten.add.Tensor %35694, %35684, %int1_33222 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33223 = torch.constant.int 1
%35696 = torch.aten.add.Tensor %35695, %35687, %int1_33223 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33224 = torch.constant.int 1
%35697 = torch.aten.add.Tensor %35696, %35690, %int1_33224 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35698 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33225 = arith.constant 1 : index
%dim_33226 = tensor.dim %35698, %c1_33225 : tensor<4x?x4096xf16>
%35699 = flow.tensor.transfer %35698 : tensor<4x?x4096xf16>{%dim_33226} to #hal.device.promise<@__device_1>
%35700 = torch_c.from_builtin_tensor %35699 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35701 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33227 = arith.constant 1 : index
%dim_33228 = tensor.dim %35701, %c1_33227 : tensor<4x?x4096xf16>
%35702 = flow.tensor.transfer %35701 : tensor<4x?x4096xf16>{%dim_33228} to #hal.device.promise<@__device_1>
%35703 = torch_c.from_builtin_tensor %35702 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35703, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35704 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33229 = arith.constant 1 : index
%dim_33230 = tensor.dim %35704, %c1_33229 : tensor<4x?x4096xf16>
%35705 = flow.tensor.transfer %35704 : tensor<4x?x4096xf16>{%dim_33230} to #hal.device.promise<@__device_1>
%35706 = torch_c.from_builtin_tensor %35705 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35707 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33231 = arith.constant 1 : index
%dim_33232 = tensor.dim %35707, %c1_33231 : tensor<4x?x4096xf16>
%35708 = flow.tensor.transfer %35707 : tensor<4x?x4096xf16>{%dim_33232} to #hal.device.promise<@__device_1>
%35709 = torch_c.from_builtin_tensor %35708 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35709, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35710 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33233 = arith.constant 1 : index
%dim_33234 = tensor.dim %35710, %c1_33233 : tensor<4x?x4096xf16>
%35711 = flow.tensor.transfer %35710 : tensor<4x?x4096xf16>{%dim_33234} to #hal.device.promise<@__device_1>
%35712 = torch_c.from_builtin_tensor %35711 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35713 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33235 = arith.constant 1 : index
%dim_33236 = tensor.dim %35713, %c1_33235 : tensor<4x?x4096xf16>
%35714 = flow.tensor.transfer %35713 : tensor<4x?x4096xf16>{%dim_33236} to #hal.device.promise<@__device_1>
%35715 = torch_c.from_builtin_tensor %35714 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35716 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33237 = arith.constant 1 : index
%dim_33238 = tensor.dim %35716, %c1_33237 : tensor<4x?x4096xf16>
%35717 = flow.tensor.transfer %35716 : tensor<4x?x4096xf16>{%dim_33238} to #hal.device.promise<@__device_1>
%35718 = torch_c.from_builtin_tensor %35717 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33239 = torch.constant.int 1
%35719 = torch.aten.add.Tensor %35700, %35627, %int1_33239 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33240 = torch.constant.int 1
%35720 = torch.aten.add.Tensor %35719, %35703, %int1_33240 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33241 = torch.constant.int 1
%35721 = torch.aten.add.Tensor %35720, %35706, %int1_33241 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33242 = torch.constant.int 1
%35722 = torch.aten.add.Tensor %35721, %35709, %int1_33242 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33243 = torch.constant.int 1
%35723 = torch.aten.add.Tensor %35722, %35712, %int1_33243 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33244 = torch.constant.int 1
%35724 = torch.aten.add.Tensor %35723, %35715, %int1_33244 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33245 = torch.constant.int 1
%35725 = torch.aten.add.Tensor %35724, %35718, %int1_33245 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35726 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33246 = arith.constant 1 : index
%dim_33247 = tensor.dim %35726, %c1_33246 : tensor<4x?x4096xf16>
%35727 = flow.tensor.transfer %35726 : tensor<4x?x4096xf16>{%dim_33247} to #hal.device.promise<@__device_2>
%35728 = torch_c.from_builtin_tensor %35727 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35728, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35729 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33248 = arith.constant 1 : index
%dim_33249 = tensor.dim %35729, %c1_33248 : tensor<4x?x4096xf16>
%35730 = flow.tensor.transfer %35729 : tensor<4x?x4096xf16>{%dim_33249} to #hal.device.promise<@__device_2>
%35731 = torch_c.from_builtin_tensor %35730 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35732 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33250 = arith.constant 1 : index
%dim_33251 = tensor.dim %35732, %c1_33250 : tensor<4x?x4096xf16>
%35733 = flow.tensor.transfer %35732 : tensor<4x?x4096xf16>{%dim_33251} to #hal.device.promise<@__device_2>
%35734 = torch_c.from_builtin_tensor %35733 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35734, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35735 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33252 = arith.constant 1 : index
%dim_33253 = tensor.dim %35735, %c1_33252 : tensor<4x?x4096xf16>
%35736 = flow.tensor.transfer %35735 : tensor<4x?x4096xf16>{%dim_33253} to #hal.device.promise<@__device_2>
%35737 = torch_c.from_builtin_tensor %35736 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35738 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33254 = arith.constant 1 : index
%dim_33255 = tensor.dim %35738, %c1_33254 : tensor<4x?x4096xf16>
%35739 = flow.tensor.transfer %35738 : tensor<4x?x4096xf16>{%dim_33255} to #hal.device.promise<@__device_2>
%35740 = torch_c.from_builtin_tensor %35739 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35740, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35741 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33256 = arith.constant 1 : index
%dim_33257 = tensor.dim %35741, %c1_33256 : tensor<4x?x4096xf16>
%35742 = flow.tensor.transfer %35741 : tensor<4x?x4096xf16>{%dim_33257} to #hal.device.promise<@__device_2>
%35743 = torch_c.from_builtin_tensor %35742 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35744 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33258 = arith.constant 1 : index
%dim_33259 = tensor.dim %35744, %c1_33258 : tensor<4x?x4096xf16>
%35745 = flow.tensor.transfer %35744 : tensor<4x?x4096xf16>{%dim_33259} to #hal.device.promise<@__device_2>
%35746 = torch_c.from_builtin_tensor %35745 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35746, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33260 = torch.constant.int 1
%35747 = torch.aten.add.Tensor %35728, %35731, %int1_33260 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33261 = torch.constant.int 1
%35748 = torch.aten.add.Tensor %35747, %35634, %int1_33261 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33262 = torch.constant.int 1
%35749 = torch.aten.add.Tensor %35748, %35734, %int1_33262 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33263 = torch.constant.int 1
%35750 = torch.aten.add.Tensor %35749, %35737, %int1_33263 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33264 = torch.constant.int 1
%35751 = torch.aten.add.Tensor %35750, %35740, %int1_33264 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33265 = torch.constant.int 1
%35752 = torch.aten.add.Tensor %35751, %35743, %int1_33265 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33266 = torch.constant.int 1
%35753 = torch.aten.add.Tensor %35752, %35746, %int1_33266 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35754 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33267 = arith.constant 1 : index
%dim_33268 = tensor.dim %35754, %c1_33267 : tensor<4x?x4096xf16>
%35755 = flow.tensor.transfer %35754 : tensor<4x?x4096xf16>{%dim_33268} to #hal.device.promise<@__device_3>
%35756 = torch_c.from_builtin_tensor %35755 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35757 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33269 = arith.constant 1 : index
%dim_33270 = tensor.dim %35757, %c1_33269 : tensor<4x?x4096xf16>
%35758 = flow.tensor.transfer %35757 : tensor<4x?x4096xf16>{%dim_33270} to #hal.device.promise<@__device_3>
%35759 = torch_c.from_builtin_tensor %35758 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35760 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33271 = arith.constant 1 : index
%dim_33272 = tensor.dim %35760, %c1_33271 : tensor<4x?x4096xf16>
%35761 = flow.tensor.transfer %35760 : tensor<4x?x4096xf16>{%dim_33272} to #hal.device.promise<@__device_3>
%35762 = torch_c.from_builtin_tensor %35761 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35763 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33273 = arith.constant 1 : index
%dim_33274 = tensor.dim %35763, %c1_33273 : tensor<4x?x4096xf16>
%35764 = flow.tensor.transfer %35763 : tensor<4x?x4096xf16>{%dim_33274} to #hal.device.promise<@__device_3>
%35765 = torch_c.from_builtin_tensor %35764 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35766 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33275 = arith.constant 1 : index
%dim_33276 = tensor.dim %35766, %c1_33275 : tensor<4x?x4096xf16>
%35767 = flow.tensor.transfer %35766 : tensor<4x?x4096xf16>{%dim_33276} to #hal.device.promise<@__device_3>
%35768 = torch_c.from_builtin_tensor %35767 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35769 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33277 = arith.constant 1 : index
%dim_33278 = tensor.dim %35769, %c1_33277 : tensor<4x?x4096xf16>
%35770 = flow.tensor.transfer %35769 : tensor<4x?x4096xf16>{%dim_33278} to #hal.device.promise<@__device_3>
%35771 = torch_c.from_builtin_tensor %35770 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35772 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33279 = arith.constant 1 : index
%dim_33280 = tensor.dim %35772, %c1_33279 : tensor<4x?x4096xf16>
%35773 = flow.tensor.transfer %35772 : tensor<4x?x4096xf16>{%dim_33280} to #hal.device.promise<@__device_3>
%35774 = torch_c.from_builtin_tensor %35773 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33281 = torch.constant.int 1
%35775 = torch.aten.add.Tensor %35756, %35759, %int1_33281 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33282 = torch.constant.int 1
%35776 = torch.aten.add.Tensor %35775, %35762, %int1_33282 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33283 = torch.constant.int 1
%35777 = torch.aten.add.Tensor %35776, %35641, %int1_33283 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33284 = torch.constant.int 1
%35778 = torch.aten.add.Tensor %35777, %35765, %int1_33284 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33285 = torch.constant.int 1
%35779 = torch.aten.add.Tensor %35778, %35768, %int1_33285 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35779, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33286 = torch.constant.int 1
%35780 = torch.aten.add.Tensor %35779, %35771, %int1_33286 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33287 = torch.constant.int 1
%35781 = torch.aten.add.Tensor %35780, %35774, %int1_33287 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35782 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33288 = arith.constant 1 : index
%dim_33289 = tensor.dim %35782, %c1_33288 : tensor<4x?x4096xf16>
%35783 = flow.tensor.transfer %35782 : tensor<4x?x4096xf16>{%dim_33289} to #hal.device.promise<@__device_4>
%35784 = torch_c.from_builtin_tensor %35783 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35785 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33290 = arith.constant 1 : index
%dim_33291 = tensor.dim %35785, %c1_33290 : tensor<4x?x4096xf16>
%35786 = flow.tensor.transfer %35785 : tensor<4x?x4096xf16>{%dim_33291} to #hal.device.promise<@__device_4>
%35787 = torch_c.from_builtin_tensor %35786 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35788 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33292 = arith.constant 1 : index
%dim_33293 = tensor.dim %35788, %c1_33292 : tensor<4x?x4096xf16>
%35789 = flow.tensor.transfer %35788 : tensor<4x?x4096xf16>{%dim_33293} to #hal.device.promise<@__device_4>
%35790 = torch_c.from_builtin_tensor %35789 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35791 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33294 = arith.constant 1 : index
%dim_33295 = tensor.dim %35791, %c1_33294 : tensor<4x?x4096xf16>
%35792 = flow.tensor.transfer %35791 : tensor<4x?x4096xf16>{%dim_33295} to #hal.device.promise<@__device_4>
%35793 = torch_c.from_builtin_tensor %35792 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35794 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33296 = arith.constant 1 : index
%dim_33297 = tensor.dim %35794, %c1_33296 : tensor<4x?x4096xf16>
%35795 = flow.tensor.transfer %35794 : tensor<4x?x4096xf16>{%dim_33297} to #hal.device.promise<@__device_4>
%35796 = torch_c.from_builtin_tensor %35795 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35797 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33298 = arith.constant 1 : index
%dim_33299 = tensor.dim %35797, %c1_33298 : tensor<4x?x4096xf16>
%35798 = flow.tensor.transfer %35797 : tensor<4x?x4096xf16>{%dim_33299} to #hal.device.promise<@__device_4>
%35799 = torch_c.from_builtin_tensor %35798 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35800 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33300 = arith.constant 1 : index
%dim_33301 = tensor.dim %35800, %c1_33300 : tensor<4x?x4096xf16>
%35801 = flow.tensor.transfer %35800 : tensor<4x?x4096xf16>{%dim_33301} to #hal.device.promise<@__device_4>
%35802 = torch_c.from_builtin_tensor %35801 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33302 = torch.constant.int 1
%35803 = torch.aten.add.Tensor %35784, %35787, %int1_33302 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33303 = torch.constant.int 1
%35804 = torch.aten.add.Tensor %35803, %35790, %int1_33303 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33304 = torch.constant.int 1
%35805 = torch.aten.add.Tensor %35804, %35793, %int1_33304 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33305 = torch.constant.int 1
%35806 = torch.aten.add.Tensor %35805, %35648, %int1_33305 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33306 = torch.constant.int 1
%35807 = torch.aten.add.Tensor %35806, %35796, %int1_33306 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33307 = torch.constant.int 1
%35808 = torch.aten.add.Tensor %35807, %35799, %int1_33307 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33308 = torch.constant.int 1
%35809 = torch.aten.add.Tensor %35808, %35802, %int1_33308 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35810 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33309 = arith.constant 1 : index
%dim_33310 = tensor.dim %35810, %c1_33309 : tensor<4x?x4096xf16>
%35811 = flow.tensor.transfer %35810 : tensor<4x?x4096xf16>{%dim_33310} to #hal.device.promise<@__device_5>
%35812 = torch_c.from_builtin_tensor %35811 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35813 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33311 = arith.constant 1 : index
%dim_33312 = tensor.dim %35813, %c1_33311 : tensor<4x?x4096xf16>
%35814 = flow.tensor.transfer %35813 : tensor<4x?x4096xf16>{%dim_33312} to #hal.device.promise<@__device_5>
%35815 = torch_c.from_builtin_tensor %35814 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35816 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33313 = arith.constant 1 : index
%dim_33314 = tensor.dim %35816, %c1_33313 : tensor<4x?x4096xf16>
%35817 = flow.tensor.transfer %35816 : tensor<4x?x4096xf16>{%dim_33314} to #hal.device.promise<@__device_5>
%35818 = torch_c.from_builtin_tensor %35817 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35819 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33315 = arith.constant 1 : index
%dim_33316 = tensor.dim %35819, %c1_33315 : tensor<4x?x4096xf16>
%35820 = flow.tensor.transfer %35819 : tensor<4x?x4096xf16>{%dim_33316} to #hal.device.promise<@__device_5>
%35821 = torch_c.from_builtin_tensor %35820 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35822 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33317 = arith.constant 1 : index
%dim_33318 = tensor.dim %35822, %c1_33317 : tensor<4x?x4096xf16>
%35823 = flow.tensor.transfer %35822 : tensor<4x?x4096xf16>{%dim_33318} to #hal.device.promise<@__device_5>
%35824 = torch_c.from_builtin_tensor %35823 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35825 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33319 = arith.constant 1 : index
%dim_33320 = tensor.dim %35825, %c1_33319 : tensor<4x?x4096xf16>
%35826 = flow.tensor.transfer %35825 : tensor<4x?x4096xf16>{%dim_33320} to #hal.device.promise<@__device_5>
%35827 = torch_c.from_builtin_tensor %35826 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35828 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33321 = arith.constant 1 : index
%dim_33322 = tensor.dim %35828, %c1_33321 : tensor<4x?x4096xf16>
%35829 = flow.tensor.transfer %35828 : tensor<4x?x4096xf16>{%dim_33322} to #hal.device.promise<@__device_5>
%35830 = torch_c.from_builtin_tensor %35829 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33323 = torch.constant.int 1
%35831 = torch.aten.add.Tensor %35812, %35815, %int1_33323 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33324 = torch.constant.int 1
%35832 = torch.aten.add.Tensor %35831, %35818, %int1_33324 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33325 = torch.constant.int 1
%35833 = torch.aten.add.Tensor %35832, %35821, %int1_33325 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33326 = torch.constant.int 1
%35834 = torch.aten.add.Tensor %35833, %35824, %int1_33326 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33327 = torch.constant.int 1
%35835 = torch.aten.add.Tensor %35834, %35655, %int1_33327 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33328 = torch.constant.int 1
%35836 = torch.aten.add.Tensor %35835, %35827, %int1_33328 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35836, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33329 = torch.constant.int 1
%35837 = torch.aten.add.Tensor %35836, %35830, %int1_33329 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35838 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33330 = arith.constant 1 : index
%dim_33331 = tensor.dim %35838, %c1_33330 : tensor<4x?x4096xf16>
%35839 = flow.tensor.transfer %35838 : tensor<4x?x4096xf16>{%dim_33331} to #hal.device.promise<@__device_6>
%35840 = torch_c.from_builtin_tensor %35839 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35840, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35841 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33332 = arith.constant 1 : index
%dim_33333 = tensor.dim %35841, %c1_33332 : tensor<4x?x4096xf16>
%35842 = flow.tensor.transfer %35841 : tensor<4x?x4096xf16>{%dim_33333} to #hal.device.promise<@__device_6>
%35843 = torch_c.from_builtin_tensor %35842 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35844 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33334 = arith.constant 1 : index
%dim_33335 = tensor.dim %35844, %c1_33334 : tensor<4x?x4096xf16>
%35845 = flow.tensor.transfer %35844 : tensor<4x?x4096xf16>{%dim_33335} to #hal.device.promise<@__device_6>
%35846 = torch_c.from_builtin_tensor %35845 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35846, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35847 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33336 = arith.constant 1 : index
%dim_33337 = tensor.dim %35847, %c1_33336 : tensor<4x?x4096xf16>
%35848 = flow.tensor.transfer %35847 : tensor<4x?x4096xf16>{%dim_33337} to #hal.device.promise<@__device_6>
%35849 = torch_c.from_builtin_tensor %35848 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35850 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33338 = arith.constant 1 : index
%dim_33339 = tensor.dim %35850, %c1_33338 : tensor<4x?x4096xf16>
%35851 = flow.tensor.transfer %35850 : tensor<4x?x4096xf16>{%dim_33339} to #hal.device.promise<@__device_6>
%35852 = torch_c.from_builtin_tensor %35851 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35853 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33340 = arith.constant 1 : index
%dim_33341 = tensor.dim %35853, %c1_33340 : tensor<4x?x4096xf16>
%35854 = flow.tensor.transfer %35853 : tensor<4x?x4096xf16>{%dim_33341} to #hal.device.promise<@__device_6>
%35855 = torch_c.from_builtin_tensor %35854 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35856 = torch_c.to_builtin_tensor %35669 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33342 = arith.constant 1 : index
%dim_33343 = tensor.dim %35856, %c1_33342 : tensor<4x?x4096xf16>
%35857 = flow.tensor.transfer %35856 : tensor<4x?x4096xf16>{%dim_33343} to #hal.device.promise<@__device_6>
%35858 = torch_c.from_builtin_tensor %35857 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35858, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33344 = torch.constant.int 1
%35859 = torch.aten.add.Tensor %35840, %35843, %int1_33344 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33345 = torch.constant.int 1
%35860 = torch.aten.add.Tensor %35859, %35846, %int1_33345 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33346 = torch.constant.int 1
%35861 = torch.aten.add.Tensor %35860, %35849, %int1_33346 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35861, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33347 = torch.constant.int 1
%35862 = torch.aten.add.Tensor %35861, %35852, %int1_33347 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33348 = torch.constant.int 1
%35863 = torch.aten.add.Tensor %35862, %35855, %int1_33348 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33349 = torch.constant.int 1
%35864 = torch.aten.add.Tensor %35863, %35662, %int1_33349 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33350 = torch.constant.int 1
%35865 = torch.aten.add.Tensor %35864, %35858, %int1_33350 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35866 = torch_c.to_builtin_tensor %35620 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33351 = arith.constant 1 : index
%dim_33352 = tensor.dim %35866, %c1_33351 : tensor<4x?x4096xf16>
%35867 = flow.tensor.transfer %35866 : tensor<4x?x4096xf16>{%dim_33352} to #hal.device.promise<@__device_7>
%35868 = torch_c.from_builtin_tensor %35867 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35869 = torch_c.to_builtin_tensor %35627 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33353 = arith.constant 1 : index
%dim_33354 = tensor.dim %35869, %c1_33353 : tensor<4x?x4096xf16>
%35870 = flow.tensor.transfer %35869 : tensor<4x?x4096xf16>{%dim_33354} to #hal.device.promise<@__device_7>
%35871 = torch_c.from_builtin_tensor %35870 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35872 = torch_c.to_builtin_tensor %35634 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33355 = arith.constant 1 : index
%dim_33356 = tensor.dim %35872, %c1_33355 : tensor<4x?x4096xf16>
%35873 = flow.tensor.transfer %35872 : tensor<4x?x4096xf16>{%dim_33356} to #hal.device.promise<@__device_7>
%35874 = torch_c.from_builtin_tensor %35873 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35875 = torch_c.to_builtin_tensor %35641 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33357 = arith.constant 1 : index
%dim_33358 = tensor.dim %35875, %c1_33357 : tensor<4x?x4096xf16>
%35876 = flow.tensor.transfer %35875 : tensor<4x?x4096xf16>{%dim_33358} to #hal.device.promise<@__device_7>
%35877 = torch_c.from_builtin_tensor %35876 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35878 = torch_c.to_builtin_tensor %35648 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33359 = arith.constant 1 : index
%dim_33360 = tensor.dim %35878, %c1_33359 : tensor<4x?x4096xf16>
%35879 = flow.tensor.transfer %35878 : tensor<4x?x4096xf16>{%dim_33360} to #hal.device.promise<@__device_7>
%35880 = torch_c.from_builtin_tensor %35879 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35881 = torch_c.to_builtin_tensor %35655 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33361 = arith.constant 1 : index
%dim_33362 = tensor.dim %35881, %c1_33361 : tensor<4x?x4096xf16>
%35882 = flow.tensor.transfer %35881 : tensor<4x?x4096xf16>{%dim_33362} to #hal.device.promise<@__device_7>
%35883 = torch_c.from_builtin_tensor %35882 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%35884 = torch_c.to_builtin_tensor %35662 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_33363 = arith.constant 1 : index
%dim_33364 = tensor.dim %35884, %c1_33363 : tensor<4x?x4096xf16>
%35885 = flow.tensor.transfer %35884 : tensor<4x?x4096xf16>{%dim_33364} to #hal.device.promise<@__device_7>
%35886 = torch_c.from_builtin_tensor %35885 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33365 = torch.constant.int 1
%35887 = torch.aten.add.Tensor %35868, %35871, %int1_33365 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33366 = torch.constant.int 1
%35888 = torch.aten.add.Tensor %35887, %35874, %int1_33366 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33367 = torch.constant.int 1
%35889 = torch.aten.add.Tensor %35888, %35877, %int1_33367 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33368 = torch.constant.int 1
%35890 = torch.aten.add.Tensor %35889, %35880, %int1_33368 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33369 = torch.constant.int 1
%35891 = torch.aten.add.Tensor %35890, %35883, %int1_33369 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33370 = torch.constant.int 1
%35892 = torch.aten.add.Tensor %35891, %35886, %int1_33370 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33371 = torch.constant.int 1
%35893 = torch.aten.add.Tensor %35892, %35669, %int1_33371 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33372 = torch.constant.int 1
%35894 = torch.aten.add.Tensor %35374, %35697, %int1_33372 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33373 = torch.constant.int 1
%35895 = torch.aten.add.Tensor %35375, %35725, %int1_33373 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33374 = torch.constant.int 1
%35896 = torch.aten.add.Tensor %35376, %35753, %int1_33374 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33375 = torch.constant.int 1
%35897 = torch.aten.add.Tensor %35377, %35781, %int1_33375 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33376 = torch.constant.int 1
%35898 = torch.aten.add.Tensor %35378, %35809, %int1_33376 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33377 = torch.constant.int 1
%35899 = torch.aten.add.Tensor %35379, %35837, %int1_33377 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33378 = torch.constant.int 1
%35900 = torch.aten.add.Tensor %35380, %35865, %int1_33378 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35900, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33379 = torch.constant.int 1
%35901 = torch.aten.add.Tensor %35381, %35893, %int1_33379 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_33380 = torch.constant.int 6
%35902 = torch.prims.convert_element_type %35894, %int6_33380 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33381 = torch.constant.int 6
%35903 = torch.prims.convert_element_type %35895, %int6_33381 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33382 = torch.constant.int 6
%35904 = torch.prims.convert_element_type %35896, %int6_33382 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35904, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33383 = torch.constant.int 6
%35905 = torch.prims.convert_element_type %35897, %int6_33383 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33384 = torch.constant.int 6
%35906 = torch.prims.convert_element_type %35898, %int6_33384 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35906, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33385 = torch.constant.int 6
%35907 = torch.prims.convert_element_type %35899, %int6_33385 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33386 = torch.constant.int 6
%35908 = torch.prims.convert_element_type %35900, %int6_33386 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_33387 = torch.constant.int 6
%35909 = torch.prims.convert_element_type %35901, %int6_33387 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33388 = torch.constant.int 2
%35910 = torch.aten.pow.Tensor_Scalar %35902, %int2_33388 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35910, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33389 = torch.constant.int 2
%35911 = torch.aten.pow.Tensor_Scalar %35903, %int2_33389 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33390 = torch.constant.int 2
%35912 = torch.aten.pow.Tensor_Scalar %35904, %int2_33390 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35912, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33391 = torch.constant.int 2
%35913 = torch.aten.pow.Tensor_Scalar %35905, %int2_33391 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33392 = torch.constant.int 2
%35914 = torch.aten.pow.Tensor_Scalar %35906, %int2_33392 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33393 = torch.constant.int 2
%35915 = torch.aten.pow.Tensor_Scalar %35907, %int2_33393 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33394 = torch.constant.int 2
%35916 = torch.aten.pow.Tensor_Scalar %35908, %int2_33394 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35916, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_33395 = torch.constant.int 2
%35917 = torch.aten.pow.Tensor_Scalar %35909, %int2_33395 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_33396 = torch.constant.int -1
%35918 = torch.prim.ListConstruct %int-1_33396 : (!torch.int) -> !torch.list<int>
%true_33397 = torch.constant.bool true
%none_33398 = torch.constant.none
%35919 = torch.aten.mean.dim %35910, %35918, %true_33397, %none_33398 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33399 = torch.constant.int -1
%35920 = torch.prim.ListConstruct %int-1_33399 : (!torch.int) -> !torch.list<int>
%true_33400 = torch.constant.bool true
%none_33401 = torch.constant.none
%35921 = torch.aten.mean.dim %35911, %35920, %true_33400, %none_33401 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33402 = torch.constant.int -1
%35922 = torch.prim.ListConstruct %int-1_33402 : (!torch.int) -> !torch.list<int>
%true_33403 = torch.constant.bool true
%none_33404 = torch.constant.none
%35923 = torch.aten.mean.dim %35912, %35922, %true_33403, %none_33404 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33405 = torch.constant.int -1
%35924 = torch.prim.ListConstruct %int-1_33405 : (!torch.int) -> !torch.list<int>
%true_33406 = torch.constant.bool true
%none_33407 = torch.constant.none
%35925 = torch.aten.mean.dim %35913, %35924, %true_33406, %none_33407 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33408 = torch.constant.int -1
%35926 = torch.prim.ListConstruct %int-1_33408 : (!torch.int) -> !torch.list<int>
%true_33409 = torch.constant.bool true
%none_33410 = torch.constant.none
%35927 = torch.aten.mean.dim %35914, %35926, %true_33409, %none_33410 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33411 = torch.constant.int -1
%35928 = torch.prim.ListConstruct %int-1_33411 : (!torch.int) -> !torch.list<int>
%true_33412 = torch.constant.bool true
%none_33413 = torch.constant.none
%35929 = torch.aten.mean.dim %35915, %35928, %true_33412, %none_33413 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33414 = torch.constant.int -1
%35930 = torch.prim.ListConstruct %int-1_33414 : (!torch.int) -> !torch.list<int>
%true_33415 = torch.constant.bool true
%none_33416 = torch.constant.none
%35931 = torch.aten.mean.dim %35916, %35930, %true_33415, %none_33416 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_33417 = torch.constant.int -1
%35932 = torch.prim.ListConstruct %int-1_33417 : (!torch.int) -> !torch.list<int>
%true_33418 = torch.constant.bool true
%none_33419 = torch.constant.none
%35933 = torch.aten.mean.dim %35917, %35932, %true_33418, %none_33419 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33420 = torch.constant.float 9.9999997473787516E-6
%int1_33421 = torch.constant.int 1
%35934 = torch.aten.add.Scalar %35919, %float9.999990e-06_33420, %int1_33421 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33422 = torch.constant.float 9.9999997473787516E-6
%int1_33423 = torch.constant.int 1
%35935 = torch.aten.add.Scalar %35921, %float9.999990e-06_33422, %int1_33423 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33424 = torch.constant.float 9.9999997473787516E-6
%int1_33425 = torch.constant.int 1
%35936 = torch.aten.add.Scalar %35923, %float9.999990e-06_33424, %int1_33425 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33426 = torch.constant.float 9.9999997473787516E-6
%int1_33427 = torch.constant.int 1
%35937 = torch.aten.add.Scalar %35925, %float9.999990e-06_33426, %int1_33427 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35937, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33428 = torch.constant.float 9.9999997473787516E-6
%int1_33429 = torch.constant.int 1
%35938 = torch.aten.add.Scalar %35927, %float9.999990e-06_33428, %int1_33429 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33430 = torch.constant.float 9.9999997473787516E-6
%int1_33431 = torch.constant.int 1
%35939 = torch.aten.add.Scalar %35929, %float9.999990e-06_33430, %int1_33431 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33432 = torch.constant.float 9.9999997473787516E-6
%int1_33433 = torch.constant.int 1
%35940 = torch.aten.add.Scalar %35931, %float9.999990e-06_33432, %int1_33433 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_33434 = torch.constant.float 9.9999997473787516E-6
%int1_33435 = torch.constant.int 1
%35941 = torch.aten.add.Scalar %35933, %float9.999990e-06_33434, %int1_33435 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35941, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35942 = torch.aten.rsqrt %35934 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35943 = torch.aten.rsqrt %35935 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35943, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35944 = torch.aten.rsqrt %35936 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35945 = torch.aten.rsqrt %35937 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35946 = torch.aten.rsqrt %35938 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35947 = torch.aten.rsqrt %35939 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35947, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35948 = torch.aten.rsqrt %35940 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35949 = torch.aten.rsqrt %35941 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %35949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%35950 = torch.aten.mul.Tensor %35902, %35942 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35951 = torch.aten.mul.Tensor %35903, %35943 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35952 = torch.aten.mul.Tensor %35904, %35944 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35953 = torch.aten.mul.Tensor %35905, %35945 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35954 = torch.aten.mul.Tensor %35906, %35946 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35955 = torch.aten.mul.Tensor %35907, %35947 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35956 = torch.aten.mul.Tensor %35908, %35948 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35957 = torch.aten.mul.Tensor %35909, %35949 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35957, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35958 = torch.aten.mul.Tensor %1304, %35950 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35959 = torch.aten.mul.Tensor %1305, %35951 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35959, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35960 = torch.aten.mul.Tensor %1306, %35952 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35960, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35961 = torch.aten.mul.Tensor %1307, %35953 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35962 = torch.aten.mul.Tensor %1308, %35954 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35963 = torch.aten.mul.Tensor %1309, %35955 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35963, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35964 = torch.aten.mul.Tensor %1310, %35956 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%35965 = torch.aten.mul.Tensor %1311, %35957 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %35965, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_33436 = torch.constant.int 5
%35966 = torch.prims.convert_element_type %35958, %int5_33436 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35966, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33437 = torch.constant.int 5
%35967 = torch.prims.convert_element_type %35959, %int5_33437 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33438 = torch.constant.int 5
%35968 = torch.prims.convert_element_type %35960, %int5_33438 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33439 = torch.constant.int 5
%35969 = torch.prims.convert_element_type %35961, %int5_33439 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35969, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33440 = torch.constant.int 5
%35970 = torch.prims.convert_element_type %35962, %int5_33440 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33441 = torch.constant.int 5
%35971 = torch.prims.convert_element_type %35963, %int5_33441 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35971, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33442 = torch.constant.int 5
%35972 = torch.prims.convert_element_type %35964, %int5_33442 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35972, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_33443 = torch.constant.int 5
%35973 = torch.prims.convert_element_type %35965, %int5_33443 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %35973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_33444 = torch.constant.int 1
%int0_33445 = torch.constant.int 0
%35974 = torch.prim.ListConstruct %int1_33444, %int0_33445 : (!torch.int, !torch.int) -> !torch.list<int>
%35975 = torch.aten.permute %1312, %35974 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33446 = torch.constant.int 1
%int0_33447 = torch.constant.int 0
%35976 = torch.prim.ListConstruct %int1_33446, %int0_33447 : (!torch.int, !torch.int) -> !torch.list<int>
%35977 = torch.aten.permute %1313, %35976 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33448 = torch.constant.int 1
%int0_33449 = torch.constant.int 0
%35978 = torch.prim.ListConstruct %int1_33448, %int0_33449 : (!torch.int, !torch.int) -> !torch.list<int>
%35979 = torch.aten.permute %1314, %35978 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33450 = torch.constant.int 1
%int0_33451 = torch.constant.int 0
%35980 = torch.prim.ListConstruct %int1_33450, %int0_33451 : (!torch.int, !torch.int) -> !torch.list<int>
%35981 = torch.aten.permute %1315, %35980 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33452 = torch.constant.int 1
%int0_33453 = torch.constant.int 0
%35982 = torch.prim.ListConstruct %int1_33452, %int0_33453 : (!torch.int, !torch.int) -> !torch.list<int>
%35983 = torch.aten.permute %1316, %35982 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33454 = torch.constant.int 1
%int0_33455 = torch.constant.int 0
%35984 = torch.prim.ListConstruct %int1_33454, %int0_33455 : (!torch.int, !torch.int) -> !torch.list<int>
%35985 = torch.aten.permute %1317, %35984 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33456 = torch.constant.int 1
%int0_33457 = torch.constant.int 0
%35986 = torch.prim.ListConstruct %int1_33456, %int0_33457 : (!torch.int, !torch.int) -> !torch.list<int>
%35987 = torch.aten.permute %1318, %35986 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_33458 = torch.constant.int 1
%int0_33459 = torch.constant.int 0
%35988 = torch.prim.ListConstruct %int1_33458, %int0_33459 : (!torch.int, !torch.int) -> !torch.list<int>
%35989 = torch.aten.permute %1319, %35988 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_33460 = torch.constant.int 4
%35990 = torch.aten.mul.int %int4_33460, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33461 = torch.constant.int 4096
%35991 = torch.prim.ListConstruct %35990, %int4096_33461 : (!torch.int, !torch.int) -> !torch.list<int>
%35992 = torch.aten.view %35966, %35991 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35992, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35993 = torch.aten.mm %35992, %35975 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35993, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33462 = torch.constant.int 4
%int512_33463 = torch.constant.int 512
%35994 = torch.prim.ListConstruct %int4_33462, %2482, %int512_33463 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%35995 = torch.aten.view %35993, %35994 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %35995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33464 = torch.constant.int 4
%35996 = torch.aten.mul.int %int4_33464, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33465 = torch.constant.int 4096
%35997 = torch.prim.ListConstruct %35996, %int4096_33465 : (!torch.int, !torch.int) -> !torch.list<int>
%35998 = torch.aten.view %35967, %35997 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %35998, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%35999 = torch.aten.mm %35998, %35977 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %35999, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33466 = torch.constant.int 4
%int512_33467 = torch.constant.int 512
%36000 = torch.prim.ListConstruct %int4_33466, %2482, %int512_33467 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36001 = torch.aten.view %35999, %36000 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33468 = torch.constant.int 4
%36002 = torch.aten.mul.int %int4_33468, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33469 = torch.constant.int 4096
%36003 = torch.prim.ListConstruct %36002, %int4096_33469 : (!torch.int, !torch.int) -> !torch.list<int>
%36004 = torch.aten.view %35968, %36003 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36004, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36005 = torch.aten.mm %36004, %35979 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36005, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33470 = torch.constant.int 4
%int512_33471 = torch.constant.int 512
%36006 = torch.prim.ListConstruct %int4_33470, %2482, %int512_33471 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36007 = torch.aten.view %36005, %36006 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33472 = torch.constant.int 4
%36008 = torch.aten.mul.int %int4_33472, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33473 = torch.constant.int 4096
%36009 = torch.prim.ListConstruct %36008, %int4096_33473 : (!torch.int, !torch.int) -> !torch.list<int>
%36010 = torch.aten.view %35969, %36009 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36010, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36011 = torch.aten.mm %36010, %35981 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36011, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33474 = torch.constant.int 4
%int512_33475 = torch.constant.int 512
%36012 = torch.prim.ListConstruct %int4_33474, %2482, %int512_33475 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36013 = torch.aten.view %36011, %36012 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33476 = torch.constant.int 4
%36014 = torch.aten.mul.int %int4_33476, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33477 = torch.constant.int 4096
%36015 = torch.prim.ListConstruct %36014, %int4096_33477 : (!torch.int, !torch.int) -> !torch.list<int>
%36016 = torch.aten.view %35970, %36015 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36016, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36017 = torch.aten.mm %36016, %35983 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36017, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33478 = torch.constant.int 4
%int512_33479 = torch.constant.int 512
%36018 = torch.prim.ListConstruct %int4_33478, %2482, %int512_33479 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36019 = torch.aten.view %36017, %36018 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33480 = torch.constant.int 4
%36020 = torch.aten.mul.int %int4_33480, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33481 = torch.constant.int 4096
%36021 = torch.prim.ListConstruct %36020, %int4096_33481 : (!torch.int, !torch.int) -> !torch.list<int>
%36022 = torch.aten.view %35971, %36021 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36022, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36023 = torch.aten.mm %36022, %35985 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36023, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33482 = torch.constant.int 4
%int512_33483 = torch.constant.int 512
%36024 = torch.prim.ListConstruct %int4_33482, %2482, %int512_33483 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36025 = torch.aten.view %36023, %36024 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33484 = torch.constant.int 4
%36026 = torch.aten.mul.int %int4_33484, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33485 = torch.constant.int 4096
%36027 = torch.prim.ListConstruct %36026, %int4096_33485 : (!torch.int, !torch.int) -> !torch.list<int>
%36028 = torch.aten.view %35972, %36027 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36028, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36029 = torch.aten.mm %36028, %35987 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36029, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33486 = torch.constant.int 4
%int512_33487 = torch.constant.int 512
%36030 = torch.prim.ListConstruct %int4_33486, %2482, %int512_33487 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36031 = torch.aten.view %36029, %36030 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_33488 = torch.constant.int 4
%36032 = torch.aten.mul.int %int4_33488, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33489 = torch.constant.int 4096
%36033 = torch.prim.ListConstruct %36032, %int4096_33489 : (!torch.int, !torch.int) -> !torch.list<int>
%36034 = torch.aten.view %35973, %36033 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36034, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36035 = torch.aten.mm %36034, %35989 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36035, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_33490 = torch.constant.int 4
%int512_33491 = torch.constant.int 512
%36036 = torch.prim.ListConstruct %int4_33490, %2482, %int512_33491 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36037 = torch.aten.view %36035, %36036 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_33492 = torch.constant.int 1
%int0_33493 = torch.constant.int 0
%36038 = torch.prim.ListConstruct %int1_33492, %int0_33493 : (!torch.int, !torch.int) -> !torch.list<int>
%36039 = torch.aten.permute %1320, %36038 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33494 = torch.constant.int 1
%int0_33495 = torch.constant.int 0
%36040 = torch.prim.ListConstruct %int1_33494, %int0_33495 : (!torch.int, !torch.int) -> !torch.list<int>
%36041 = torch.aten.permute %1321, %36040 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33496 = torch.constant.int 1
%int0_33497 = torch.constant.int 0
%36042 = torch.prim.ListConstruct %int1_33496, %int0_33497 : (!torch.int, !torch.int) -> !torch.list<int>
%36043 = torch.aten.permute %1322, %36042 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33498 = torch.constant.int 1
%int0_33499 = torch.constant.int 0
%36044 = torch.prim.ListConstruct %int1_33498, %int0_33499 : (!torch.int, !torch.int) -> !torch.list<int>
%36045 = torch.aten.permute %1323, %36044 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33500 = torch.constant.int 1
%int0_33501 = torch.constant.int 0
%36046 = torch.prim.ListConstruct %int1_33500, %int0_33501 : (!torch.int, !torch.int) -> !torch.list<int>
%36047 = torch.aten.permute %1324, %36046 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33502 = torch.constant.int 1
%int0_33503 = torch.constant.int 0
%36048 = torch.prim.ListConstruct %int1_33502, %int0_33503 : (!torch.int, !torch.int) -> !torch.list<int>
%36049 = torch.aten.permute %1325, %36048 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33504 = torch.constant.int 1
%int0_33505 = torch.constant.int 0
%36050 = torch.prim.ListConstruct %int1_33504, %int0_33505 : (!torch.int, !torch.int) -> !torch.list<int>
%36051 = torch.aten.permute %1326, %36050 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33506 = torch.constant.int 1
%int0_33507 = torch.constant.int 0
%36052 = torch.prim.ListConstruct %int1_33506, %int0_33507 : (!torch.int, !torch.int) -> !torch.list<int>
%36053 = torch.aten.permute %1327, %36052 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_33508 = torch.constant.int 4
%36054 = torch.aten.mul.int %int4_33508, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33509 = torch.constant.int 4096
%36055 = torch.prim.ListConstruct %36054, %int4096_33509 : (!torch.int, !torch.int) -> !torch.list<int>
%36056 = torch.aten.view %35966, %36055 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36056, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36057 = torch.aten.mm %36056, %36039 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36057, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33510 = torch.constant.int 4
%int128_33511 = torch.constant.int 128
%36058 = torch.prim.ListConstruct %int4_33510, %2482, %int128_33511 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36059 = torch.aten.view %36057, %36058 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33512 = torch.constant.int 4
%36060 = torch.aten.mul.int %int4_33512, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33513 = torch.constant.int 4096
%36061 = torch.prim.ListConstruct %36060, %int4096_33513 : (!torch.int, !torch.int) -> !torch.list<int>
%36062 = torch.aten.view %35967, %36061 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36062, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36063 = torch.aten.mm %36062, %36041 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36063, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33514 = torch.constant.int 4
%int128_33515 = torch.constant.int 128
%36064 = torch.prim.ListConstruct %int4_33514, %2482, %int128_33515 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36065 = torch.aten.view %36063, %36064 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33516 = torch.constant.int 4
%36066 = torch.aten.mul.int %int4_33516, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33517 = torch.constant.int 4096
%36067 = torch.prim.ListConstruct %36066, %int4096_33517 : (!torch.int, !torch.int) -> !torch.list<int>
%36068 = torch.aten.view %35968, %36067 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36068, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36069 = torch.aten.mm %36068, %36043 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36069, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33518 = torch.constant.int 4
%int128_33519 = torch.constant.int 128
%36070 = torch.prim.ListConstruct %int4_33518, %2482, %int128_33519 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36071 = torch.aten.view %36069, %36070 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36071, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33520 = torch.constant.int 4
%36072 = torch.aten.mul.int %int4_33520, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33521 = torch.constant.int 4096
%36073 = torch.prim.ListConstruct %36072, %int4096_33521 : (!torch.int, !torch.int) -> !torch.list<int>
%36074 = torch.aten.view %35969, %36073 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36074, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36075 = torch.aten.mm %36074, %36045 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36075, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33522 = torch.constant.int 4
%int128_33523 = torch.constant.int 128
%36076 = torch.prim.ListConstruct %int4_33522, %2482, %int128_33523 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36077 = torch.aten.view %36075, %36076 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33524 = torch.constant.int 4
%36078 = torch.aten.mul.int %int4_33524, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33525 = torch.constant.int 4096
%36079 = torch.prim.ListConstruct %36078, %int4096_33525 : (!torch.int, !torch.int) -> !torch.list<int>
%36080 = torch.aten.view %35970, %36079 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36080, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36081 = torch.aten.mm %36080, %36047 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36081, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33526 = torch.constant.int 4
%int128_33527 = torch.constant.int 128
%36082 = torch.prim.ListConstruct %int4_33526, %2482, %int128_33527 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36083 = torch.aten.view %36081, %36082 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36083, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33528 = torch.constant.int 4
%36084 = torch.aten.mul.int %int4_33528, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33529 = torch.constant.int 4096
%36085 = torch.prim.ListConstruct %36084, %int4096_33529 : (!torch.int, !torch.int) -> !torch.list<int>
%36086 = torch.aten.view %35971, %36085 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36086, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36087 = torch.aten.mm %36086, %36049 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36087, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33530 = torch.constant.int 4
%int128_33531 = torch.constant.int 128
%36088 = torch.prim.ListConstruct %int4_33530, %2482, %int128_33531 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36089 = torch.aten.view %36087, %36088 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33532 = torch.constant.int 4
%36090 = torch.aten.mul.int %int4_33532, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33533 = torch.constant.int 4096
%36091 = torch.prim.ListConstruct %36090, %int4096_33533 : (!torch.int, !torch.int) -> !torch.list<int>
%36092 = torch.aten.view %35972, %36091 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36092, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36093 = torch.aten.mm %36092, %36051 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36093, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33534 = torch.constant.int 4
%int128_33535 = torch.constant.int 128
%36094 = torch.prim.ListConstruct %int4_33534, %2482, %int128_33535 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36095 = torch.aten.view %36093, %36094 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33536 = torch.constant.int 4
%36096 = torch.aten.mul.int %int4_33536, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33537 = torch.constant.int 4096
%36097 = torch.prim.ListConstruct %36096, %int4096_33537 : (!torch.int, !torch.int) -> !torch.list<int>
%36098 = torch.aten.view %35973, %36097 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36098, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36099 = torch.aten.mm %36098, %36053 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36099, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33538 = torch.constant.int 4
%int128_33539 = torch.constant.int 128
%36100 = torch.prim.ListConstruct %int4_33538, %2482, %int128_33539 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36101 = torch.aten.view %36099, %36100 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_33540 = torch.constant.int 1
%int0_33541 = torch.constant.int 0
%36102 = torch.prim.ListConstruct %int1_33540, %int0_33541 : (!torch.int, !torch.int) -> !torch.list<int>
%36103 = torch.aten.permute %1328, %36102 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33542 = torch.constant.int 1
%int0_33543 = torch.constant.int 0
%36104 = torch.prim.ListConstruct %int1_33542, %int0_33543 : (!torch.int, !torch.int) -> !torch.list<int>
%36105 = torch.aten.permute %1329, %36104 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33544 = torch.constant.int 1
%int0_33545 = torch.constant.int 0
%36106 = torch.prim.ListConstruct %int1_33544, %int0_33545 : (!torch.int, !torch.int) -> !torch.list<int>
%36107 = torch.aten.permute %1330, %36106 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33546 = torch.constant.int 1
%int0_33547 = torch.constant.int 0
%36108 = torch.prim.ListConstruct %int1_33546, %int0_33547 : (!torch.int, !torch.int) -> !torch.list<int>
%36109 = torch.aten.permute %1331, %36108 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33548 = torch.constant.int 1
%int0_33549 = torch.constant.int 0
%36110 = torch.prim.ListConstruct %int1_33548, %int0_33549 : (!torch.int, !torch.int) -> !torch.list<int>
%36111 = torch.aten.permute %1332, %36110 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33550 = torch.constant.int 1
%int0_33551 = torch.constant.int 0
%36112 = torch.prim.ListConstruct %int1_33550, %int0_33551 : (!torch.int, !torch.int) -> !torch.list<int>
%36113 = torch.aten.permute %1333, %36112 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33552 = torch.constant.int 1
%int0_33553 = torch.constant.int 0
%36114 = torch.prim.ListConstruct %int1_33552, %int0_33553 : (!torch.int, !torch.int) -> !torch.list<int>
%36115 = torch.aten.permute %1334, %36114 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_33554 = torch.constant.int 1
%int0_33555 = torch.constant.int 0
%36116 = torch.prim.ListConstruct %int1_33554, %int0_33555 : (!torch.int, !torch.int) -> !torch.list<int>
%36117 = torch.aten.permute %1335, %36116 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_33556 = torch.constant.int 4
%36118 = torch.aten.mul.int %int4_33556, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33557 = torch.constant.int 4096
%36119 = torch.prim.ListConstruct %36118, %int4096_33557 : (!torch.int, !torch.int) -> !torch.list<int>
%36120 = torch.aten.view %35966, %36119 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36120, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36121 = torch.aten.mm %36120, %36103 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36121, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33558 = torch.constant.int 4
%int128_33559 = torch.constant.int 128
%36122 = torch.prim.ListConstruct %int4_33558, %2482, %int128_33559 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36123 = torch.aten.view %36121, %36122 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33560 = torch.constant.int 4
%36124 = torch.aten.mul.int %int4_33560, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33561 = torch.constant.int 4096
%36125 = torch.prim.ListConstruct %36124, %int4096_33561 : (!torch.int, !torch.int) -> !torch.list<int>
%36126 = torch.aten.view %35967, %36125 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36126, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36127 = torch.aten.mm %36126, %36105 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36127, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33562 = torch.constant.int 4
%int128_33563 = torch.constant.int 128
%36128 = torch.prim.ListConstruct %int4_33562, %2482, %int128_33563 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36129 = torch.aten.view %36127, %36128 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33564 = torch.constant.int 4
%36130 = torch.aten.mul.int %int4_33564, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33565 = torch.constant.int 4096
%36131 = torch.prim.ListConstruct %36130, %int4096_33565 : (!torch.int, !torch.int) -> !torch.list<int>
%36132 = torch.aten.view %35968, %36131 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36132, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36133 = torch.aten.mm %36132, %36107 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36133, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33566 = torch.constant.int 4
%int128_33567 = torch.constant.int 128
%36134 = torch.prim.ListConstruct %int4_33566, %2482, %int128_33567 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36135 = torch.aten.view %36133, %36134 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33568 = torch.constant.int 4
%36136 = torch.aten.mul.int %int4_33568, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33569 = torch.constant.int 4096
%36137 = torch.prim.ListConstruct %36136, %int4096_33569 : (!torch.int, !torch.int) -> !torch.list<int>
%36138 = torch.aten.view %35969, %36137 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36138, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36139 = torch.aten.mm %36138, %36109 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36139, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33570 = torch.constant.int 4
%int128_33571 = torch.constant.int 128
%36140 = torch.prim.ListConstruct %int4_33570, %2482, %int128_33571 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36141 = torch.aten.view %36139, %36140 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33572 = torch.constant.int 4
%36142 = torch.aten.mul.int %int4_33572, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33573 = torch.constant.int 4096
%36143 = torch.prim.ListConstruct %36142, %int4096_33573 : (!torch.int, !torch.int) -> !torch.list<int>
%36144 = torch.aten.view %35970, %36143 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36144, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36145 = torch.aten.mm %36144, %36111 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36145, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33574 = torch.constant.int 4
%int128_33575 = torch.constant.int 128
%36146 = torch.prim.ListConstruct %int4_33574, %2482, %int128_33575 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36147 = torch.aten.view %36145, %36146 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33576 = torch.constant.int 4
%36148 = torch.aten.mul.int %int4_33576, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33577 = torch.constant.int 4096
%36149 = torch.prim.ListConstruct %36148, %int4096_33577 : (!torch.int, !torch.int) -> !torch.list<int>
%36150 = torch.aten.view %35971, %36149 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36150, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36151 = torch.aten.mm %36150, %36113 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36151, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33578 = torch.constant.int 4
%int128_33579 = torch.constant.int 128
%36152 = torch.prim.ListConstruct %int4_33578, %2482, %int128_33579 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36153 = torch.aten.view %36151, %36152 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33580 = torch.constant.int 4
%36154 = torch.aten.mul.int %int4_33580, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33581 = torch.constant.int 4096
%36155 = torch.prim.ListConstruct %36154, %int4096_33581 : (!torch.int, !torch.int) -> !torch.list<int>
%36156 = torch.aten.view %35972, %36155 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36156, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36157 = torch.aten.mm %36156, %36115 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36157, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33582 = torch.constant.int 4
%int128_33583 = torch.constant.int 128
%36158 = torch.prim.ListConstruct %int4_33582, %2482, %int128_33583 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36159 = torch.aten.view %36157, %36158 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33584 = torch.constant.int 4
%36160 = torch.aten.mul.int %int4_33584, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_33585 = torch.constant.int 4096
%36161 = torch.prim.ListConstruct %36160, %int4096_33585 : (!torch.int, !torch.int) -> !torch.list<int>
%36162 = torch.aten.view %35973, %36161 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36162, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%36163 = torch.aten.mm %36162, %36117 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %36163, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_33586 = torch.constant.int 4
%int128_33587 = torch.constant.int 128
%36164 = torch.prim.ListConstruct %int4_33586, %2482, %int128_33587 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36165 = torch.aten.view %36163, %36164 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %36165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_33588 = torch.constant.int 4
%int4_33589 = torch.constant.int 4
%int128_33590 = torch.constant.int 128
%36166 = torch.prim.ListConstruct %int4_33588, %2482, %int4_33589, %int128_33590 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36167 = torch.aten.view %35995, %36166 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33591 = torch.constant.int 4
%int4_33592 = torch.constant.int 4
%int128_33593 = torch.constant.int 128
%36168 = torch.prim.ListConstruct %int4_33591, %2482, %int4_33592, %int128_33593 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36169 = torch.aten.view %36001, %36168 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33594 = torch.constant.int 4
%int4_33595 = torch.constant.int 4
%int128_33596 = torch.constant.int 128
%36170 = torch.prim.ListConstruct %int4_33594, %2482, %int4_33595, %int128_33596 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36171 = torch.aten.view %36007, %36170 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33597 = torch.constant.int 4
%int4_33598 = torch.constant.int 4
%int128_33599 = torch.constant.int 128
%36172 = torch.prim.ListConstruct %int4_33597, %2482, %int4_33598, %int128_33599 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36173 = torch.aten.view %36013, %36172 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33600 = torch.constant.int 4
%int4_33601 = torch.constant.int 4
%int128_33602 = torch.constant.int 128
%36174 = torch.prim.ListConstruct %int4_33600, %2482, %int4_33601, %int128_33602 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36175 = torch.aten.view %36019, %36174 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33603 = torch.constant.int 4
%int4_33604 = torch.constant.int 4
%int128_33605 = torch.constant.int 128
%36176 = torch.prim.ListConstruct %int4_33603, %2482, %int4_33604, %int128_33605 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36177 = torch.aten.view %36025, %36176 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33606 = torch.constant.int 4
%int4_33607 = torch.constant.int 4
%int128_33608 = torch.constant.int 128
%36178 = torch.prim.ListConstruct %int4_33606, %2482, %int4_33607, %int128_33608 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36179 = torch.aten.view %36031, %36178 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36179, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33609 = torch.constant.int 4
%int4_33610 = torch.constant.int 4
%int128_33611 = torch.constant.int 128
%36180 = torch.prim.ListConstruct %int4_33609, %2482, %int4_33610, %int128_33611 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36181 = torch.aten.view %36037, %36180 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_33612 = torch.constant.int 4
%int1_33613 = torch.constant.int 1
%int128_33614 = torch.constant.int 128
%36182 = torch.prim.ListConstruct %int4_33612, %2482, %int1_33613, %int128_33614 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36183 = torch.aten.view %36059, %36182 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33615 = torch.constant.int 4
%int1_33616 = torch.constant.int 1
%int128_33617 = torch.constant.int 128
%36184 = torch.prim.ListConstruct %int4_33615, %2482, %int1_33616, %int128_33617 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36185 = torch.aten.view %36065, %36184 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36185, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33618 = torch.constant.int 4
%int1_33619 = torch.constant.int 1
%int128_33620 = torch.constant.int 128
%36186 = torch.prim.ListConstruct %int4_33618, %2482, %int1_33619, %int128_33620 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36187 = torch.aten.view %36071, %36186 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33621 = torch.constant.int 4
%int1_33622 = torch.constant.int 1
%int128_33623 = torch.constant.int 128
%36188 = torch.prim.ListConstruct %int4_33621, %2482, %int1_33622, %int128_33623 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36189 = torch.aten.view %36077, %36188 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36189, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33624 = torch.constant.int 4
%int1_33625 = torch.constant.int 1
%int128_33626 = torch.constant.int 128
%36190 = torch.prim.ListConstruct %int4_33624, %2482, %int1_33625, %int128_33626 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36191 = torch.aten.view %36083, %36190 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36191, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33627 = torch.constant.int 4
%int1_33628 = torch.constant.int 1
%int128_33629 = torch.constant.int 128
%36192 = torch.prim.ListConstruct %int4_33627, %2482, %int1_33628, %int128_33629 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36193 = torch.aten.view %36089, %36192 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33630 = torch.constant.int 4
%int1_33631 = torch.constant.int 1
%int128_33632 = torch.constant.int 128
%36194 = torch.prim.ListConstruct %int4_33630, %2482, %int1_33631, %int128_33632 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36195 = torch.aten.view %36095, %36194 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36195, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33633 = torch.constant.int 4
%int1_33634 = torch.constant.int 1
%int128_33635 = torch.constant.int 128
%36196 = torch.prim.ListConstruct %int4_33633, %2482, %int1_33634, %int128_33635 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36197 = torch.aten.view %36101, %36196 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33636 = torch.constant.int 4
%int1_33637 = torch.constant.int 1
%int128_33638 = torch.constant.int 128
%36198 = torch.prim.ListConstruct %int4_33636, %2482, %int1_33637, %int128_33638 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36199 = torch.aten.view %36123, %36198 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33639 = torch.constant.int 4
%int1_33640 = torch.constant.int 1
%int128_33641 = torch.constant.int 128
%36200 = torch.prim.ListConstruct %int4_33639, %2482, %int1_33640, %int128_33641 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36201 = torch.aten.view %36129, %36200 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33642 = torch.constant.int 4
%int1_33643 = torch.constant.int 1
%int128_33644 = torch.constant.int 128
%36202 = torch.prim.ListConstruct %int4_33642, %2482, %int1_33643, %int128_33644 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36203 = torch.aten.view %36135, %36202 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33645 = torch.constant.int 4
%int1_33646 = torch.constant.int 1
%int128_33647 = torch.constant.int 128
%36204 = torch.prim.ListConstruct %int4_33645, %2482, %int1_33646, %int128_33647 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36205 = torch.aten.view %36141, %36204 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33648 = torch.constant.int 4
%int1_33649 = torch.constant.int 1
%int128_33650 = torch.constant.int 128
%36206 = torch.prim.ListConstruct %int4_33648, %2482, %int1_33649, %int128_33650 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36207 = torch.aten.view %36147, %36206 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36207, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33651 = torch.constant.int 4
%int1_33652 = torch.constant.int 1
%int128_33653 = torch.constant.int 128
%36208 = torch.prim.ListConstruct %int4_33651, %2482, %int1_33652, %int128_33653 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36209 = torch.aten.view %36153, %36208 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33654 = torch.constant.int 4
%int1_33655 = torch.constant.int 1
%int128_33656 = torch.constant.int 128
%36210 = torch.prim.ListConstruct %int4_33654, %2482, %int1_33655, %int128_33656 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36211 = torch.aten.view %36159, %36210 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36211, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_33657 = torch.constant.int 4
%int1_33658 = torch.constant.int 1
%int128_33659 = torch.constant.int 128
%36212 = torch.prim.ListConstruct %int4_33657, %2482, %int1_33658, %int128_33659 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36213 = torch.aten.view %36165, %36212 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_33660 = torch.constant.int 131072
%none_33661 = torch.constant.none
%none_33662 = torch.constant.none
%cpu_33663 = torch.constant.device "cpu"
%false_33664 = torch.constant.bool false
%36214 = torch.aten.arange %int131072_33660, %none_33661, %none_33662, %cpu_33663, %false_33664 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_33665 = torch.constant.int 0
%int128_33666 = torch.constant.int 128
%int2_33667 = torch.constant.int 2
%none_33668 = torch.constant.none
%none_33669 = torch.constant.none
%cpu_33670 = torch.constant.device "cpu"
%false_33671 = torch.constant.bool false
%36215 = torch.aten.arange.start_step %int0_33665, %int128_33666, %int2_33667, %none_33668, %none_33669, %cpu_33670, %false_33671 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_33672 = torch.constant.int 0
%int0_33673 = torch.constant.int 0
%int64_33674 = torch.constant.int 64
%int1_33675 = torch.constant.int 1
%36216 = torch.aten.slice.Tensor %36215, %int0_33672, %int0_33673, %int64_33674, %int1_33675 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_33676 = torch.constant.int 6
%36217 = torch.prims.convert_element_type %36216, %int6_33676 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_33677 = torch.constant.int 128
%36218 = torch.aten.div.Scalar %36217, %int128_33677 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_33678 = torch.constant.float 5.000000e+05
%36219 = torch.aten.pow.Scalar %float5.000000e05_33678, %36218 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%36220 = torch.aten.reciprocal %36219 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_33679 = torch.constant.float 1.000000e+00
%36221 = torch.aten.mul.Scalar %36220, %float1.000000e00_33679 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_33680 = torch.constant.int 131072
%int1_33681 = torch.constant.int 1
%36222 = torch.prim.ListConstruct %int131072_33680, %int1_33681 : (!torch.int, !torch.int) -> !torch.list<int>
%36223 = torch.aten.view %36214, %36222 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%36224 = torch.aten.mul.Tensor %36223, %36221 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%36225 = torch.aten.cos %36224 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%36226 = torch.aten.sin %36224 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%36227 = torch.aten.complex %36225, %36226 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%36228 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36229 = flow.tensor.transfer %36228 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%36230 = torch_c.from_builtin_tensor %36229 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36231 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36232 = flow.tensor.transfer %36231 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%36233 = torch_c.from_builtin_tensor %36232 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36234 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36235 = flow.tensor.transfer %36234 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%36236 = torch_c.from_builtin_tensor %36235 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36237 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36238 = flow.tensor.transfer %36237 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%36239 = torch_c.from_builtin_tensor %36238 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36240 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36241 = flow.tensor.transfer %36240 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%36242 = torch_c.from_builtin_tensor %36241 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36243 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36244 = flow.tensor.transfer %36243 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%36245 = torch_c.from_builtin_tensor %36244 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36246 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36247 = flow.tensor.transfer %36246 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%36248 = torch_c.from_builtin_tensor %36247 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36249 = torch_c.to_builtin_tensor %36227 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36250 = flow.tensor.transfer %36249 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%36251 = torch_c.from_builtin_tensor %36250 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_33682 = torch.constant.int 1
%36252 = torch.aten.size.int %35995, %int1_33682 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33683 = torch.constant.int 0
%36253 = torch.aten.add.int %int0_33683, %36252 : !torch.int, !torch.int -> !torch.int
%int0_33684 = torch.constant.int 0
%int0_33685 = torch.constant.int 0
%int1_33686 = torch.constant.int 1
%36254 = torch.aten.slice.Tensor %36230, %int0_33684, %int0_33685, %36253, %int1_33686 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36254, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33687 = torch.constant.int 1
%int0_33688 = torch.constant.int 0
%int9223372036854775807_33689 = torch.constant.int 9223372036854775807
%int1_33690 = torch.constant.int 1
%36255 = torch.aten.slice.Tensor %36254, %int1_33687, %int0_33688, %int9223372036854775807_33689, %int1_33690 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36255, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33691 = torch.constant.int 0
%36256 = torch.aten.unsqueeze %36255, %int0_33691 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36256, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33692 = torch.constant.int 2
%36257 = torch.aten.unsqueeze %36256, %int2_33692 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36257, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33693 = torch.constant.int 3
%int0_33694 = torch.constant.int 0
%int9223372036854775807_33695 = torch.constant.int 9223372036854775807
%int1_33696 = torch.constant.int 1
%36258 = torch.aten.slice.Tensor %36257, %int3_33693, %int0_33694, %int9223372036854775807_33695, %int1_33696 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36258, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36259 = torch_c.to_builtin_tensor %36167 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33697 = arith.constant 1 : index
%dim_33698 = tensor.dim %36259, %c1_33697 : tensor<4x?x4x128xf16>
%36260 = flow.tensor.bitcast %36259 : tensor<4x?x4x128xf16>{%dim_33698} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33698}
%36261 = torch_c.from_builtin_tensor %36260 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36261, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36262 = torch.aten.mul.Tensor %36261, %36258 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36263 = torch_c.to_builtin_tensor %36262 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33699 = arith.constant 1 : index
%dim_33700 = tensor.dim %36263, %c1_33699 : tensor<4x?x4x64xcomplex<f32>>
%36264 = flow.tensor.bitcast %36263 : tensor<4x?x4x64xcomplex<f32>>{%dim_33700} -> tensor<4x?x4x128xf32>{%dim_33700}
%36265 = torch_c.from_builtin_tensor %36264 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36265, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33701 = torch.constant.int 5
%36266 = torch.prims.convert_element_type %36265, %int5_33701 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33702 = torch.constant.int 1
%36267 = torch.aten.size.int %36001, %int1_33702 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33703 = torch.constant.int 0
%36268 = torch.aten.add.int %int0_33703, %36267 : !torch.int, !torch.int -> !torch.int
%int0_33704 = torch.constant.int 0
%int0_33705 = torch.constant.int 0
%int1_33706 = torch.constant.int 1
%36269 = torch.aten.slice.Tensor %36233, %int0_33704, %int0_33705, %36268, %int1_33706 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36269, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33707 = torch.constant.int 1
%int0_33708 = torch.constant.int 0
%int9223372036854775807_33709 = torch.constant.int 9223372036854775807
%int1_33710 = torch.constant.int 1
%36270 = torch.aten.slice.Tensor %36269, %int1_33707, %int0_33708, %int9223372036854775807_33709, %int1_33710 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36270, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33711 = torch.constant.int 0
%36271 = torch.aten.unsqueeze %36270, %int0_33711 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36271, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33712 = torch.constant.int 2
%36272 = torch.aten.unsqueeze %36271, %int2_33712 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36272, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33713 = torch.constant.int 3
%int0_33714 = torch.constant.int 0
%int9223372036854775807_33715 = torch.constant.int 9223372036854775807
%int1_33716 = torch.constant.int 1
%36273 = torch.aten.slice.Tensor %36272, %int3_33713, %int0_33714, %int9223372036854775807_33715, %int1_33716 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36273, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36274 = torch_c.to_builtin_tensor %36169 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33717 = arith.constant 1 : index
%dim_33718 = tensor.dim %36274, %c1_33717 : tensor<4x?x4x128xf16>
%36275 = flow.tensor.bitcast %36274 : tensor<4x?x4x128xf16>{%dim_33718} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33718}
%36276 = torch_c.from_builtin_tensor %36275 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36277 = torch.aten.mul.Tensor %36276, %36273 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36278 = torch_c.to_builtin_tensor %36277 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33719 = arith.constant 1 : index
%dim_33720 = tensor.dim %36278, %c1_33719 : tensor<4x?x4x64xcomplex<f32>>
%36279 = flow.tensor.bitcast %36278 : tensor<4x?x4x64xcomplex<f32>>{%dim_33720} -> tensor<4x?x4x128xf32>{%dim_33720}
%36280 = torch_c.from_builtin_tensor %36279 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33721 = torch.constant.int 5
%36281 = torch.prims.convert_element_type %36280, %int5_33721 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33722 = torch.constant.int 1
%36282 = torch.aten.size.int %36007, %int1_33722 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33723 = torch.constant.int 0
%36283 = torch.aten.add.int %int0_33723, %36282 : !torch.int, !torch.int -> !torch.int
%int0_33724 = torch.constant.int 0
%int0_33725 = torch.constant.int 0
%int1_33726 = torch.constant.int 1
%36284 = torch.aten.slice.Tensor %36236, %int0_33724, %int0_33725, %36283, %int1_33726 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36284, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33727 = torch.constant.int 1
%int0_33728 = torch.constant.int 0
%int9223372036854775807_33729 = torch.constant.int 9223372036854775807
%int1_33730 = torch.constant.int 1
%36285 = torch.aten.slice.Tensor %36284, %int1_33727, %int0_33728, %int9223372036854775807_33729, %int1_33730 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36285, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33731 = torch.constant.int 0
%36286 = torch.aten.unsqueeze %36285, %int0_33731 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36286, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33732 = torch.constant.int 2
%36287 = torch.aten.unsqueeze %36286, %int2_33732 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36287, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33733 = torch.constant.int 3
%int0_33734 = torch.constant.int 0
%int9223372036854775807_33735 = torch.constant.int 9223372036854775807
%int1_33736 = torch.constant.int 1
%36288 = torch.aten.slice.Tensor %36287, %int3_33733, %int0_33734, %int9223372036854775807_33735, %int1_33736 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36288, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36289 = torch_c.to_builtin_tensor %36171 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33737 = arith.constant 1 : index
%dim_33738 = tensor.dim %36289, %c1_33737 : tensor<4x?x4x128xf16>
%36290 = flow.tensor.bitcast %36289 : tensor<4x?x4x128xf16>{%dim_33738} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33738}
%36291 = torch_c.from_builtin_tensor %36290 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36292 = torch.aten.mul.Tensor %36291, %36288 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36293 = torch_c.to_builtin_tensor %36292 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33739 = arith.constant 1 : index
%dim_33740 = tensor.dim %36293, %c1_33739 : tensor<4x?x4x64xcomplex<f32>>
%36294 = flow.tensor.bitcast %36293 : tensor<4x?x4x64xcomplex<f32>>{%dim_33740} -> tensor<4x?x4x128xf32>{%dim_33740}
%36295 = torch_c.from_builtin_tensor %36294 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33741 = torch.constant.int 5
%36296 = torch.prims.convert_element_type %36295, %int5_33741 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33742 = torch.constant.int 1
%36297 = torch.aten.size.int %36013, %int1_33742 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33743 = torch.constant.int 0
%36298 = torch.aten.add.int %int0_33743, %36297 : !torch.int, !torch.int -> !torch.int
%int0_33744 = torch.constant.int 0
%int0_33745 = torch.constant.int 0
%int1_33746 = torch.constant.int 1
%36299 = torch.aten.slice.Tensor %36239, %int0_33744, %int0_33745, %36298, %int1_33746 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36299, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33747 = torch.constant.int 1
%int0_33748 = torch.constant.int 0
%int9223372036854775807_33749 = torch.constant.int 9223372036854775807
%int1_33750 = torch.constant.int 1
%36300 = torch.aten.slice.Tensor %36299, %int1_33747, %int0_33748, %int9223372036854775807_33749, %int1_33750 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36300, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33751 = torch.constant.int 0
%36301 = torch.aten.unsqueeze %36300, %int0_33751 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36301, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33752 = torch.constant.int 2
%36302 = torch.aten.unsqueeze %36301, %int2_33752 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36302, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33753 = torch.constant.int 3
%int0_33754 = torch.constant.int 0
%int9223372036854775807_33755 = torch.constant.int 9223372036854775807
%int1_33756 = torch.constant.int 1
%36303 = torch.aten.slice.Tensor %36302, %int3_33753, %int0_33754, %int9223372036854775807_33755, %int1_33756 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36303, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36304 = torch_c.to_builtin_tensor %36173 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33757 = arith.constant 1 : index
%dim_33758 = tensor.dim %36304, %c1_33757 : tensor<4x?x4x128xf16>
%36305 = flow.tensor.bitcast %36304 : tensor<4x?x4x128xf16>{%dim_33758} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33758}
%36306 = torch_c.from_builtin_tensor %36305 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36307 = torch.aten.mul.Tensor %36306, %36303 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36308 = torch_c.to_builtin_tensor %36307 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33759 = arith.constant 1 : index
%dim_33760 = tensor.dim %36308, %c1_33759 : tensor<4x?x4x64xcomplex<f32>>
%36309 = flow.tensor.bitcast %36308 : tensor<4x?x4x64xcomplex<f32>>{%dim_33760} -> tensor<4x?x4x128xf32>{%dim_33760}
%36310 = torch_c.from_builtin_tensor %36309 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33761 = torch.constant.int 5
%36311 = torch.prims.convert_element_type %36310, %int5_33761 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33762 = torch.constant.int 1
%36312 = torch.aten.size.int %36019, %int1_33762 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33763 = torch.constant.int 0
%36313 = torch.aten.add.int %int0_33763, %36312 : !torch.int, !torch.int -> !torch.int
%int0_33764 = torch.constant.int 0
%int0_33765 = torch.constant.int 0
%int1_33766 = torch.constant.int 1
%36314 = torch.aten.slice.Tensor %36242, %int0_33764, %int0_33765, %36313, %int1_33766 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36314, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33767 = torch.constant.int 1
%int0_33768 = torch.constant.int 0
%int9223372036854775807_33769 = torch.constant.int 9223372036854775807
%int1_33770 = torch.constant.int 1
%36315 = torch.aten.slice.Tensor %36314, %int1_33767, %int0_33768, %int9223372036854775807_33769, %int1_33770 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36315, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33771 = torch.constant.int 0
%36316 = torch.aten.unsqueeze %36315, %int0_33771 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36316, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33772 = torch.constant.int 2
%36317 = torch.aten.unsqueeze %36316, %int2_33772 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36317, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33773 = torch.constant.int 3
%int0_33774 = torch.constant.int 0
%int9223372036854775807_33775 = torch.constant.int 9223372036854775807
%int1_33776 = torch.constant.int 1
%36318 = torch.aten.slice.Tensor %36317, %int3_33773, %int0_33774, %int9223372036854775807_33775, %int1_33776 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36318, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36319 = torch_c.to_builtin_tensor %36175 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33777 = arith.constant 1 : index
%dim_33778 = tensor.dim %36319, %c1_33777 : tensor<4x?x4x128xf16>
%36320 = flow.tensor.bitcast %36319 : tensor<4x?x4x128xf16>{%dim_33778} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33778}
%36321 = torch_c.from_builtin_tensor %36320 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36321, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36322 = torch.aten.mul.Tensor %36321, %36318 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36322, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36323 = torch_c.to_builtin_tensor %36322 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33779 = arith.constant 1 : index
%dim_33780 = tensor.dim %36323, %c1_33779 : tensor<4x?x4x64xcomplex<f32>>
%36324 = flow.tensor.bitcast %36323 : tensor<4x?x4x64xcomplex<f32>>{%dim_33780} -> tensor<4x?x4x128xf32>{%dim_33780}
%36325 = torch_c.from_builtin_tensor %36324 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33781 = torch.constant.int 5
%36326 = torch.prims.convert_element_type %36325, %int5_33781 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33782 = torch.constant.int 1
%36327 = torch.aten.size.int %36025, %int1_33782 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33783 = torch.constant.int 0
%36328 = torch.aten.add.int %int0_33783, %36327 : !torch.int, !torch.int -> !torch.int
%int0_33784 = torch.constant.int 0
%int0_33785 = torch.constant.int 0
%int1_33786 = torch.constant.int 1
%36329 = torch.aten.slice.Tensor %36245, %int0_33784, %int0_33785, %36328, %int1_33786 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36329, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33787 = torch.constant.int 1
%int0_33788 = torch.constant.int 0
%int9223372036854775807_33789 = torch.constant.int 9223372036854775807
%int1_33790 = torch.constant.int 1
%36330 = torch.aten.slice.Tensor %36329, %int1_33787, %int0_33788, %int9223372036854775807_33789, %int1_33790 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36330, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33791 = torch.constant.int 0
%36331 = torch.aten.unsqueeze %36330, %int0_33791 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36331, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33792 = torch.constant.int 2
%36332 = torch.aten.unsqueeze %36331, %int2_33792 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36332, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33793 = torch.constant.int 3
%int0_33794 = torch.constant.int 0
%int9223372036854775807_33795 = torch.constant.int 9223372036854775807
%int1_33796 = torch.constant.int 1
%36333 = torch.aten.slice.Tensor %36332, %int3_33793, %int0_33794, %int9223372036854775807_33795, %int1_33796 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36333, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36334 = torch_c.to_builtin_tensor %36177 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33797 = arith.constant 1 : index
%dim_33798 = tensor.dim %36334, %c1_33797 : tensor<4x?x4x128xf16>
%36335 = flow.tensor.bitcast %36334 : tensor<4x?x4x128xf16>{%dim_33798} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33798}
%36336 = torch_c.from_builtin_tensor %36335 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36337 = torch.aten.mul.Tensor %36336, %36333 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36337, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36338 = torch_c.to_builtin_tensor %36337 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33799 = arith.constant 1 : index
%dim_33800 = tensor.dim %36338, %c1_33799 : tensor<4x?x4x64xcomplex<f32>>
%36339 = flow.tensor.bitcast %36338 : tensor<4x?x4x64xcomplex<f32>>{%dim_33800} -> tensor<4x?x4x128xf32>{%dim_33800}
%36340 = torch_c.from_builtin_tensor %36339 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33801 = torch.constant.int 5
%36341 = torch.prims.convert_element_type %36340, %int5_33801 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33802 = torch.constant.int 1
%36342 = torch.aten.size.int %36031, %int1_33802 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33803 = torch.constant.int 0
%36343 = torch.aten.add.int %int0_33803, %36342 : !torch.int, !torch.int -> !torch.int
%int0_33804 = torch.constant.int 0
%int0_33805 = torch.constant.int 0
%int1_33806 = torch.constant.int 1
%36344 = torch.aten.slice.Tensor %36248, %int0_33804, %int0_33805, %36343, %int1_33806 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36344, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33807 = torch.constant.int 1
%int0_33808 = torch.constant.int 0
%int9223372036854775807_33809 = torch.constant.int 9223372036854775807
%int1_33810 = torch.constant.int 1
%36345 = torch.aten.slice.Tensor %36344, %int1_33807, %int0_33808, %int9223372036854775807_33809, %int1_33810 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36345, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33811 = torch.constant.int 0
%36346 = torch.aten.unsqueeze %36345, %int0_33811 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36346, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33812 = torch.constant.int 2
%36347 = torch.aten.unsqueeze %36346, %int2_33812 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36347, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33813 = torch.constant.int 3
%int0_33814 = torch.constant.int 0
%int9223372036854775807_33815 = torch.constant.int 9223372036854775807
%int1_33816 = torch.constant.int 1
%36348 = torch.aten.slice.Tensor %36347, %int3_33813, %int0_33814, %int9223372036854775807_33815, %int1_33816 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36348, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36349 = torch_c.to_builtin_tensor %36179 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33817 = arith.constant 1 : index
%dim_33818 = tensor.dim %36349, %c1_33817 : tensor<4x?x4x128xf16>
%36350 = flow.tensor.bitcast %36349 : tensor<4x?x4x128xf16>{%dim_33818} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33818}
%36351 = torch_c.from_builtin_tensor %36350 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36351, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36352 = torch.aten.mul.Tensor %36351, %36348 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36352, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36353 = torch_c.to_builtin_tensor %36352 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33819 = arith.constant 1 : index
%dim_33820 = tensor.dim %36353, %c1_33819 : tensor<4x?x4x64xcomplex<f32>>
%36354 = flow.tensor.bitcast %36353 : tensor<4x?x4x64xcomplex<f32>>{%dim_33820} -> tensor<4x?x4x128xf32>{%dim_33820}
%36355 = torch_c.from_builtin_tensor %36354 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33821 = torch.constant.int 5
%36356 = torch.prims.convert_element_type %36355, %int5_33821 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_33822 = torch.constant.int 1
%36357 = torch.aten.size.int %36037, %int1_33822 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_33823 = torch.constant.int 0
%36358 = torch.aten.add.int %int0_33823, %36357 : !torch.int, !torch.int -> !torch.int
%int0_33824 = torch.constant.int 0
%int0_33825 = torch.constant.int 0
%int1_33826 = torch.constant.int 1
%36359 = torch.aten.slice.Tensor %36251, %int0_33824, %int0_33825, %36358, %int1_33826 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36359, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33827 = torch.constant.int 1
%int0_33828 = torch.constant.int 0
%int9223372036854775807_33829 = torch.constant.int 9223372036854775807
%int1_33830 = torch.constant.int 1
%36360 = torch.aten.slice.Tensor %36359, %int1_33827, %int0_33828, %int9223372036854775807_33829, %int1_33830 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36360, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33831 = torch.constant.int 0
%36361 = torch.aten.unsqueeze %36360, %int0_33831 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36361, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33832 = torch.constant.int 2
%36362 = torch.aten.unsqueeze %36361, %int2_33832 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36362, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33833 = torch.constant.int 3
%int0_33834 = torch.constant.int 0
%int9223372036854775807_33835 = torch.constant.int 9223372036854775807
%int1_33836 = torch.constant.int 1
%36363 = torch.aten.slice.Tensor %36362, %int3_33833, %int0_33834, %int9223372036854775807_33835, %int1_33836 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36363, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36364 = torch_c.to_builtin_tensor %36181 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_33837 = arith.constant 1 : index
%dim_33838 = tensor.dim %36364, %c1_33837 : tensor<4x?x4x128xf16>
%36365 = flow.tensor.bitcast %36364 : tensor<4x?x4x128xf16>{%dim_33838} -> tensor<4x?x4x64xcomplex<f16>>{%dim_33838}
%36366 = torch_c.from_builtin_tensor %36365 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %36366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%36367 = torch.aten.mul.Tensor %36366, %36363 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %36367, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%36368 = torch_c.to_builtin_tensor %36367 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_33839 = arith.constant 1 : index
%dim_33840 = tensor.dim %36368, %c1_33839 : tensor<4x?x4x64xcomplex<f32>>
%36369 = flow.tensor.bitcast %36368 : tensor<4x?x4x64xcomplex<f32>>{%dim_33840} -> tensor<4x?x4x128xf32>{%dim_33840}
%36370 = torch_c.from_builtin_tensor %36369 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %36370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_33841 = torch.constant.int 5
%36371 = torch.prims.convert_element_type %36370, %int5_33841 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_33842 = torch.constant.int 131072
%none_33843 = torch.constant.none
%none_33844 = torch.constant.none
%cpu_33845 = torch.constant.device "cpu"
%false_33846 = torch.constant.bool false
%36372 = torch.aten.arange %int131072_33842, %none_33843, %none_33844, %cpu_33845, %false_33846 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_33847 = torch.constant.int 0
%int128_33848 = torch.constant.int 128
%int2_33849 = torch.constant.int 2
%none_33850 = torch.constant.none
%none_33851 = torch.constant.none
%cpu_33852 = torch.constant.device "cpu"
%false_33853 = torch.constant.bool false
%36373 = torch.aten.arange.start_step %int0_33847, %int128_33848, %int2_33849, %none_33850, %none_33851, %cpu_33852, %false_33853 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_33854 = torch.constant.int 0
%int0_33855 = torch.constant.int 0
%int64_33856 = torch.constant.int 64
%int1_33857 = torch.constant.int 1
%36374 = torch.aten.slice.Tensor %36373, %int0_33854, %int0_33855, %int64_33856, %int1_33857 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_33858 = torch.constant.int 6
%36375 = torch.prims.convert_element_type %36374, %int6_33858 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_33859 = torch.constant.int 128
%36376 = torch.aten.div.Scalar %36375, %int128_33859 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_33860 = torch.constant.float 5.000000e+05
%36377 = torch.aten.pow.Scalar %float5.000000e05_33860, %36376 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%36378 = torch.aten.reciprocal %36377 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_33861 = torch.constant.float 1.000000e+00
%36379 = torch.aten.mul.Scalar %36378, %float1.000000e00_33861 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_33862 = torch.constant.int 131072
%int1_33863 = torch.constant.int 1
%36380 = torch.prim.ListConstruct %int131072_33862, %int1_33863 : (!torch.int, !torch.int) -> !torch.list<int>
%36381 = torch.aten.view %36372, %36380 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%36382 = torch.aten.mul.Tensor %36381, %36379 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%36383 = torch.aten.cos %36382 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%36384 = torch.aten.sin %36382 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%36385 = torch.aten.complex %36383, %36384 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%36386 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36387 = flow.tensor.transfer %36386 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%36388 = torch_c.from_builtin_tensor %36387 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36389 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36390 = flow.tensor.transfer %36389 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%36391 = torch_c.from_builtin_tensor %36390 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36392 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36393 = flow.tensor.transfer %36392 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%36394 = torch_c.from_builtin_tensor %36393 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36395 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36396 = flow.tensor.transfer %36395 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%36397 = torch_c.from_builtin_tensor %36396 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36398 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36399 = flow.tensor.transfer %36398 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%36400 = torch_c.from_builtin_tensor %36399 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36401 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36402 = flow.tensor.transfer %36401 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%36403 = torch_c.from_builtin_tensor %36402 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36404 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36405 = flow.tensor.transfer %36404 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%36406 = torch_c.from_builtin_tensor %36405 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%36407 = torch_c.to_builtin_tensor %36385 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%36408 = flow.tensor.transfer %36407 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%36409 = torch_c.from_builtin_tensor %36408 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_33864 = torch.constant.int 1
%36410 = torch.aten.size.int %36059, %int1_33864 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33865 = torch.constant.int 0
%36411 = torch.aten.add.int %int0_33865, %36410 : !torch.int, !torch.int -> !torch.int
%int0_33866 = torch.constant.int 0
%int0_33867 = torch.constant.int 0
%int1_33868 = torch.constant.int 1
%36412 = torch.aten.slice.Tensor %36388, %int0_33866, %int0_33867, %36411, %int1_33868 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36412, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33869 = torch.constant.int 1
%int0_33870 = torch.constant.int 0
%int9223372036854775807_33871 = torch.constant.int 9223372036854775807
%int1_33872 = torch.constant.int 1
%36413 = torch.aten.slice.Tensor %36412, %int1_33869, %int0_33870, %int9223372036854775807_33871, %int1_33872 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36413, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33873 = torch.constant.int 0
%36414 = torch.aten.unsqueeze %36413, %int0_33873 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36414, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33874 = torch.constant.int 2
%36415 = torch.aten.unsqueeze %36414, %int2_33874 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36415, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33875 = torch.constant.int 3
%int0_33876 = torch.constant.int 0
%int9223372036854775807_33877 = torch.constant.int 9223372036854775807
%int1_33878 = torch.constant.int 1
%36416 = torch.aten.slice.Tensor %36415, %int3_33875, %int0_33876, %int9223372036854775807_33877, %int1_33878 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36416, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36417 = torch_c.to_builtin_tensor %36183 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33879 = arith.constant 1 : index
%dim_33880 = tensor.dim %36417, %c1_33879 : tensor<4x?x1x128xf16>
%36418 = flow.tensor.bitcast %36417 : tensor<4x?x1x128xf16>{%dim_33880} -> tensor<4x?x1x64xcomplex<f16>>{%dim_33880}
%36419 = torch_c.from_builtin_tensor %36418 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36420 = torch.aten.mul.Tensor %36419, %36416 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36421 = torch_c.to_builtin_tensor %36420 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_33881 = arith.constant 1 : index
%dim_33882 = tensor.dim %36421, %c1_33881 : tensor<4x?x1x64xcomplex<f32>>
%36422 = flow.tensor.bitcast %36421 : tensor<4x?x1x64xcomplex<f32>>{%dim_33882} -> tensor<4x?x1x128xf32>{%dim_33882}
%36423 = torch_c.from_builtin_tensor %36422 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36423, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_33883 = torch.constant.int 5
%36424 = torch.prims.convert_element_type %36423, %int5_33883 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36424, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_33884 = torch.constant.int 1
%36425 = torch.aten.size.int %36065, %int1_33884 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33885 = torch.constant.int 0
%36426 = torch.aten.add.int %int0_33885, %36425 : !torch.int, !torch.int -> !torch.int
%int0_33886 = torch.constant.int 0
%int0_33887 = torch.constant.int 0
%int1_33888 = torch.constant.int 1
%36427 = torch.aten.slice.Tensor %36391, %int0_33886, %int0_33887, %36426, %int1_33888 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36427, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33889 = torch.constant.int 1
%int0_33890 = torch.constant.int 0
%int9223372036854775807_33891 = torch.constant.int 9223372036854775807
%int1_33892 = torch.constant.int 1
%36428 = torch.aten.slice.Tensor %36427, %int1_33889, %int0_33890, %int9223372036854775807_33891, %int1_33892 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36428, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33893 = torch.constant.int 0
%36429 = torch.aten.unsqueeze %36428, %int0_33893 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36429, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33894 = torch.constant.int 2
%36430 = torch.aten.unsqueeze %36429, %int2_33894 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36430, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33895 = torch.constant.int 3
%int0_33896 = torch.constant.int 0
%int9223372036854775807_33897 = torch.constant.int 9223372036854775807
%int1_33898 = torch.constant.int 1
%36431 = torch.aten.slice.Tensor %36430, %int3_33895, %int0_33896, %int9223372036854775807_33897, %int1_33898 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36431, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36432 = torch_c.to_builtin_tensor %36185 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33899 = arith.constant 1 : index
%dim_33900 = tensor.dim %36432, %c1_33899 : tensor<4x?x1x128xf16>
%36433 = flow.tensor.bitcast %36432 : tensor<4x?x1x128xf16>{%dim_33900} -> tensor<4x?x1x64xcomplex<f16>>{%dim_33900}
%36434 = torch_c.from_builtin_tensor %36433 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36435 = torch.aten.mul.Tensor %36434, %36431 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36435, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36436 = torch_c.to_builtin_tensor %36435 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_33901 = arith.constant 1 : index
%dim_33902 = tensor.dim %36436, %c1_33901 : tensor<4x?x1x64xcomplex<f32>>
%36437 = flow.tensor.bitcast %36436 : tensor<4x?x1x64xcomplex<f32>>{%dim_33902} -> tensor<4x?x1x128xf32>{%dim_33902}
%36438 = torch_c.from_builtin_tensor %36437 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_33903 = torch.constant.int 5
%36439 = torch.prims.convert_element_type %36438, %int5_33903 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36439, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_33904 = torch.constant.int 1
%36440 = torch.aten.size.int %36071, %int1_33904 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33905 = torch.constant.int 0
%36441 = torch.aten.add.int %int0_33905, %36440 : !torch.int, !torch.int -> !torch.int
%int0_33906 = torch.constant.int 0
%int0_33907 = torch.constant.int 0
%int1_33908 = torch.constant.int 1
%36442 = torch.aten.slice.Tensor %36394, %int0_33906, %int0_33907, %36441, %int1_33908 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36442, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33909 = torch.constant.int 1
%int0_33910 = torch.constant.int 0
%int9223372036854775807_33911 = torch.constant.int 9223372036854775807
%int1_33912 = torch.constant.int 1
%36443 = torch.aten.slice.Tensor %36442, %int1_33909, %int0_33910, %int9223372036854775807_33911, %int1_33912 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36443, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33913 = torch.constant.int 0
%36444 = torch.aten.unsqueeze %36443, %int0_33913 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36444, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33914 = torch.constant.int 2
%36445 = torch.aten.unsqueeze %36444, %int2_33914 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36445, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33915 = torch.constant.int 3
%int0_33916 = torch.constant.int 0
%int9223372036854775807_33917 = torch.constant.int 9223372036854775807
%int1_33918 = torch.constant.int 1
%36446 = torch.aten.slice.Tensor %36445, %int3_33915, %int0_33916, %int9223372036854775807_33917, %int1_33918 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36446, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36447 = torch_c.to_builtin_tensor %36187 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33919 = arith.constant 1 : index
%dim_33920 = tensor.dim %36447, %c1_33919 : tensor<4x?x1x128xf16>
%36448 = flow.tensor.bitcast %36447 : tensor<4x?x1x128xf16>{%dim_33920} -> tensor<4x?x1x64xcomplex<f16>>{%dim_33920}
%36449 = torch_c.from_builtin_tensor %36448 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36449, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36450 = torch.aten.mul.Tensor %36449, %36446 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36451 = torch_c.to_builtin_tensor %36450 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_33921 = arith.constant 1 : index
%dim_33922 = tensor.dim %36451, %c1_33921 : tensor<4x?x1x64xcomplex<f32>>
%36452 = flow.tensor.bitcast %36451 : tensor<4x?x1x64xcomplex<f32>>{%dim_33922} -> tensor<4x?x1x128xf32>{%dim_33922}
%36453 = torch_c.from_builtin_tensor %36452 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_33923 = torch.constant.int 5
%36454 = torch.prims.convert_element_type %36453, %int5_33923 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_33924 = torch.constant.int 1
%36455 = torch.aten.size.int %36077, %int1_33924 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33925 = torch.constant.int 0
%36456 = torch.aten.add.int %int0_33925, %36455 : !torch.int, !torch.int -> !torch.int
%int0_33926 = torch.constant.int 0
%int0_33927 = torch.constant.int 0
%int1_33928 = torch.constant.int 1
%36457 = torch.aten.slice.Tensor %36397, %int0_33926, %int0_33927, %36456, %int1_33928 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36457, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33929 = torch.constant.int 1
%int0_33930 = torch.constant.int 0
%int9223372036854775807_33931 = torch.constant.int 9223372036854775807
%int1_33932 = torch.constant.int 1
%36458 = torch.aten.slice.Tensor %36457, %int1_33929, %int0_33930, %int9223372036854775807_33931, %int1_33932 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36458, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33933 = torch.constant.int 0
%36459 = torch.aten.unsqueeze %36458, %int0_33933 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36459, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33934 = torch.constant.int 2
%36460 = torch.aten.unsqueeze %36459, %int2_33934 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36460, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33935 = torch.constant.int 3
%int0_33936 = torch.constant.int 0
%int9223372036854775807_33937 = torch.constant.int 9223372036854775807
%int1_33938 = torch.constant.int 1
%36461 = torch.aten.slice.Tensor %36460, %int3_33935, %int0_33936, %int9223372036854775807_33937, %int1_33938 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36461, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36462 = torch_c.to_builtin_tensor %36189 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33939 = arith.constant 1 : index
%dim_33940 = tensor.dim %36462, %c1_33939 : tensor<4x?x1x128xf16>
%36463 = flow.tensor.bitcast %36462 : tensor<4x?x1x128xf16>{%dim_33940} -> tensor<4x?x1x64xcomplex<f16>>{%dim_33940}
%36464 = torch_c.from_builtin_tensor %36463 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36464, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36465 = torch.aten.mul.Tensor %36464, %36461 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36466 = torch_c.to_builtin_tensor %36465 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_33941 = arith.constant 1 : index
%dim_33942 = tensor.dim %36466, %c1_33941 : tensor<4x?x1x64xcomplex<f32>>
%36467 = flow.tensor.bitcast %36466 : tensor<4x?x1x64xcomplex<f32>>{%dim_33942} -> tensor<4x?x1x128xf32>{%dim_33942}
%36468 = torch_c.from_builtin_tensor %36467 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_33943 = torch.constant.int 5
%36469 = torch.prims.convert_element_type %36468, %int5_33943 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_33944 = torch.constant.int 1
%36470 = torch.aten.size.int %36083, %int1_33944 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33945 = torch.constant.int 0
%36471 = torch.aten.add.int %int0_33945, %36470 : !torch.int, !torch.int -> !torch.int
%int0_33946 = torch.constant.int 0
%int0_33947 = torch.constant.int 0
%int1_33948 = torch.constant.int 1
%36472 = torch.aten.slice.Tensor %36400, %int0_33946, %int0_33947, %36471, %int1_33948 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36472, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33949 = torch.constant.int 1
%int0_33950 = torch.constant.int 0
%int9223372036854775807_33951 = torch.constant.int 9223372036854775807
%int1_33952 = torch.constant.int 1
%36473 = torch.aten.slice.Tensor %36472, %int1_33949, %int0_33950, %int9223372036854775807_33951, %int1_33952 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36473, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33953 = torch.constant.int 0
%36474 = torch.aten.unsqueeze %36473, %int0_33953 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36474, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33954 = torch.constant.int 2
%36475 = torch.aten.unsqueeze %36474, %int2_33954 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36475, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33955 = torch.constant.int 3
%int0_33956 = torch.constant.int 0
%int9223372036854775807_33957 = torch.constant.int 9223372036854775807
%int1_33958 = torch.constant.int 1
%36476 = torch.aten.slice.Tensor %36475, %int3_33955, %int0_33956, %int9223372036854775807_33957, %int1_33958 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36476, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36477 = torch_c.to_builtin_tensor %36191 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33959 = arith.constant 1 : index
%dim_33960 = tensor.dim %36477, %c1_33959 : tensor<4x?x1x128xf16>
%36478 = flow.tensor.bitcast %36477 : tensor<4x?x1x128xf16>{%dim_33960} -> tensor<4x?x1x64xcomplex<f16>>{%dim_33960}
%36479 = torch_c.from_builtin_tensor %36478 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36479, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36480 = torch.aten.mul.Tensor %36479, %36476 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36480, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36481 = torch_c.to_builtin_tensor %36480 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_33961 = arith.constant 1 : index
%dim_33962 = tensor.dim %36481, %c1_33961 : tensor<4x?x1x64xcomplex<f32>>
%36482 = flow.tensor.bitcast %36481 : tensor<4x?x1x64xcomplex<f32>>{%dim_33962} -> tensor<4x?x1x128xf32>{%dim_33962}
%36483 = torch_c.from_builtin_tensor %36482 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36483, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_33963 = torch.constant.int 5
%36484 = torch.prims.convert_element_type %36483, %int5_33963 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_33964 = torch.constant.int 1
%36485 = torch.aten.size.int %36089, %int1_33964 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33965 = torch.constant.int 0
%36486 = torch.aten.add.int %int0_33965, %36485 : !torch.int, !torch.int -> !torch.int
%int0_33966 = torch.constant.int 0
%int0_33967 = torch.constant.int 0
%int1_33968 = torch.constant.int 1
%36487 = torch.aten.slice.Tensor %36403, %int0_33966, %int0_33967, %36486, %int1_33968 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36487, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33969 = torch.constant.int 1
%int0_33970 = torch.constant.int 0
%int9223372036854775807_33971 = torch.constant.int 9223372036854775807
%int1_33972 = torch.constant.int 1
%36488 = torch.aten.slice.Tensor %36487, %int1_33969, %int0_33970, %int9223372036854775807_33971, %int1_33972 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36488, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33973 = torch.constant.int 0
%36489 = torch.aten.unsqueeze %36488, %int0_33973 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36489, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33974 = torch.constant.int 2
%36490 = torch.aten.unsqueeze %36489, %int2_33974 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36490, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33975 = torch.constant.int 3
%int0_33976 = torch.constant.int 0
%int9223372036854775807_33977 = torch.constant.int 9223372036854775807
%int1_33978 = torch.constant.int 1
%36491 = torch.aten.slice.Tensor %36490, %int3_33975, %int0_33976, %int9223372036854775807_33977, %int1_33978 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36491, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36492 = torch_c.to_builtin_tensor %36193 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33979 = arith.constant 1 : index
%dim_33980 = tensor.dim %36492, %c1_33979 : tensor<4x?x1x128xf16>
%36493 = flow.tensor.bitcast %36492 : tensor<4x?x1x128xf16>{%dim_33980} -> tensor<4x?x1x64xcomplex<f16>>{%dim_33980}
%36494 = torch_c.from_builtin_tensor %36493 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36494, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36495 = torch.aten.mul.Tensor %36494, %36491 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36496 = torch_c.to_builtin_tensor %36495 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_33981 = arith.constant 1 : index
%dim_33982 = tensor.dim %36496, %c1_33981 : tensor<4x?x1x64xcomplex<f32>>
%36497 = flow.tensor.bitcast %36496 : tensor<4x?x1x64xcomplex<f32>>{%dim_33982} -> tensor<4x?x1x128xf32>{%dim_33982}
%36498 = torch_c.from_builtin_tensor %36497 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_33983 = torch.constant.int 5
%36499 = torch.prims.convert_element_type %36498, %int5_33983 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_33984 = torch.constant.int 1
%36500 = torch.aten.size.int %36095, %int1_33984 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_33985 = torch.constant.int 0
%36501 = torch.aten.add.int %int0_33985, %36500 : !torch.int, !torch.int -> !torch.int
%int0_33986 = torch.constant.int 0
%int0_33987 = torch.constant.int 0
%int1_33988 = torch.constant.int 1
%36502 = torch.aten.slice.Tensor %36406, %int0_33986, %int0_33987, %36501, %int1_33988 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36502, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_33989 = torch.constant.int 1
%int0_33990 = torch.constant.int 0
%int9223372036854775807_33991 = torch.constant.int 9223372036854775807
%int1_33992 = torch.constant.int 1
%36503 = torch.aten.slice.Tensor %36502, %int1_33989, %int0_33990, %int9223372036854775807_33991, %int1_33992 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36503, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_33993 = torch.constant.int 0
%36504 = torch.aten.unsqueeze %36503, %int0_33993 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36504, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_33994 = torch.constant.int 2
%36505 = torch.aten.unsqueeze %36504, %int2_33994 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36505, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_33995 = torch.constant.int 3
%int0_33996 = torch.constant.int 0
%int9223372036854775807_33997 = torch.constant.int 9223372036854775807
%int1_33998 = torch.constant.int 1
%36506 = torch.aten.slice.Tensor %36505, %int3_33995, %int0_33996, %int9223372036854775807_33997, %int1_33998 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36506, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36507 = torch_c.to_builtin_tensor %36195 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_33999 = arith.constant 1 : index
%dim_34000 = tensor.dim %36507, %c1_33999 : tensor<4x?x1x128xf16>
%36508 = flow.tensor.bitcast %36507 : tensor<4x?x1x128xf16>{%dim_34000} -> tensor<4x?x1x64xcomplex<f16>>{%dim_34000}
%36509 = torch_c.from_builtin_tensor %36508 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36510 = torch.aten.mul.Tensor %36509, %36506 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36510, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36511 = torch_c.to_builtin_tensor %36510 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_34001 = arith.constant 1 : index
%dim_34002 = tensor.dim %36511, %c1_34001 : tensor<4x?x1x64xcomplex<f32>>
%36512 = flow.tensor.bitcast %36511 : tensor<4x?x1x64xcomplex<f32>>{%dim_34002} -> tensor<4x?x1x128xf32>{%dim_34002}
%36513 = torch_c.from_builtin_tensor %36512 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36513, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_34003 = torch.constant.int 5
%36514 = torch.prims.convert_element_type %36513, %int5_34003 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36514, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_34004 = torch.constant.int 1
%36515 = torch.aten.size.int %36101, %int1_34004 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_34005 = torch.constant.int 0
%36516 = torch.aten.add.int %int0_34005, %36515 : !torch.int, !torch.int -> !torch.int
%int0_34006 = torch.constant.int 0
%int0_34007 = torch.constant.int 0
%int1_34008 = torch.constant.int 1
%36517 = torch.aten.slice.Tensor %36409, %int0_34006, %int0_34007, %36516, %int1_34008 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36517, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_34009 = torch.constant.int 1
%int0_34010 = torch.constant.int 0
%int9223372036854775807_34011 = torch.constant.int 9223372036854775807
%int1_34012 = torch.constant.int 1
%36518 = torch.aten.slice.Tensor %36517, %int1_34009, %int0_34010, %int9223372036854775807_34011, %int1_34012 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %36518, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_34013 = torch.constant.int 0
%36519 = torch.aten.unsqueeze %36518, %int0_34013 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %36519, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_34014 = torch.constant.int 2
%36520 = torch.aten.unsqueeze %36519, %int2_34014 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36520, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_34015 = torch.constant.int 3
%int0_34016 = torch.constant.int 0
%int9223372036854775807_34017 = torch.constant.int 9223372036854775807
%int1_34018 = torch.constant.int 1
%36521 = torch.aten.slice.Tensor %36520, %int3_34015, %int0_34016, %int9223372036854775807_34017, %int1_34018 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36521, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%36522 = torch_c.to_builtin_tensor %36197 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_34019 = arith.constant 1 : index
%dim_34020 = tensor.dim %36522, %c1_34019 : tensor<4x?x1x128xf16>
%36523 = flow.tensor.bitcast %36522 : tensor<4x?x1x128xf16>{%dim_34020} -> tensor<4x?x1x64xcomplex<f16>>{%dim_34020}
%36524 = torch_c.from_builtin_tensor %36523 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %36524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%36525 = torch.aten.mul.Tensor %36524, %36521 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %36525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%36526 = torch_c.to_builtin_tensor %36525 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_34021 = arith.constant 1 : index
%dim_34022 = tensor.dim %36526, %c1_34021 : tensor<4x?x1x64xcomplex<f32>>
%36527 = flow.tensor.bitcast %36526 : tensor<4x?x1x64xcomplex<f32>>{%dim_34022} -> tensor<4x?x1x128xf32>{%dim_34022}
%36528 = torch_c.from_builtin_tensor %36527 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %36528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_34023 = torch.constant.int 5
%36529 = torch.prims.convert_element_type %36528, %int5_34023 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %36529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_34024 = torch.constant.int 64
%36530 = torch.aten.mul.Scalar %2364, %int64_34024 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36530, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34025 = torch.constant.int 64
%36531 = torch.aten.mul.Scalar %2367, %int64_34025 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36531, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34026 = torch.constant.int 64
%36532 = torch.aten.mul.Scalar %2370, %int64_34026 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36532, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34027 = torch.constant.int 64
%36533 = torch.aten.mul.Scalar %2373, %int64_34027 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36533, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34028 = torch.constant.int 64
%36534 = torch.aten.mul.Scalar %2376, %int64_34028 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36534, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34029 = torch.constant.int 64
%36535 = torch.aten.mul.Scalar %2379, %int64_34029 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36535, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34030 = torch.constant.int 64
%36536 = torch.aten.mul.Scalar %2382, %int64_34030 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36536, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_34031 = torch.constant.int 64
%36537 = torch.aten.mul.Scalar %2385, %int64_34031 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36537, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36 = torch.constant.int 36
%int1_34032 = torch.constant.int 1
%36538 = torch.aten.add.Scalar %36530, %int36, %int1_34032 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36538, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34033 = torch.constant.int 36
%int1_34034 = torch.constant.int 1
%36539 = torch.aten.add.Scalar %36531, %int36_34033, %int1_34034 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36539, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34035 = torch.constant.int 36
%int1_34036 = torch.constant.int 1
%36540 = torch.aten.add.Scalar %36532, %int36_34035, %int1_34036 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36540, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34037 = torch.constant.int 36
%int1_34038 = torch.constant.int 1
%36541 = torch.aten.add.Scalar %36533, %int36_34037, %int1_34038 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36541, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34039 = torch.constant.int 36
%int1_34040 = torch.constant.int 1
%36542 = torch.aten.add.Scalar %36534, %int36_34039, %int1_34040 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36542, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34041 = torch.constant.int 36
%int1_34042 = torch.constant.int 1
%36543 = torch.aten.add.Scalar %36535, %int36_34041, %int1_34042 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36543, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34043 = torch.constant.int 36
%int1_34044 = torch.constant.int 1
%36544 = torch.aten.add.Scalar %36536, %int36_34043, %int1_34044 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36544, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int36_34045 = torch.constant.int 36
%int1_34046 = torch.constant.int 1
%36545 = torch.aten.add.Scalar %36537, %int36_34045, %int1_34046 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36545, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_34047 = torch.constant.int 4
%int16_34048 = torch.constant.int 16
%int1_34049 = torch.constant.int 1
%int128_34050 = torch.constant.int 128
%36546 = torch.prim.ListConstruct %int4_34047, %3095, %int16_34048, %int1_34049, %int128_34050 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36547 = torch.aten.view %36424, %36546 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36547, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34051 = torch.constant.int 4
%int16_34052 = torch.constant.int 16
%int1_34053 = torch.constant.int 1
%int128_34054 = torch.constant.int 128
%36548 = torch.prim.ListConstruct %int4_34051, %3095, %int16_34052, %int1_34053, %int128_34054 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36549 = torch.aten.view %36439, %36548 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36549, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34055 = torch.constant.int 4
%int16_34056 = torch.constant.int 16
%int1_34057 = torch.constant.int 1
%int128_34058 = torch.constant.int 128
%36550 = torch.prim.ListConstruct %int4_34055, %3095, %int16_34056, %int1_34057, %int128_34058 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36551 = torch.aten.view %36454, %36550 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36551, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34059 = torch.constant.int 4
%int16_34060 = torch.constant.int 16
%int1_34061 = torch.constant.int 1
%int128_34062 = torch.constant.int 128
%36552 = torch.prim.ListConstruct %int4_34059, %3095, %int16_34060, %int1_34061, %int128_34062 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36553 = torch.aten.view %36469, %36552 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36553, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34063 = torch.constant.int 4
%int16_34064 = torch.constant.int 16
%int1_34065 = torch.constant.int 1
%int128_34066 = torch.constant.int 128
%36554 = torch.prim.ListConstruct %int4_34063, %3095, %int16_34064, %int1_34065, %int128_34066 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36555 = torch.aten.view %36484, %36554 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36555, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34067 = torch.constant.int 4
%int16_34068 = torch.constant.int 16
%int1_34069 = torch.constant.int 1
%int128_34070 = torch.constant.int 128
%36556 = torch.prim.ListConstruct %int4_34067, %3095, %int16_34068, %int1_34069, %int128_34070 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36557 = torch.aten.view %36499, %36556 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36557, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34071 = torch.constant.int 4
%int16_34072 = torch.constant.int 16
%int1_34073 = torch.constant.int 1
%int128_34074 = torch.constant.int 128
%36558 = torch.prim.ListConstruct %int4_34071, %3095, %int16_34072, %int1_34073, %int128_34074 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36559 = torch.aten.view %36514, %36558 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36559, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34075 = torch.constant.int 4
%int16_34076 = torch.constant.int 16
%int1_34077 = torch.constant.int 1
%int128_34078 = torch.constant.int 128
%36560 = torch.prim.ListConstruct %int4_34075, %3095, %int16_34076, %int1_34077, %int128_34078 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36561 = torch.aten.view %36529, %36560 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36561, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34079 = torch.constant.int 4
%36562 = torch.aten.mul.int %int4_34079, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34080 = torch.constant.int 16
%int1_34081 = torch.constant.int 1
%int128_34082 = torch.constant.int 128
%36563 = torch.prim.ListConstruct %36562, %int16_34080, %int1_34081, %int128_34082 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36564 = torch.aten.view %36547, %36563 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36564, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34083 = torch.constant.int 4
%36565 = torch.aten.mul.int %int4_34083, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34084 = torch.constant.int 16
%int1_34085 = torch.constant.int 1
%int128_34086 = torch.constant.int 128
%36566 = torch.prim.ListConstruct %36565, %int16_34084, %int1_34085, %int128_34086 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36567 = torch.aten.view %36549, %36566 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36567, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34087 = torch.constant.int 4
%36568 = torch.aten.mul.int %int4_34087, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34088 = torch.constant.int 16
%int1_34089 = torch.constant.int 1
%int128_34090 = torch.constant.int 128
%36569 = torch.prim.ListConstruct %36568, %int16_34088, %int1_34089, %int128_34090 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36570 = torch.aten.view %36551, %36569 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36570, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34091 = torch.constant.int 4
%36571 = torch.aten.mul.int %int4_34091, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34092 = torch.constant.int 16
%int1_34093 = torch.constant.int 1
%int128_34094 = torch.constant.int 128
%36572 = torch.prim.ListConstruct %36571, %int16_34092, %int1_34093, %int128_34094 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36573 = torch.aten.view %36553, %36572 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36573, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34095 = torch.constant.int 4
%36574 = torch.aten.mul.int %int4_34095, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34096 = torch.constant.int 16
%int1_34097 = torch.constant.int 1
%int128_34098 = torch.constant.int 128
%36575 = torch.prim.ListConstruct %36574, %int16_34096, %int1_34097, %int128_34098 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36576 = torch.aten.view %36555, %36575 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36576, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34099 = torch.constant.int 4
%36577 = torch.aten.mul.int %int4_34099, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34100 = torch.constant.int 16
%int1_34101 = torch.constant.int 1
%int128_34102 = torch.constant.int 128
%36578 = torch.prim.ListConstruct %36577, %int16_34100, %int1_34101, %int128_34102 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36579 = torch.aten.view %36557, %36578 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36579, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34103 = torch.constant.int 4
%36580 = torch.aten.mul.int %int4_34103, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34104 = torch.constant.int 16
%int1_34105 = torch.constant.int 1
%int128_34106 = torch.constant.int 128
%36581 = torch.prim.ListConstruct %36580, %int16_34104, %int1_34105, %int128_34106 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36582 = torch.aten.view %36559, %36581 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36582, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34107 = torch.constant.int 4
%36583 = torch.aten.mul.int %int4_34107, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34108 = torch.constant.int 16
%int1_34109 = torch.constant.int 1
%int128_34110 = torch.constant.int 128
%36584 = torch.prim.ListConstruct %36583, %int16_34108, %int1_34109, %int128_34110 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36585 = torch.aten.view %36561, %36584 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36585, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34111 = torch.constant.int 4
%36586 = torch.aten.mul.int %int4_34111, %3095 : !torch.int, !torch.int -> !torch.int
%36587 = torch.prim.ListConstruct %36586 : (!torch.int) -> !torch.list<int>
%36588 = torch.aten.view %36538, %36587 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36588, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34112 = torch.constant.int 4
%36589 = torch.aten.mul.int %int4_34112, %3095 : !torch.int, !torch.int -> !torch.int
%36590 = torch.prim.ListConstruct %36589 : (!torch.int) -> !torch.list<int>
%36591 = torch.aten.view %36539, %36590 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36591, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34113 = torch.constant.int 4
%36592 = torch.aten.mul.int %int4_34113, %3095 : !torch.int, !torch.int -> !torch.int
%36593 = torch.prim.ListConstruct %36592 : (!torch.int) -> !torch.list<int>
%36594 = torch.aten.view %36540, %36593 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36594, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34114 = torch.constant.int 4
%36595 = torch.aten.mul.int %int4_34114, %3095 : !torch.int, !torch.int -> !torch.int
%36596 = torch.prim.ListConstruct %36595 : (!torch.int) -> !torch.list<int>
%36597 = torch.aten.view %36541, %36596 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36597, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34115 = torch.constant.int 4
%36598 = torch.aten.mul.int %int4_34115, %3095 : !torch.int, !torch.int -> !torch.int
%36599 = torch.prim.ListConstruct %36598 : (!torch.int) -> !torch.list<int>
%36600 = torch.aten.view %36542, %36599 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36600, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34116 = torch.constant.int 4
%36601 = torch.aten.mul.int %int4_34116, %3095 : !torch.int, !torch.int -> !torch.int
%36602 = torch.prim.ListConstruct %36601 : (!torch.int) -> !torch.list<int>
%36603 = torch.aten.view %36543, %36602 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36603, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34117 = torch.constant.int 4
%36604 = torch.aten.mul.int %int4_34117, %3095 : !torch.int, !torch.int -> !torch.int
%36605 = torch.prim.ListConstruct %36604 : (!torch.int) -> !torch.list<int>
%36606 = torch.aten.view %36544, %36605 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36606, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34118 = torch.constant.int 4
%36607 = torch.aten.mul.int %int4_34118, %3095 : !torch.int, !torch.int -> !torch.int
%36608 = torch.prim.ListConstruct %36607 : (!torch.int) -> !torch.list<int>
%36609 = torch.aten.view %36545, %36608 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36609, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34119 = torch.constant.int 4
%int16_34120 = torch.constant.int 16
%int1_34121 = torch.constant.int 1
%int128_34122 = torch.constant.int 128
%36610 = torch.prim.ListConstruct %int4_34119, %3095, %int16_34120, %int1_34121, %int128_34122 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36611 = torch.aten.view %36199, %36610 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36611, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34123 = torch.constant.int 4
%int16_34124 = torch.constant.int 16
%int1_34125 = torch.constant.int 1
%int128_34126 = torch.constant.int 128
%36612 = torch.prim.ListConstruct %int4_34123, %3095, %int16_34124, %int1_34125, %int128_34126 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36613 = torch.aten.view %36201, %36612 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36613, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34127 = torch.constant.int 4
%int16_34128 = torch.constant.int 16
%int1_34129 = torch.constant.int 1
%int128_34130 = torch.constant.int 128
%36614 = torch.prim.ListConstruct %int4_34127, %3095, %int16_34128, %int1_34129, %int128_34130 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36615 = torch.aten.view %36203, %36614 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36615, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34131 = torch.constant.int 4
%int16_34132 = torch.constant.int 16
%int1_34133 = torch.constant.int 1
%int128_34134 = torch.constant.int 128
%36616 = torch.prim.ListConstruct %int4_34131, %3095, %int16_34132, %int1_34133, %int128_34134 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36617 = torch.aten.view %36205, %36616 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36617, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34135 = torch.constant.int 4
%int16_34136 = torch.constant.int 16
%int1_34137 = torch.constant.int 1
%int128_34138 = torch.constant.int 128
%36618 = torch.prim.ListConstruct %int4_34135, %3095, %int16_34136, %int1_34137, %int128_34138 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36619 = torch.aten.view %36207, %36618 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36619, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34139 = torch.constant.int 4
%int16_34140 = torch.constant.int 16
%int1_34141 = torch.constant.int 1
%int128_34142 = torch.constant.int 128
%36620 = torch.prim.ListConstruct %int4_34139, %3095, %int16_34140, %int1_34141, %int128_34142 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36621 = torch.aten.view %36209, %36620 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36621, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34143 = torch.constant.int 4
%int16_34144 = torch.constant.int 16
%int1_34145 = torch.constant.int 1
%int128_34146 = torch.constant.int 128
%36622 = torch.prim.ListConstruct %int4_34143, %3095, %int16_34144, %int1_34145, %int128_34146 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36623 = torch.aten.view %36211, %36622 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36623, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34147 = torch.constant.int 4
%int16_34148 = torch.constant.int 16
%int1_34149 = torch.constant.int 1
%int128_34150 = torch.constant.int 128
%36624 = torch.prim.ListConstruct %int4_34147, %3095, %int16_34148, %int1_34149, %int128_34150 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36625 = torch.aten.view %36213, %36624 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %36625, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_34151 = torch.constant.int 4
%36626 = torch.aten.mul.int %int4_34151, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34152 = torch.constant.int 16
%int1_34153 = torch.constant.int 1
%int128_34154 = torch.constant.int 128
%36627 = torch.prim.ListConstruct %36626, %int16_34152, %int1_34153, %int128_34154 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36628 = torch.aten.view %36611, %36627 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36628, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34155 = torch.constant.int 4
%36629 = torch.aten.mul.int %int4_34155, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34156 = torch.constant.int 16
%int1_34157 = torch.constant.int 1
%int128_34158 = torch.constant.int 128
%36630 = torch.prim.ListConstruct %36629, %int16_34156, %int1_34157, %int128_34158 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36631 = torch.aten.view %36613, %36630 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36631, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34159 = torch.constant.int 4
%36632 = torch.aten.mul.int %int4_34159, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34160 = torch.constant.int 16
%int1_34161 = torch.constant.int 1
%int128_34162 = torch.constant.int 128
%36633 = torch.prim.ListConstruct %36632, %int16_34160, %int1_34161, %int128_34162 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36634 = torch.aten.view %36615, %36633 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36634, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34163 = torch.constant.int 4
%36635 = torch.aten.mul.int %int4_34163, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34164 = torch.constant.int 16
%int1_34165 = torch.constant.int 1
%int128_34166 = torch.constant.int 128
%36636 = torch.prim.ListConstruct %36635, %int16_34164, %int1_34165, %int128_34166 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36637 = torch.aten.view %36617, %36636 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36637, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34167 = torch.constant.int 4
%36638 = torch.aten.mul.int %int4_34167, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34168 = torch.constant.int 16
%int1_34169 = torch.constant.int 1
%int128_34170 = torch.constant.int 128
%36639 = torch.prim.ListConstruct %36638, %int16_34168, %int1_34169, %int128_34170 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36640 = torch.aten.view %36619, %36639 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36640, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34171 = torch.constant.int 4
%36641 = torch.aten.mul.int %int4_34171, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34172 = torch.constant.int 16
%int1_34173 = torch.constant.int 1
%int128_34174 = torch.constant.int 128
%36642 = torch.prim.ListConstruct %36641, %int16_34172, %int1_34173, %int128_34174 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36643 = torch.aten.view %36621, %36642 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36643, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34175 = torch.constant.int 4
%36644 = torch.aten.mul.int %int4_34175, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34176 = torch.constant.int 16
%int1_34177 = torch.constant.int 1
%int128_34178 = torch.constant.int 128
%36645 = torch.prim.ListConstruct %36644, %int16_34176, %int1_34177, %int128_34178 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36646 = torch.aten.view %36623, %36645 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36646, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_34179 = torch.constant.int 4
%36647 = torch.aten.mul.int %int4_34179, %3095 : !torch.int, !torch.int -> !torch.int
%int16_34180 = torch.constant.int 16
%int1_34181 = torch.constant.int 1
%int128_34182 = torch.constant.int 128
%36648 = torch.prim.ListConstruct %36647, %int16_34180, %int1_34181, %int128_34182 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36649 = torch.aten.view %36625, %36648 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36649, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_34183 = torch.constant.int 1
%int1_34184 = torch.constant.int 1
%36650 = torch.aten.add.Scalar %36538, %int1_34183, %int1_34184 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36650, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34185 = torch.constant.int 1
%int1_34186 = torch.constant.int 1
%36651 = torch.aten.add.Scalar %36539, %int1_34185, %int1_34186 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36651, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34187 = torch.constant.int 1
%int1_34188 = torch.constant.int 1
%36652 = torch.aten.add.Scalar %36540, %int1_34187, %int1_34188 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36652, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34189 = torch.constant.int 1
%int1_34190 = torch.constant.int 1
%36653 = torch.aten.add.Scalar %36541, %int1_34189, %int1_34190 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36653, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34191 = torch.constant.int 1
%int1_34192 = torch.constant.int 1
%36654 = torch.aten.add.Scalar %36542, %int1_34191, %int1_34192 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36654, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34193 = torch.constant.int 1
%int1_34194 = torch.constant.int 1
%36655 = torch.aten.add.Scalar %36543, %int1_34193, %int1_34194 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36655, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34195 = torch.constant.int 1
%int1_34196 = torch.constant.int 1
%36656 = torch.aten.add.Scalar %36544, %int1_34195, %int1_34196 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36656, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_34197 = torch.constant.int 1
%int1_34198 = torch.constant.int 1
%36657 = torch.aten.add.Scalar %36545, %int1_34197, %int1_34198 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %36657, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_34199 = torch.constant.int 4
%36658 = torch.aten.mul.int %int4_34199, %3095 : !torch.int, !torch.int -> !torch.int
%36659 = torch.prim.ListConstruct %36658 : (!torch.int) -> !torch.list<int>
%36660 = torch.aten.view %36650, %36659 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36660, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34200 = torch.constant.int 4
%36661 = torch.aten.mul.int %int4_34200, %3095 : !torch.int, !torch.int -> !torch.int
%36662 = torch.prim.ListConstruct %36661 : (!torch.int) -> !torch.list<int>
%36663 = torch.aten.view %36651, %36662 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36663, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34201 = torch.constant.int 4
%36664 = torch.aten.mul.int %int4_34201, %3095 : !torch.int, !torch.int -> !torch.int
%36665 = torch.prim.ListConstruct %36664 : (!torch.int) -> !torch.list<int>
%36666 = torch.aten.view %36652, %36665 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36666, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34202 = torch.constant.int 4
%36667 = torch.aten.mul.int %int4_34202, %3095 : !torch.int, !torch.int -> !torch.int
%36668 = torch.prim.ListConstruct %36667 : (!torch.int) -> !torch.list<int>
%36669 = torch.aten.view %36653, %36668 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36669, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34203 = torch.constant.int 4
%36670 = torch.aten.mul.int %int4_34203, %3095 : !torch.int, !torch.int -> !torch.int
%36671 = torch.prim.ListConstruct %36670 : (!torch.int) -> !torch.list<int>
%36672 = torch.aten.view %36654, %36671 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36672, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34204 = torch.constant.int 4
%36673 = torch.aten.mul.int %int4_34204, %3095 : !torch.int, !torch.int -> !torch.int
%36674 = torch.prim.ListConstruct %36673 : (!torch.int) -> !torch.list<int>
%36675 = torch.aten.view %36655, %36674 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36675, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34205 = torch.constant.int 4
%36676 = torch.aten.mul.int %int4_34205, %3095 : !torch.int, !torch.int -> !torch.int
%36677 = torch.prim.ListConstruct %36676 : (!torch.int) -> !torch.list<int>
%36678 = torch.aten.view %36656, %36677 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36678, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_34206 = torch.constant.int 4
%36679 = torch.aten.mul.int %int4_34206, %3095 : !torch.int, !torch.int -> !torch.int
%36680 = torch.prim.ListConstruct %36679 : (!torch.int) -> !torch.list<int>
%36681 = torch.aten.view %36657, %36680 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36681, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%36682 = torch.prim.ListConstruct %36588, %36660 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34207 = torch.constant.int 0
%36683 = torch.aten.cat %36682, %int0_34207 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36683, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36684 = torch.prim.ListConstruct %36591, %36663 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34208 = torch.constant.int 0
%36685 = torch.aten.cat %36684, %int0_34208 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36685, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36686 = torch.prim.ListConstruct %36594, %36666 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34209 = torch.constant.int 0
%36687 = torch.aten.cat %36686, %int0_34209 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36687, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36688 = torch.prim.ListConstruct %36597, %36669 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34210 = torch.constant.int 0
%36689 = torch.aten.cat %36688, %int0_34210 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36689, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36690 = torch.prim.ListConstruct %36600, %36672 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34211 = torch.constant.int 0
%36691 = torch.aten.cat %36690, %int0_34211 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36691, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36692 = torch.prim.ListConstruct %36603, %36675 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34212 = torch.constant.int 0
%36693 = torch.aten.cat %36692, %int0_34212 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36693, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36694 = torch.prim.ListConstruct %36606, %36678 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34213 = torch.constant.int 0
%36695 = torch.aten.cat %36694, %int0_34213 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36695, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36696 = torch.prim.ListConstruct %36609, %36681 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_34214 = torch.constant.int 0
%36697 = torch.aten.cat %36696, %int0_34214 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %36697, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%36698 = torch.prim.ListConstruct %36564, %36628 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34215 = torch.constant.int 0
%36699 = torch.aten.cat %36698, %int0_34215 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36699, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36700 = torch.prim.ListConstruct %36567, %36631 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34216 = torch.constant.int 0
%36701 = torch.aten.cat %36700, %int0_34216 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36701, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36702 = torch.prim.ListConstruct %36570, %36634 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34217 = torch.constant.int 0
%36703 = torch.aten.cat %36702, %int0_34217 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36703, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36704 = torch.prim.ListConstruct %36573, %36637 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34218 = torch.constant.int 0
%36705 = torch.aten.cat %36704, %int0_34218 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36705, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36706 = torch.prim.ListConstruct %36576, %36640 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34219 = torch.constant.int 0
%36707 = torch.aten.cat %36706, %int0_34219 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36707, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36708 = torch.prim.ListConstruct %36579, %36643 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34220 = torch.constant.int 0
%36709 = torch.aten.cat %36708, %int0_34220 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36709, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36710 = torch.prim.ListConstruct %36582, %36646 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34221 = torch.constant.int 0
%36711 = torch.aten.cat %36710, %int0_34221 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36711, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36712 = torch.prim.ListConstruct %36585, %36649 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_34222 = torch.constant.int 0
%36713 = torch.aten.cat %36712, %int0_34222 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36713, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34223 = torch.constant.int 32
%int2_34224 = torch.constant.int 2
%int16_34225 = torch.constant.int 16
%int1_34226 = torch.constant.int 1
%int128_34227 = torch.constant.int 128
%36714 = torch.prim.ListConstruct %3023, %int32_34223, %int2_34224, %int16_34225, %int1_34226, %int128_34227 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36715 = torch.aten.view %34864, %36714 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36715, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34228 = torch.constant.int 32
%36716 = torch.aten.mul.int %3023, %int32_34228 : !torch.int, !torch.int -> !torch.int
%int2_34229 = torch.constant.int 2
%36717 = torch.aten.mul.int %36716, %int2_34229 : !torch.int, !torch.int -> !torch.int
%int16_34230 = torch.constant.int 16
%int1_34231 = torch.constant.int 1
%int128_34232 = torch.constant.int 128
%36718 = torch.prim.ListConstruct %36717, %int16_34230, %int1_34231, %int128_34232 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36719 = torch.aten.view %36715, %36718 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36719, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36720 = torch.prim.ListConstruct %36683 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34233 = torch.constant.bool false
%36721 = torch.aten.index_put %36719, %36720, %36699, %false_34233 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36721, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34234 = torch.constant.int 32
%int2_34235 = torch.constant.int 2
%int16_34236 = torch.constant.int 16
%int1_34237 = torch.constant.int 1
%int128_34238 = torch.constant.int 128
%36722 = torch.prim.ListConstruct %3023, %int32_34234, %int2_34235, %int16_34236, %int1_34237, %int128_34238 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36723 = torch.aten.view %36721, %36722 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36723, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34239 = torch.constant.int 131072
%36724 = torch.prim.ListConstruct %3023, %int131072_34239 : (!torch.int, !torch.int) -> !torch.list<int>
%36725 = torch.aten.view %36723, %36724 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36725, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34240 = torch.constant.int 32
%int2_34241 = torch.constant.int 2
%int16_34242 = torch.constant.int 16
%int1_34243 = torch.constant.int 1
%int128_34244 = torch.constant.int 128
%36726 = torch.prim.ListConstruct %3026, %int32_34240, %int2_34241, %int16_34242, %int1_34243, %int128_34244 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36727 = torch.aten.view %34876, %36726 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36727, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34245 = torch.constant.int 32
%36728 = torch.aten.mul.int %3026, %int32_34245 : !torch.int, !torch.int -> !torch.int
%int2_34246 = torch.constant.int 2
%36729 = torch.aten.mul.int %36728, %int2_34246 : !torch.int, !torch.int -> !torch.int
%int16_34247 = torch.constant.int 16
%int1_34248 = torch.constant.int 1
%int128_34249 = torch.constant.int 128
%36730 = torch.prim.ListConstruct %36729, %int16_34247, %int1_34248, %int128_34249 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36731 = torch.aten.view %36727, %36730 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36731, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36732 = torch.prim.ListConstruct %36685 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34250 = torch.constant.bool false
%36733 = torch.aten.index_put %36731, %36732, %36701, %false_34250 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36733, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34251 = torch.constant.int 32
%int2_34252 = torch.constant.int 2
%int16_34253 = torch.constant.int 16
%int1_34254 = torch.constant.int 1
%int128_34255 = torch.constant.int 128
%36734 = torch.prim.ListConstruct %3026, %int32_34251, %int2_34252, %int16_34253, %int1_34254, %int128_34255 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36735 = torch.aten.view %36733, %36734 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36735, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34256 = torch.constant.int 131072
%36736 = torch.prim.ListConstruct %3026, %int131072_34256 : (!torch.int, !torch.int) -> !torch.list<int>
%36737 = torch.aten.view %36735, %36736 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36737, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34257 = torch.constant.int 32
%int2_34258 = torch.constant.int 2
%int16_34259 = torch.constant.int 16
%int1_34260 = torch.constant.int 1
%int128_34261 = torch.constant.int 128
%36738 = torch.prim.ListConstruct %3029, %int32_34257, %int2_34258, %int16_34259, %int1_34260, %int128_34261 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36739 = torch.aten.view %34888, %36738 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36739, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34262 = torch.constant.int 32
%36740 = torch.aten.mul.int %3029, %int32_34262 : !torch.int, !torch.int -> !torch.int
%int2_34263 = torch.constant.int 2
%36741 = torch.aten.mul.int %36740, %int2_34263 : !torch.int, !torch.int -> !torch.int
%int16_34264 = torch.constant.int 16
%int1_34265 = torch.constant.int 1
%int128_34266 = torch.constant.int 128
%36742 = torch.prim.ListConstruct %36741, %int16_34264, %int1_34265, %int128_34266 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36743 = torch.aten.view %36739, %36742 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36743, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36744 = torch.prim.ListConstruct %36687 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34267 = torch.constant.bool false
%36745 = torch.aten.index_put %36743, %36744, %36703, %false_34267 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36745, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34268 = torch.constant.int 32
%int2_34269 = torch.constant.int 2
%int16_34270 = torch.constant.int 16
%int1_34271 = torch.constant.int 1
%int128_34272 = torch.constant.int 128
%36746 = torch.prim.ListConstruct %3029, %int32_34268, %int2_34269, %int16_34270, %int1_34271, %int128_34272 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36747 = torch.aten.view %36745, %36746 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36747, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34273 = torch.constant.int 131072
%36748 = torch.prim.ListConstruct %3029, %int131072_34273 : (!torch.int, !torch.int) -> !torch.list<int>
%36749 = torch.aten.view %36747, %36748 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36749, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34274 = torch.constant.int 32
%int2_34275 = torch.constant.int 2
%int16_34276 = torch.constant.int 16
%int1_34277 = torch.constant.int 1
%int128_34278 = torch.constant.int 128
%36750 = torch.prim.ListConstruct %3032, %int32_34274, %int2_34275, %int16_34276, %int1_34277, %int128_34278 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36751 = torch.aten.view %34900, %36750 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36751, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34279 = torch.constant.int 32
%36752 = torch.aten.mul.int %3032, %int32_34279 : !torch.int, !torch.int -> !torch.int
%int2_34280 = torch.constant.int 2
%36753 = torch.aten.mul.int %36752, %int2_34280 : !torch.int, !torch.int -> !torch.int
%int16_34281 = torch.constant.int 16
%int1_34282 = torch.constant.int 1
%int128_34283 = torch.constant.int 128
%36754 = torch.prim.ListConstruct %36753, %int16_34281, %int1_34282, %int128_34283 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36755 = torch.aten.view %36751, %36754 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36755, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36756 = torch.prim.ListConstruct %36689 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34284 = torch.constant.bool false
%36757 = torch.aten.index_put %36755, %36756, %36705, %false_34284 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36757, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34285 = torch.constant.int 32
%int2_34286 = torch.constant.int 2
%int16_34287 = torch.constant.int 16
%int1_34288 = torch.constant.int 1
%int128_34289 = torch.constant.int 128
%36758 = torch.prim.ListConstruct %3032, %int32_34285, %int2_34286, %int16_34287, %int1_34288, %int128_34289 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36759 = torch.aten.view %36757, %36758 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36759, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34290 = torch.constant.int 131072
%36760 = torch.prim.ListConstruct %3032, %int131072_34290 : (!torch.int, !torch.int) -> !torch.list<int>
%36761 = torch.aten.view %36759, %36760 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36761, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34291 = torch.constant.int 32
%int2_34292 = torch.constant.int 2
%int16_34293 = torch.constant.int 16
%int1_34294 = torch.constant.int 1
%int128_34295 = torch.constant.int 128
%36762 = torch.prim.ListConstruct %3035, %int32_34291, %int2_34292, %int16_34293, %int1_34294, %int128_34295 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36763 = torch.aten.view %34912, %36762 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36763, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34296 = torch.constant.int 32
%36764 = torch.aten.mul.int %3035, %int32_34296 : !torch.int, !torch.int -> !torch.int
%int2_34297 = torch.constant.int 2
%36765 = torch.aten.mul.int %36764, %int2_34297 : !torch.int, !torch.int -> !torch.int
%int16_34298 = torch.constant.int 16
%int1_34299 = torch.constant.int 1
%int128_34300 = torch.constant.int 128
%36766 = torch.prim.ListConstruct %36765, %int16_34298, %int1_34299, %int128_34300 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36767 = torch.aten.view %36763, %36766 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36767, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36768 = torch.prim.ListConstruct %36691 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34301 = torch.constant.bool false
%36769 = torch.aten.index_put %36767, %36768, %36707, %false_34301 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36769, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34302 = torch.constant.int 32
%int2_34303 = torch.constant.int 2
%int16_34304 = torch.constant.int 16
%int1_34305 = torch.constant.int 1
%int128_34306 = torch.constant.int 128
%36770 = torch.prim.ListConstruct %3035, %int32_34302, %int2_34303, %int16_34304, %int1_34305, %int128_34306 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36771 = torch.aten.view %36769, %36770 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36771, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34307 = torch.constant.int 131072
%36772 = torch.prim.ListConstruct %3035, %int131072_34307 : (!torch.int, !torch.int) -> !torch.list<int>
%36773 = torch.aten.view %36771, %36772 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36773, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34308 = torch.constant.int 32
%int2_34309 = torch.constant.int 2
%int16_34310 = torch.constant.int 16
%int1_34311 = torch.constant.int 1
%int128_34312 = torch.constant.int 128
%36774 = torch.prim.ListConstruct %3038, %int32_34308, %int2_34309, %int16_34310, %int1_34311, %int128_34312 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36775 = torch.aten.view %34924, %36774 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36775, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34313 = torch.constant.int 32
%36776 = torch.aten.mul.int %3038, %int32_34313 : !torch.int, !torch.int -> !torch.int
%int2_34314 = torch.constant.int 2
%36777 = torch.aten.mul.int %36776, %int2_34314 : !torch.int, !torch.int -> !torch.int
%int16_34315 = torch.constant.int 16
%int1_34316 = torch.constant.int 1
%int128_34317 = torch.constant.int 128
%36778 = torch.prim.ListConstruct %36777, %int16_34315, %int1_34316, %int128_34317 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36779 = torch.aten.view %36775, %36778 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36779, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36780 = torch.prim.ListConstruct %36693 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34318 = torch.constant.bool false
%36781 = torch.aten.index_put %36779, %36780, %36709, %false_34318 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36781, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34319 = torch.constant.int 32
%int2_34320 = torch.constant.int 2
%int16_34321 = torch.constant.int 16
%int1_34322 = torch.constant.int 1
%int128_34323 = torch.constant.int 128
%36782 = torch.prim.ListConstruct %3038, %int32_34319, %int2_34320, %int16_34321, %int1_34322, %int128_34323 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36783 = torch.aten.view %36781, %36782 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36783, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34324 = torch.constant.int 131072
%36784 = torch.prim.ListConstruct %3038, %int131072_34324 : (!torch.int, !torch.int) -> !torch.list<int>
%36785 = torch.aten.view %36783, %36784 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36785, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34325 = torch.constant.int 32
%int2_34326 = torch.constant.int 2
%int16_34327 = torch.constant.int 16
%int1_34328 = torch.constant.int 1
%int128_34329 = torch.constant.int 128
%36786 = torch.prim.ListConstruct %3041, %int32_34325, %int2_34326, %int16_34327, %int1_34328, %int128_34329 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36787 = torch.aten.view %34936, %36786 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36787, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34330 = torch.constant.int 32
%36788 = torch.aten.mul.int %3041, %int32_34330 : !torch.int, !torch.int -> !torch.int
%int2_34331 = torch.constant.int 2
%36789 = torch.aten.mul.int %36788, %int2_34331 : !torch.int, !torch.int -> !torch.int
%int16_34332 = torch.constant.int 16
%int1_34333 = torch.constant.int 1
%int128_34334 = torch.constant.int 128
%36790 = torch.prim.ListConstruct %36789, %int16_34332, %int1_34333, %int128_34334 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36791 = torch.aten.view %36787, %36790 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36791, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36792 = torch.prim.ListConstruct %36695 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34335 = torch.constant.bool false
%36793 = torch.aten.index_put %36791, %36792, %36711, %false_34335 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36793, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34336 = torch.constant.int 32
%int2_34337 = torch.constant.int 2
%int16_34338 = torch.constant.int 16
%int1_34339 = torch.constant.int 1
%int128_34340 = torch.constant.int 128
%36794 = torch.prim.ListConstruct %3041, %int32_34336, %int2_34337, %int16_34338, %int1_34339, %int128_34340 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36795 = torch.aten.view %36793, %36794 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36795, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34341 = torch.constant.int 131072
%36796 = torch.prim.ListConstruct %3041, %int131072_34341 : (!torch.int, !torch.int) -> !torch.list<int>
%36797 = torch.aten.view %36795, %36796 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36797, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_34342 = torch.constant.int 32
%int2_34343 = torch.constant.int 2
%int16_34344 = torch.constant.int 16
%int1_34345 = torch.constant.int 1
%int128_34346 = torch.constant.int 128
%36798 = torch.prim.ListConstruct %3044, %int32_34342, %int2_34343, %int16_34344, %int1_34345, %int128_34346 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36799 = torch.aten.view %34948, %36798 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36799, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_34347 = torch.constant.int 32
%36800 = torch.aten.mul.int %3044, %int32_34347 : !torch.int, !torch.int -> !torch.int
%int2_34348 = torch.constant.int 2
%36801 = torch.aten.mul.int %36800, %int2_34348 : !torch.int, !torch.int -> !torch.int
%int16_34349 = torch.constant.int 16
%int1_34350 = torch.constant.int 1
%int128_34351 = torch.constant.int 128
%36802 = torch.prim.ListConstruct %36801, %int16_34349, %int1_34350, %int128_34351 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36803 = torch.aten.view %36799, %36802 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36803, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%36804 = torch.prim.ListConstruct %36697 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_34352 = torch.constant.bool false
%36805 = torch.aten.index_put %36803, %36804, %36713, %false_34352 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %36805, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_34353 = torch.constant.int 32
%int2_34354 = torch.constant.int 2
%int16_34355 = torch.constant.int 16
%int1_34356 = torch.constant.int 1
%int128_34357 = torch.constant.int 128
%36806 = torch.prim.ListConstruct %3044, %int32_34353, %int2_34354, %int16_34355, %int1_34356, %int128_34357 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36807 = torch.aten.view %36805, %36806 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %36807, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_34358 = torch.constant.int 131072
%36808 = torch.prim.ListConstruct %3044, %int131072_34358 : (!torch.int, !torch.int) -> !torch.list<int>
%36809 = torch.aten.view %36807, %36808 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %36809, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_34359 = torch.constant.int -2
%36810 = torch.aten.unsqueeze %36424, %int-2_34359 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34360 = torch.constant.int -2
%36811 = torch.aten.unsqueeze %36439, %int-2_34360 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34361 = torch.constant.int -2
%36812 = torch.aten.unsqueeze %36454, %int-2_34361 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34362 = torch.constant.int -2
%36813 = torch.aten.unsqueeze %36469, %int-2_34362 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34363 = torch.constant.int -2
%36814 = torch.aten.unsqueeze %36484, %int-2_34363 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34364 = torch.constant.int -2
%36815 = torch.aten.unsqueeze %36499, %int-2_34364 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34365 = torch.constant.int -2
%36816 = torch.aten.unsqueeze %36514, %int-2_34365 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34366 = torch.constant.int -2
%36817 = torch.aten.unsqueeze %36529, %int-2_34366 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_34367 = torch.constant.int 4
%int1_34368 = torch.constant.int 1
%int4_34369 = torch.constant.int 4
%int128_34370 = torch.constant.int 128
%36818 = torch.prim.ListConstruct %int4_34367, %36410, %int1_34368, %int4_34369, %int128_34370 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34371 = torch.constant.bool false
%36819 = torch.aten.expand %36810, %36818, %false_34371 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34372 = torch.constant.int 4
%int1_34373 = torch.constant.int 1
%int4_34374 = torch.constant.int 4
%int128_34375 = torch.constant.int 128
%36820 = torch.prim.ListConstruct %int4_34372, %36410, %int1_34373, %int4_34374, %int128_34375 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34376 = torch.constant.bool false
%36821 = torch.aten.expand %36811, %36820, %false_34376 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34377 = torch.constant.int 4
%int1_34378 = torch.constant.int 1
%int4_34379 = torch.constant.int 4
%int128_34380 = torch.constant.int 128
%36822 = torch.prim.ListConstruct %int4_34377, %36410, %int1_34378, %int4_34379, %int128_34380 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34381 = torch.constant.bool false
%36823 = torch.aten.expand %36812, %36822, %false_34381 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34382 = torch.constant.int 4
%int1_34383 = torch.constant.int 1
%int4_34384 = torch.constant.int 4
%int128_34385 = torch.constant.int 128
%36824 = torch.prim.ListConstruct %int4_34382, %36410, %int1_34383, %int4_34384, %int128_34385 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34386 = torch.constant.bool false
%36825 = torch.aten.expand %36813, %36824, %false_34386 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34387 = torch.constant.int 4
%int1_34388 = torch.constant.int 1
%int4_34389 = torch.constant.int 4
%int128_34390 = torch.constant.int 128
%36826 = torch.prim.ListConstruct %int4_34387, %36410, %int1_34388, %int4_34389, %int128_34390 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34391 = torch.constant.bool false
%36827 = torch.aten.expand %36814, %36826, %false_34391 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34392 = torch.constant.int 4
%int1_34393 = torch.constant.int 1
%int4_34394 = torch.constant.int 4
%int128_34395 = torch.constant.int 128
%36828 = torch.prim.ListConstruct %int4_34392, %36410, %int1_34393, %int4_34394, %int128_34395 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34396 = torch.constant.bool false
%36829 = torch.aten.expand %36815, %36828, %false_34396 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34397 = torch.constant.int 4
%int1_34398 = torch.constant.int 1
%int4_34399 = torch.constant.int 4
%int128_34400 = torch.constant.int 128
%36830 = torch.prim.ListConstruct %int4_34397, %36410, %int1_34398, %int4_34399, %int128_34400 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34401 = torch.constant.bool false
%36831 = torch.aten.expand %36816, %36830, %false_34401 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34402 = torch.constant.int 4
%int1_34403 = torch.constant.int 1
%int4_34404 = torch.constant.int 4
%int128_34405 = torch.constant.int 128
%36832 = torch.prim.ListConstruct %int4_34402, %36410, %int1_34403, %int4_34404, %int128_34405 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34406 = torch.constant.bool false
%36833 = torch.aten.expand %36817, %36832, %false_34406 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34407 = torch.constant.int 4
%int4_34408 = torch.constant.int 4
%int128_34409 = torch.constant.int 128
%36834 = torch.prim.ListConstruct %int4_34407, %36410, %int4_34408, %int128_34409 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36835 = torch.aten.view %36819, %36834 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34410 = torch.constant.int 4
%int4_34411 = torch.constant.int 4
%int128_34412 = torch.constant.int 128
%36836 = torch.prim.ListConstruct %int4_34410, %36410, %int4_34411, %int128_34412 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36837 = torch.aten.view %36821, %36836 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36837, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34413 = torch.constant.int 4
%int4_34414 = torch.constant.int 4
%int128_34415 = torch.constant.int 128
%36838 = torch.prim.ListConstruct %int4_34413, %36410, %int4_34414, %int128_34415 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36839 = torch.aten.view %36823, %36838 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36839, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34416 = torch.constant.int 4
%int4_34417 = torch.constant.int 4
%int128_34418 = torch.constant.int 128
%36840 = torch.prim.ListConstruct %int4_34416, %36410, %int4_34417, %int128_34418 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36841 = torch.aten.view %36825, %36840 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34419 = torch.constant.int 4
%int4_34420 = torch.constant.int 4
%int128_34421 = torch.constant.int 128
%36842 = torch.prim.ListConstruct %int4_34419, %36410, %int4_34420, %int128_34421 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36843 = torch.aten.view %36827, %36842 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36843, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34422 = torch.constant.int 4
%int4_34423 = torch.constant.int 4
%int128_34424 = torch.constant.int 128
%36844 = torch.prim.ListConstruct %int4_34422, %36410, %int4_34423, %int128_34424 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36845 = torch.aten.view %36829, %36844 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34425 = torch.constant.int 4
%int4_34426 = torch.constant.int 4
%int128_34427 = torch.constant.int 128
%36846 = torch.prim.ListConstruct %int4_34425, %36410, %int4_34426, %int128_34427 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36847 = torch.aten.view %36831, %36846 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34428 = torch.constant.int 4
%int4_34429 = torch.constant.int 4
%int128_34430 = torch.constant.int 128
%36848 = torch.prim.ListConstruct %int4_34428, %36410, %int4_34429, %int128_34430 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36849 = torch.aten.view %36833, %36848 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36849, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_34431 = torch.constant.int -2
%36850 = torch.aten.unsqueeze %36199, %int-2_34431 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36850, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34432 = torch.constant.int -2
%36851 = torch.aten.unsqueeze %36201, %int-2_34432 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34433 = torch.constant.int -2
%36852 = torch.aten.unsqueeze %36203, %int-2_34433 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36852, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34434 = torch.constant.int -2
%36853 = torch.aten.unsqueeze %36205, %int-2_34434 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34435 = torch.constant.int -2
%36854 = torch.aten.unsqueeze %36207, %int-2_34435 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36854, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34436 = torch.constant.int -2
%36855 = torch.aten.unsqueeze %36209, %int-2_34436 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36855, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34437 = torch.constant.int -2
%36856 = torch.aten.unsqueeze %36211, %int-2_34437 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_34438 = torch.constant.int -2
%36857 = torch.aten.unsqueeze %36213, %int-2_34438 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %36857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_34439 = torch.constant.int 1
%36858 = torch.aten.size.int %36123, %int1_34439 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_34440 = torch.constant.int 4
%int1_34441 = torch.constant.int 1
%int4_34442 = torch.constant.int 4
%int128_34443 = torch.constant.int 128
%36859 = torch.prim.ListConstruct %int4_34440, %36858, %int1_34441, %int4_34442, %int128_34443 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34444 = torch.constant.bool false
%36860 = torch.aten.expand %36850, %36859, %false_34444 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36860, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34445 = torch.constant.int 4
%int1_34446 = torch.constant.int 1
%int4_34447 = torch.constant.int 4
%int128_34448 = torch.constant.int 128
%36861 = torch.prim.ListConstruct %int4_34445, %36858, %int1_34446, %int4_34447, %int128_34448 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34449 = torch.constant.bool false
%36862 = torch.aten.expand %36851, %36861, %false_34449 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34450 = torch.constant.int 4
%int1_34451 = torch.constant.int 1
%int4_34452 = torch.constant.int 4
%int128_34453 = torch.constant.int 128
%36863 = torch.prim.ListConstruct %int4_34450, %36858, %int1_34451, %int4_34452, %int128_34453 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34454 = torch.constant.bool false
%36864 = torch.aten.expand %36852, %36863, %false_34454 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36864, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34455 = torch.constant.int 4
%int1_34456 = torch.constant.int 1
%int4_34457 = torch.constant.int 4
%int128_34458 = torch.constant.int 128
%36865 = torch.prim.ListConstruct %int4_34455, %36858, %int1_34456, %int4_34457, %int128_34458 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34459 = torch.constant.bool false
%36866 = torch.aten.expand %36853, %36865, %false_34459 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36866, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34460 = torch.constant.int 4
%int1_34461 = torch.constant.int 1
%int4_34462 = torch.constant.int 4
%int128_34463 = torch.constant.int 128
%36867 = torch.prim.ListConstruct %int4_34460, %36858, %int1_34461, %int4_34462, %int128_34463 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34464 = torch.constant.bool false
%36868 = torch.aten.expand %36854, %36867, %false_34464 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34465 = torch.constant.int 4
%int1_34466 = torch.constant.int 1
%int4_34467 = torch.constant.int 4
%int128_34468 = torch.constant.int 128
%36869 = torch.prim.ListConstruct %int4_34465, %36858, %int1_34466, %int4_34467, %int128_34468 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34469 = torch.constant.bool false
%36870 = torch.aten.expand %36855, %36869, %false_34469 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36870, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34470 = torch.constant.int 4
%int1_34471 = torch.constant.int 1
%int4_34472 = torch.constant.int 4
%int128_34473 = torch.constant.int 128
%36871 = torch.prim.ListConstruct %int4_34470, %36858, %int1_34471, %int4_34472, %int128_34473 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34474 = torch.constant.bool false
%36872 = torch.aten.expand %36856, %36871, %false_34474 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36872, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34475 = torch.constant.int 4
%int1_34476 = torch.constant.int 1
%int4_34477 = torch.constant.int 4
%int128_34478 = torch.constant.int 128
%36873 = torch.prim.ListConstruct %int4_34475, %36858, %int1_34476, %int4_34477, %int128_34478 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_34479 = torch.constant.bool false
%36874 = torch.aten.expand %36857, %36873, %false_34479 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %36874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_34480 = torch.constant.int 4
%int4_34481 = torch.constant.int 4
%int128_34482 = torch.constant.int 128
%36875 = torch.prim.ListConstruct %int4_34480, %36858, %int4_34481, %int128_34482 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36876 = torch.aten.view %36860, %36875 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36876, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34483 = torch.constant.int 4
%int4_34484 = torch.constant.int 4
%int128_34485 = torch.constant.int 128
%36877 = torch.prim.ListConstruct %int4_34483, %36858, %int4_34484, %int128_34485 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36878 = torch.aten.view %36862, %36877 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36878, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34486 = torch.constant.int 4
%int4_34487 = torch.constant.int 4
%int128_34488 = torch.constant.int 128
%36879 = torch.prim.ListConstruct %int4_34486, %36858, %int4_34487, %int128_34488 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36880 = torch.aten.view %36864, %36879 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34489 = torch.constant.int 4
%int4_34490 = torch.constant.int 4
%int128_34491 = torch.constant.int 128
%36881 = torch.prim.ListConstruct %int4_34489, %36858, %int4_34490, %int128_34491 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36882 = torch.aten.view %36866, %36881 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36882, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34492 = torch.constant.int 4
%int4_34493 = torch.constant.int 4
%int128_34494 = torch.constant.int 128
%36883 = torch.prim.ListConstruct %int4_34492, %36858, %int4_34493, %int128_34494 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36884 = torch.aten.view %36868, %36883 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36884, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34495 = torch.constant.int 4
%int4_34496 = torch.constant.int 4
%int128_34497 = torch.constant.int 128
%36885 = torch.prim.ListConstruct %int4_34495, %36858, %int4_34496, %int128_34497 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36886 = torch.aten.view %36870, %36885 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34498 = torch.constant.int 4
%int4_34499 = torch.constant.int 4
%int128_34500 = torch.constant.int 128
%36887 = torch.prim.ListConstruct %int4_34498, %36858, %int4_34499, %int128_34500 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36888 = torch.aten.view %36872, %36887 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36888, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34501 = torch.constant.int 4
%int4_34502 = torch.constant.int 4
%int128_34503 = torch.constant.int 128
%36889 = torch.prim.ListConstruct %int4_34501, %36858, %int4_34502, %int128_34503 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36890 = torch.aten.view %36874, %36889 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36890, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34504 = torch.constant.int 1
%int2_34505 = torch.constant.int 2
%36891 = torch.aten.transpose.int %36266, %int1_34504, %int2_34505 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36891, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34506 = torch.constant.int 1
%int2_34507 = torch.constant.int 2
%36892 = torch.aten.transpose.int %36281, %int1_34506, %int2_34507 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36892, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34508 = torch.constant.int 1
%int2_34509 = torch.constant.int 2
%36893 = torch.aten.transpose.int %36296, %int1_34508, %int2_34509 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36893, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34510 = torch.constant.int 1
%int2_34511 = torch.constant.int 2
%36894 = torch.aten.transpose.int %36311, %int1_34510, %int2_34511 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36894, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34512 = torch.constant.int 1
%int2_34513 = torch.constant.int 2
%36895 = torch.aten.transpose.int %36326, %int1_34512, %int2_34513 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36895, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34514 = torch.constant.int 1
%int2_34515 = torch.constant.int 2
%36896 = torch.aten.transpose.int %36341, %int1_34514, %int2_34515 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36896, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34516 = torch.constant.int 1
%int2_34517 = torch.constant.int 2
%36897 = torch.aten.transpose.int %36356, %int1_34516, %int2_34517 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36897, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34518 = torch.constant.int 1
%int2_34519 = torch.constant.int 2
%36898 = torch.aten.transpose.int %36371, %int1_34518, %int2_34519 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36898, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34520 = torch.constant.int 1
%int2_34521 = torch.constant.int 2
%36899 = torch.aten.transpose.int %36835, %int1_34520, %int2_34521 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36899, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34522 = torch.constant.int 1
%int2_34523 = torch.constant.int 2
%36900 = torch.aten.transpose.int %36837, %int1_34522, %int2_34523 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36900, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34524 = torch.constant.int 1
%int2_34525 = torch.constant.int 2
%36901 = torch.aten.transpose.int %36839, %int1_34524, %int2_34525 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36901, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34526 = torch.constant.int 1
%int2_34527 = torch.constant.int 2
%36902 = torch.aten.transpose.int %36841, %int1_34526, %int2_34527 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36902, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34528 = torch.constant.int 1
%int2_34529 = torch.constant.int 2
%36903 = torch.aten.transpose.int %36843, %int1_34528, %int2_34529 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36903, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34530 = torch.constant.int 1
%int2_34531 = torch.constant.int 2
%36904 = torch.aten.transpose.int %36845, %int1_34530, %int2_34531 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36904, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34532 = torch.constant.int 1
%int2_34533 = torch.constant.int 2
%36905 = torch.aten.transpose.int %36847, %int1_34532, %int2_34533 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36905, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34534 = torch.constant.int 1
%int2_34535 = torch.constant.int 2
%36906 = torch.aten.transpose.int %36849, %int1_34534, %int2_34535 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36906, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34536 = torch.constant.int 1
%int2_34537 = torch.constant.int 2
%36907 = torch.aten.transpose.int %36876, %int1_34536, %int2_34537 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36907, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34538 = torch.constant.int 1
%int2_34539 = torch.constant.int 2
%36908 = torch.aten.transpose.int %36878, %int1_34538, %int2_34539 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36908, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34540 = torch.constant.int 1
%int2_34541 = torch.constant.int 2
%36909 = torch.aten.transpose.int %36880, %int1_34540, %int2_34541 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36909, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34542 = torch.constant.int 1
%int2_34543 = torch.constant.int 2
%36910 = torch.aten.transpose.int %36882, %int1_34542, %int2_34543 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36910, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34544 = torch.constant.int 1
%int2_34545 = torch.constant.int 2
%36911 = torch.aten.transpose.int %36884, %int1_34544, %int2_34545 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36911, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34546 = torch.constant.int 1
%int2_34547 = torch.constant.int 2
%36912 = torch.aten.transpose.int %36886, %int1_34546, %int2_34547 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36912, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34548 = torch.constant.int 1
%int2_34549 = torch.constant.int 2
%36913 = torch.aten.transpose.int %36888, %int1_34548, %int2_34549 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36913, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34550 = torch.constant.int 1
%int2_34551 = torch.constant.int 2
%36914 = torch.aten.transpose.int %36890, %int1_34550, %int2_34551 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %36914, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34552 = torch.constant.float 0.000000e+00
%true_34553 = torch.constant.bool true
%none_34554 = torch.constant.none
%none_34555 = torch.constant.none
%36915:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36891, %36899, %36907, %float0.000000e00_34552, %true_34553, %none_34554, %none_34555) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36915#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34556 = torch.constant.float 0.000000e+00
%true_34557 = torch.constant.bool true
%none_34558 = torch.constant.none
%none_34559 = torch.constant.none
%36916:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36892, %36900, %36908, %float0.000000e00_34556, %true_34557, %none_34558, %none_34559) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36916#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34560 = torch.constant.float 0.000000e+00
%true_34561 = torch.constant.bool true
%none_34562 = torch.constant.none
%none_34563 = torch.constant.none
%36917:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36893, %36901, %36909, %float0.000000e00_34560, %true_34561, %none_34562, %none_34563) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36917#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34564 = torch.constant.float 0.000000e+00
%true_34565 = torch.constant.bool true
%none_34566 = torch.constant.none
%none_34567 = torch.constant.none
%36918:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36894, %36902, %36910, %float0.000000e00_34564, %true_34565, %none_34566, %none_34567) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36918#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34568 = torch.constant.float 0.000000e+00
%true_34569 = torch.constant.bool true
%none_34570 = torch.constant.none
%none_34571 = torch.constant.none
%36919:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36895, %36903, %36911, %float0.000000e00_34568, %true_34569, %none_34570, %none_34571) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36919#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34572 = torch.constant.float 0.000000e+00
%true_34573 = torch.constant.bool true
%none_34574 = torch.constant.none
%none_34575 = torch.constant.none
%36920:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36896, %36904, %36912, %float0.000000e00_34572, %true_34573, %none_34574, %none_34575) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36920#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34576 = torch.constant.float 0.000000e+00
%true_34577 = torch.constant.bool true
%none_34578 = torch.constant.none
%none_34579 = torch.constant.none
%36921:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36897, %36905, %36913, %float0.000000e00_34576, %true_34577, %none_34578, %none_34579) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36921#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_34580 = torch.constant.float 0.000000e+00
%true_34581 = torch.constant.bool true
%none_34582 = torch.constant.none
%none_34583 = torch.constant.none
%36922:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%36898, %36906, %36914, %float0.000000e00_34580, %true_34581, %none_34582, %none_34583) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %36922#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_34584 = torch.constant.int 1
%int2_34585 = torch.constant.int 2
%36923 = torch.aten.transpose.int %36915#0, %int1_34584, %int2_34585 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34586 = torch.constant.int 1
%int2_34587 = torch.constant.int 2
%36924 = torch.aten.transpose.int %36916#0, %int1_34586, %int2_34587 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34588 = torch.constant.int 1
%int2_34589 = torch.constant.int 2
%36925 = torch.aten.transpose.int %36917#0, %int1_34588, %int2_34589 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34590 = torch.constant.int 1
%int2_34591 = torch.constant.int 2
%36926 = torch.aten.transpose.int %36918#0, %int1_34590, %int2_34591 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34592 = torch.constant.int 1
%int2_34593 = torch.constant.int 2
%36927 = torch.aten.transpose.int %36919#0, %int1_34592, %int2_34593 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34594 = torch.constant.int 1
%int2_34595 = torch.constant.int 2
%36928 = torch.aten.transpose.int %36920#0, %int1_34594, %int2_34595 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36928, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34596 = torch.constant.int 1
%int2_34597 = torch.constant.int 2
%36929 = torch.aten.transpose.int %36921#0, %int1_34596, %int2_34597 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_34598 = torch.constant.int 1
%int2_34599 = torch.constant.int 2
%36930 = torch.aten.transpose.int %36922#0, %int1_34598, %int2_34599 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %36930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_34600 = torch.constant.int 4
%int512_34601 = torch.constant.int 512
%36931 = torch.prim.ListConstruct %int4_34600, %36252, %int512_34601 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36932 = torch.aten.view %36923, %36931 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34602 = torch.constant.int 4
%int512_34603 = torch.constant.int 512
%36933 = torch.prim.ListConstruct %int4_34602, %36267, %int512_34603 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36934 = torch.aten.view %36924, %36933 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36934, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34604 = torch.constant.int 4
%int512_34605 = torch.constant.int 512
%36935 = torch.prim.ListConstruct %int4_34604, %36282, %int512_34605 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36936 = torch.aten.view %36925, %36935 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34606 = torch.constant.int 4
%int512_34607 = torch.constant.int 512
%36937 = torch.prim.ListConstruct %int4_34606, %36297, %int512_34607 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36938 = torch.aten.view %36926, %36937 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34608 = torch.constant.int 4
%int512_34609 = torch.constant.int 512
%36939 = torch.prim.ListConstruct %int4_34608, %36312, %int512_34609 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36940 = torch.aten.view %36927, %36939 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36940, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34610 = torch.constant.int 4
%int512_34611 = torch.constant.int 512
%36941 = torch.prim.ListConstruct %int4_34610, %36327, %int512_34611 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36942 = torch.aten.view %36928, %36941 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34612 = torch.constant.int 4
%int512_34613 = torch.constant.int 512
%36943 = torch.prim.ListConstruct %int4_34612, %36342, %int512_34613 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36944 = torch.aten.view %36929, %36943 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_34614 = torch.constant.int 4
%int512_34615 = torch.constant.int 512
%36945 = torch.prim.ListConstruct %int4_34614, %36357, %int512_34615 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36946 = torch.aten.view %36930, %36945 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %36946, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_34616 = torch.constant.int 1
%int0_34617 = torch.constant.int 0
%36947 = torch.prim.ListConstruct %int1_34616, %int0_34617 : (!torch.int, !torch.int) -> !torch.list<int>
%36948 = torch.aten.permute %1336, %36947 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34618 = torch.constant.int 1
%int0_34619 = torch.constant.int 0
%36949 = torch.prim.ListConstruct %int1_34618, %int0_34619 : (!torch.int, !torch.int) -> !torch.list<int>
%36950 = torch.aten.permute %1337, %36949 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34620 = torch.constant.int 1
%int0_34621 = torch.constant.int 0
%36951 = torch.prim.ListConstruct %int1_34620, %int0_34621 : (!torch.int, !torch.int) -> !torch.list<int>
%36952 = torch.aten.permute %1338, %36951 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34622 = torch.constant.int 1
%int0_34623 = torch.constant.int 0
%36953 = torch.prim.ListConstruct %int1_34622, %int0_34623 : (!torch.int, !torch.int) -> !torch.list<int>
%36954 = torch.aten.permute %1339, %36953 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34624 = torch.constant.int 1
%int0_34625 = torch.constant.int 0
%36955 = torch.prim.ListConstruct %int1_34624, %int0_34625 : (!torch.int, !torch.int) -> !torch.list<int>
%36956 = torch.aten.permute %1340, %36955 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34626 = torch.constant.int 1
%int0_34627 = torch.constant.int 0
%36957 = torch.prim.ListConstruct %int1_34626, %int0_34627 : (!torch.int, !torch.int) -> !torch.list<int>
%36958 = torch.aten.permute %1341, %36957 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34628 = torch.constant.int 1
%int0_34629 = torch.constant.int 0
%36959 = torch.prim.ListConstruct %int1_34628, %int0_34629 : (!torch.int, !torch.int) -> !torch.list<int>
%36960 = torch.aten.permute %1342, %36959 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_34630 = torch.constant.int 1
%int0_34631 = torch.constant.int 0
%36961 = torch.prim.ListConstruct %int1_34630, %int0_34631 : (!torch.int, !torch.int) -> !torch.list<int>
%36962 = torch.aten.permute %1343, %36961 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_34632 = torch.constant.int 4
%36963 = torch.aten.mul.int %int4_34632, %36252 : !torch.int, !torch.int -> !torch.int
%int512_34633 = torch.constant.int 512
%36964 = torch.prim.ListConstruct %36963, %int512_34633 : (!torch.int, !torch.int) -> !torch.list<int>
%36965 = torch.aten.view %36932, %36964 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36965, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%36966 = torch.aten.mm %36965, %36948 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36966, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34634 = torch.constant.int 4
%int4096_34635 = torch.constant.int 4096
%36967 = torch.prim.ListConstruct %int4_34634, %36252, %int4096_34635 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36968 = torch.aten.view %36966, %36967 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %36968, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34636 = torch.constant.int 4
%36969 = torch.aten.mul.int %int4_34636, %36267 : !torch.int, !torch.int -> !torch.int
%int512_34637 = torch.constant.int 512
%36970 = torch.prim.ListConstruct %36969, %int512_34637 : (!torch.int, !torch.int) -> !torch.list<int>
%36971 = torch.aten.view %36934, %36970 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36971, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%36972 = torch.aten.mm %36971, %36950 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36972, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34638 = torch.constant.int 4
%int4096_34639 = torch.constant.int 4096
%36973 = torch.prim.ListConstruct %int4_34638, %36267, %int4096_34639 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36974 = torch.aten.view %36972, %36973 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %36974, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34640 = torch.constant.int 4
%36975 = torch.aten.mul.int %int4_34640, %36282 : !torch.int, !torch.int -> !torch.int
%int512_34641 = torch.constant.int 512
%36976 = torch.prim.ListConstruct %36975, %int512_34641 : (!torch.int, !torch.int) -> !torch.list<int>
%36977 = torch.aten.view %36936, %36976 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36977, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%36978 = torch.aten.mm %36977, %36952 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36978, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34642 = torch.constant.int 4
%int4096_34643 = torch.constant.int 4096
%36979 = torch.prim.ListConstruct %int4_34642, %36282, %int4096_34643 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36980 = torch.aten.view %36978, %36979 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %36980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34644 = torch.constant.int 4
%36981 = torch.aten.mul.int %int4_34644, %36297 : !torch.int, !torch.int -> !torch.int
%int512_34645 = torch.constant.int 512
%36982 = torch.prim.ListConstruct %36981, %int512_34645 : (!torch.int, !torch.int) -> !torch.list<int>
%36983 = torch.aten.view %36938, %36982 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36983, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%36984 = torch.aten.mm %36983, %36954 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36984, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34646 = torch.constant.int 4
%int4096_34647 = torch.constant.int 4096
%36985 = torch.prim.ListConstruct %int4_34646, %36297, %int4096_34647 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36986 = torch.aten.view %36984, %36985 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %36986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34648 = torch.constant.int 4
%36987 = torch.aten.mul.int %int4_34648, %36312 : !torch.int, !torch.int -> !torch.int
%int512_34649 = torch.constant.int 512
%36988 = torch.prim.ListConstruct %36987, %int512_34649 : (!torch.int, !torch.int) -> !torch.list<int>
%36989 = torch.aten.view %36940, %36988 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36989, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%36990 = torch.aten.mm %36989, %36956 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36990, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34650 = torch.constant.int 4
%int4096_34651 = torch.constant.int 4096
%36991 = torch.prim.ListConstruct %int4_34650, %36312, %int4096_34651 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36992 = torch.aten.view %36990, %36991 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %36992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34652 = torch.constant.int 4
%36993 = torch.aten.mul.int %int4_34652, %36327 : !torch.int, !torch.int -> !torch.int
%int512_34653 = torch.constant.int 512
%36994 = torch.prim.ListConstruct %36993, %int512_34653 : (!torch.int, !torch.int) -> !torch.list<int>
%36995 = torch.aten.view %36942, %36994 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %36995, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%36996 = torch.aten.mm %36995, %36958 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %36996, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34654 = torch.constant.int 4
%int4096_34655 = torch.constant.int 4096
%36997 = torch.prim.ListConstruct %int4_34654, %36327, %int4096_34655 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%36998 = torch.aten.view %36996, %36997 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %36998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34656 = torch.constant.int 4
%36999 = torch.aten.mul.int %int4_34656, %36342 : !torch.int, !torch.int -> !torch.int
%int512_34657 = torch.constant.int 512
%37000 = torch.prim.ListConstruct %36999, %int512_34657 : (!torch.int, !torch.int) -> !torch.list<int>
%37001 = torch.aten.view %36944, %37000 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37001, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%37002 = torch.aten.mm %37001, %36960 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37002, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34658 = torch.constant.int 4
%int4096_34659 = torch.constant.int 4096
%37003 = torch.prim.ListConstruct %int4_34658, %36342, %int4096_34659 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37004 = torch.aten.view %37002, %37003 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_34660 = torch.constant.int 4
%37005 = torch.aten.mul.int %int4_34660, %36357 : !torch.int, !torch.int -> !torch.int
%int512_34661 = torch.constant.int 512
%37006 = torch.prim.ListConstruct %37005, %int512_34661 : (!torch.int, !torch.int) -> !torch.list<int>
%37007 = torch.aten.view %36946, %37006 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37007, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%37008 = torch.aten.mm %37007, %36962 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37008, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_34662 = torch.constant.int 4
%int4096_34663 = torch.constant.int 4096
%37009 = torch.prim.ListConstruct %int4_34662, %36357, %int4096_34663 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37010 = torch.aten.view %37008, %37009 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37011 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34664 = arith.constant 1 : index
%dim_34665 = tensor.dim %37011, %c1_34664 : tensor<4x?x4096xf16>
%37012 = flow.tensor.transfer %37011 : tensor<4x?x4096xf16>{%dim_34665} to #hal.device.promise<@__device_0>
%37013 = torch_c.from_builtin_tensor %37012 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37014 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34666 = arith.constant 1 : index
%dim_34667 = tensor.dim %37014, %c1_34666 : tensor<4x?x4096xf16>
%37015 = flow.tensor.transfer %37014 : tensor<4x?x4096xf16>{%dim_34667} to #hal.device.promise<@__device_0>
%37016 = torch_c.from_builtin_tensor %37015 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37016, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37017 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34668 = arith.constant 1 : index
%dim_34669 = tensor.dim %37017, %c1_34668 : tensor<4x?x4096xf16>
%37018 = flow.tensor.transfer %37017 : tensor<4x?x4096xf16>{%dim_34669} to #hal.device.promise<@__device_0>
%37019 = torch_c.from_builtin_tensor %37018 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37019, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37020 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34670 = arith.constant 1 : index
%dim_34671 = tensor.dim %37020, %c1_34670 : tensor<4x?x4096xf16>
%37021 = flow.tensor.transfer %37020 : tensor<4x?x4096xf16>{%dim_34671} to #hal.device.promise<@__device_0>
%37022 = torch_c.from_builtin_tensor %37021 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37022, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37023 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34672 = arith.constant 1 : index
%dim_34673 = tensor.dim %37023, %c1_34672 : tensor<4x?x4096xf16>
%37024 = flow.tensor.transfer %37023 : tensor<4x?x4096xf16>{%dim_34673} to #hal.device.promise<@__device_0>
%37025 = torch_c.from_builtin_tensor %37024 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37025, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37026 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34674 = arith.constant 1 : index
%dim_34675 = tensor.dim %37026, %c1_34674 : tensor<4x?x4096xf16>
%37027 = flow.tensor.transfer %37026 : tensor<4x?x4096xf16>{%dim_34675} to #hal.device.promise<@__device_0>
%37028 = torch_c.from_builtin_tensor %37027 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37029 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34676 = arith.constant 1 : index
%dim_34677 = tensor.dim %37029, %c1_34676 : tensor<4x?x4096xf16>
%37030 = flow.tensor.transfer %37029 : tensor<4x?x4096xf16>{%dim_34677} to #hal.device.promise<@__device_0>
%37031 = torch_c.from_builtin_tensor %37030 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37031, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34678 = torch.constant.int 1
%37032 = torch.aten.add.Tensor %36968, %37013, %int1_34678 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34679 = torch.constant.int 1
%37033 = torch.aten.add.Tensor %37032, %37016, %int1_34679 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34680 = torch.constant.int 1
%37034 = torch.aten.add.Tensor %37033, %37019, %int1_34680 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34681 = torch.constant.int 1
%37035 = torch.aten.add.Tensor %37034, %37022, %int1_34681 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34682 = torch.constant.int 1
%37036 = torch.aten.add.Tensor %37035, %37025, %int1_34682 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34683 = torch.constant.int 1
%37037 = torch.aten.add.Tensor %37036, %37028, %int1_34683 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34684 = torch.constant.int 1
%37038 = torch.aten.add.Tensor %37037, %37031, %int1_34684 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37039 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34685 = arith.constant 1 : index
%dim_34686 = tensor.dim %37039, %c1_34685 : tensor<4x?x4096xf16>
%37040 = flow.tensor.transfer %37039 : tensor<4x?x4096xf16>{%dim_34686} to #hal.device.promise<@__device_1>
%37041 = torch_c.from_builtin_tensor %37040 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37041, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37042 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34687 = arith.constant 1 : index
%dim_34688 = tensor.dim %37042, %c1_34687 : tensor<4x?x4096xf16>
%37043 = flow.tensor.transfer %37042 : tensor<4x?x4096xf16>{%dim_34688} to #hal.device.promise<@__device_1>
%37044 = torch_c.from_builtin_tensor %37043 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37045 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34689 = arith.constant 1 : index
%dim_34690 = tensor.dim %37045, %c1_34689 : tensor<4x?x4096xf16>
%37046 = flow.tensor.transfer %37045 : tensor<4x?x4096xf16>{%dim_34690} to #hal.device.promise<@__device_1>
%37047 = torch_c.from_builtin_tensor %37046 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37048 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34691 = arith.constant 1 : index
%dim_34692 = tensor.dim %37048, %c1_34691 : tensor<4x?x4096xf16>
%37049 = flow.tensor.transfer %37048 : tensor<4x?x4096xf16>{%dim_34692} to #hal.device.promise<@__device_1>
%37050 = torch_c.from_builtin_tensor %37049 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37051 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34693 = arith.constant 1 : index
%dim_34694 = tensor.dim %37051, %c1_34693 : tensor<4x?x4096xf16>
%37052 = flow.tensor.transfer %37051 : tensor<4x?x4096xf16>{%dim_34694} to #hal.device.promise<@__device_1>
%37053 = torch_c.from_builtin_tensor %37052 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37053, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37054 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34695 = arith.constant 1 : index
%dim_34696 = tensor.dim %37054, %c1_34695 : tensor<4x?x4096xf16>
%37055 = flow.tensor.transfer %37054 : tensor<4x?x4096xf16>{%dim_34696} to #hal.device.promise<@__device_1>
%37056 = torch_c.from_builtin_tensor %37055 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37057 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34697 = arith.constant 1 : index
%dim_34698 = tensor.dim %37057, %c1_34697 : tensor<4x?x4096xf16>
%37058 = flow.tensor.transfer %37057 : tensor<4x?x4096xf16>{%dim_34698} to #hal.device.promise<@__device_1>
%37059 = torch_c.from_builtin_tensor %37058 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34699 = torch.constant.int 1
%37060 = torch.aten.add.Tensor %37041, %36974, %int1_34699 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34700 = torch.constant.int 1
%37061 = torch.aten.add.Tensor %37060, %37044, %int1_34700 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34701 = torch.constant.int 1
%37062 = torch.aten.add.Tensor %37061, %37047, %int1_34701 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34702 = torch.constant.int 1
%37063 = torch.aten.add.Tensor %37062, %37050, %int1_34702 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34703 = torch.constant.int 1
%37064 = torch.aten.add.Tensor %37063, %37053, %int1_34703 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34704 = torch.constant.int 1
%37065 = torch.aten.add.Tensor %37064, %37056, %int1_34704 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34705 = torch.constant.int 1
%37066 = torch.aten.add.Tensor %37065, %37059, %int1_34705 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37067 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34706 = arith.constant 1 : index
%dim_34707 = tensor.dim %37067, %c1_34706 : tensor<4x?x4096xf16>
%37068 = flow.tensor.transfer %37067 : tensor<4x?x4096xf16>{%dim_34707} to #hal.device.promise<@__device_2>
%37069 = torch_c.from_builtin_tensor %37068 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37069, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37070 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34708 = arith.constant 1 : index
%dim_34709 = tensor.dim %37070, %c1_34708 : tensor<4x?x4096xf16>
%37071 = flow.tensor.transfer %37070 : tensor<4x?x4096xf16>{%dim_34709} to #hal.device.promise<@__device_2>
%37072 = torch_c.from_builtin_tensor %37071 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37073 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34710 = arith.constant 1 : index
%dim_34711 = tensor.dim %37073, %c1_34710 : tensor<4x?x4096xf16>
%37074 = flow.tensor.transfer %37073 : tensor<4x?x4096xf16>{%dim_34711} to #hal.device.promise<@__device_2>
%37075 = torch_c.from_builtin_tensor %37074 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37075, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37076 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34712 = arith.constant 1 : index
%dim_34713 = tensor.dim %37076, %c1_34712 : tensor<4x?x4096xf16>
%37077 = flow.tensor.transfer %37076 : tensor<4x?x4096xf16>{%dim_34713} to #hal.device.promise<@__device_2>
%37078 = torch_c.from_builtin_tensor %37077 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37079 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34714 = arith.constant 1 : index
%dim_34715 = tensor.dim %37079, %c1_34714 : tensor<4x?x4096xf16>
%37080 = flow.tensor.transfer %37079 : tensor<4x?x4096xf16>{%dim_34715} to #hal.device.promise<@__device_2>
%37081 = torch_c.from_builtin_tensor %37080 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37081, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37082 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34716 = arith.constant 1 : index
%dim_34717 = tensor.dim %37082, %c1_34716 : tensor<4x?x4096xf16>
%37083 = flow.tensor.transfer %37082 : tensor<4x?x4096xf16>{%dim_34717} to #hal.device.promise<@__device_2>
%37084 = torch_c.from_builtin_tensor %37083 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37084, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37085 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34718 = arith.constant 1 : index
%dim_34719 = tensor.dim %37085, %c1_34718 : tensor<4x?x4096xf16>
%37086 = flow.tensor.transfer %37085 : tensor<4x?x4096xf16>{%dim_34719} to #hal.device.promise<@__device_2>
%37087 = torch_c.from_builtin_tensor %37086 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37087, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34720 = torch.constant.int 1
%37088 = torch.aten.add.Tensor %37069, %37072, %int1_34720 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34721 = torch.constant.int 1
%37089 = torch.aten.add.Tensor %37088, %36980, %int1_34721 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34722 = torch.constant.int 1
%37090 = torch.aten.add.Tensor %37089, %37075, %int1_34722 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34723 = torch.constant.int 1
%37091 = torch.aten.add.Tensor %37090, %37078, %int1_34723 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34724 = torch.constant.int 1
%37092 = torch.aten.add.Tensor %37091, %37081, %int1_34724 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34725 = torch.constant.int 1
%37093 = torch.aten.add.Tensor %37092, %37084, %int1_34725 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34726 = torch.constant.int 1
%37094 = torch.aten.add.Tensor %37093, %37087, %int1_34726 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37095 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34727 = arith.constant 1 : index
%dim_34728 = tensor.dim %37095, %c1_34727 : tensor<4x?x4096xf16>
%37096 = flow.tensor.transfer %37095 : tensor<4x?x4096xf16>{%dim_34728} to #hal.device.promise<@__device_3>
%37097 = torch_c.from_builtin_tensor %37096 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37098 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34729 = arith.constant 1 : index
%dim_34730 = tensor.dim %37098, %c1_34729 : tensor<4x?x4096xf16>
%37099 = flow.tensor.transfer %37098 : tensor<4x?x4096xf16>{%dim_34730} to #hal.device.promise<@__device_3>
%37100 = torch_c.from_builtin_tensor %37099 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37101 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34731 = arith.constant 1 : index
%dim_34732 = tensor.dim %37101, %c1_34731 : tensor<4x?x4096xf16>
%37102 = flow.tensor.transfer %37101 : tensor<4x?x4096xf16>{%dim_34732} to #hal.device.promise<@__device_3>
%37103 = torch_c.from_builtin_tensor %37102 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37104 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34733 = arith.constant 1 : index
%dim_34734 = tensor.dim %37104, %c1_34733 : tensor<4x?x4096xf16>
%37105 = flow.tensor.transfer %37104 : tensor<4x?x4096xf16>{%dim_34734} to #hal.device.promise<@__device_3>
%37106 = torch_c.from_builtin_tensor %37105 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37107 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34735 = arith.constant 1 : index
%dim_34736 = tensor.dim %37107, %c1_34735 : tensor<4x?x4096xf16>
%37108 = flow.tensor.transfer %37107 : tensor<4x?x4096xf16>{%dim_34736} to #hal.device.promise<@__device_3>
%37109 = torch_c.from_builtin_tensor %37108 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37110 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34737 = arith.constant 1 : index
%dim_34738 = tensor.dim %37110, %c1_34737 : tensor<4x?x4096xf16>
%37111 = flow.tensor.transfer %37110 : tensor<4x?x4096xf16>{%dim_34738} to #hal.device.promise<@__device_3>
%37112 = torch_c.from_builtin_tensor %37111 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37113 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34739 = arith.constant 1 : index
%dim_34740 = tensor.dim %37113, %c1_34739 : tensor<4x?x4096xf16>
%37114 = flow.tensor.transfer %37113 : tensor<4x?x4096xf16>{%dim_34740} to #hal.device.promise<@__device_3>
%37115 = torch_c.from_builtin_tensor %37114 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34741 = torch.constant.int 1
%37116 = torch.aten.add.Tensor %37097, %37100, %int1_34741 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34742 = torch.constant.int 1
%37117 = torch.aten.add.Tensor %37116, %37103, %int1_34742 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34743 = torch.constant.int 1
%37118 = torch.aten.add.Tensor %37117, %36986, %int1_34743 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34744 = torch.constant.int 1
%37119 = torch.aten.add.Tensor %37118, %37106, %int1_34744 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34745 = torch.constant.int 1
%37120 = torch.aten.add.Tensor %37119, %37109, %int1_34745 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37120, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34746 = torch.constant.int 1
%37121 = torch.aten.add.Tensor %37120, %37112, %int1_34746 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34747 = torch.constant.int 1
%37122 = torch.aten.add.Tensor %37121, %37115, %int1_34747 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37123 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34748 = arith.constant 1 : index
%dim_34749 = tensor.dim %37123, %c1_34748 : tensor<4x?x4096xf16>
%37124 = flow.tensor.transfer %37123 : tensor<4x?x4096xf16>{%dim_34749} to #hal.device.promise<@__device_4>
%37125 = torch_c.from_builtin_tensor %37124 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37126 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34750 = arith.constant 1 : index
%dim_34751 = tensor.dim %37126, %c1_34750 : tensor<4x?x4096xf16>
%37127 = flow.tensor.transfer %37126 : tensor<4x?x4096xf16>{%dim_34751} to #hal.device.promise<@__device_4>
%37128 = torch_c.from_builtin_tensor %37127 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37128, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37129 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34752 = arith.constant 1 : index
%dim_34753 = tensor.dim %37129, %c1_34752 : tensor<4x?x4096xf16>
%37130 = flow.tensor.transfer %37129 : tensor<4x?x4096xf16>{%dim_34753} to #hal.device.promise<@__device_4>
%37131 = torch_c.from_builtin_tensor %37130 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37132 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34754 = arith.constant 1 : index
%dim_34755 = tensor.dim %37132, %c1_34754 : tensor<4x?x4096xf16>
%37133 = flow.tensor.transfer %37132 : tensor<4x?x4096xf16>{%dim_34755} to #hal.device.promise<@__device_4>
%37134 = torch_c.from_builtin_tensor %37133 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37134, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37135 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34756 = arith.constant 1 : index
%dim_34757 = tensor.dim %37135, %c1_34756 : tensor<4x?x4096xf16>
%37136 = flow.tensor.transfer %37135 : tensor<4x?x4096xf16>{%dim_34757} to #hal.device.promise<@__device_4>
%37137 = torch_c.from_builtin_tensor %37136 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37138 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34758 = arith.constant 1 : index
%dim_34759 = tensor.dim %37138, %c1_34758 : tensor<4x?x4096xf16>
%37139 = flow.tensor.transfer %37138 : tensor<4x?x4096xf16>{%dim_34759} to #hal.device.promise<@__device_4>
%37140 = torch_c.from_builtin_tensor %37139 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37141 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34760 = arith.constant 1 : index
%dim_34761 = tensor.dim %37141, %c1_34760 : tensor<4x?x4096xf16>
%37142 = flow.tensor.transfer %37141 : tensor<4x?x4096xf16>{%dim_34761} to #hal.device.promise<@__device_4>
%37143 = torch_c.from_builtin_tensor %37142 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34762 = torch.constant.int 1
%37144 = torch.aten.add.Tensor %37125, %37128, %int1_34762 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34763 = torch.constant.int 1
%37145 = torch.aten.add.Tensor %37144, %37131, %int1_34763 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34764 = torch.constant.int 1
%37146 = torch.aten.add.Tensor %37145, %37134, %int1_34764 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34765 = torch.constant.int 1
%37147 = torch.aten.add.Tensor %37146, %36992, %int1_34765 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34766 = torch.constant.int 1
%37148 = torch.aten.add.Tensor %37147, %37137, %int1_34766 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34767 = torch.constant.int 1
%37149 = torch.aten.add.Tensor %37148, %37140, %int1_34767 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34768 = torch.constant.int 1
%37150 = torch.aten.add.Tensor %37149, %37143, %int1_34768 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37151 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34769 = arith.constant 1 : index
%dim_34770 = tensor.dim %37151, %c1_34769 : tensor<4x?x4096xf16>
%37152 = flow.tensor.transfer %37151 : tensor<4x?x4096xf16>{%dim_34770} to #hal.device.promise<@__device_5>
%37153 = torch_c.from_builtin_tensor %37152 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37154 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34771 = arith.constant 1 : index
%dim_34772 = tensor.dim %37154, %c1_34771 : tensor<4x?x4096xf16>
%37155 = flow.tensor.transfer %37154 : tensor<4x?x4096xf16>{%dim_34772} to #hal.device.promise<@__device_5>
%37156 = torch_c.from_builtin_tensor %37155 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37157 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34773 = arith.constant 1 : index
%dim_34774 = tensor.dim %37157, %c1_34773 : tensor<4x?x4096xf16>
%37158 = flow.tensor.transfer %37157 : tensor<4x?x4096xf16>{%dim_34774} to #hal.device.promise<@__device_5>
%37159 = torch_c.from_builtin_tensor %37158 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37160 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34775 = arith.constant 1 : index
%dim_34776 = tensor.dim %37160, %c1_34775 : tensor<4x?x4096xf16>
%37161 = flow.tensor.transfer %37160 : tensor<4x?x4096xf16>{%dim_34776} to #hal.device.promise<@__device_5>
%37162 = torch_c.from_builtin_tensor %37161 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37163 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34777 = arith.constant 1 : index
%dim_34778 = tensor.dim %37163, %c1_34777 : tensor<4x?x4096xf16>
%37164 = flow.tensor.transfer %37163 : tensor<4x?x4096xf16>{%dim_34778} to #hal.device.promise<@__device_5>
%37165 = torch_c.from_builtin_tensor %37164 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37166 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34779 = arith.constant 1 : index
%dim_34780 = tensor.dim %37166, %c1_34779 : tensor<4x?x4096xf16>
%37167 = flow.tensor.transfer %37166 : tensor<4x?x4096xf16>{%dim_34780} to #hal.device.promise<@__device_5>
%37168 = torch_c.from_builtin_tensor %37167 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37169 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34781 = arith.constant 1 : index
%dim_34782 = tensor.dim %37169, %c1_34781 : tensor<4x?x4096xf16>
%37170 = flow.tensor.transfer %37169 : tensor<4x?x4096xf16>{%dim_34782} to #hal.device.promise<@__device_5>
%37171 = torch_c.from_builtin_tensor %37170 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34783 = torch.constant.int 1
%37172 = torch.aten.add.Tensor %37153, %37156, %int1_34783 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34784 = torch.constant.int 1
%37173 = torch.aten.add.Tensor %37172, %37159, %int1_34784 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34785 = torch.constant.int 1
%37174 = torch.aten.add.Tensor %37173, %37162, %int1_34785 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34786 = torch.constant.int 1
%37175 = torch.aten.add.Tensor %37174, %37165, %int1_34786 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34787 = torch.constant.int 1
%37176 = torch.aten.add.Tensor %37175, %36998, %int1_34787 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37176, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34788 = torch.constant.int 1
%37177 = torch.aten.add.Tensor %37176, %37168, %int1_34788 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37177, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34789 = torch.constant.int 1
%37178 = torch.aten.add.Tensor %37177, %37171, %int1_34789 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37178, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37179 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34790 = arith.constant 1 : index
%dim_34791 = tensor.dim %37179, %c1_34790 : tensor<4x?x4096xf16>
%37180 = flow.tensor.transfer %37179 : tensor<4x?x4096xf16>{%dim_34791} to #hal.device.promise<@__device_6>
%37181 = torch_c.from_builtin_tensor %37180 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37181, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37182 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34792 = arith.constant 1 : index
%dim_34793 = tensor.dim %37182, %c1_34792 : tensor<4x?x4096xf16>
%37183 = flow.tensor.transfer %37182 : tensor<4x?x4096xf16>{%dim_34793} to #hal.device.promise<@__device_6>
%37184 = torch_c.from_builtin_tensor %37183 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37184, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37185 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34794 = arith.constant 1 : index
%dim_34795 = tensor.dim %37185, %c1_34794 : tensor<4x?x4096xf16>
%37186 = flow.tensor.transfer %37185 : tensor<4x?x4096xf16>{%dim_34795} to #hal.device.promise<@__device_6>
%37187 = torch_c.from_builtin_tensor %37186 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37188 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34796 = arith.constant 1 : index
%dim_34797 = tensor.dim %37188, %c1_34796 : tensor<4x?x4096xf16>
%37189 = flow.tensor.transfer %37188 : tensor<4x?x4096xf16>{%dim_34797} to #hal.device.promise<@__device_6>
%37190 = torch_c.from_builtin_tensor %37189 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37190, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37191 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34798 = arith.constant 1 : index
%dim_34799 = tensor.dim %37191, %c1_34798 : tensor<4x?x4096xf16>
%37192 = flow.tensor.transfer %37191 : tensor<4x?x4096xf16>{%dim_34799} to #hal.device.promise<@__device_6>
%37193 = torch_c.from_builtin_tensor %37192 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37193, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37194 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34800 = arith.constant 1 : index
%dim_34801 = tensor.dim %37194, %c1_34800 : tensor<4x?x4096xf16>
%37195 = flow.tensor.transfer %37194 : tensor<4x?x4096xf16>{%dim_34801} to #hal.device.promise<@__device_6>
%37196 = torch_c.from_builtin_tensor %37195 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37196, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37197 = torch_c.to_builtin_tensor %37010 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34802 = arith.constant 1 : index
%dim_34803 = tensor.dim %37197, %c1_34802 : tensor<4x?x4096xf16>
%37198 = flow.tensor.transfer %37197 : tensor<4x?x4096xf16>{%dim_34803} to #hal.device.promise<@__device_6>
%37199 = torch_c.from_builtin_tensor %37198 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37199, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34804 = torch.constant.int 1
%37200 = torch.aten.add.Tensor %37181, %37184, %int1_34804 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37200, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34805 = torch.constant.int 1
%37201 = torch.aten.add.Tensor %37200, %37187, %int1_34805 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34806 = torch.constant.int 1
%37202 = torch.aten.add.Tensor %37201, %37190, %int1_34806 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34807 = torch.constant.int 1
%37203 = torch.aten.add.Tensor %37202, %37193, %int1_34807 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34808 = torch.constant.int 1
%37204 = torch.aten.add.Tensor %37203, %37196, %int1_34808 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37204, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34809 = torch.constant.int 1
%37205 = torch.aten.add.Tensor %37204, %37004, %int1_34809 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37205, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34810 = torch.constant.int 1
%37206 = torch.aten.add.Tensor %37205, %37199, %int1_34810 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37206, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37207 = torch_c.to_builtin_tensor %36968 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34811 = arith.constant 1 : index
%dim_34812 = tensor.dim %37207, %c1_34811 : tensor<4x?x4096xf16>
%37208 = flow.tensor.transfer %37207 : tensor<4x?x4096xf16>{%dim_34812} to #hal.device.promise<@__device_7>
%37209 = torch_c.from_builtin_tensor %37208 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37210 = torch_c.to_builtin_tensor %36974 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34813 = arith.constant 1 : index
%dim_34814 = tensor.dim %37210, %c1_34813 : tensor<4x?x4096xf16>
%37211 = flow.tensor.transfer %37210 : tensor<4x?x4096xf16>{%dim_34814} to #hal.device.promise<@__device_7>
%37212 = torch_c.from_builtin_tensor %37211 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37213 = torch_c.to_builtin_tensor %36980 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34815 = arith.constant 1 : index
%dim_34816 = tensor.dim %37213, %c1_34815 : tensor<4x?x4096xf16>
%37214 = flow.tensor.transfer %37213 : tensor<4x?x4096xf16>{%dim_34816} to #hal.device.promise<@__device_7>
%37215 = torch_c.from_builtin_tensor %37214 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37216 = torch_c.to_builtin_tensor %36986 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34817 = arith.constant 1 : index
%dim_34818 = tensor.dim %37216, %c1_34817 : tensor<4x?x4096xf16>
%37217 = flow.tensor.transfer %37216 : tensor<4x?x4096xf16>{%dim_34818} to #hal.device.promise<@__device_7>
%37218 = torch_c.from_builtin_tensor %37217 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37218, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37219 = torch_c.to_builtin_tensor %36992 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34819 = arith.constant 1 : index
%dim_34820 = tensor.dim %37219, %c1_34819 : tensor<4x?x4096xf16>
%37220 = flow.tensor.transfer %37219 : tensor<4x?x4096xf16>{%dim_34820} to #hal.device.promise<@__device_7>
%37221 = torch_c.from_builtin_tensor %37220 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37222 = torch_c.to_builtin_tensor %36998 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34821 = arith.constant 1 : index
%dim_34822 = tensor.dim %37222, %c1_34821 : tensor<4x?x4096xf16>
%37223 = flow.tensor.transfer %37222 : tensor<4x?x4096xf16>{%dim_34822} to #hal.device.promise<@__device_7>
%37224 = torch_c.from_builtin_tensor %37223 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37224, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37225 = torch_c.to_builtin_tensor %37004 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_34823 = arith.constant 1 : index
%dim_34824 = tensor.dim %37225, %c1_34823 : tensor<4x?x4096xf16>
%37226 = flow.tensor.transfer %37225 : tensor<4x?x4096xf16>{%dim_34824} to #hal.device.promise<@__device_7>
%37227 = torch_c.from_builtin_tensor %37226 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34825 = torch.constant.int 1
%37228 = torch.aten.add.Tensor %37209, %37212, %int1_34825 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34826 = torch.constant.int 1
%37229 = torch.aten.add.Tensor %37228, %37215, %int1_34826 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37229, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34827 = torch.constant.int 1
%37230 = torch.aten.add.Tensor %37229, %37218, %int1_34827 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37230, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34828 = torch.constant.int 1
%37231 = torch.aten.add.Tensor %37230, %37221, %int1_34828 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34829 = torch.constant.int 1
%37232 = torch.aten.add.Tensor %37231, %37224, %int1_34829 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34830 = torch.constant.int 1
%37233 = torch.aten.add.Tensor %37232, %37227, %int1_34830 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34831 = torch.constant.int 1
%37234 = torch.aten.add.Tensor %37233, %37010, %int1_34831 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37234, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34832 = torch.constant.int 1
%37235 = torch.aten.add.Tensor %35894, %37038, %int1_34832 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37235, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34833 = torch.constant.int 1
%37236 = torch.aten.add.Tensor %35895, %37066, %int1_34833 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37236, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34834 = torch.constant.int 1
%37237 = torch.aten.add.Tensor %35896, %37094, %int1_34834 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37237, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34835 = torch.constant.int 1
%37238 = torch.aten.add.Tensor %35897, %37122, %int1_34835 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37238, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34836 = torch.constant.int 1
%37239 = torch.aten.add.Tensor %35898, %37150, %int1_34836 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34837 = torch.constant.int 1
%37240 = torch.aten.add.Tensor %35899, %37178, %int1_34837 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34838 = torch.constant.int 1
%37241 = torch.aten.add.Tensor %35900, %37206, %int1_34838 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34839 = torch.constant.int 1
%37242 = torch.aten.add.Tensor %35901, %37234, %int1_34839 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_34840 = torch.constant.int 6
%37243 = torch.prims.convert_element_type %37235, %int6_34840 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34841 = torch.constant.int 6
%37244 = torch.prims.convert_element_type %37236, %int6_34841 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34842 = torch.constant.int 6
%37245 = torch.prims.convert_element_type %37237, %int6_34842 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34843 = torch.constant.int 6
%37246 = torch.prims.convert_element_type %37238, %int6_34843 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34844 = torch.constant.int 6
%37247 = torch.prims.convert_element_type %37239, %int6_34844 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34845 = torch.constant.int 6
%37248 = torch.prims.convert_element_type %37240, %int6_34845 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37248, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34846 = torch.constant.int 6
%37249 = torch.prims.convert_element_type %37241, %int6_34846 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37249, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_34847 = torch.constant.int 6
%37250 = torch.prims.convert_element_type %37242, %int6_34847 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37250, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34848 = torch.constant.int 2
%37251 = torch.aten.pow.Tensor_Scalar %37243, %int2_34848 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37251, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34849 = torch.constant.int 2
%37252 = torch.aten.pow.Tensor_Scalar %37244, %int2_34849 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37252, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34850 = torch.constant.int 2
%37253 = torch.aten.pow.Tensor_Scalar %37245, %int2_34850 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37253, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34851 = torch.constant.int 2
%37254 = torch.aten.pow.Tensor_Scalar %37246, %int2_34851 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37254, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34852 = torch.constant.int 2
%37255 = torch.aten.pow.Tensor_Scalar %37247, %int2_34852 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37255, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34853 = torch.constant.int 2
%37256 = torch.aten.pow.Tensor_Scalar %37248, %int2_34853 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37256, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34854 = torch.constant.int 2
%37257 = torch.aten.pow.Tensor_Scalar %37249, %int2_34854 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37257, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_34855 = torch.constant.int 2
%37258 = torch.aten.pow.Tensor_Scalar %37250, %int2_34855 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37258, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_34856 = torch.constant.int -1
%37259 = torch.prim.ListConstruct %int-1_34856 : (!torch.int) -> !torch.list<int>
%true_34857 = torch.constant.bool true
%none_34858 = torch.constant.none
%37260 = torch.aten.mean.dim %37251, %37259, %true_34857, %none_34858 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37260, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34859 = torch.constant.int -1
%37261 = torch.prim.ListConstruct %int-1_34859 : (!torch.int) -> !torch.list<int>
%true_34860 = torch.constant.bool true
%none_34861 = torch.constant.none
%37262 = torch.aten.mean.dim %37252, %37261, %true_34860, %none_34861 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37262, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34862 = torch.constant.int -1
%37263 = torch.prim.ListConstruct %int-1_34862 : (!torch.int) -> !torch.list<int>
%true_34863 = torch.constant.bool true
%none_34864 = torch.constant.none
%37264 = torch.aten.mean.dim %37253, %37263, %true_34863, %none_34864 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37264, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34865 = torch.constant.int -1
%37265 = torch.prim.ListConstruct %int-1_34865 : (!torch.int) -> !torch.list<int>
%true_34866 = torch.constant.bool true
%none_34867 = torch.constant.none
%37266 = torch.aten.mean.dim %37254, %37265, %true_34866, %none_34867 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37266, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34868 = torch.constant.int -1
%37267 = torch.prim.ListConstruct %int-1_34868 : (!torch.int) -> !torch.list<int>
%true_34869 = torch.constant.bool true
%none_34870 = torch.constant.none
%37268 = torch.aten.mean.dim %37255, %37267, %true_34869, %none_34870 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37268, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34871 = torch.constant.int -1
%37269 = torch.prim.ListConstruct %int-1_34871 : (!torch.int) -> !torch.list<int>
%true_34872 = torch.constant.bool true
%none_34873 = torch.constant.none
%37270 = torch.aten.mean.dim %37256, %37269, %true_34872, %none_34873 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37270, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34874 = torch.constant.int -1
%37271 = torch.prim.ListConstruct %int-1_34874 : (!torch.int) -> !torch.list<int>
%true_34875 = torch.constant.bool true
%none_34876 = torch.constant.none
%37272 = torch.aten.mean.dim %37257, %37271, %true_34875, %none_34876 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37272, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_34877 = torch.constant.int -1
%37273 = torch.prim.ListConstruct %int-1_34877 : (!torch.int) -> !torch.list<int>
%true_34878 = torch.constant.bool true
%none_34879 = torch.constant.none
%37274 = torch.aten.mean.dim %37258, %37273, %true_34878, %none_34879 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37274, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34880 = torch.constant.float 9.9999997473787516E-6
%int1_34881 = torch.constant.int 1
%37275 = torch.aten.add.Scalar %37260, %float9.999990e-06_34880, %int1_34881 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34882 = torch.constant.float 9.9999997473787516E-6
%int1_34883 = torch.constant.int 1
%37276 = torch.aten.add.Scalar %37262, %float9.999990e-06_34882, %int1_34883 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37276, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34884 = torch.constant.float 9.9999997473787516E-6
%int1_34885 = torch.constant.int 1
%37277 = torch.aten.add.Scalar %37264, %float9.999990e-06_34884, %int1_34885 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37277, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34886 = torch.constant.float 9.9999997473787516E-6
%int1_34887 = torch.constant.int 1
%37278 = torch.aten.add.Scalar %37266, %float9.999990e-06_34886, %int1_34887 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37278, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34888 = torch.constant.float 9.9999997473787516E-6
%int1_34889 = torch.constant.int 1
%37279 = torch.aten.add.Scalar %37268, %float9.999990e-06_34888, %int1_34889 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37279, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34890 = torch.constant.float 9.9999997473787516E-6
%int1_34891 = torch.constant.int 1
%37280 = torch.aten.add.Scalar %37270, %float9.999990e-06_34890, %int1_34891 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34892 = torch.constant.float 9.9999997473787516E-6
%int1_34893 = torch.constant.int 1
%37281 = torch.aten.add.Scalar %37272, %float9.999990e-06_34892, %int1_34893 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_34894 = torch.constant.float 9.9999997473787516E-6
%int1_34895 = torch.constant.int 1
%37282 = torch.aten.add.Scalar %37274, %float9.999990e-06_34894, %int1_34895 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37282, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37283 = torch.aten.rsqrt %37275 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37283, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37284 = torch.aten.rsqrt %37276 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37285 = torch.aten.rsqrt %37277 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37286 = torch.aten.rsqrt %37278 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37286, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37287 = torch.aten.rsqrt %37279 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37288 = torch.aten.rsqrt %37280 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37288, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37289 = torch.aten.rsqrt %37281 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37289, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37290 = torch.aten.rsqrt %37282 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37290, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37291 = torch.aten.mul.Tensor %37243, %37283 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37291, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37292 = torch.aten.mul.Tensor %37244, %37284 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37292, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37293 = torch.aten.mul.Tensor %37245, %37285 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37294 = torch.aten.mul.Tensor %37246, %37286 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37294, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37295 = torch.aten.mul.Tensor %37247, %37287 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37296 = torch.aten.mul.Tensor %37248, %37288 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37297 = torch.aten.mul.Tensor %37249, %37289 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37297, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37298 = torch.aten.mul.Tensor %37250, %37290 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37298, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37299 = torch.aten.mul.Tensor %1344, %37291 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37300 = torch.aten.mul.Tensor %1345, %37292 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37301 = torch.aten.mul.Tensor %1346, %37293 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37301, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37302 = torch.aten.mul.Tensor %1347, %37294 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37302, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37303 = torch.aten.mul.Tensor %1348, %37295 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37303, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37304 = torch.aten.mul.Tensor %1349, %37296 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37304, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37305 = torch.aten.mul.Tensor %1350, %37297 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37306 = torch.aten.mul.Tensor %1351, %37298 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37306, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_34896 = torch.constant.int 5
%37307 = torch.prims.convert_element_type %37299, %int5_34896 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37307, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34897 = torch.constant.int 5
%37308 = torch.prims.convert_element_type %37300, %int5_34897 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37308, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34898 = torch.constant.int 5
%37309 = torch.prims.convert_element_type %37301, %int5_34898 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37309, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34899 = torch.constant.int 5
%37310 = torch.prims.convert_element_type %37302, %int5_34899 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34900 = torch.constant.int 5
%37311 = torch.prims.convert_element_type %37303, %int5_34900 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34901 = torch.constant.int 5
%37312 = torch.prims.convert_element_type %37304, %int5_34901 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34902 = torch.constant.int 5
%37313 = torch.prims.convert_element_type %37305, %int5_34902 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_34903 = torch.constant.int 5
%37314 = torch.prims.convert_element_type %37306, %int5_34903 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_34904 = torch.constant.int 1
%int0_34905 = torch.constant.int 0
%37315 = torch.prim.ListConstruct %int1_34904, %int0_34905 : (!torch.int, !torch.int) -> !torch.list<int>
%37316 = torch.aten.permute %1352, %37315 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34906 = torch.constant.int 1
%int0_34907 = torch.constant.int 0
%37317 = torch.prim.ListConstruct %int1_34906, %int0_34907 : (!torch.int, !torch.int) -> !torch.list<int>
%37318 = torch.aten.permute %1353, %37317 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34908 = torch.constant.int 1
%int0_34909 = torch.constant.int 0
%37319 = torch.prim.ListConstruct %int1_34908, %int0_34909 : (!torch.int, !torch.int) -> !torch.list<int>
%37320 = torch.aten.permute %1354, %37319 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34910 = torch.constant.int 1
%int0_34911 = torch.constant.int 0
%37321 = torch.prim.ListConstruct %int1_34910, %int0_34911 : (!torch.int, !torch.int) -> !torch.list<int>
%37322 = torch.aten.permute %1355, %37321 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34912 = torch.constant.int 1
%int0_34913 = torch.constant.int 0
%37323 = torch.prim.ListConstruct %int1_34912, %int0_34913 : (!torch.int, !torch.int) -> !torch.list<int>
%37324 = torch.aten.permute %1356, %37323 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34914 = torch.constant.int 1
%int0_34915 = torch.constant.int 0
%37325 = torch.prim.ListConstruct %int1_34914, %int0_34915 : (!torch.int, !torch.int) -> !torch.list<int>
%37326 = torch.aten.permute %1357, %37325 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34916 = torch.constant.int 1
%int0_34917 = torch.constant.int 0
%37327 = torch.prim.ListConstruct %int1_34916, %int0_34917 : (!torch.int, !torch.int) -> !torch.list<int>
%37328 = torch.aten.permute %1358, %37327 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34918 = torch.constant.int 1
%int0_34919 = torch.constant.int 0
%37329 = torch.prim.ListConstruct %int1_34918, %int0_34919 : (!torch.int, !torch.int) -> !torch.list<int>
%37330 = torch.aten.permute %1359, %37329 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_34920 = torch.constant.int 4
%37331 = torch.aten.mul.int %int4_34920, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34921 = torch.constant.int 4096
%37332 = torch.prim.ListConstruct %37331, %int4096_34921 : (!torch.int, !torch.int) -> !torch.list<int>
%37333 = torch.aten.view %37307, %37332 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37333, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37334 = torch.aten.mm %37333, %37316 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37334, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34922 = torch.constant.int 4
%int1792_34923 = torch.constant.int 1792
%37335 = torch.prim.ListConstruct %int4_34922, %2482, %int1792_34923 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37336 = torch.aten.view %37334, %37335 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37336, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34924 = torch.constant.int 4
%37337 = torch.aten.mul.int %int4_34924, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34925 = torch.constant.int 4096
%37338 = torch.prim.ListConstruct %37337, %int4096_34925 : (!torch.int, !torch.int) -> !torch.list<int>
%37339 = torch.aten.view %37308, %37338 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37339, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37340 = torch.aten.mm %37339, %37318 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37340, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34926 = torch.constant.int 4
%int1792_34927 = torch.constant.int 1792
%37341 = torch.prim.ListConstruct %int4_34926, %2482, %int1792_34927 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37342 = torch.aten.view %37340, %37341 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34928 = torch.constant.int 4
%37343 = torch.aten.mul.int %int4_34928, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34929 = torch.constant.int 4096
%37344 = torch.prim.ListConstruct %37343, %int4096_34929 : (!torch.int, !torch.int) -> !torch.list<int>
%37345 = torch.aten.view %37309, %37344 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37345, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37346 = torch.aten.mm %37345, %37320 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37346, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34930 = torch.constant.int 4
%int1792_34931 = torch.constant.int 1792
%37347 = torch.prim.ListConstruct %int4_34930, %2482, %int1792_34931 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37348 = torch.aten.view %37346, %37347 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37348, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34932 = torch.constant.int 4
%37349 = torch.aten.mul.int %int4_34932, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34933 = torch.constant.int 4096
%37350 = torch.prim.ListConstruct %37349, %int4096_34933 : (!torch.int, !torch.int) -> !torch.list<int>
%37351 = torch.aten.view %37310, %37350 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37351, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37352 = torch.aten.mm %37351, %37322 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37352, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34934 = torch.constant.int 4
%int1792_34935 = torch.constant.int 1792
%37353 = torch.prim.ListConstruct %int4_34934, %2482, %int1792_34935 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37354 = torch.aten.view %37352, %37353 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37354, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34936 = torch.constant.int 4
%37355 = torch.aten.mul.int %int4_34936, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34937 = torch.constant.int 4096
%37356 = torch.prim.ListConstruct %37355, %int4096_34937 : (!torch.int, !torch.int) -> !torch.list<int>
%37357 = torch.aten.view %37311, %37356 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37357, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37358 = torch.aten.mm %37357, %37324 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37358, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34938 = torch.constant.int 4
%int1792_34939 = torch.constant.int 1792
%37359 = torch.prim.ListConstruct %int4_34938, %2482, %int1792_34939 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37360 = torch.aten.view %37358, %37359 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34940 = torch.constant.int 4
%37361 = torch.aten.mul.int %int4_34940, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34941 = torch.constant.int 4096
%37362 = torch.prim.ListConstruct %37361, %int4096_34941 : (!torch.int, !torch.int) -> !torch.list<int>
%37363 = torch.aten.view %37312, %37362 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37363, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37364 = torch.aten.mm %37363, %37326 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37364, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34942 = torch.constant.int 4
%int1792_34943 = torch.constant.int 1792
%37365 = torch.prim.ListConstruct %int4_34942, %2482, %int1792_34943 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37366 = torch.aten.view %37364, %37365 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37366, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34944 = torch.constant.int 4
%37367 = torch.aten.mul.int %int4_34944, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34945 = torch.constant.int 4096
%37368 = torch.prim.ListConstruct %37367, %int4096_34945 : (!torch.int, !torch.int) -> !torch.list<int>
%37369 = torch.aten.view %37313, %37368 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37369, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37370 = torch.aten.mm %37369, %37328 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37370, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34946 = torch.constant.int 4
%int1792_34947 = torch.constant.int 1792
%37371 = torch.prim.ListConstruct %int4_34946, %2482, %int1792_34947 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37372 = torch.aten.view %37370, %37371 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37372, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34948 = torch.constant.int 4
%37373 = torch.aten.mul.int %int4_34948, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34949 = torch.constant.int 4096
%37374 = torch.prim.ListConstruct %37373, %int4096_34949 : (!torch.int, !torch.int) -> !torch.list<int>
%37375 = torch.aten.view %37314, %37374 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37375, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37376 = torch.aten.mm %37375, %37330 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37376, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34950 = torch.constant.int 4
%int1792_34951 = torch.constant.int 1792
%37377 = torch.prim.ListConstruct %int4_34950, %2482, %int1792_34951 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37378 = torch.aten.view %37376, %37377 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37378, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37379 = torch.aten.silu %37336 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37379, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37380 = torch.aten.silu %37342 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37380, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37381 = torch.aten.silu %37348 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37381, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37382 = torch.aten.silu %37354 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37382, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37383 = torch.aten.silu %37360 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37383, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37384 = torch.aten.silu %37366 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37385 = torch.aten.silu %37372 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37386 = torch.aten.silu %37378 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_34952 = torch.constant.int 1
%int0_34953 = torch.constant.int 0
%37387 = torch.prim.ListConstruct %int1_34952, %int0_34953 : (!torch.int, !torch.int) -> !torch.list<int>
%37388 = torch.aten.permute %1360, %37387 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34954 = torch.constant.int 1
%int0_34955 = torch.constant.int 0
%37389 = torch.prim.ListConstruct %int1_34954, %int0_34955 : (!torch.int, !torch.int) -> !torch.list<int>
%37390 = torch.aten.permute %1361, %37389 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34956 = torch.constant.int 1
%int0_34957 = torch.constant.int 0
%37391 = torch.prim.ListConstruct %int1_34956, %int0_34957 : (!torch.int, !torch.int) -> !torch.list<int>
%37392 = torch.aten.permute %1362, %37391 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34958 = torch.constant.int 1
%int0_34959 = torch.constant.int 0
%37393 = torch.prim.ListConstruct %int1_34958, %int0_34959 : (!torch.int, !torch.int) -> !torch.list<int>
%37394 = torch.aten.permute %1363, %37393 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34960 = torch.constant.int 1
%int0_34961 = torch.constant.int 0
%37395 = torch.prim.ListConstruct %int1_34960, %int0_34961 : (!torch.int, !torch.int) -> !torch.list<int>
%37396 = torch.aten.permute %1364, %37395 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34962 = torch.constant.int 1
%int0_34963 = torch.constant.int 0
%37397 = torch.prim.ListConstruct %int1_34962, %int0_34963 : (!torch.int, !torch.int) -> !torch.list<int>
%37398 = torch.aten.permute %1365, %37397 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34964 = torch.constant.int 1
%int0_34965 = torch.constant.int 0
%37399 = torch.prim.ListConstruct %int1_34964, %int0_34965 : (!torch.int, !torch.int) -> !torch.list<int>
%37400 = torch.aten.permute %1366, %37399 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_34966 = torch.constant.int 1
%int0_34967 = torch.constant.int 0
%37401 = torch.prim.ListConstruct %int1_34966, %int0_34967 : (!torch.int, !torch.int) -> !torch.list<int>
%37402 = torch.aten.permute %1367, %37401 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_34968 = torch.constant.int 4
%37403 = torch.aten.mul.int %int4_34968, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34969 = torch.constant.int 4096
%37404 = torch.prim.ListConstruct %37403, %int4096_34969 : (!torch.int, !torch.int) -> !torch.list<int>
%37405 = torch.aten.view %37307, %37404 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37405, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37406 = torch.aten.mm %37405, %37388 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37406, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34970 = torch.constant.int 4
%int1792_34971 = torch.constant.int 1792
%37407 = torch.prim.ListConstruct %int4_34970, %2482, %int1792_34971 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37408 = torch.aten.view %37406, %37407 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37408, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34972 = torch.constant.int 4
%37409 = torch.aten.mul.int %int4_34972, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34973 = torch.constant.int 4096
%37410 = torch.prim.ListConstruct %37409, %int4096_34973 : (!torch.int, !torch.int) -> !torch.list<int>
%37411 = torch.aten.view %37308, %37410 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37411, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37412 = torch.aten.mm %37411, %37390 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37412, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34974 = torch.constant.int 4
%int1792_34975 = torch.constant.int 1792
%37413 = torch.prim.ListConstruct %int4_34974, %2482, %int1792_34975 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37414 = torch.aten.view %37412, %37413 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34976 = torch.constant.int 4
%37415 = torch.aten.mul.int %int4_34976, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34977 = torch.constant.int 4096
%37416 = torch.prim.ListConstruct %37415, %int4096_34977 : (!torch.int, !torch.int) -> !torch.list<int>
%37417 = torch.aten.view %37309, %37416 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37417, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37418 = torch.aten.mm %37417, %37392 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37418, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34978 = torch.constant.int 4
%int1792_34979 = torch.constant.int 1792
%37419 = torch.prim.ListConstruct %int4_34978, %2482, %int1792_34979 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37420 = torch.aten.view %37418, %37419 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37420, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34980 = torch.constant.int 4
%37421 = torch.aten.mul.int %int4_34980, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34981 = torch.constant.int 4096
%37422 = torch.prim.ListConstruct %37421, %int4096_34981 : (!torch.int, !torch.int) -> !torch.list<int>
%37423 = torch.aten.view %37310, %37422 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37423, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37424 = torch.aten.mm %37423, %37394 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37424, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34982 = torch.constant.int 4
%int1792_34983 = torch.constant.int 1792
%37425 = torch.prim.ListConstruct %int4_34982, %2482, %int1792_34983 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37426 = torch.aten.view %37424, %37425 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37426, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34984 = torch.constant.int 4
%37427 = torch.aten.mul.int %int4_34984, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34985 = torch.constant.int 4096
%37428 = torch.prim.ListConstruct %37427, %int4096_34985 : (!torch.int, !torch.int) -> !torch.list<int>
%37429 = torch.aten.view %37311, %37428 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37429, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37430 = torch.aten.mm %37429, %37396 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37430, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34986 = torch.constant.int 4
%int1792_34987 = torch.constant.int 1792
%37431 = torch.prim.ListConstruct %int4_34986, %2482, %int1792_34987 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37432 = torch.aten.view %37430, %37431 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37432, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34988 = torch.constant.int 4
%37433 = torch.aten.mul.int %int4_34988, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34989 = torch.constant.int 4096
%37434 = torch.prim.ListConstruct %37433, %int4096_34989 : (!torch.int, !torch.int) -> !torch.list<int>
%37435 = torch.aten.view %37312, %37434 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37435, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37436 = torch.aten.mm %37435, %37398 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37436, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34990 = torch.constant.int 4
%int1792_34991 = torch.constant.int 1792
%37437 = torch.prim.ListConstruct %int4_34990, %2482, %int1792_34991 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37438 = torch.aten.view %37436, %37437 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37438, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34992 = torch.constant.int 4
%37439 = torch.aten.mul.int %int4_34992, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34993 = torch.constant.int 4096
%37440 = torch.prim.ListConstruct %37439, %int4096_34993 : (!torch.int, !torch.int) -> !torch.list<int>
%37441 = torch.aten.view %37313, %37440 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37441, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37442 = torch.aten.mm %37441, %37400 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37442, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34994 = torch.constant.int 4
%int1792_34995 = torch.constant.int 1792
%37443 = torch.prim.ListConstruct %int4_34994, %2482, %int1792_34995 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37444 = torch.aten.view %37442, %37443 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_34996 = torch.constant.int 4
%37445 = torch.aten.mul.int %int4_34996, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_34997 = torch.constant.int 4096
%37446 = torch.prim.ListConstruct %37445, %int4096_34997 : (!torch.int, !torch.int) -> !torch.list<int>
%37447 = torch.aten.view %37314, %37446 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37447, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37448 = torch.aten.mm %37447, %37402 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37448, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_34998 = torch.constant.int 4
%int1792_34999 = torch.constant.int 1792
%37449 = torch.prim.ListConstruct %int4_34998, %2482, %int1792_34999 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37450 = torch.aten.view %37448, %37449 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37451 = torch.aten.mul.Tensor %37379, %37408 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37451, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37452 = torch.aten.mul.Tensor %37380, %37414 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37452, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37453 = torch.aten.mul.Tensor %37381, %37420 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37454 = torch.aten.mul.Tensor %37382, %37426 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37454, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37455 = torch.aten.mul.Tensor %37383, %37432 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37455, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37456 = torch.aten.mul.Tensor %37384, %37438 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37457 = torch.aten.mul.Tensor %37385, %37444 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37457, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%37458 = torch.aten.mul.Tensor %37386, %37450 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %37458, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_35000 = torch.constant.int 1
%int0_35001 = torch.constant.int 0
%37459 = torch.prim.ListConstruct %int1_35000, %int0_35001 : (!torch.int, !torch.int) -> !torch.list<int>
%37460 = torch.aten.permute %1368, %37459 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35002 = torch.constant.int 1
%int0_35003 = torch.constant.int 0
%37461 = torch.prim.ListConstruct %int1_35002, %int0_35003 : (!torch.int, !torch.int) -> !torch.list<int>
%37462 = torch.aten.permute %1369, %37461 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35004 = torch.constant.int 1
%int0_35005 = torch.constant.int 0
%37463 = torch.prim.ListConstruct %int1_35004, %int0_35005 : (!torch.int, !torch.int) -> !torch.list<int>
%37464 = torch.aten.permute %1370, %37463 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35006 = torch.constant.int 1
%int0_35007 = torch.constant.int 0
%37465 = torch.prim.ListConstruct %int1_35006, %int0_35007 : (!torch.int, !torch.int) -> !torch.list<int>
%37466 = torch.aten.permute %1371, %37465 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35008 = torch.constant.int 1
%int0_35009 = torch.constant.int 0
%37467 = torch.prim.ListConstruct %int1_35008, %int0_35009 : (!torch.int, !torch.int) -> !torch.list<int>
%37468 = torch.aten.permute %1372, %37467 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35010 = torch.constant.int 1
%int0_35011 = torch.constant.int 0
%37469 = torch.prim.ListConstruct %int1_35010, %int0_35011 : (!torch.int, !torch.int) -> !torch.list<int>
%37470 = torch.aten.permute %1373, %37469 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35012 = torch.constant.int 1
%int0_35013 = torch.constant.int 0
%37471 = torch.prim.ListConstruct %int1_35012, %int0_35013 : (!torch.int, !torch.int) -> !torch.list<int>
%37472 = torch.aten.permute %1374, %37471 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35014 = torch.constant.int 1
%int0_35015 = torch.constant.int 0
%37473 = torch.prim.ListConstruct %int1_35014, %int0_35015 : (!torch.int, !torch.int) -> !torch.list<int>
%37474 = torch.aten.permute %1375, %37473 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_35016 = torch.constant.int 1
%37475 = torch.aten.size.int %37336, %int1_35016 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35017 = torch.constant.int 4
%37476 = torch.aten.mul.int %int4_35017, %37475 : !torch.int, !torch.int -> !torch.int
%int1792_35018 = torch.constant.int 1792
%37477 = torch.prim.ListConstruct %37476, %int1792_35018 : (!torch.int, !torch.int) -> !torch.list<int>
%37478 = torch.aten.view %37451, %37477 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37478, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37479 = torch.aten.mm %37478, %37460 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37479, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35019 = torch.constant.int 4
%int4096_35020 = torch.constant.int 4096
%37480 = torch.prim.ListConstruct %int4_35019, %37475, %int4096_35020 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37481 = torch.aten.view %37479, %37480 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35021 = torch.constant.int 1
%37482 = torch.aten.size.int %37342, %int1_35021 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35022 = torch.constant.int 4
%37483 = torch.aten.mul.int %int4_35022, %37482 : !torch.int, !torch.int -> !torch.int
%int1792_35023 = torch.constant.int 1792
%37484 = torch.prim.ListConstruct %37483, %int1792_35023 : (!torch.int, !torch.int) -> !torch.list<int>
%37485 = torch.aten.view %37452, %37484 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37485, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37486 = torch.aten.mm %37485, %37462 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37486, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35024 = torch.constant.int 4
%int4096_35025 = torch.constant.int 4096
%37487 = torch.prim.ListConstruct %int4_35024, %37482, %int4096_35025 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37488 = torch.aten.view %37486, %37487 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37488, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35026 = torch.constant.int 1
%37489 = torch.aten.size.int %37348, %int1_35026 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35027 = torch.constant.int 4
%37490 = torch.aten.mul.int %int4_35027, %37489 : !torch.int, !torch.int -> !torch.int
%int1792_35028 = torch.constant.int 1792
%37491 = torch.prim.ListConstruct %37490, %int1792_35028 : (!torch.int, !torch.int) -> !torch.list<int>
%37492 = torch.aten.view %37453, %37491 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37492, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37493 = torch.aten.mm %37492, %37464 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37493, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35029 = torch.constant.int 4
%int4096_35030 = torch.constant.int 4096
%37494 = torch.prim.ListConstruct %int4_35029, %37489, %int4096_35030 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37495 = torch.aten.view %37493, %37494 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37495, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35031 = torch.constant.int 1
%37496 = torch.aten.size.int %37354, %int1_35031 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35032 = torch.constant.int 4
%37497 = torch.aten.mul.int %int4_35032, %37496 : !torch.int, !torch.int -> !torch.int
%int1792_35033 = torch.constant.int 1792
%37498 = torch.prim.ListConstruct %37497, %int1792_35033 : (!torch.int, !torch.int) -> !torch.list<int>
%37499 = torch.aten.view %37454, %37498 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37499, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37500 = torch.aten.mm %37499, %37466 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37500, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35034 = torch.constant.int 4
%int4096_35035 = torch.constant.int 4096
%37501 = torch.prim.ListConstruct %int4_35034, %37496, %int4096_35035 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37502 = torch.aten.view %37500, %37501 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35036 = torch.constant.int 1
%37503 = torch.aten.size.int %37360, %int1_35036 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35037 = torch.constant.int 4
%37504 = torch.aten.mul.int %int4_35037, %37503 : !torch.int, !torch.int -> !torch.int
%int1792_35038 = torch.constant.int 1792
%37505 = torch.prim.ListConstruct %37504, %int1792_35038 : (!torch.int, !torch.int) -> !torch.list<int>
%37506 = torch.aten.view %37455, %37505 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37506, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37507 = torch.aten.mm %37506, %37468 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37507, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35039 = torch.constant.int 4
%int4096_35040 = torch.constant.int 4096
%37508 = torch.prim.ListConstruct %int4_35039, %37503, %int4096_35040 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37509 = torch.aten.view %37507, %37508 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35041 = torch.constant.int 1
%37510 = torch.aten.size.int %37366, %int1_35041 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35042 = torch.constant.int 4
%37511 = torch.aten.mul.int %int4_35042, %37510 : !torch.int, !torch.int -> !torch.int
%int1792_35043 = torch.constant.int 1792
%37512 = torch.prim.ListConstruct %37511, %int1792_35043 : (!torch.int, !torch.int) -> !torch.list<int>
%37513 = torch.aten.view %37456, %37512 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37513, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37514 = torch.aten.mm %37513, %37470 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37514, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35044 = torch.constant.int 4
%int4096_35045 = torch.constant.int 4096
%37515 = torch.prim.ListConstruct %int4_35044, %37510, %int4096_35045 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37516 = torch.aten.view %37514, %37515 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37516, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35046 = torch.constant.int 1
%37517 = torch.aten.size.int %37372, %int1_35046 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35047 = torch.constant.int 4
%37518 = torch.aten.mul.int %int4_35047, %37517 : !torch.int, !torch.int -> !torch.int
%int1792_35048 = torch.constant.int 1792
%37519 = torch.prim.ListConstruct %37518, %int1792_35048 : (!torch.int, !torch.int) -> !torch.list<int>
%37520 = torch.aten.view %37457, %37519 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37520, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37521 = torch.aten.mm %37520, %37472 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37521, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35049 = torch.constant.int 4
%int4096_35050 = torch.constant.int 4096
%37522 = torch.prim.ListConstruct %int4_35049, %37517, %int4096_35050 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37523 = torch.aten.view %37521, %37522 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37523, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35051 = torch.constant.int 1
%37524 = torch.aten.size.int %37378, %int1_35051 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_35052 = torch.constant.int 4
%37525 = torch.aten.mul.int %int4_35052, %37524 : !torch.int, !torch.int -> !torch.int
%int1792_35053 = torch.constant.int 1792
%37526 = torch.prim.ListConstruct %37525, %int1792_35053 : (!torch.int, !torch.int) -> !torch.list<int>
%37527 = torch.aten.view %37458, %37526 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %37527, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%37528 = torch.aten.mm %37527, %37474 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37528, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_35054 = torch.constant.int 4
%int4096_35055 = torch.constant.int 4096
%37529 = torch.prim.ListConstruct %int4_35054, %37524, %int4096_35055 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37530 = torch.aten.view %37528, %37529 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37531 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35056 = arith.constant 1 : index
%dim_35057 = tensor.dim %37531, %c1_35056 : tensor<4x?x4096xf16>
%37532 = flow.tensor.transfer %37531 : tensor<4x?x4096xf16>{%dim_35057} to #hal.device.promise<@__device_0>
%37533 = torch_c.from_builtin_tensor %37532 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37533, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37534 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35058 = arith.constant 1 : index
%dim_35059 = tensor.dim %37534, %c1_35058 : tensor<4x?x4096xf16>
%37535 = flow.tensor.transfer %37534 : tensor<4x?x4096xf16>{%dim_35059} to #hal.device.promise<@__device_0>
%37536 = torch_c.from_builtin_tensor %37535 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37536, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37537 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35060 = arith.constant 1 : index
%dim_35061 = tensor.dim %37537, %c1_35060 : tensor<4x?x4096xf16>
%37538 = flow.tensor.transfer %37537 : tensor<4x?x4096xf16>{%dim_35061} to #hal.device.promise<@__device_0>
%37539 = torch_c.from_builtin_tensor %37538 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37539, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37540 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35062 = arith.constant 1 : index
%dim_35063 = tensor.dim %37540, %c1_35062 : tensor<4x?x4096xf16>
%37541 = flow.tensor.transfer %37540 : tensor<4x?x4096xf16>{%dim_35063} to #hal.device.promise<@__device_0>
%37542 = torch_c.from_builtin_tensor %37541 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37542, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37543 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35064 = arith.constant 1 : index
%dim_35065 = tensor.dim %37543, %c1_35064 : tensor<4x?x4096xf16>
%37544 = flow.tensor.transfer %37543 : tensor<4x?x4096xf16>{%dim_35065} to #hal.device.promise<@__device_0>
%37545 = torch_c.from_builtin_tensor %37544 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37545, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37546 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35066 = arith.constant 1 : index
%dim_35067 = tensor.dim %37546, %c1_35066 : tensor<4x?x4096xf16>
%37547 = flow.tensor.transfer %37546 : tensor<4x?x4096xf16>{%dim_35067} to #hal.device.promise<@__device_0>
%37548 = torch_c.from_builtin_tensor %37547 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37548, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37549 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35068 = arith.constant 1 : index
%dim_35069 = tensor.dim %37549, %c1_35068 : tensor<4x?x4096xf16>
%37550 = flow.tensor.transfer %37549 : tensor<4x?x4096xf16>{%dim_35069} to #hal.device.promise<@__device_0>
%37551 = torch_c.from_builtin_tensor %37550 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37551, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35070 = torch.constant.int 1
%37552 = torch.aten.add.Tensor %37481, %37533, %int1_35070 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35071 = torch.constant.int 1
%37553 = torch.aten.add.Tensor %37552, %37536, %int1_35071 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35072 = torch.constant.int 1
%37554 = torch.aten.add.Tensor %37553, %37539, %int1_35072 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35073 = torch.constant.int 1
%37555 = torch.aten.add.Tensor %37554, %37542, %int1_35073 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35074 = torch.constant.int 1
%37556 = torch.aten.add.Tensor %37555, %37545, %int1_35074 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35075 = torch.constant.int 1
%37557 = torch.aten.add.Tensor %37556, %37548, %int1_35075 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35076 = torch.constant.int 1
%37558 = torch.aten.add.Tensor %37557, %37551, %int1_35076 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37559 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35077 = arith.constant 1 : index
%dim_35078 = tensor.dim %37559, %c1_35077 : tensor<4x?x4096xf16>
%37560 = flow.tensor.transfer %37559 : tensor<4x?x4096xf16>{%dim_35078} to #hal.device.promise<@__device_1>
%37561 = torch_c.from_builtin_tensor %37560 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37561, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37562 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35079 = arith.constant 1 : index
%dim_35080 = tensor.dim %37562, %c1_35079 : tensor<4x?x4096xf16>
%37563 = flow.tensor.transfer %37562 : tensor<4x?x4096xf16>{%dim_35080} to #hal.device.promise<@__device_1>
%37564 = torch_c.from_builtin_tensor %37563 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37564, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37565 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35081 = arith.constant 1 : index
%dim_35082 = tensor.dim %37565, %c1_35081 : tensor<4x?x4096xf16>
%37566 = flow.tensor.transfer %37565 : tensor<4x?x4096xf16>{%dim_35082} to #hal.device.promise<@__device_1>
%37567 = torch_c.from_builtin_tensor %37566 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37567, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37568 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35083 = arith.constant 1 : index
%dim_35084 = tensor.dim %37568, %c1_35083 : tensor<4x?x4096xf16>
%37569 = flow.tensor.transfer %37568 : tensor<4x?x4096xf16>{%dim_35084} to #hal.device.promise<@__device_1>
%37570 = torch_c.from_builtin_tensor %37569 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37570, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37571 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35085 = arith.constant 1 : index
%dim_35086 = tensor.dim %37571, %c1_35085 : tensor<4x?x4096xf16>
%37572 = flow.tensor.transfer %37571 : tensor<4x?x4096xf16>{%dim_35086} to #hal.device.promise<@__device_1>
%37573 = torch_c.from_builtin_tensor %37572 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37573, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37574 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35087 = arith.constant 1 : index
%dim_35088 = tensor.dim %37574, %c1_35087 : tensor<4x?x4096xf16>
%37575 = flow.tensor.transfer %37574 : tensor<4x?x4096xf16>{%dim_35088} to #hal.device.promise<@__device_1>
%37576 = torch_c.from_builtin_tensor %37575 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37576, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37577 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35089 = arith.constant 1 : index
%dim_35090 = tensor.dim %37577, %c1_35089 : tensor<4x?x4096xf16>
%37578 = flow.tensor.transfer %37577 : tensor<4x?x4096xf16>{%dim_35090} to #hal.device.promise<@__device_1>
%37579 = torch_c.from_builtin_tensor %37578 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37579, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35091 = torch.constant.int 1
%37580 = torch.aten.add.Tensor %37561, %37488, %int1_35091 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35092 = torch.constant.int 1
%37581 = torch.aten.add.Tensor %37580, %37564, %int1_35092 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35093 = torch.constant.int 1
%37582 = torch.aten.add.Tensor %37581, %37567, %int1_35093 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35094 = torch.constant.int 1
%37583 = torch.aten.add.Tensor %37582, %37570, %int1_35094 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35095 = torch.constant.int 1
%37584 = torch.aten.add.Tensor %37583, %37573, %int1_35095 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35096 = torch.constant.int 1
%37585 = torch.aten.add.Tensor %37584, %37576, %int1_35096 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35097 = torch.constant.int 1
%37586 = torch.aten.add.Tensor %37585, %37579, %int1_35097 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37587 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35098 = arith.constant 1 : index
%dim_35099 = tensor.dim %37587, %c1_35098 : tensor<4x?x4096xf16>
%37588 = flow.tensor.transfer %37587 : tensor<4x?x4096xf16>{%dim_35099} to #hal.device.promise<@__device_2>
%37589 = torch_c.from_builtin_tensor %37588 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37589, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37590 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35100 = arith.constant 1 : index
%dim_35101 = tensor.dim %37590, %c1_35100 : tensor<4x?x4096xf16>
%37591 = flow.tensor.transfer %37590 : tensor<4x?x4096xf16>{%dim_35101} to #hal.device.promise<@__device_2>
%37592 = torch_c.from_builtin_tensor %37591 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37592, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37593 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35102 = arith.constant 1 : index
%dim_35103 = tensor.dim %37593, %c1_35102 : tensor<4x?x4096xf16>
%37594 = flow.tensor.transfer %37593 : tensor<4x?x4096xf16>{%dim_35103} to #hal.device.promise<@__device_2>
%37595 = torch_c.from_builtin_tensor %37594 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37595, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37596 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35104 = arith.constant 1 : index
%dim_35105 = tensor.dim %37596, %c1_35104 : tensor<4x?x4096xf16>
%37597 = flow.tensor.transfer %37596 : tensor<4x?x4096xf16>{%dim_35105} to #hal.device.promise<@__device_2>
%37598 = torch_c.from_builtin_tensor %37597 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37598, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37599 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35106 = arith.constant 1 : index
%dim_35107 = tensor.dim %37599, %c1_35106 : tensor<4x?x4096xf16>
%37600 = flow.tensor.transfer %37599 : tensor<4x?x4096xf16>{%dim_35107} to #hal.device.promise<@__device_2>
%37601 = torch_c.from_builtin_tensor %37600 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37601, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37602 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35108 = arith.constant 1 : index
%dim_35109 = tensor.dim %37602, %c1_35108 : tensor<4x?x4096xf16>
%37603 = flow.tensor.transfer %37602 : tensor<4x?x4096xf16>{%dim_35109} to #hal.device.promise<@__device_2>
%37604 = torch_c.from_builtin_tensor %37603 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37604, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37605 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35110 = arith.constant 1 : index
%dim_35111 = tensor.dim %37605, %c1_35110 : tensor<4x?x4096xf16>
%37606 = flow.tensor.transfer %37605 : tensor<4x?x4096xf16>{%dim_35111} to #hal.device.promise<@__device_2>
%37607 = torch_c.from_builtin_tensor %37606 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37607, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35112 = torch.constant.int 1
%37608 = torch.aten.add.Tensor %37589, %37592, %int1_35112 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35113 = torch.constant.int 1
%37609 = torch.aten.add.Tensor %37608, %37495, %int1_35113 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35114 = torch.constant.int 1
%37610 = torch.aten.add.Tensor %37609, %37595, %int1_35114 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35115 = torch.constant.int 1
%37611 = torch.aten.add.Tensor %37610, %37598, %int1_35115 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35116 = torch.constant.int 1
%37612 = torch.aten.add.Tensor %37611, %37601, %int1_35116 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35117 = torch.constant.int 1
%37613 = torch.aten.add.Tensor %37612, %37604, %int1_35117 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35118 = torch.constant.int 1
%37614 = torch.aten.add.Tensor %37613, %37607, %int1_35118 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37615 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35119 = arith.constant 1 : index
%dim_35120 = tensor.dim %37615, %c1_35119 : tensor<4x?x4096xf16>
%37616 = flow.tensor.transfer %37615 : tensor<4x?x4096xf16>{%dim_35120} to #hal.device.promise<@__device_3>
%37617 = torch_c.from_builtin_tensor %37616 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37618 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35121 = arith.constant 1 : index
%dim_35122 = tensor.dim %37618, %c1_35121 : tensor<4x?x4096xf16>
%37619 = flow.tensor.transfer %37618 : tensor<4x?x4096xf16>{%dim_35122} to #hal.device.promise<@__device_3>
%37620 = torch_c.from_builtin_tensor %37619 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37621 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35123 = arith.constant 1 : index
%dim_35124 = tensor.dim %37621, %c1_35123 : tensor<4x?x4096xf16>
%37622 = flow.tensor.transfer %37621 : tensor<4x?x4096xf16>{%dim_35124} to #hal.device.promise<@__device_3>
%37623 = torch_c.from_builtin_tensor %37622 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37624 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35125 = arith.constant 1 : index
%dim_35126 = tensor.dim %37624, %c1_35125 : tensor<4x?x4096xf16>
%37625 = flow.tensor.transfer %37624 : tensor<4x?x4096xf16>{%dim_35126} to #hal.device.promise<@__device_3>
%37626 = torch_c.from_builtin_tensor %37625 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37627 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35127 = arith.constant 1 : index
%dim_35128 = tensor.dim %37627, %c1_35127 : tensor<4x?x4096xf16>
%37628 = flow.tensor.transfer %37627 : tensor<4x?x4096xf16>{%dim_35128} to #hal.device.promise<@__device_3>
%37629 = torch_c.from_builtin_tensor %37628 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37630 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35129 = arith.constant 1 : index
%dim_35130 = tensor.dim %37630, %c1_35129 : tensor<4x?x4096xf16>
%37631 = flow.tensor.transfer %37630 : tensor<4x?x4096xf16>{%dim_35130} to #hal.device.promise<@__device_3>
%37632 = torch_c.from_builtin_tensor %37631 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37633 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35131 = arith.constant 1 : index
%dim_35132 = tensor.dim %37633, %c1_35131 : tensor<4x?x4096xf16>
%37634 = flow.tensor.transfer %37633 : tensor<4x?x4096xf16>{%dim_35132} to #hal.device.promise<@__device_3>
%37635 = torch_c.from_builtin_tensor %37634 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35133 = torch.constant.int 1
%37636 = torch.aten.add.Tensor %37617, %37620, %int1_35133 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35134 = torch.constant.int 1
%37637 = torch.aten.add.Tensor %37636, %37623, %int1_35134 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35135 = torch.constant.int 1
%37638 = torch.aten.add.Tensor %37637, %37502, %int1_35135 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35136 = torch.constant.int 1
%37639 = torch.aten.add.Tensor %37638, %37626, %int1_35136 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35137 = torch.constant.int 1
%37640 = torch.aten.add.Tensor %37639, %37629, %int1_35137 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37640, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35138 = torch.constant.int 1
%37641 = torch.aten.add.Tensor %37640, %37632, %int1_35138 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35139 = torch.constant.int 1
%37642 = torch.aten.add.Tensor %37641, %37635, %int1_35139 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37642, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37643 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35140 = arith.constant 1 : index
%dim_35141 = tensor.dim %37643, %c1_35140 : tensor<4x?x4096xf16>
%37644 = flow.tensor.transfer %37643 : tensor<4x?x4096xf16>{%dim_35141} to #hal.device.promise<@__device_4>
%37645 = torch_c.from_builtin_tensor %37644 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37646 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35142 = arith.constant 1 : index
%dim_35143 = tensor.dim %37646, %c1_35142 : tensor<4x?x4096xf16>
%37647 = flow.tensor.transfer %37646 : tensor<4x?x4096xf16>{%dim_35143} to #hal.device.promise<@__device_4>
%37648 = torch_c.from_builtin_tensor %37647 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37648, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37649 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35144 = arith.constant 1 : index
%dim_35145 = tensor.dim %37649, %c1_35144 : tensor<4x?x4096xf16>
%37650 = flow.tensor.transfer %37649 : tensor<4x?x4096xf16>{%dim_35145} to #hal.device.promise<@__device_4>
%37651 = torch_c.from_builtin_tensor %37650 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37652 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35146 = arith.constant 1 : index
%dim_35147 = tensor.dim %37652, %c1_35146 : tensor<4x?x4096xf16>
%37653 = flow.tensor.transfer %37652 : tensor<4x?x4096xf16>{%dim_35147} to #hal.device.promise<@__device_4>
%37654 = torch_c.from_builtin_tensor %37653 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37654, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37655 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35148 = arith.constant 1 : index
%dim_35149 = tensor.dim %37655, %c1_35148 : tensor<4x?x4096xf16>
%37656 = flow.tensor.transfer %37655 : tensor<4x?x4096xf16>{%dim_35149} to #hal.device.promise<@__device_4>
%37657 = torch_c.from_builtin_tensor %37656 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37658 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35150 = arith.constant 1 : index
%dim_35151 = tensor.dim %37658, %c1_35150 : tensor<4x?x4096xf16>
%37659 = flow.tensor.transfer %37658 : tensor<4x?x4096xf16>{%dim_35151} to #hal.device.promise<@__device_4>
%37660 = torch_c.from_builtin_tensor %37659 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37661 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35152 = arith.constant 1 : index
%dim_35153 = tensor.dim %37661, %c1_35152 : tensor<4x?x4096xf16>
%37662 = flow.tensor.transfer %37661 : tensor<4x?x4096xf16>{%dim_35153} to #hal.device.promise<@__device_4>
%37663 = torch_c.from_builtin_tensor %37662 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35154 = torch.constant.int 1
%37664 = torch.aten.add.Tensor %37645, %37648, %int1_35154 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35155 = torch.constant.int 1
%37665 = torch.aten.add.Tensor %37664, %37651, %int1_35155 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35156 = torch.constant.int 1
%37666 = torch.aten.add.Tensor %37665, %37654, %int1_35156 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35157 = torch.constant.int 1
%37667 = torch.aten.add.Tensor %37666, %37509, %int1_35157 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35158 = torch.constant.int 1
%37668 = torch.aten.add.Tensor %37667, %37657, %int1_35158 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35159 = torch.constant.int 1
%37669 = torch.aten.add.Tensor %37668, %37660, %int1_35159 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35160 = torch.constant.int 1
%37670 = torch.aten.add.Tensor %37669, %37663, %int1_35160 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37671 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35161 = arith.constant 1 : index
%dim_35162 = tensor.dim %37671, %c1_35161 : tensor<4x?x4096xf16>
%37672 = flow.tensor.transfer %37671 : tensor<4x?x4096xf16>{%dim_35162} to #hal.device.promise<@__device_5>
%37673 = torch_c.from_builtin_tensor %37672 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37674 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35163 = arith.constant 1 : index
%dim_35164 = tensor.dim %37674, %c1_35163 : tensor<4x?x4096xf16>
%37675 = flow.tensor.transfer %37674 : tensor<4x?x4096xf16>{%dim_35164} to #hal.device.promise<@__device_5>
%37676 = torch_c.from_builtin_tensor %37675 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37677 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35165 = arith.constant 1 : index
%dim_35166 = tensor.dim %37677, %c1_35165 : tensor<4x?x4096xf16>
%37678 = flow.tensor.transfer %37677 : tensor<4x?x4096xf16>{%dim_35166} to #hal.device.promise<@__device_5>
%37679 = torch_c.from_builtin_tensor %37678 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37680 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35167 = arith.constant 1 : index
%dim_35168 = tensor.dim %37680, %c1_35167 : tensor<4x?x4096xf16>
%37681 = flow.tensor.transfer %37680 : tensor<4x?x4096xf16>{%dim_35168} to #hal.device.promise<@__device_5>
%37682 = torch_c.from_builtin_tensor %37681 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37683 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35169 = arith.constant 1 : index
%dim_35170 = tensor.dim %37683, %c1_35169 : tensor<4x?x4096xf16>
%37684 = flow.tensor.transfer %37683 : tensor<4x?x4096xf16>{%dim_35170} to #hal.device.promise<@__device_5>
%37685 = torch_c.from_builtin_tensor %37684 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37686 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35171 = arith.constant 1 : index
%dim_35172 = tensor.dim %37686, %c1_35171 : tensor<4x?x4096xf16>
%37687 = flow.tensor.transfer %37686 : tensor<4x?x4096xf16>{%dim_35172} to #hal.device.promise<@__device_5>
%37688 = torch_c.from_builtin_tensor %37687 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37689 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35173 = arith.constant 1 : index
%dim_35174 = tensor.dim %37689, %c1_35173 : tensor<4x?x4096xf16>
%37690 = flow.tensor.transfer %37689 : tensor<4x?x4096xf16>{%dim_35174} to #hal.device.promise<@__device_5>
%37691 = torch_c.from_builtin_tensor %37690 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35175 = torch.constant.int 1
%37692 = torch.aten.add.Tensor %37673, %37676, %int1_35175 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35176 = torch.constant.int 1
%37693 = torch.aten.add.Tensor %37692, %37679, %int1_35176 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35177 = torch.constant.int 1
%37694 = torch.aten.add.Tensor %37693, %37682, %int1_35177 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35178 = torch.constant.int 1
%37695 = torch.aten.add.Tensor %37694, %37685, %int1_35178 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35179 = torch.constant.int 1
%37696 = torch.aten.add.Tensor %37695, %37516, %int1_35179 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35180 = torch.constant.int 1
%37697 = torch.aten.add.Tensor %37696, %37688, %int1_35180 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37697, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35181 = torch.constant.int 1
%37698 = torch.aten.add.Tensor %37697, %37691, %int1_35181 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37699 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35182 = arith.constant 1 : index
%dim_35183 = tensor.dim %37699, %c1_35182 : tensor<4x?x4096xf16>
%37700 = flow.tensor.transfer %37699 : tensor<4x?x4096xf16>{%dim_35183} to #hal.device.promise<@__device_6>
%37701 = torch_c.from_builtin_tensor %37700 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37701, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37702 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35184 = arith.constant 1 : index
%dim_35185 = tensor.dim %37702, %c1_35184 : tensor<4x?x4096xf16>
%37703 = flow.tensor.transfer %37702 : tensor<4x?x4096xf16>{%dim_35185} to #hal.device.promise<@__device_6>
%37704 = torch_c.from_builtin_tensor %37703 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37705 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35186 = arith.constant 1 : index
%dim_35187 = tensor.dim %37705, %c1_35186 : tensor<4x?x4096xf16>
%37706 = flow.tensor.transfer %37705 : tensor<4x?x4096xf16>{%dim_35187} to #hal.device.promise<@__device_6>
%37707 = torch_c.from_builtin_tensor %37706 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37707, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37708 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35188 = arith.constant 1 : index
%dim_35189 = tensor.dim %37708, %c1_35188 : tensor<4x?x4096xf16>
%37709 = flow.tensor.transfer %37708 : tensor<4x?x4096xf16>{%dim_35189} to #hal.device.promise<@__device_6>
%37710 = torch_c.from_builtin_tensor %37709 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37711 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35190 = arith.constant 1 : index
%dim_35191 = tensor.dim %37711, %c1_35190 : tensor<4x?x4096xf16>
%37712 = flow.tensor.transfer %37711 : tensor<4x?x4096xf16>{%dim_35191} to #hal.device.promise<@__device_6>
%37713 = torch_c.from_builtin_tensor %37712 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37714 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35192 = arith.constant 1 : index
%dim_35193 = tensor.dim %37714, %c1_35192 : tensor<4x?x4096xf16>
%37715 = flow.tensor.transfer %37714 : tensor<4x?x4096xf16>{%dim_35193} to #hal.device.promise<@__device_6>
%37716 = torch_c.from_builtin_tensor %37715 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37717 = torch_c.to_builtin_tensor %37530 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35194 = arith.constant 1 : index
%dim_35195 = tensor.dim %37717, %c1_35194 : tensor<4x?x4096xf16>
%37718 = flow.tensor.transfer %37717 : tensor<4x?x4096xf16>{%dim_35195} to #hal.device.promise<@__device_6>
%37719 = torch_c.from_builtin_tensor %37718 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37719, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35196 = torch.constant.int 1
%37720 = torch.aten.add.Tensor %37701, %37704, %int1_35196 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37720, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35197 = torch.constant.int 1
%37721 = torch.aten.add.Tensor %37720, %37707, %int1_35197 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35198 = torch.constant.int 1
%37722 = torch.aten.add.Tensor %37721, %37710, %int1_35198 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37722, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35199 = torch.constant.int 1
%37723 = torch.aten.add.Tensor %37722, %37713, %int1_35199 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35200 = torch.constant.int 1
%37724 = torch.aten.add.Tensor %37723, %37716, %int1_35200 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37724, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35201 = torch.constant.int 1
%37725 = torch.aten.add.Tensor %37724, %37523, %int1_35201 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35202 = torch.constant.int 1
%37726 = torch.aten.add.Tensor %37725, %37719, %int1_35202 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37726, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37727 = torch_c.to_builtin_tensor %37481 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35203 = arith.constant 1 : index
%dim_35204 = tensor.dim %37727, %c1_35203 : tensor<4x?x4096xf16>
%37728 = flow.tensor.transfer %37727 : tensor<4x?x4096xf16>{%dim_35204} to #hal.device.promise<@__device_7>
%37729 = torch_c.from_builtin_tensor %37728 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37730 = torch_c.to_builtin_tensor %37488 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35205 = arith.constant 1 : index
%dim_35206 = tensor.dim %37730, %c1_35205 : tensor<4x?x4096xf16>
%37731 = flow.tensor.transfer %37730 : tensor<4x?x4096xf16>{%dim_35206} to #hal.device.promise<@__device_7>
%37732 = torch_c.from_builtin_tensor %37731 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37732, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37733 = torch_c.to_builtin_tensor %37495 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35207 = arith.constant 1 : index
%dim_35208 = tensor.dim %37733, %c1_35207 : tensor<4x?x4096xf16>
%37734 = flow.tensor.transfer %37733 : tensor<4x?x4096xf16>{%dim_35208} to #hal.device.promise<@__device_7>
%37735 = torch_c.from_builtin_tensor %37734 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37736 = torch_c.to_builtin_tensor %37502 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35209 = arith.constant 1 : index
%dim_35210 = tensor.dim %37736, %c1_35209 : tensor<4x?x4096xf16>
%37737 = flow.tensor.transfer %37736 : tensor<4x?x4096xf16>{%dim_35210} to #hal.device.promise<@__device_7>
%37738 = torch_c.from_builtin_tensor %37737 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37738, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37739 = torch_c.to_builtin_tensor %37509 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35211 = arith.constant 1 : index
%dim_35212 = tensor.dim %37739, %c1_35211 : tensor<4x?x4096xf16>
%37740 = flow.tensor.transfer %37739 : tensor<4x?x4096xf16>{%dim_35212} to #hal.device.promise<@__device_7>
%37741 = torch_c.from_builtin_tensor %37740 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37742 = torch_c.to_builtin_tensor %37516 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35213 = arith.constant 1 : index
%dim_35214 = tensor.dim %37742, %c1_35213 : tensor<4x?x4096xf16>
%37743 = flow.tensor.transfer %37742 : tensor<4x?x4096xf16>{%dim_35214} to #hal.device.promise<@__device_7>
%37744 = torch_c.from_builtin_tensor %37743 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37744, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%37745 = torch_c.to_builtin_tensor %37523 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_35215 = arith.constant 1 : index
%dim_35216 = tensor.dim %37745, %c1_35215 : tensor<4x?x4096xf16>
%37746 = flow.tensor.transfer %37745 : tensor<4x?x4096xf16>{%dim_35216} to #hal.device.promise<@__device_7>
%37747 = torch_c.from_builtin_tensor %37746 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35217 = torch.constant.int 1
%37748 = torch.aten.add.Tensor %37729, %37732, %int1_35217 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37748, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35218 = torch.constant.int 1
%37749 = torch.aten.add.Tensor %37748, %37735, %int1_35218 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35219 = torch.constant.int 1
%37750 = torch.aten.add.Tensor %37749, %37738, %int1_35219 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37750, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35220 = torch.constant.int 1
%37751 = torch.aten.add.Tensor %37750, %37741, %int1_35220 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35221 = torch.constant.int 1
%37752 = torch.aten.add.Tensor %37751, %37744, %int1_35221 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37752, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35222 = torch.constant.int 1
%37753 = torch.aten.add.Tensor %37752, %37747, %int1_35222 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35223 = torch.constant.int 1
%37754 = torch.aten.add.Tensor %37753, %37530, %int1_35223 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37754, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35224 = torch.constant.int 1
%37755 = torch.aten.add.Tensor %37235, %37558, %int1_35224 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37755, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35225 = torch.constant.int 1
%37756 = torch.aten.add.Tensor %37236, %37586, %int1_35225 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37756, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35226 = torch.constant.int 1
%37757 = torch.aten.add.Tensor %37237, %37614, %int1_35226 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37757, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35227 = torch.constant.int 1
%37758 = torch.aten.add.Tensor %37238, %37642, %int1_35227 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37758, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35228 = torch.constant.int 1
%37759 = torch.aten.add.Tensor %37239, %37670, %int1_35228 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35229 = torch.constant.int 1
%37760 = torch.aten.add.Tensor %37240, %37698, %int1_35229 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37760, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35230 = torch.constant.int 1
%37761 = torch.aten.add.Tensor %37241, %37726, %int1_35230 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37761, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35231 = torch.constant.int 1
%37762 = torch.aten.add.Tensor %37242, %37754, %int1_35231 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37762, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_35232 = torch.constant.int 6
%37763 = torch.prims.convert_element_type %37755, %int6_35232 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37763, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35233 = torch.constant.int 6
%37764 = torch.prims.convert_element_type %37756, %int6_35233 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37764, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35234 = torch.constant.int 6
%37765 = torch.prims.convert_element_type %37757, %int6_35234 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37765, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35235 = torch.constant.int 6
%37766 = torch.prims.convert_element_type %37758, %int6_35235 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37766, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35236 = torch.constant.int 6
%37767 = torch.prims.convert_element_type %37759, %int6_35236 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37767, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35237 = torch.constant.int 6
%37768 = torch.prims.convert_element_type %37760, %int6_35237 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37768, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35238 = torch.constant.int 6
%37769 = torch.prims.convert_element_type %37761, %int6_35238 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37769, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_35239 = torch.constant.int 6
%37770 = torch.prims.convert_element_type %37762, %int6_35239 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37770, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35240 = torch.constant.int 2
%37771 = torch.aten.pow.Tensor_Scalar %37763, %int2_35240 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37771, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35241 = torch.constant.int 2
%37772 = torch.aten.pow.Tensor_Scalar %37764, %int2_35241 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37772, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35242 = torch.constant.int 2
%37773 = torch.aten.pow.Tensor_Scalar %37765, %int2_35242 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37773, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35243 = torch.constant.int 2
%37774 = torch.aten.pow.Tensor_Scalar %37766, %int2_35243 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37774, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35244 = torch.constant.int 2
%37775 = torch.aten.pow.Tensor_Scalar %37767, %int2_35244 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37775, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35245 = torch.constant.int 2
%37776 = torch.aten.pow.Tensor_Scalar %37768, %int2_35245 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37776, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35246 = torch.constant.int 2
%37777 = torch.aten.pow.Tensor_Scalar %37769, %int2_35246 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37777, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_35247 = torch.constant.int 2
%37778 = torch.aten.pow.Tensor_Scalar %37770, %int2_35247 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37778, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_35248 = torch.constant.int -1
%37779 = torch.prim.ListConstruct %int-1_35248 : (!torch.int) -> !torch.list<int>
%true_35249 = torch.constant.bool true
%none_35250 = torch.constant.none
%37780 = torch.aten.mean.dim %37771, %37779, %true_35249, %none_35250 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37780, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35251 = torch.constant.int -1
%37781 = torch.prim.ListConstruct %int-1_35251 : (!torch.int) -> !torch.list<int>
%true_35252 = torch.constant.bool true
%none_35253 = torch.constant.none
%37782 = torch.aten.mean.dim %37772, %37781, %true_35252, %none_35253 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37782, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35254 = torch.constant.int -1
%37783 = torch.prim.ListConstruct %int-1_35254 : (!torch.int) -> !torch.list<int>
%true_35255 = torch.constant.bool true
%none_35256 = torch.constant.none
%37784 = torch.aten.mean.dim %37773, %37783, %true_35255, %none_35256 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35257 = torch.constant.int -1
%37785 = torch.prim.ListConstruct %int-1_35257 : (!torch.int) -> !torch.list<int>
%true_35258 = torch.constant.bool true
%none_35259 = torch.constant.none
%37786 = torch.aten.mean.dim %37774, %37785, %true_35258, %none_35259 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35260 = torch.constant.int -1
%37787 = torch.prim.ListConstruct %int-1_35260 : (!torch.int) -> !torch.list<int>
%true_35261 = torch.constant.bool true
%none_35262 = torch.constant.none
%37788 = torch.aten.mean.dim %37775, %37787, %true_35261, %none_35262 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35263 = torch.constant.int -1
%37789 = torch.prim.ListConstruct %int-1_35263 : (!torch.int) -> !torch.list<int>
%true_35264 = torch.constant.bool true
%none_35265 = torch.constant.none
%37790 = torch.aten.mean.dim %37776, %37789, %true_35264, %none_35265 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35266 = torch.constant.int -1
%37791 = torch.prim.ListConstruct %int-1_35266 : (!torch.int) -> !torch.list<int>
%true_35267 = torch.constant.bool true
%none_35268 = torch.constant.none
%37792 = torch.aten.mean.dim %37777, %37791, %true_35267, %none_35268 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37792, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_35269 = torch.constant.int -1
%37793 = torch.prim.ListConstruct %int-1_35269 : (!torch.int) -> !torch.list<int>
%true_35270 = torch.constant.bool true
%none_35271 = torch.constant.none
%37794 = torch.aten.mean.dim %37778, %37793, %true_35270, %none_35271 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37794, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35272 = torch.constant.float 9.9999997473787516E-6
%int1_35273 = torch.constant.int 1
%37795 = torch.aten.add.Scalar %37780, %float9.999990e-06_35272, %int1_35273 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35274 = torch.constant.float 9.9999997473787516E-6
%int1_35275 = torch.constant.int 1
%37796 = torch.aten.add.Scalar %37782, %float9.999990e-06_35274, %int1_35275 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37796, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35276 = torch.constant.float 9.9999997473787516E-6
%int1_35277 = torch.constant.int 1
%37797 = torch.aten.add.Scalar %37784, %float9.999990e-06_35276, %int1_35277 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35278 = torch.constant.float 9.9999997473787516E-6
%int1_35279 = torch.constant.int 1
%37798 = torch.aten.add.Scalar %37786, %float9.999990e-06_35278, %int1_35279 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37798, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35280 = torch.constant.float 9.9999997473787516E-6
%int1_35281 = torch.constant.int 1
%37799 = torch.aten.add.Scalar %37788, %float9.999990e-06_35280, %int1_35281 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35282 = torch.constant.float 9.9999997473787516E-6
%int1_35283 = torch.constant.int 1
%37800 = torch.aten.add.Scalar %37790, %float9.999990e-06_35282, %int1_35283 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37800, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35284 = torch.constant.float 9.9999997473787516E-6
%int1_35285 = torch.constant.int 1
%37801 = torch.aten.add.Scalar %37792, %float9.999990e-06_35284, %int1_35285 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_35286 = torch.constant.float 9.9999997473787516E-6
%int1_35287 = torch.constant.int 1
%37802 = torch.aten.add.Scalar %37794, %float9.999990e-06_35286, %int1_35287 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37802, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37803 = torch.aten.rsqrt %37795 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37804 = torch.aten.rsqrt %37796 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37804, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37805 = torch.aten.rsqrt %37797 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37806 = torch.aten.rsqrt %37798 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37806, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37807 = torch.aten.rsqrt %37799 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37808 = torch.aten.rsqrt %37800 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37808, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37809 = torch.aten.rsqrt %37801 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37809, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37810 = torch.aten.rsqrt %37802 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %37810, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%37811 = torch.aten.mul.Tensor %37763, %37803 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37812 = torch.aten.mul.Tensor %37764, %37804 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37812, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37813 = torch.aten.mul.Tensor %37765, %37805 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37813, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37814 = torch.aten.mul.Tensor %37766, %37806 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37814, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37815 = torch.aten.mul.Tensor %37767, %37807 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37815, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37816 = torch.aten.mul.Tensor %37768, %37808 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37816, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37817 = torch.aten.mul.Tensor %37769, %37809 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37818 = torch.aten.mul.Tensor %37770, %37810 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37818, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37819 = torch.aten.mul.Tensor %1376, %37811 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37819, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37820 = torch.aten.mul.Tensor %1377, %37812 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37820, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37821 = torch.aten.mul.Tensor %1378, %37813 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37821, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37822 = torch.aten.mul.Tensor %1379, %37814 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37822, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37823 = torch.aten.mul.Tensor %1380, %37815 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37824 = torch.aten.mul.Tensor %1381, %37816 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37824, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37825 = torch.aten.mul.Tensor %1382, %37817 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37825, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%37826 = torch.aten.mul.Tensor %1383, %37818 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %37826, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_35288 = torch.constant.int 5
%37827 = torch.prims.convert_element_type %37819, %int5_35288 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37827, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35289 = torch.constant.int 5
%37828 = torch.prims.convert_element_type %37820, %int5_35289 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37828, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35290 = torch.constant.int 5
%37829 = torch.prims.convert_element_type %37821, %int5_35290 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35291 = torch.constant.int 5
%37830 = torch.prims.convert_element_type %37822, %int5_35291 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37830, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35292 = torch.constant.int 5
%37831 = torch.prims.convert_element_type %37823, %int5_35292 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37831, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35293 = torch.constant.int 5
%37832 = torch.prims.convert_element_type %37824, %int5_35293 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37832, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35294 = torch.constant.int 5
%37833 = torch.prims.convert_element_type %37825, %int5_35294 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37833, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_35295 = torch.constant.int 5
%37834 = torch.prims.convert_element_type %37826, %int5_35295 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %37834, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_35296 = torch.constant.int 1
%int0_35297 = torch.constant.int 0
%37835 = torch.prim.ListConstruct %int1_35296, %int0_35297 : (!torch.int, !torch.int) -> !torch.list<int>
%37836 = torch.aten.permute %1384, %37835 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35298 = torch.constant.int 1
%int0_35299 = torch.constant.int 0
%37837 = torch.prim.ListConstruct %int1_35298, %int0_35299 : (!torch.int, !torch.int) -> !torch.list<int>
%37838 = torch.aten.permute %1385, %37837 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35300 = torch.constant.int 1
%int0_35301 = torch.constant.int 0
%37839 = torch.prim.ListConstruct %int1_35300, %int0_35301 : (!torch.int, !torch.int) -> !torch.list<int>
%37840 = torch.aten.permute %1386, %37839 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35302 = torch.constant.int 1
%int0_35303 = torch.constant.int 0
%37841 = torch.prim.ListConstruct %int1_35302, %int0_35303 : (!torch.int, !torch.int) -> !torch.list<int>
%37842 = torch.aten.permute %1387, %37841 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35304 = torch.constant.int 1
%int0_35305 = torch.constant.int 0
%37843 = torch.prim.ListConstruct %int1_35304, %int0_35305 : (!torch.int, !torch.int) -> !torch.list<int>
%37844 = torch.aten.permute %1388, %37843 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35306 = torch.constant.int 1
%int0_35307 = torch.constant.int 0
%37845 = torch.prim.ListConstruct %int1_35306, %int0_35307 : (!torch.int, !torch.int) -> !torch.list<int>
%37846 = torch.aten.permute %1389, %37845 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35308 = torch.constant.int 1
%int0_35309 = torch.constant.int 0
%37847 = torch.prim.ListConstruct %int1_35308, %int0_35309 : (!torch.int, !torch.int) -> !torch.list<int>
%37848 = torch.aten.permute %1390, %37847 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_35310 = torch.constant.int 1
%int0_35311 = torch.constant.int 0
%37849 = torch.prim.ListConstruct %int1_35310, %int0_35311 : (!torch.int, !torch.int) -> !torch.list<int>
%37850 = torch.aten.permute %1391, %37849 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_35312 = torch.constant.int 4
%37851 = torch.aten.mul.int %int4_35312, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35313 = torch.constant.int 4096
%37852 = torch.prim.ListConstruct %37851, %int4096_35313 : (!torch.int, !torch.int) -> !torch.list<int>
%37853 = torch.aten.view %37827, %37852 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37853, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37854 = torch.aten.mm %37853, %37836 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37854, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35314 = torch.constant.int 4
%int512_35315 = torch.constant.int 512
%37855 = torch.prim.ListConstruct %int4_35314, %2482, %int512_35315 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37856 = torch.aten.view %37854, %37855 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37856, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35316 = torch.constant.int 4
%37857 = torch.aten.mul.int %int4_35316, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35317 = torch.constant.int 4096
%37858 = torch.prim.ListConstruct %37857, %int4096_35317 : (!torch.int, !torch.int) -> !torch.list<int>
%37859 = torch.aten.view %37828, %37858 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37859, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37860 = torch.aten.mm %37859, %37838 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37860, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35318 = torch.constant.int 4
%int512_35319 = torch.constant.int 512
%37861 = torch.prim.ListConstruct %int4_35318, %2482, %int512_35319 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37862 = torch.aten.view %37860, %37861 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37862, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35320 = torch.constant.int 4
%37863 = torch.aten.mul.int %int4_35320, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35321 = torch.constant.int 4096
%37864 = torch.prim.ListConstruct %37863, %int4096_35321 : (!torch.int, !torch.int) -> !torch.list<int>
%37865 = torch.aten.view %37829, %37864 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37865, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37866 = torch.aten.mm %37865, %37840 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37866, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35322 = torch.constant.int 4
%int512_35323 = torch.constant.int 512
%37867 = torch.prim.ListConstruct %int4_35322, %2482, %int512_35323 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37868 = torch.aten.view %37866, %37867 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37868, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35324 = torch.constant.int 4
%37869 = torch.aten.mul.int %int4_35324, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35325 = torch.constant.int 4096
%37870 = torch.prim.ListConstruct %37869, %int4096_35325 : (!torch.int, !torch.int) -> !torch.list<int>
%37871 = torch.aten.view %37830, %37870 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37871, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37872 = torch.aten.mm %37871, %37842 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37872, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35326 = torch.constant.int 4
%int512_35327 = torch.constant.int 512
%37873 = torch.prim.ListConstruct %int4_35326, %2482, %int512_35327 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37874 = torch.aten.view %37872, %37873 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35328 = torch.constant.int 4
%37875 = torch.aten.mul.int %int4_35328, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35329 = torch.constant.int 4096
%37876 = torch.prim.ListConstruct %37875, %int4096_35329 : (!torch.int, !torch.int) -> !torch.list<int>
%37877 = torch.aten.view %37831, %37876 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37877, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37878 = torch.aten.mm %37877, %37844 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37878, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35330 = torch.constant.int 4
%int512_35331 = torch.constant.int 512
%37879 = torch.prim.ListConstruct %int4_35330, %2482, %int512_35331 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37880 = torch.aten.view %37878, %37879 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35332 = torch.constant.int 4
%37881 = torch.aten.mul.int %int4_35332, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35333 = torch.constant.int 4096
%37882 = torch.prim.ListConstruct %37881, %int4096_35333 : (!torch.int, !torch.int) -> !torch.list<int>
%37883 = torch.aten.view %37832, %37882 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37883, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37884 = torch.aten.mm %37883, %37846 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37884, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35334 = torch.constant.int 4
%int512_35335 = torch.constant.int 512
%37885 = torch.prim.ListConstruct %int4_35334, %2482, %int512_35335 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37886 = torch.aten.view %37884, %37885 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35336 = torch.constant.int 4
%37887 = torch.aten.mul.int %int4_35336, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35337 = torch.constant.int 4096
%37888 = torch.prim.ListConstruct %37887, %int4096_35337 : (!torch.int, !torch.int) -> !torch.list<int>
%37889 = torch.aten.view %37833, %37888 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37889, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37890 = torch.aten.mm %37889, %37848 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37890, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35338 = torch.constant.int 4
%int512_35339 = torch.constant.int 512
%37891 = torch.prim.ListConstruct %int4_35338, %2482, %int512_35339 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37892 = torch.aten.view %37890, %37891 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_35340 = torch.constant.int 4
%37893 = torch.aten.mul.int %int4_35340, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35341 = torch.constant.int 4096
%37894 = torch.prim.ListConstruct %37893, %int4096_35341 : (!torch.int, !torch.int) -> !torch.list<int>
%37895 = torch.aten.view %37834, %37894 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37895, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37896 = torch.aten.mm %37895, %37850 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %37896, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_35342 = torch.constant.int 4
%int512_35343 = torch.constant.int 512
%37897 = torch.prim.ListConstruct %int4_35342, %2482, %int512_35343 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37898 = torch.aten.view %37896, %37897 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %37898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_35344 = torch.constant.int 1
%int0_35345 = torch.constant.int 0
%37899 = torch.prim.ListConstruct %int1_35344, %int0_35345 : (!torch.int, !torch.int) -> !torch.list<int>
%37900 = torch.aten.permute %1392, %37899 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35346 = torch.constant.int 1
%int0_35347 = torch.constant.int 0
%37901 = torch.prim.ListConstruct %int1_35346, %int0_35347 : (!torch.int, !torch.int) -> !torch.list<int>
%37902 = torch.aten.permute %1393, %37901 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35348 = torch.constant.int 1
%int0_35349 = torch.constant.int 0
%37903 = torch.prim.ListConstruct %int1_35348, %int0_35349 : (!torch.int, !torch.int) -> !torch.list<int>
%37904 = torch.aten.permute %1394, %37903 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35350 = torch.constant.int 1
%int0_35351 = torch.constant.int 0
%37905 = torch.prim.ListConstruct %int1_35350, %int0_35351 : (!torch.int, !torch.int) -> !torch.list<int>
%37906 = torch.aten.permute %1395, %37905 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35352 = torch.constant.int 1
%int0_35353 = torch.constant.int 0
%37907 = torch.prim.ListConstruct %int1_35352, %int0_35353 : (!torch.int, !torch.int) -> !torch.list<int>
%37908 = torch.aten.permute %1396, %37907 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35354 = torch.constant.int 1
%int0_35355 = torch.constant.int 0
%37909 = torch.prim.ListConstruct %int1_35354, %int0_35355 : (!torch.int, !torch.int) -> !torch.list<int>
%37910 = torch.aten.permute %1397, %37909 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35356 = torch.constant.int 1
%int0_35357 = torch.constant.int 0
%37911 = torch.prim.ListConstruct %int1_35356, %int0_35357 : (!torch.int, !torch.int) -> !torch.list<int>
%37912 = torch.aten.permute %1398, %37911 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35358 = torch.constant.int 1
%int0_35359 = torch.constant.int 0
%37913 = torch.prim.ListConstruct %int1_35358, %int0_35359 : (!torch.int, !torch.int) -> !torch.list<int>
%37914 = torch.aten.permute %1399, %37913 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_35360 = torch.constant.int 4
%37915 = torch.aten.mul.int %int4_35360, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35361 = torch.constant.int 4096
%37916 = torch.prim.ListConstruct %37915, %int4096_35361 : (!torch.int, !torch.int) -> !torch.list<int>
%37917 = torch.aten.view %37827, %37916 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37917, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37918 = torch.aten.mm %37917, %37900 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37918, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35362 = torch.constant.int 4
%int128_35363 = torch.constant.int 128
%37919 = torch.prim.ListConstruct %int4_35362, %2482, %int128_35363 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37920 = torch.aten.view %37918, %37919 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35364 = torch.constant.int 4
%37921 = torch.aten.mul.int %int4_35364, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35365 = torch.constant.int 4096
%37922 = torch.prim.ListConstruct %37921, %int4096_35365 : (!torch.int, !torch.int) -> !torch.list<int>
%37923 = torch.aten.view %37828, %37922 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37923, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37924 = torch.aten.mm %37923, %37902 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37924, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35366 = torch.constant.int 4
%int128_35367 = torch.constant.int 128
%37925 = torch.prim.ListConstruct %int4_35366, %2482, %int128_35367 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37926 = torch.aten.view %37924, %37925 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35368 = torch.constant.int 4
%37927 = torch.aten.mul.int %int4_35368, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35369 = torch.constant.int 4096
%37928 = torch.prim.ListConstruct %37927, %int4096_35369 : (!torch.int, !torch.int) -> !torch.list<int>
%37929 = torch.aten.view %37829, %37928 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37929, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37930 = torch.aten.mm %37929, %37904 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37930, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35370 = torch.constant.int 4
%int128_35371 = torch.constant.int 128
%37931 = torch.prim.ListConstruct %int4_35370, %2482, %int128_35371 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37932 = torch.aten.view %37930, %37931 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37932, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35372 = torch.constant.int 4
%37933 = torch.aten.mul.int %int4_35372, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35373 = torch.constant.int 4096
%37934 = torch.prim.ListConstruct %37933, %int4096_35373 : (!torch.int, !torch.int) -> !torch.list<int>
%37935 = torch.aten.view %37830, %37934 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37935, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37936 = torch.aten.mm %37935, %37906 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37936, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35374 = torch.constant.int 4
%int128_35375 = torch.constant.int 128
%37937 = torch.prim.ListConstruct %int4_35374, %2482, %int128_35375 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37938 = torch.aten.view %37936, %37937 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37938, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35376 = torch.constant.int 4
%37939 = torch.aten.mul.int %int4_35376, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35377 = torch.constant.int 4096
%37940 = torch.prim.ListConstruct %37939, %int4096_35377 : (!torch.int, !torch.int) -> !torch.list<int>
%37941 = torch.aten.view %37831, %37940 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37941, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37942 = torch.aten.mm %37941, %37908 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37942, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35378 = torch.constant.int 4
%int128_35379 = torch.constant.int 128
%37943 = torch.prim.ListConstruct %int4_35378, %2482, %int128_35379 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37944 = torch.aten.view %37942, %37943 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37944, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35380 = torch.constant.int 4
%37945 = torch.aten.mul.int %int4_35380, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35381 = torch.constant.int 4096
%37946 = torch.prim.ListConstruct %37945, %int4096_35381 : (!torch.int, !torch.int) -> !torch.list<int>
%37947 = torch.aten.view %37832, %37946 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37947, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37948 = torch.aten.mm %37947, %37910 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37948, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35382 = torch.constant.int 4
%int128_35383 = torch.constant.int 128
%37949 = torch.prim.ListConstruct %int4_35382, %2482, %int128_35383 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37950 = torch.aten.view %37948, %37949 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35384 = torch.constant.int 4
%37951 = torch.aten.mul.int %int4_35384, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35385 = torch.constant.int 4096
%37952 = torch.prim.ListConstruct %37951, %int4096_35385 : (!torch.int, !torch.int) -> !torch.list<int>
%37953 = torch.aten.view %37833, %37952 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37953, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37954 = torch.aten.mm %37953, %37912 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37954, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35386 = torch.constant.int 4
%int128_35387 = torch.constant.int 128
%37955 = torch.prim.ListConstruct %int4_35386, %2482, %int128_35387 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37956 = torch.aten.view %37954, %37955 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37956, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35388 = torch.constant.int 4
%37957 = torch.aten.mul.int %int4_35388, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35389 = torch.constant.int 4096
%37958 = torch.prim.ListConstruct %37957, %int4096_35389 : (!torch.int, !torch.int) -> !torch.list<int>
%37959 = torch.aten.view %37834, %37958 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37959, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37960 = torch.aten.mm %37959, %37914 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37960, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35390 = torch.constant.int 4
%int128_35391 = torch.constant.int 128
%37961 = torch.prim.ListConstruct %int4_35390, %2482, %int128_35391 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37962 = torch.aten.view %37960, %37961 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37962, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_35392 = torch.constant.int 1
%int0_35393 = torch.constant.int 0
%37963 = torch.prim.ListConstruct %int1_35392, %int0_35393 : (!torch.int, !torch.int) -> !torch.list<int>
%37964 = torch.aten.permute %1400, %37963 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35394 = torch.constant.int 1
%int0_35395 = torch.constant.int 0
%37965 = torch.prim.ListConstruct %int1_35394, %int0_35395 : (!torch.int, !torch.int) -> !torch.list<int>
%37966 = torch.aten.permute %1401, %37965 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35396 = torch.constant.int 1
%int0_35397 = torch.constant.int 0
%37967 = torch.prim.ListConstruct %int1_35396, %int0_35397 : (!torch.int, !torch.int) -> !torch.list<int>
%37968 = torch.aten.permute %1402, %37967 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35398 = torch.constant.int 1
%int0_35399 = torch.constant.int 0
%37969 = torch.prim.ListConstruct %int1_35398, %int0_35399 : (!torch.int, !torch.int) -> !torch.list<int>
%37970 = torch.aten.permute %1403, %37969 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35400 = torch.constant.int 1
%int0_35401 = torch.constant.int 0
%37971 = torch.prim.ListConstruct %int1_35400, %int0_35401 : (!torch.int, !torch.int) -> !torch.list<int>
%37972 = torch.aten.permute %1404, %37971 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35402 = torch.constant.int 1
%int0_35403 = torch.constant.int 0
%37973 = torch.prim.ListConstruct %int1_35402, %int0_35403 : (!torch.int, !torch.int) -> !torch.list<int>
%37974 = torch.aten.permute %1405, %37973 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35404 = torch.constant.int 1
%int0_35405 = torch.constant.int 0
%37975 = torch.prim.ListConstruct %int1_35404, %int0_35405 : (!torch.int, !torch.int) -> !torch.list<int>
%37976 = torch.aten.permute %1406, %37975 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_35406 = torch.constant.int 1
%int0_35407 = torch.constant.int 0
%37977 = torch.prim.ListConstruct %int1_35406, %int0_35407 : (!torch.int, !torch.int) -> !torch.list<int>
%37978 = torch.aten.permute %1407, %37977 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_35408 = torch.constant.int 4
%37979 = torch.aten.mul.int %int4_35408, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35409 = torch.constant.int 4096
%37980 = torch.prim.ListConstruct %37979, %int4096_35409 : (!torch.int, !torch.int) -> !torch.list<int>
%37981 = torch.aten.view %37827, %37980 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37981, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37982 = torch.aten.mm %37981, %37964 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37982, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35410 = torch.constant.int 4
%int128_35411 = torch.constant.int 128
%37983 = torch.prim.ListConstruct %int4_35410, %2482, %int128_35411 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37984 = torch.aten.view %37982, %37983 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35412 = torch.constant.int 4
%37985 = torch.aten.mul.int %int4_35412, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35413 = torch.constant.int 4096
%37986 = torch.prim.ListConstruct %37985, %int4096_35413 : (!torch.int, !torch.int) -> !torch.list<int>
%37987 = torch.aten.view %37828, %37986 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37987, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37988 = torch.aten.mm %37987, %37966 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37988, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35414 = torch.constant.int 4
%int128_35415 = torch.constant.int 128
%37989 = torch.prim.ListConstruct %int4_35414, %2482, %int128_35415 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37990 = torch.aten.view %37988, %37989 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37990, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35416 = torch.constant.int 4
%37991 = torch.aten.mul.int %int4_35416, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35417 = torch.constant.int 4096
%37992 = torch.prim.ListConstruct %37991, %int4096_35417 : (!torch.int, !torch.int) -> !torch.list<int>
%37993 = torch.aten.view %37829, %37992 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37993, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%37994 = torch.aten.mm %37993, %37968 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %37994, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35418 = torch.constant.int 4
%int128_35419 = torch.constant.int 128
%37995 = torch.prim.ListConstruct %int4_35418, %2482, %int128_35419 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%37996 = torch.aten.view %37994, %37995 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %37996, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35420 = torch.constant.int 4
%37997 = torch.aten.mul.int %int4_35420, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35421 = torch.constant.int 4096
%37998 = torch.prim.ListConstruct %37997, %int4096_35421 : (!torch.int, !torch.int) -> !torch.list<int>
%37999 = torch.aten.view %37830, %37998 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %37999, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%38000 = torch.aten.mm %37999, %37970 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %38000, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35422 = torch.constant.int 4
%int128_35423 = torch.constant.int 128
%38001 = torch.prim.ListConstruct %int4_35422, %2482, %int128_35423 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38002 = torch.aten.view %38000, %38001 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %38002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35424 = torch.constant.int 4
%38003 = torch.aten.mul.int %int4_35424, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35425 = torch.constant.int 4096
%38004 = torch.prim.ListConstruct %38003, %int4096_35425 : (!torch.int, !torch.int) -> !torch.list<int>
%38005 = torch.aten.view %37831, %38004 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38005, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%38006 = torch.aten.mm %38005, %37972 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %38006, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35426 = torch.constant.int 4
%int128_35427 = torch.constant.int 128
%38007 = torch.prim.ListConstruct %int4_35426, %2482, %int128_35427 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38008 = torch.aten.view %38006, %38007 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %38008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35428 = torch.constant.int 4
%38009 = torch.aten.mul.int %int4_35428, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35429 = torch.constant.int 4096
%38010 = torch.prim.ListConstruct %38009, %int4096_35429 : (!torch.int, !torch.int) -> !torch.list<int>
%38011 = torch.aten.view %37832, %38010 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38011, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%38012 = torch.aten.mm %38011, %37974 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %38012, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35430 = torch.constant.int 4
%int128_35431 = torch.constant.int 128
%38013 = torch.prim.ListConstruct %int4_35430, %2482, %int128_35431 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38014 = torch.aten.view %38012, %38013 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %38014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35432 = torch.constant.int 4
%38015 = torch.aten.mul.int %int4_35432, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35433 = torch.constant.int 4096
%38016 = torch.prim.ListConstruct %38015, %int4096_35433 : (!torch.int, !torch.int) -> !torch.list<int>
%38017 = torch.aten.view %37833, %38016 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38017, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%38018 = torch.aten.mm %38017, %37976 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %38018, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35434 = torch.constant.int 4
%int128_35435 = torch.constant.int 128
%38019 = torch.prim.ListConstruct %int4_35434, %2482, %int128_35435 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38020 = torch.aten.view %38018, %38019 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %38020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35436 = torch.constant.int 4
%38021 = torch.aten.mul.int %int4_35436, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_35437 = torch.constant.int 4096
%38022 = torch.prim.ListConstruct %38021, %int4096_35437 : (!torch.int, !torch.int) -> !torch.list<int>
%38023 = torch.aten.view %37834, %38022 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38023, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%38024 = torch.aten.mm %38023, %37978 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %38024, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_35438 = torch.constant.int 4
%int128_35439 = torch.constant.int 128
%38025 = torch.prim.ListConstruct %int4_35438, %2482, %int128_35439 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38026 = torch.aten.view %38024, %38025 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %38026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_35440 = torch.constant.int 4
%int4_35441 = torch.constant.int 4
%int128_35442 = torch.constant.int 128
%38027 = torch.prim.ListConstruct %int4_35440, %2482, %int4_35441, %int128_35442 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38028 = torch.aten.view %37856, %38027 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35443 = torch.constant.int 4
%int4_35444 = torch.constant.int 4
%int128_35445 = torch.constant.int 128
%38029 = torch.prim.ListConstruct %int4_35443, %2482, %int4_35444, %int128_35445 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38030 = torch.aten.view %37862, %38029 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38030, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35446 = torch.constant.int 4
%int4_35447 = torch.constant.int 4
%int128_35448 = torch.constant.int 128
%38031 = torch.prim.ListConstruct %int4_35446, %2482, %int4_35447, %int128_35448 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38032 = torch.aten.view %37868, %38031 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35449 = torch.constant.int 4
%int4_35450 = torch.constant.int 4
%int128_35451 = torch.constant.int 128
%38033 = torch.prim.ListConstruct %int4_35449, %2482, %int4_35450, %int128_35451 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38034 = torch.aten.view %37874, %38033 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35452 = torch.constant.int 4
%int4_35453 = torch.constant.int 4
%int128_35454 = torch.constant.int 128
%38035 = torch.prim.ListConstruct %int4_35452, %2482, %int4_35453, %int128_35454 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38036 = torch.aten.view %37880, %38035 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35455 = torch.constant.int 4
%int4_35456 = torch.constant.int 4
%int128_35457 = torch.constant.int 128
%38037 = torch.prim.ListConstruct %int4_35455, %2482, %int4_35456, %int128_35457 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38038 = torch.aten.view %37886, %38037 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35458 = torch.constant.int 4
%int4_35459 = torch.constant.int 4
%int128_35460 = torch.constant.int 128
%38039 = torch.prim.ListConstruct %int4_35458, %2482, %int4_35459, %int128_35460 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38040 = torch.aten.view %37892, %38039 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38040, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35461 = torch.constant.int 4
%int4_35462 = torch.constant.int 4
%int128_35463 = torch.constant.int 128
%38041 = torch.prim.ListConstruct %int4_35461, %2482, %int4_35462, %int128_35463 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38042 = torch.aten.view %37898, %38041 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_35464 = torch.constant.int 4
%int1_35465 = torch.constant.int 1
%int128_35466 = torch.constant.int 128
%38043 = torch.prim.ListConstruct %int4_35464, %2482, %int1_35465, %int128_35466 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38044 = torch.aten.view %37920, %38043 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35467 = torch.constant.int 4
%int1_35468 = torch.constant.int 1
%int128_35469 = torch.constant.int 128
%38045 = torch.prim.ListConstruct %int4_35467, %2482, %int1_35468, %int128_35469 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38046 = torch.aten.view %37926, %38045 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38046, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35470 = torch.constant.int 4
%int1_35471 = torch.constant.int 1
%int128_35472 = torch.constant.int 128
%38047 = torch.prim.ListConstruct %int4_35470, %2482, %int1_35471, %int128_35472 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38048 = torch.aten.view %37932, %38047 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35473 = torch.constant.int 4
%int1_35474 = torch.constant.int 1
%int128_35475 = torch.constant.int 128
%38049 = torch.prim.ListConstruct %int4_35473, %2482, %int1_35474, %int128_35475 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38050 = torch.aten.view %37938, %38049 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38050, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35476 = torch.constant.int 4
%int1_35477 = torch.constant.int 1
%int128_35478 = torch.constant.int 128
%38051 = torch.prim.ListConstruct %int4_35476, %2482, %int1_35477, %int128_35478 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38052 = torch.aten.view %37944, %38051 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38052, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35479 = torch.constant.int 4
%int1_35480 = torch.constant.int 1
%int128_35481 = torch.constant.int 128
%38053 = torch.prim.ListConstruct %int4_35479, %2482, %int1_35480, %int128_35481 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38054 = torch.aten.view %37950, %38053 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35482 = torch.constant.int 4
%int1_35483 = torch.constant.int 1
%int128_35484 = torch.constant.int 128
%38055 = torch.prim.ListConstruct %int4_35482, %2482, %int1_35483, %int128_35484 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38056 = torch.aten.view %37956, %38055 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38056, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35485 = torch.constant.int 4
%int1_35486 = torch.constant.int 1
%int128_35487 = torch.constant.int 128
%38057 = torch.prim.ListConstruct %int4_35485, %2482, %int1_35486, %int128_35487 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38058 = torch.aten.view %37962, %38057 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35488 = torch.constant.int 4
%int1_35489 = torch.constant.int 1
%int128_35490 = torch.constant.int 128
%38059 = torch.prim.ListConstruct %int4_35488, %2482, %int1_35489, %int128_35490 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38060 = torch.aten.view %37984, %38059 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35491 = torch.constant.int 4
%int1_35492 = torch.constant.int 1
%int128_35493 = torch.constant.int 128
%38061 = torch.prim.ListConstruct %int4_35491, %2482, %int1_35492, %int128_35493 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38062 = torch.aten.view %37990, %38061 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35494 = torch.constant.int 4
%int1_35495 = torch.constant.int 1
%int128_35496 = torch.constant.int 128
%38063 = torch.prim.ListConstruct %int4_35494, %2482, %int1_35495, %int128_35496 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38064 = torch.aten.view %37996, %38063 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35497 = torch.constant.int 4
%int1_35498 = torch.constant.int 1
%int128_35499 = torch.constant.int 128
%38065 = torch.prim.ListConstruct %int4_35497, %2482, %int1_35498, %int128_35499 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38066 = torch.aten.view %38002, %38065 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35500 = torch.constant.int 4
%int1_35501 = torch.constant.int 1
%int128_35502 = torch.constant.int 128
%38067 = torch.prim.ListConstruct %int4_35500, %2482, %int1_35501, %int128_35502 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38068 = torch.aten.view %38008, %38067 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38068, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35503 = torch.constant.int 4
%int1_35504 = torch.constant.int 1
%int128_35505 = torch.constant.int 128
%38069 = torch.prim.ListConstruct %int4_35503, %2482, %int1_35504, %int128_35505 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38070 = torch.aten.view %38014, %38069 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35506 = torch.constant.int 4
%int1_35507 = torch.constant.int 1
%int128_35508 = torch.constant.int 128
%38071 = torch.prim.ListConstruct %int4_35506, %2482, %int1_35507, %int128_35508 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38072 = torch.aten.view %38020, %38071 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38072, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_35509 = torch.constant.int 4
%int1_35510 = torch.constant.int 1
%int128_35511 = torch.constant.int 128
%38073 = torch.prim.ListConstruct %int4_35509, %2482, %int1_35510, %int128_35511 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38074 = torch.aten.view %38026, %38073 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_35512 = torch.constant.int 131072
%none_35513 = torch.constant.none
%none_35514 = torch.constant.none
%cpu_35515 = torch.constant.device "cpu"
%false_35516 = torch.constant.bool false
%38075 = torch.aten.arange %int131072_35512, %none_35513, %none_35514, %cpu_35515, %false_35516 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_35517 = torch.constant.int 0
%int128_35518 = torch.constant.int 128
%int2_35519 = torch.constant.int 2
%none_35520 = torch.constant.none
%none_35521 = torch.constant.none
%cpu_35522 = torch.constant.device "cpu"
%false_35523 = torch.constant.bool false
%38076 = torch.aten.arange.start_step %int0_35517, %int128_35518, %int2_35519, %none_35520, %none_35521, %cpu_35522, %false_35523 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_35524 = torch.constant.int 0
%int0_35525 = torch.constant.int 0
%int64_35526 = torch.constant.int 64
%int1_35527 = torch.constant.int 1
%38077 = torch.aten.slice.Tensor %38076, %int0_35524, %int0_35525, %int64_35526, %int1_35527 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_35528 = torch.constant.int 6
%38078 = torch.prims.convert_element_type %38077, %int6_35528 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_35529 = torch.constant.int 128
%38079 = torch.aten.div.Scalar %38078, %int128_35529 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_35530 = torch.constant.float 5.000000e+05
%38080 = torch.aten.pow.Scalar %float5.000000e05_35530, %38079 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%38081 = torch.aten.reciprocal %38080 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_35531 = torch.constant.float 1.000000e+00
%38082 = torch.aten.mul.Scalar %38081, %float1.000000e00_35531 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_35532 = torch.constant.int 131072
%int1_35533 = torch.constant.int 1
%38083 = torch.prim.ListConstruct %int131072_35532, %int1_35533 : (!torch.int, !torch.int) -> !torch.list<int>
%38084 = torch.aten.view %38075, %38083 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%38085 = torch.aten.mul.Tensor %38084, %38082 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%38086 = torch.aten.cos %38085 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%38087 = torch.aten.sin %38085 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%38088 = torch.aten.complex %38086, %38087 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%38089 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38090 = flow.tensor.transfer %38089 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%38091 = torch_c.from_builtin_tensor %38090 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38092 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38093 = flow.tensor.transfer %38092 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%38094 = torch_c.from_builtin_tensor %38093 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38095 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38096 = flow.tensor.transfer %38095 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%38097 = torch_c.from_builtin_tensor %38096 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38098 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38099 = flow.tensor.transfer %38098 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%38100 = torch_c.from_builtin_tensor %38099 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38101 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38102 = flow.tensor.transfer %38101 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%38103 = torch_c.from_builtin_tensor %38102 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38104 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38105 = flow.tensor.transfer %38104 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%38106 = torch_c.from_builtin_tensor %38105 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38107 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38108 = flow.tensor.transfer %38107 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%38109 = torch_c.from_builtin_tensor %38108 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38110 = torch_c.to_builtin_tensor %38088 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38111 = flow.tensor.transfer %38110 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%38112 = torch_c.from_builtin_tensor %38111 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_35534 = torch.constant.int 1
%38113 = torch.aten.size.int %37856, %int1_35534 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35535 = torch.constant.int 0
%38114 = torch.aten.add.int %int0_35535, %38113 : !torch.int, !torch.int -> !torch.int
%int0_35536 = torch.constant.int 0
%int0_35537 = torch.constant.int 0
%int1_35538 = torch.constant.int 1
%38115 = torch.aten.slice.Tensor %38091, %int0_35536, %int0_35537, %38114, %int1_35538 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38115, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35539 = torch.constant.int 1
%int0_35540 = torch.constant.int 0
%int9223372036854775807_35541 = torch.constant.int 9223372036854775807
%int1_35542 = torch.constant.int 1
%38116 = torch.aten.slice.Tensor %38115, %int1_35539, %int0_35540, %int9223372036854775807_35541, %int1_35542 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38116, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35543 = torch.constant.int 0
%38117 = torch.aten.unsqueeze %38116, %int0_35543 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38117, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35544 = torch.constant.int 2
%38118 = torch.aten.unsqueeze %38117, %int2_35544 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38118, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35545 = torch.constant.int 3
%int0_35546 = torch.constant.int 0
%int9223372036854775807_35547 = torch.constant.int 9223372036854775807
%int1_35548 = torch.constant.int 1
%38119 = torch.aten.slice.Tensor %38118, %int3_35545, %int0_35546, %int9223372036854775807_35547, %int1_35548 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38119, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38120 = torch_c.to_builtin_tensor %38028 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35549 = arith.constant 1 : index
%dim_35550 = tensor.dim %38120, %c1_35549 : tensor<4x?x4x128xf16>
%38121 = flow.tensor.bitcast %38120 : tensor<4x?x4x128xf16>{%dim_35550} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35550}
%38122 = torch_c.from_builtin_tensor %38121 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38122, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38123 = torch.aten.mul.Tensor %38122, %38119 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38124 = torch_c.to_builtin_tensor %38123 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35551 = arith.constant 1 : index
%dim_35552 = tensor.dim %38124, %c1_35551 : tensor<4x?x4x64xcomplex<f32>>
%38125 = flow.tensor.bitcast %38124 : tensor<4x?x4x64xcomplex<f32>>{%dim_35552} -> tensor<4x?x4x128xf32>{%dim_35552}
%38126 = torch_c.from_builtin_tensor %38125 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38126, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35553 = torch.constant.int 5
%38127 = torch.prims.convert_element_type %38126, %int5_35553 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35554 = torch.constant.int 1
%38128 = torch.aten.size.int %37862, %int1_35554 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35555 = torch.constant.int 0
%38129 = torch.aten.add.int %int0_35555, %38128 : !torch.int, !torch.int -> !torch.int
%int0_35556 = torch.constant.int 0
%int0_35557 = torch.constant.int 0
%int1_35558 = torch.constant.int 1
%38130 = torch.aten.slice.Tensor %38094, %int0_35556, %int0_35557, %38129, %int1_35558 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38130, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35559 = torch.constant.int 1
%int0_35560 = torch.constant.int 0
%int9223372036854775807_35561 = torch.constant.int 9223372036854775807
%int1_35562 = torch.constant.int 1
%38131 = torch.aten.slice.Tensor %38130, %int1_35559, %int0_35560, %int9223372036854775807_35561, %int1_35562 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38131, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35563 = torch.constant.int 0
%38132 = torch.aten.unsqueeze %38131, %int0_35563 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38132, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35564 = torch.constant.int 2
%38133 = torch.aten.unsqueeze %38132, %int2_35564 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38133, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35565 = torch.constant.int 3
%int0_35566 = torch.constant.int 0
%int9223372036854775807_35567 = torch.constant.int 9223372036854775807
%int1_35568 = torch.constant.int 1
%38134 = torch.aten.slice.Tensor %38133, %int3_35565, %int0_35566, %int9223372036854775807_35567, %int1_35568 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38134, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38135 = torch_c.to_builtin_tensor %38030 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35569 = arith.constant 1 : index
%dim_35570 = tensor.dim %38135, %c1_35569 : tensor<4x?x4x128xf16>
%38136 = flow.tensor.bitcast %38135 : tensor<4x?x4x128xf16>{%dim_35570} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35570}
%38137 = torch_c.from_builtin_tensor %38136 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38138 = torch.aten.mul.Tensor %38137, %38134 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38139 = torch_c.to_builtin_tensor %38138 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35571 = arith.constant 1 : index
%dim_35572 = tensor.dim %38139, %c1_35571 : tensor<4x?x4x64xcomplex<f32>>
%38140 = flow.tensor.bitcast %38139 : tensor<4x?x4x64xcomplex<f32>>{%dim_35572} -> tensor<4x?x4x128xf32>{%dim_35572}
%38141 = torch_c.from_builtin_tensor %38140 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35573 = torch.constant.int 5
%38142 = torch.prims.convert_element_type %38141, %int5_35573 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35574 = torch.constant.int 1
%38143 = torch.aten.size.int %37868, %int1_35574 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35575 = torch.constant.int 0
%38144 = torch.aten.add.int %int0_35575, %38143 : !torch.int, !torch.int -> !torch.int
%int0_35576 = torch.constant.int 0
%int0_35577 = torch.constant.int 0
%int1_35578 = torch.constant.int 1
%38145 = torch.aten.slice.Tensor %38097, %int0_35576, %int0_35577, %38144, %int1_35578 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38145, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35579 = torch.constant.int 1
%int0_35580 = torch.constant.int 0
%int9223372036854775807_35581 = torch.constant.int 9223372036854775807
%int1_35582 = torch.constant.int 1
%38146 = torch.aten.slice.Tensor %38145, %int1_35579, %int0_35580, %int9223372036854775807_35581, %int1_35582 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38146, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35583 = torch.constant.int 0
%38147 = torch.aten.unsqueeze %38146, %int0_35583 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38147, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35584 = torch.constant.int 2
%38148 = torch.aten.unsqueeze %38147, %int2_35584 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38148, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35585 = torch.constant.int 3
%int0_35586 = torch.constant.int 0
%int9223372036854775807_35587 = torch.constant.int 9223372036854775807
%int1_35588 = torch.constant.int 1
%38149 = torch.aten.slice.Tensor %38148, %int3_35585, %int0_35586, %int9223372036854775807_35587, %int1_35588 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38149, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38150 = torch_c.to_builtin_tensor %38032 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35589 = arith.constant 1 : index
%dim_35590 = tensor.dim %38150, %c1_35589 : tensor<4x?x4x128xf16>
%38151 = flow.tensor.bitcast %38150 : tensor<4x?x4x128xf16>{%dim_35590} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35590}
%38152 = torch_c.from_builtin_tensor %38151 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38153 = torch.aten.mul.Tensor %38152, %38149 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38154 = torch_c.to_builtin_tensor %38153 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35591 = arith.constant 1 : index
%dim_35592 = tensor.dim %38154, %c1_35591 : tensor<4x?x4x64xcomplex<f32>>
%38155 = flow.tensor.bitcast %38154 : tensor<4x?x4x64xcomplex<f32>>{%dim_35592} -> tensor<4x?x4x128xf32>{%dim_35592}
%38156 = torch_c.from_builtin_tensor %38155 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35593 = torch.constant.int 5
%38157 = torch.prims.convert_element_type %38156, %int5_35593 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35594 = torch.constant.int 1
%38158 = torch.aten.size.int %37874, %int1_35594 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35595 = torch.constant.int 0
%38159 = torch.aten.add.int %int0_35595, %38158 : !torch.int, !torch.int -> !torch.int
%int0_35596 = torch.constant.int 0
%int0_35597 = torch.constant.int 0
%int1_35598 = torch.constant.int 1
%38160 = torch.aten.slice.Tensor %38100, %int0_35596, %int0_35597, %38159, %int1_35598 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38160, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35599 = torch.constant.int 1
%int0_35600 = torch.constant.int 0
%int9223372036854775807_35601 = torch.constant.int 9223372036854775807
%int1_35602 = torch.constant.int 1
%38161 = torch.aten.slice.Tensor %38160, %int1_35599, %int0_35600, %int9223372036854775807_35601, %int1_35602 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38161, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35603 = torch.constant.int 0
%38162 = torch.aten.unsqueeze %38161, %int0_35603 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38162, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35604 = torch.constant.int 2
%38163 = torch.aten.unsqueeze %38162, %int2_35604 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38163, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35605 = torch.constant.int 3
%int0_35606 = torch.constant.int 0
%int9223372036854775807_35607 = torch.constant.int 9223372036854775807
%int1_35608 = torch.constant.int 1
%38164 = torch.aten.slice.Tensor %38163, %int3_35605, %int0_35606, %int9223372036854775807_35607, %int1_35608 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38164, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38165 = torch_c.to_builtin_tensor %38034 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35609 = arith.constant 1 : index
%dim_35610 = tensor.dim %38165, %c1_35609 : tensor<4x?x4x128xf16>
%38166 = flow.tensor.bitcast %38165 : tensor<4x?x4x128xf16>{%dim_35610} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35610}
%38167 = torch_c.from_builtin_tensor %38166 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38168 = torch.aten.mul.Tensor %38167, %38164 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38169 = torch_c.to_builtin_tensor %38168 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35611 = arith.constant 1 : index
%dim_35612 = tensor.dim %38169, %c1_35611 : tensor<4x?x4x64xcomplex<f32>>
%38170 = flow.tensor.bitcast %38169 : tensor<4x?x4x64xcomplex<f32>>{%dim_35612} -> tensor<4x?x4x128xf32>{%dim_35612}
%38171 = torch_c.from_builtin_tensor %38170 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35613 = torch.constant.int 5
%38172 = torch.prims.convert_element_type %38171, %int5_35613 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35614 = torch.constant.int 1
%38173 = torch.aten.size.int %37880, %int1_35614 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35615 = torch.constant.int 0
%38174 = torch.aten.add.int %int0_35615, %38173 : !torch.int, !torch.int -> !torch.int
%int0_35616 = torch.constant.int 0
%int0_35617 = torch.constant.int 0
%int1_35618 = torch.constant.int 1
%38175 = torch.aten.slice.Tensor %38103, %int0_35616, %int0_35617, %38174, %int1_35618 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38175, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35619 = torch.constant.int 1
%int0_35620 = torch.constant.int 0
%int9223372036854775807_35621 = torch.constant.int 9223372036854775807
%int1_35622 = torch.constant.int 1
%38176 = torch.aten.slice.Tensor %38175, %int1_35619, %int0_35620, %int9223372036854775807_35621, %int1_35622 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38176, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35623 = torch.constant.int 0
%38177 = torch.aten.unsqueeze %38176, %int0_35623 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38177, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35624 = torch.constant.int 2
%38178 = torch.aten.unsqueeze %38177, %int2_35624 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38178, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35625 = torch.constant.int 3
%int0_35626 = torch.constant.int 0
%int9223372036854775807_35627 = torch.constant.int 9223372036854775807
%int1_35628 = torch.constant.int 1
%38179 = torch.aten.slice.Tensor %38178, %int3_35625, %int0_35626, %int9223372036854775807_35627, %int1_35628 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38179, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38180 = torch_c.to_builtin_tensor %38036 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35629 = arith.constant 1 : index
%dim_35630 = tensor.dim %38180, %c1_35629 : tensor<4x?x4x128xf16>
%38181 = flow.tensor.bitcast %38180 : tensor<4x?x4x128xf16>{%dim_35630} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35630}
%38182 = torch_c.from_builtin_tensor %38181 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38182, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38183 = torch.aten.mul.Tensor %38182, %38179 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38183, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38184 = torch_c.to_builtin_tensor %38183 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35631 = arith.constant 1 : index
%dim_35632 = tensor.dim %38184, %c1_35631 : tensor<4x?x4x64xcomplex<f32>>
%38185 = flow.tensor.bitcast %38184 : tensor<4x?x4x64xcomplex<f32>>{%dim_35632} -> tensor<4x?x4x128xf32>{%dim_35632}
%38186 = torch_c.from_builtin_tensor %38185 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38186, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35633 = torch.constant.int 5
%38187 = torch.prims.convert_element_type %38186, %int5_35633 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38187, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35634 = torch.constant.int 1
%38188 = torch.aten.size.int %37886, %int1_35634 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35635 = torch.constant.int 0
%38189 = torch.aten.add.int %int0_35635, %38188 : !torch.int, !torch.int -> !torch.int
%int0_35636 = torch.constant.int 0
%int0_35637 = torch.constant.int 0
%int1_35638 = torch.constant.int 1
%38190 = torch.aten.slice.Tensor %38106, %int0_35636, %int0_35637, %38189, %int1_35638 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38190, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35639 = torch.constant.int 1
%int0_35640 = torch.constant.int 0
%int9223372036854775807_35641 = torch.constant.int 9223372036854775807
%int1_35642 = torch.constant.int 1
%38191 = torch.aten.slice.Tensor %38190, %int1_35639, %int0_35640, %int9223372036854775807_35641, %int1_35642 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38191, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35643 = torch.constant.int 0
%38192 = torch.aten.unsqueeze %38191, %int0_35643 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38192, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35644 = torch.constant.int 2
%38193 = torch.aten.unsqueeze %38192, %int2_35644 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38193, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35645 = torch.constant.int 3
%int0_35646 = torch.constant.int 0
%int9223372036854775807_35647 = torch.constant.int 9223372036854775807
%int1_35648 = torch.constant.int 1
%38194 = torch.aten.slice.Tensor %38193, %int3_35645, %int0_35646, %int9223372036854775807_35647, %int1_35648 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38194, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38195 = torch_c.to_builtin_tensor %38038 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35649 = arith.constant 1 : index
%dim_35650 = tensor.dim %38195, %c1_35649 : tensor<4x?x4x128xf16>
%38196 = flow.tensor.bitcast %38195 : tensor<4x?x4x128xf16>{%dim_35650} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35650}
%38197 = torch_c.from_builtin_tensor %38196 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38198 = torch.aten.mul.Tensor %38197, %38194 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38198, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38199 = torch_c.to_builtin_tensor %38198 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35651 = arith.constant 1 : index
%dim_35652 = tensor.dim %38199, %c1_35651 : tensor<4x?x4x64xcomplex<f32>>
%38200 = flow.tensor.bitcast %38199 : tensor<4x?x4x64xcomplex<f32>>{%dim_35652} -> tensor<4x?x4x128xf32>{%dim_35652}
%38201 = torch_c.from_builtin_tensor %38200 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38201, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35653 = torch.constant.int 5
%38202 = torch.prims.convert_element_type %38201, %int5_35653 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38202, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35654 = torch.constant.int 1
%38203 = torch.aten.size.int %37892, %int1_35654 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35655 = torch.constant.int 0
%38204 = torch.aten.add.int %int0_35655, %38203 : !torch.int, !torch.int -> !torch.int
%int0_35656 = torch.constant.int 0
%int0_35657 = torch.constant.int 0
%int1_35658 = torch.constant.int 1
%38205 = torch.aten.slice.Tensor %38109, %int0_35656, %int0_35657, %38204, %int1_35658 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38205, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35659 = torch.constant.int 1
%int0_35660 = torch.constant.int 0
%int9223372036854775807_35661 = torch.constant.int 9223372036854775807
%int1_35662 = torch.constant.int 1
%38206 = torch.aten.slice.Tensor %38205, %int1_35659, %int0_35660, %int9223372036854775807_35661, %int1_35662 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38206, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35663 = torch.constant.int 0
%38207 = torch.aten.unsqueeze %38206, %int0_35663 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38207, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35664 = torch.constant.int 2
%38208 = torch.aten.unsqueeze %38207, %int2_35664 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38208, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35665 = torch.constant.int 3
%int0_35666 = torch.constant.int 0
%int9223372036854775807_35667 = torch.constant.int 9223372036854775807
%int1_35668 = torch.constant.int 1
%38209 = torch.aten.slice.Tensor %38208, %int3_35665, %int0_35666, %int9223372036854775807_35667, %int1_35668 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38209, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38210 = torch_c.to_builtin_tensor %38040 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35669 = arith.constant 1 : index
%dim_35670 = tensor.dim %38210, %c1_35669 : tensor<4x?x4x128xf16>
%38211 = flow.tensor.bitcast %38210 : tensor<4x?x4x128xf16>{%dim_35670} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35670}
%38212 = torch_c.from_builtin_tensor %38211 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38212, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38213 = torch.aten.mul.Tensor %38212, %38209 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38213, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38214 = torch_c.to_builtin_tensor %38213 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35671 = arith.constant 1 : index
%dim_35672 = tensor.dim %38214, %c1_35671 : tensor<4x?x4x64xcomplex<f32>>
%38215 = flow.tensor.bitcast %38214 : tensor<4x?x4x64xcomplex<f32>>{%dim_35672} -> tensor<4x?x4x128xf32>{%dim_35672}
%38216 = torch_c.from_builtin_tensor %38215 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38216, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35673 = torch.constant.int 5
%38217 = torch.prims.convert_element_type %38216, %int5_35673 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38217, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_35674 = torch.constant.int 1
%38218 = torch.aten.size.int %37898, %int1_35674 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_35675 = torch.constant.int 0
%38219 = torch.aten.add.int %int0_35675, %38218 : !torch.int, !torch.int -> !torch.int
%int0_35676 = torch.constant.int 0
%int0_35677 = torch.constant.int 0
%int1_35678 = torch.constant.int 1
%38220 = torch.aten.slice.Tensor %38112, %int0_35676, %int0_35677, %38219, %int1_35678 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38220, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35679 = torch.constant.int 1
%int0_35680 = torch.constant.int 0
%int9223372036854775807_35681 = torch.constant.int 9223372036854775807
%int1_35682 = torch.constant.int 1
%38221 = torch.aten.slice.Tensor %38220, %int1_35679, %int0_35680, %int9223372036854775807_35681, %int1_35682 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38221, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35683 = torch.constant.int 0
%38222 = torch.aten.unsqueeze %38221, %int0_35683 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38222, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35684 = torch.constant.int 2
%38223 = torch.aten.unsqueeze %38222, %int2_35684 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38223, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35685 = torch.constant.int 3
%int0_35686 = torch.constant.int 0
%int9223372036854775807_35687 = torch.constant.int 9223372036854775807
%int1_35688 = torch.constant.int 1
%38224 = torch.aten.slice.Tensor %38223, %int3_35685, %int0_35686, %int9223372036854775807_35687, %int1_35688 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38224, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38225 = torch_c.to_builtin_tensor %38042 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_35689 = arith.constant 1 : index
%dim_35690 = tensor.dim %38225, %c1_35689 : tensor<4x?x4x128xf16>
%38226 = flow.tensor.bitcast %38225 : tensor<4x?x4x128xf16>{%dim_35690} -> tensor<4x?x4x64xcomplex<f16>>{%dim_35690}
%38227 = torch_c.from_builtin_tensor %38226 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %38227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%38228 = torch.aten.mul.Tensor %38227, %38224 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %38228, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%38229 = torch_c.to_builtin_tensor %38228 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_35691 = arith.constant 1 : index
%dim_35692 = tensor.dim %38229, %c1_35691 : tensor<4x?x4x64xcomplex<f32>>
%38230 = flow.tensor.bitcast %38229 : tensor<4x?x4x64xcomplex<f32>>{%dim_35692} -> tensor<4x?x4x128xf32>{%dim_35692}
%38231 = torch_c.from_builtin_tensor %38230 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %38231, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_35693 = torch.constant.int 5
%38232 = torch.prims.convert_element_type %38231, %int5_35693 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38232, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_35694 = torch.constant.int 131072
%none_35695 = torch.constant.none
%none_35696 = torch.constant.none
%cpu_35697 = torch.constant.device "cpu"
%false_35698 = torch.constant.bool false
%38233 = torch.aten.arange %int131072_35694, %none_35695, %none_35696, %cpu_35697, %false_35698 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_35699 = torch.constant.int 0
%int128_35700 = torch.constant.int 128
%int2_35701 = torch.constant.int 2
%none_35702 = torch.constant.none
%none_35703 = torch.constant.none
%cpu_35704 = torch.constant.device "cpu"
%false_35705 = torch.constant.bool false
%38234 = torch.aten.arange.start_step %int0_35699, %int128_35700, %int2_35701, %none_35702, %none_35703, %cpu_35704, %false_35705 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_35706 = torch.constant.int 0
%int0_35707 = torch.constant.int 0
%int64_35708 = torch.constant.int 64
%int1_35709 = torch.constant.int 1
%38235 = torch.aten.slice.Tensor %38234, %int0_35706, %int0_35707, %int64_35708, %int1_35709 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_35710 = torch.constant.int 6
%38236 = torch.prims.convert_element_type %38235, %int6_35710 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_35711 = torch.constant.int 128
%38237 = torch.aten.div.Scalar %38236, %int128_35711 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_35712 = torch.constant.float 5.000000e+05
%38238 = torch.aten.pow.Scalar %float5.000000e05_35712, %38237 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%38239 = torch.aten.reciprocal %38238 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_35713 = torch.constant.float 1.000000e+00
%38240 = torch.aten.mul.Scalar %38239, %float1.000000e00_35713 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_35714 = torch.constant.int 131072
%int1_35715 = torch.constant.int 1
%38241 = torch.prim.ListConstruct %int131072_35714, %int1_35715 : (!torch.int, !torch.int) -> !torch.list<int>
%38242 = torch.aten.view %38233, %38241 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%38243 = torch.aten.mul.Tensor %38242, %38240 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%38244 = torch.aten.cos %38243 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%38245 = torch.aten.sin %38243 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%38246 = torch.aten.complex %38244, %38245 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%38247 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38248 = flow.tensor.transfer %38247 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%38249 = torch_c.from_builtin_tensor %38248 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38250 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38251 = flow.tensor.transfer %38250 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%38252 = torch_c.from_builtin_tensor %38251 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38253 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38254 = flow.tensor.transfer %38253 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%38255 = torch_c.from_builtin_tensor %38254 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38256 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38257 = flow.tensor.transfer %38256 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%38258 = torch_c.from_builtin_tensor %38257 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38259 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38260 = flow.tensor.transfer %38259 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%38261 = torch_c.from_builtin_tensor %38260 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38262 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38263 = flow.tensor.transfer %38262 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%38264 = torch_c.from_builtin_tensor %38263 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38265 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38266 = flow.tensor.transfer %38265 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%38267 = torch_c.from_builtin_tensor %38266 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%38268 = torch_c.to_builtin_tensor %38246 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%38269 = flow.tensor.transfer %38268 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%38270 = torch_c.from_builtin_tensor %38269 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_35716 = torch.constant.int 1
%38271 = torch.aten.size.int %37920, %int1_35716 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35717 = torch.constant.int 0
%38272 = torch.aten.add.int %int0_35717, %38271 : !torch.int, !torch.int -> !torch.int
%int0_35718 = torch.constant.int 0
%int0_35719 = torch.constant.int 0
%int1_35720 = torch.constant.int 1
%38273 = torch.aten.slice.Tensor %38249, %int0_35718, %int0_35719, %38272, %int1_35720 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38273, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35721 = torch.constant.int 1
%int0_35722 = torch.constant.int 0
%int9223372036854775807_35723 = torch.constant.int 9223372036854775807
%int1_35724 = torch.constant.int 1
%38274 = torch.aten.slice.Tensor %38273, %int1_35721, %int0_35722, %int9223372036854775807_35723, %int1_35724 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38274, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35725 = torch.constant.int 0
%38275 = torch.aten.unsqueeze %38274, %int0_35725 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38275, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35726 = torch.constant.int 2
%38276 = torch.aten.unsqueeze %38275, %int2_35726 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38276, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35727 = torch.constant.int 3
%int0_35728 = torch.constant.int 0
%int9223372036854775807_35729 = torch.constant.int 9223372036854775807
%int1_35730 = torch.constant.int 1
%38277 = torch.aten.slice.Tensor %38276, %int3_35727, %int0_35728, %int9223372036854775807_35729, %int1_35730 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38277, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38278 = torch_c.to_builtin_tensor %38044 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35731 = arith.constant 1 : index
%dim_35732 = tensor.dim %38278, %c1_35731 : tensor<4x?x1x128xf16>
%38279 = flow.tensor.bitcast %38278 : tensor<4x?x1x128xf16>{%dim_35732} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35732}
%38280 = torch_c.from_builtin_tensor %38279 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38280, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38281 = torch.aten.mul.Tensor %38280, %38277 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38282 = torch_c.to_builtin_tensor %38281 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35733 = arith.constant 1 : index
%dim_35734 = tensor.dim %38282, %c1_35733 : tensor<4x?x1x64xcomplex<f32>>
%38283 = flow.tensor.bitcast %38282 : tensor<4x?x1x64xcomplex<f32>>{%dim_35734} -> tensor<4x?x1x128xf32>{%dim_35734}
%38284 = torch_c.from_builtin_tensor %38283 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38284, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35735 = torch.constant.int 5
%38285 = torch.prims.convert_element_type %38284, %int5_35735 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38285, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35736 = torch.constant.int 1
%38286 = torch.aten.size.int %37926, %int1_35736 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35737 = torch.constant.int 0
%38287 = torch.aten.add.int %int0_35737, %38286 : !torch.int, !torch.int -> !torch.int
%int0_35738 = torch.constant.int 0
%int0_35739 = torch.constant.int 0
%int1_35740 = torch.constant.int 1
%38288 = torch.aten.slice.Tensor %38252, %int0_35738, %int0_35739, %38287, %int1_35740 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38288, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35741 = torch.constant.int 1
%int0_35742 = torch.constant.int 0
%int9223372036854775807_35743 = torch.constant.int 9223372036854775807
%int1_35744 = torch.constant.int 1
%38289 = torch.aten.slice.Tensor %38288, %int1_35741, %int0_35742, %int9223372036854775807_35743, %int1_35744 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38289, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35745 = torch.constant.int 0
%38290 = torch.aten.unsqueeze %38289, %int0_35745 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38290, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35746 = torch.constant.int 2
%38291 = torch.aten.unsqueeze %38290, %int2_35746 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38291, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35747 = torch.constant.int 3
%int0_35748 = torch.constant.int 0
%int9223372036854775807_35749 = torch.constant.int 9223372036854775807
%int1_35750 = torch.constant.int 1
%38292 = torch.aten.slice.Tensor %38291, %int3_35747, %int0_35748, %int9223372036854775807_35749, %int1_35750 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38292, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38293 = torch_c.to_builtin_tensor %38046 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35751 = arith.constant 1 : index
%dim_35752 = tensor.dim %38293, %c1_35751 : tensor<4x?x1x128xf16>
%38294 = flow.tensor.bitcast %38293 : tensor<4x?x1x128xf16>{%dim_35752} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35752}
%38295 = torch_c.from_builtin_tensor %38294 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38295, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38296 = torch.aten.mul.Tensor %38295, %38292 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38296, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38297 = torch_c.to_builtin_tensor %38296 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35753 = arith.constant 1 : index
%dim_35754 = tensor.dim %38297, %c1_35753 : tensor<4x?x1x64xcomplex<f32>>
%38298 = flow.tensor.bitcast %38297 : tensor<4x?x1x64xcomplex<f32>>{%dim_35754} -> tensor<4x?x1x128xf32>{%dim_35754}
%38299 = torch_c.from_builtin_tensor %38298 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35755 = torch.constant.int 5
%38300 = torch.prims.convert_element_type %38299, %int5_35755 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38300, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35756 = torch.constant.int 1
%38301 = torch.aten.size.int %37932, %int1_35756 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35757 = torch.constant.int 0
%38302 = torch.aten.add.int %int0_35757, %38301 : !torch.int, !torch.int -> !torch.int
%int0_35758 = torch.constant.int 0
%int0_35759 = torch.constant.int 0
%int1_35760 = torch.constant.int 1
%38303 = torch.aten.slice.Tensor %38255, %int0_35758, %int0_35759, %38302, %int1_35760 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38303, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35761 = torch.constant.int 1
%int0_35762 = torch.constant.int 0
%int9223372036854775807_35763 = torch.constant.int 9223372036854775807
%int1_35764 = torch.constant.int 1
%38304 = torch.aten.slice.Tensor %38303, %int1_35761, %int0_35762, %int9223372036854775807_35763, %int1_35764 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38304, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35765 = torch.constant.int 0
%38305 = torch.aten.unsqueeze %38304, %int0_35765 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38305, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35766 = torch.constant.int 2
%38306 = torch.aten.unsqueeze %38305, %int2_35766 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38306, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35767 = torch.constant.int 3
%int0_35768 = torch.constant.int 0
%int9223372036854775807_35769 = torch.constant.int 9223372036854775807
%int1_35770 = torch.constant.int 1
%38307 = torch.aten.slice.Tensor %38306, %int3_35767, %int0_35768, %int9223372036854775807_35769, %int1_35770 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38307, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38308 = torch_c.to_builtin_tensor %38048 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35771 = arith.constant 1 : index
%dim_35772 = tensor.dim %38308, %c1_35771 : tensor<4x?x1x128xf16>
%38309 = flow.tensor.bitcast %38308 : tensor<4x?x1x128xf16>{%dim_35772} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35772}
%38310 = torch_c.from_builtin_tensor %38309 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38310, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38311 = torch.aten.mul.Tensor %38310, %38307 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38312 = torch_c.to_builtin_tensor %38311 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35773 = arith.constant 1 : index
%dim_35774 = tensor.dim %38312, %c1_35773 : tensor<4x?x1x64xcomplex<f32>>
%38313 = flow.tensor.bitcast %38312 : tensor<4x?x1x64xcomplex<f32>>{%dim_35774} -> tensor<4x?x1x128xf32>{%dim_35774}
%38314 = torch_c.from_builtin_tensor %38313 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35775 = torch.constant.int 5
%38315 = torch.prims.convert_element_type %38314, %int5_35775 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35776 = torch.constant.int 1
%38316 = torch.aten.size.int %37938, %int1_35776 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35777 = torch.constant.int 0
%38317 = torch.aten.add.int %int0_35777, %38316 : !torch.int, !torch.int -> !torch.int
%int0_35778 = torch.constant.int 0
%int0_35779 = torch.constant.int 0
%int1_35780 = torch.constant.int 1
%38318 = torch.aten.slice.Tensor %38258, %int0_35778, %int0_35779, %38317, %int1_35780 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38318, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35781 = torch.constant.int 1
%int0_35782 = torch.constant.int 0
%int9223372036854775807_35783 = torch.constant.int 9223372036854775807
%int1_35784 = torch.constant.int 1
%38319 = torch.aten.slice.Tensor %38318, %int1_35781, %int0_35782, %int9223372036854775807_35783, %int1_35784 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38319, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35785 = torch.constant.int 0
%38320 = torch.aten.unsqueeze %38319, %int0_35785 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38320, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35786 = torch.constant.int 2
%38321 = torch.aten.unsqueeze %38320, %int2_35786 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38321, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35787 = torch.constant.int 3
%int0_35788 = torch.constant.int 0
%int9223372036854775807_35789 = torch.constant.int 9223372036854775807
%int1_35790 = torch.constant.int 1
%38322 = torch.aten.slice.Tensor %38321, %int3_35787, %int0_35788, %int9223372036854775807_35789, %int1_35790 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38322, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38323 = torch_c.to_builtin_tensor %38050 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35791 = arith.constant 1 : index
%dim_35792 = tensor.dim %38323, %c1_35791 : tensor<4x?x1x128xf16>
%38324 = flow.tensor.bitcast %38323 : tensor<4x?x1x128xf16>{%dim_35792} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35792}
%38325 = torch_c.from_builtin_tensor %38324 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38325, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38326 = torch.aten.mul.Tensor %38325, %38322 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38326, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38327 = torch_c.to_builtin_tensor %38326 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35793 = arith.constant 1 : index
%dim_35794 = tensor.dim %38327, %c1_35793 : tensor<4x?x1x64xcomplex<f32>>
%38328 = flow.tensor.bitcast %38327 : tensor<4x?x1x64xcomplex<f32>>{%dim_35794} -> tensor<4x?x1x128xf32>{%dim_35794}
%38329 = torch_c.from_builtin_tensor %38328 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38329, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35795 = torch.constant.int 5
%38330 = torch.prims.convert_element_type %38329, %int5_35795 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38330, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35796 = torch.constant.int 1
%38331 = torch.aten.size.int %37944, %int1_35796 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35797 = torch.constant.int 0
%38332 = torch.aten.add.int %int0_35797, %38331 : !torch.int, !torch.int -> !torch.int
%int0_35798 = torch.constant.int 0
%int0_35799 = torch.constant.int 0
%int1_35800 = torch.constant.int 1
%38333 = torch.aten.slice.Tensor %38261, %int0_35798, %int0_35799, %38332, %int1_35800 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38333, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35801 = torch.constant.int 1
%int0_35802 = torch.constant.int 0
%int9223372036854775807_35803 = torch.constant.int 9223372036854775807
%int1_35804 = torch.constant.int 1
%38334 = torch.aten.slice.Tensor %38333, %int1_35801, %int0_35802, %int9223372036854775807_35803, %int1_35804 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38334, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35805 = torch.constant.int 0
%38335 = torch.aten.unsqueeze %38334, %int0_35805 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38335, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35806 = torch.constant.int 2
%38336 = torch.aten.unsqueeze %38335, %int2_35806 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38336, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35807 = torch.constant.int 3
%int0_35808 = torch.constant.int 0
%int9223372036854775807_35809 = torch.constant.int 9223372036854775807
%int1_35810 = torch.constant.int 1
%38337 = torch.aten.slice.Tensor %38336, %int3_35807, %int0_35808, %int9223372036854775807_35809, %int1_35810 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38337, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38338 = torch_c.to_builtin_tensor %38052 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35811 = arith.constant 1 : index
%dim_35812 = tensor.dim %38338, %c1_35811 : tensor<4x?x1x128xf16>
%38339 = flow.tensor.bitcast %38338 : tensor<4x?x1x128xf16>{%dim_35812} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35812}
%38340 = torch_c.from_builtin_tensor %38339 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38340, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38341 = torch.aten.mul.Tensor %38340, %38337 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38341, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38342 = torch_c.to_builtin_tensor %38341 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35813 = arith.constant 1 : index
%dim_35814 = tensor.dim %38342, %c1_35813 : tensor<4x?x1x64xcomplex<f32>>
%38343 = flow.tensor.bitcast %38342 : tensor<4x?x1x64xcomplex<f32>>{%dim_35814} -> tensor<4x?x1x128xf32>{%dim_35814}
%38344 = torch_c.from_builtin_tensor %38343 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38344, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35815 = torch.constant.int 5
%38345 = torch.prims.convert_element_type %38344, %int5_35815 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38345, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35816 = torch.constant.int 1
%38346 = torch.aten.size.int %37950, %int1_35816 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35817 = torch.constant.int 0
%38347 = torch.aten.add.int %int0_35817, %38346 : !torch.int, !torch.int -> !torch.int
%int0_35818 = torch.constant.int 0
%int0_35819 = torch.constant.int 0
%int1_35820 = torch.constant.int 1
%38348 = torch.aten.slice.Tensor %38264, %int0_35818, %int0_35819, %38347, %int1_35820 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38348, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35821 = torch.constant.int 1
%int0_35822 = torch.constant.int 0
%int9223372036854775807_35823 = torch.constant.int 9223372036854775807
%int1_35824 = torch.constant.int 1
%38349 = torch.aten.slice.Tensor %38348, %int1_35821, %int0_35822, %int9223372036854775807_35823, %int1_35824 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38349, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35825 = torch.constant.int 0
%38350 = torch.aten.unsqueeze %38349, %int0_35825 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38350, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35826 = torch.constant.int 2
%38351 = torch.aten.unsqueeze %38350, %int2_35826 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38351, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35827 = torch.constant.int 3
%int0_35828 = torch.constant.int 0
%int9223372036854775807_35829 = torch.constant.int 9223372036854775807
%int1_35830 = torch.constant.int 1
%38352 = torch.aten.slice.Tensor %38351, %int3_35827, %int0_35828, %int9223372036854775807_35829, %int1_35830 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38352, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38353 = torch_c.to_builtin_tensor %38054 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35831 = arith.constant 1 : index
%dim_35832 = tensor.dim %38353, %c1_35831 : tensor<4x?x1x128xf16>
%38354 = flow.tensor.bitcast %38353 : tensor<4x?x1x128xf16>{%dim_35832} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35832}
%38355 = torch_c.from_builtin_tensor %38354 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38355, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38356 = torch.aten.mul.Tensor %38355, %38352 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38357 = torch_c.to_builtin_tensor %38356 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35833 = arith.constant 1 : index
%dim_35834 = tensor.dim %38357, %c1_35833 : tensor<4x?x1x64xcomplex<f32>>
%38358 = flow.tensor.bitcast %38357 : tensor<4x?x1x64xcomplex<f32>>{%dim_35834} -> tensor<4x?x1x128xf32>{%dim_35834}
%38359 = torch_c.from_builtin_tensor %38358 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38359, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35835 = torch.constant.int 5
%38360 = torch.prims.convert_element_type %38359, %int5_35835 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38360, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35836 = torch.constant.int 1
%38361 = torch.aten.size.int %37956, %int1_35836 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35837 = torch.constant.int 0
%38362 = torch.aten.add.int %int0_35837, %38361 : !torch.int, !torch.int -> !torch.int
%int0_35838 = torch.constant.int 0
%int0_35839 = torch.constant.int 0
%int1_35840 = torch.constant.int 1
%38363 = torch.aten.slice.Tensor %38267, %int0_35838, %int0_35839, %38362, %int1_35840 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38363, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35841 = torch.constant.int 1
%int0_35842 = torch.constant.int 0
%int9223372036854775807_35843 = torch.constant.int 9223372036854775807
%int1_35844 = torch.constant.int 1
%38364 = torch.aten.slice.Tensor %38363, %int1_35841, %int0_35842, %int9223372036854775807_35843, %int1_35844 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38364, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35845 = torch.constant.int 0
%38365 = torch.aten.unsqueeze %38364, %int0_35845 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38365, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35846 = torch.constant.int 2
%38366 = torch.aten.unsqueeze %38365, %int2_35846 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38366, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35847 = torch.constant.int 3
%int0_35848 = torch.constant.int 0
%int9223372036854775807_35849 = torch.constant.int 9223372036854775807
%int1_35850 = torch.constant.int 1
%38367 = torch.aten.slice.Tensor %38366, %int3_35847, %int0_35848, %int9223372036854775807_35849, %int1_35850 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38367, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38368 = torch_c.to_builtin_tensor %38056 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35851 = arith.constant 1 : index
%dim_35852 = tensor.dim %38368, %c1_35851 : tensor<4x?x1x128xf16>
%38369 = flow.tensor.bitcast %38368 : tensor<4x?x1x128xf16>{%dim_35852} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35852}
%38370 = torch_c.from_builtin_tensor %38369 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38371 = torch.aten.mul.Tensor %38370, %38367 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38371, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38372 = torch_c.to_builtin_tensor %38371 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35853 = arith.constant 1 : index
%dim_35854 = tensor.dim %38372, %c1_35853 : tensor<4x?x1x64xcomplex<f32>>
%38373 = flow.tensor.bitcast %38372 : tensor<4x?x1x64xcomplex<f32>>{%dim_35854} -> tensor<4x?x1x128xf32>{%dim_35854}
%38374 = torch_c.from_builtin_tensor %38373 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38374, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35855 = torch.constant.int 5
%38375 = torch.prims.convert_element_type %38374, %int5_35855 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38375, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_35856 = torch.constant.int 1
%38376 = torch.aten.size.int %37962, %int1_35856 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_35857 = torch.constant.int 0
%38377 = torch.aten.add.int %int0_35857, %38376 : !torch.int, !torch.int -> !torch.int
%int0_35858 = torch.constant.int 0
%int0_35859 = torch.constant.int 0
%int1_35860 = torch.constant.int 1
%38378 = torch.aten.slice.Tensor %38270, %int0_35858, %int0_35859, %38377, %int1_35860 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38378, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_35861 = torch.constant.int 1
%int0_35862 = torch.constant.int 0
%int9223372036854775807_35863 = torch.constant.int 9223372036854775807
%int1_35864 = torch.constant.int 1
%38379 = torch.aten.slice.Tensor %38378, %int1_35861, %int0_35862, %int9223372036854775807_35863, %int1_35864 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %38379, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_35865 = torch.constant.int 0
%38380 = torch.aten.unsqueeze %38379, %int0_35865 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %38380, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_35866 = torch.constant.int 2
%38381 = torch.aten.unsqueeze %38380, %int2_35866 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38381, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_35867 = torch.constant.int 3
%int0_35868 = torch.constant.int 0
%int9223372036854775807_35869 = torch.constant.int 9223372036854775807
%int1_35870 = torch.constant.int 1
%38382 = torch.aten.slice.Tensor %38381, %int3_35867, %int0_35868, %int9223372036854775807_35869, %int1_35870 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38382, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%38383 = torch_c.to_builtin_tensor %38058 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_35871 = arith.constant 1 : index
%dim_35872 = tensor.dim %38383, %c1_35871 : tensor<4x?x1x128xf16>
%38384 = flow.tensor.bitcast %38383 : tensor<4x?x1x128xf16>{%dim_35872} -> tensor<4x?x1x64xcomplex<f16>>{%dim_35872}
%38385 = torch_c.from_builtin_tensor %38384 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %38385, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%38386 = torch.aten.mul.Tensor %38385, %38382 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %38386, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%38387 = torch_c.to_builtin_tensor %38386 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_35873 = arith.constant 1 : index
%dim_35874 = tensor.dim %38387, %c1_35873 : tensor<4x?x1x64xcomplex<f32>>
%38388 = flow.tensor.bitcast %38387 : tensor<4x?x1x64xcomplex<f32>>{%dim_35874} -> tensor<4x?x1x128xf32>{%dim_35874}
%38389 = torch_c.from_builtin_tensor %38388 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %38389, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_35875 = torch.constant.int 5
%38390 = torch.prims.convert_element_type %38389, %int5_35875 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %38390, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int64_35876 = torch.constant.int 64
%38391 = torch.aten.mul.Scalar %2364, %int64_35876 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38391, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35877 = torch.constant.int 64
%38392 = torch.aten.mul.Scalar %2367, %int64_35877 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38392, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35878 = torch.constant.int 64
%38393 = torch.aten.mul.Scalar %2370, %int64_35878 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38393, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35879 = torch.constant.int 64
%38394 = torch.aten.mul.Scalar %2373, %int64_35879 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38394, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35880 = torch.constant.int 64
%38395 = torch.aten.mul.Scalar %2376, %int64_35880 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38395, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35881 = torch.constant.int 64
%38396 = torch.aten.mul.Scalar %2379, %int64_35881 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38396, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35882 = torch.constant.int 64
%38397 = torch.aten.mul.Scalar %2382, %int64_35882 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38397, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int64_35883 = torch.constant.int 64
%38398 = torch.aten.mul.Scalar %2385, %int64_35883 : !torch.vtensor<[4,?],si64>, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38398, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38 = torch.constant.int 38
%int1_35884 = torch.constant.int 1
%38399 = torch.aten.add.Scalar %38391, %int38, %int1_35884 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38399, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35885 = torch.constant.int 38
%int1_35886 = torch.constant.int 1
%38400 = torch.aten.add.Scalar %38392, %int38_35885, %int1_35886 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38400, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35887 = torch.constant.int 38
%int1_35888 = torch.constant.int 1
%38401 = torch.aten.add.Scalar %38393, %int38_35887, %int1_35888 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38401, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35889 = torch.constant.int 38
%int1_35890 = torch.constant.int 1
%38402 = torch.aten.add.Scalar %38394, %int38_35889, %int1_35890 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38402, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35891 = torch.constant.int 38
%int1_35892 = torch.constant.int 1
%38403 = torch.aten.add.Scalar %38395, %int38_35891, %int1_35892 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38403, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35893 = torch.constant.int 38
%int1_35894 = torch.constant.int 1
%38404 = torch.aten.add.Scalar %38396, %int38_35893, %int1_35894 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38404, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35895 = torch.constant.int 38
%int1_35896 = torch.constant.int 1
%38405 = torch.aten.add.Scalar %38397, %int38_35895, %int1_35896 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38405, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int38_35897 = torch.constant.int 38
%int1_35898 = torch.constant.int 1
%38406 = torch.aten.add.Scalar %38398, %int38_35897, %int1_35898 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38406, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_35899 = torch.constant.int 4
%int16_35900 = torch.constant.int 16
%int1_35901 = torch.constant.int 1
%int128_35902 = torch.constant.int 128
%38407 = torch.prim.ListConstruct %int4_35899, %3095, %int16_35900, %int1_35901, %int128_35902 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38408 = torch.aten.view %38285, %38407 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38408, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35903 = torch.constant.int 4
%int16_35904 = torch.constant.int 16
%int1_35905 = torch.constant.int 1
%int128_35906 = torch.constant.int 128
%38409 = torch.prim.ListConstruct %int4_35903, %3095, %int16_35904, %int1_35905, %int128_35906 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38410 = torch.aten.view %38300, %38409 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38410, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35907 = torch.constant.int 4
%int16_35908 = torch.constant.int 16
%int1_35909 = torch.constant.int 1
%int128_35910 = torch.constant.int 128
%38411 = torch.prim.ListConstruct %int4_35907, %3095, %int16_35908, %int1_35909, %int128_35910 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38412 = torch.aten.view %38315, %38411 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38412, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35911 = torch.constant.int 4
%int16_35912 = torch.constant.int 16
%int1_35913 = torch.constant.int 1
%int128_35914 = torch.constant.int 128
%38413 = torch.prim.ListConstruct %int4_35911, %3095, %int16_35912, %int1_35913, %int128_35914 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38414 = torch.aten.view %38330, %38413 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38414, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35915 = torch.constant.int 4
%int16_35916 = torch.constant.int 16
%int1_35917 = torch.constant.int 1
%int128_35918 = torch.constant.int 128
%38415 = torch.prim.ListConstruct %int4_35915, %3095, %int16_35916, %int1_35917, %int128_35918 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38416 = torch.aten.view %38345, %38415 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38416, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35919 = torch.constant.int 4
%int16_35920 = torch.constant.int 16
%int1_35921 = torch.constant.int 1
%int128_35922 = torch.constant.int 128
%38417 = torch.prim.ListConstruct %int4_35919, %3095, %int16_35920, %int1_35921, %int128_35922 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38418 = torch.aten.view %38360, %38417 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38418, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35923 = torch.constant.int 4
%int16_35924 = torch.constant.int 16
%int1_35925 = torch.constant.int 1
%int128_35926 = torch.constant.int 128
%38419 = torch.prim.ListConstruct %int4_35923, %3095, %int16_35924, %int1_35925, %int128_35926 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38420 = torch.aten.view %38375, %38419 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38420, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35927 = torch.constant.int 4
%int16_35928 = torch.constant.int 16
%int1_35929 = torch.constant.int 1
%int128_35930 = torch.constant.int 128
%38421 = torch.prim.ListConstruct %int4_35927, %3095, %int16_35928, %int1_35929, %int128_35930 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38422 = torch.aten.view %38390, %38421 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38422, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35931 = torch.constant.int 4
%38423 = torch.aten.mul.int %int4_35931, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35932 = torch.constant.int 16
%int1_35933 = torch.constant.int 1
%int128_35934 = torch.constant.int 128
%38424 = torch.prim.ListConstruct %38423, %int16_35932, %int1_35933, %int128_35934 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38425 = torch.aten.view %38408, %38424 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38425, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35935 = torch.constant.int 4
%38426 = torch.aten.mul.int %int4_35935, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35936 = torch.constant.int 16
%int1_35937 = torch.constant.int 1
%int128_35938 = torch.constant.int 128
%38427 = torch.prim.ListConstruct %38426, %int16_35936, %int1_35937, %int128_35938 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38428 = torch.aten.view %38410, %38427 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38428, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35939 = torch.constant.int 4
%38429 = torch.aten.mul.int %int4_35939, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35940 = torch.constant.int 16
%int1_35941 = torch.constant.int 1
%int128_35942 = torch.constant.int 128
%38430 = torch.prim.ListConstruct %38429, %int16_35940, %int1_35941, %int128_35942 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38431 = torch.aten.view %38412, %38430 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38431, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35943 = torch.constant.int 4
%38432 = torch.aten.mul.int %int4_35943, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35944 = torch.constant.int 16
%int1_35945 = torch.constant.int 1
%int128_35946 = torch.constant.int 128
%38433 = torch.prim.ListConstruct %38432, %int16_35944, %int1_35945, %int128_35946 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38434 = torch.aten.view %38414, %38433 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38434, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35947 = torch.constant.int 4
%38435 = torch.aten.mul.int %int4_35947, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35948 = torch.constant.int 16
%int1_35949 = torch.constant.int 1
%int128_35950 = torch.constant.int 128
%38436 = torch.prim.ListConstruct %38435, %int16_35948, %int1_35949, %int128_35950 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38437 = torch.aten.view %38416, %38436 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38437, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35951 = torch.constant.int 4
%38438 = torch.aten.mul.int %int4_35951, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35952 = torch.constant.int 16
%int1_35953 = torch.constant.int 1
%int128_35954 = torch.constant.int 128
%38439 = torch.prim.ListConstruct %38438, %int16_35952, %int1_35953, %int128_35954 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38440 = torch.aten.view %38418, %38439 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38440, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35955 = torch.constant.int 4
%38441 = torch.aten.mul.int %int4_35955, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35956 = torch.constant.int 16
%int1_35957 = torch.constant.int 1
%int128_35958 = torch.constant.int 128
%38442 = torch.prim.ListConstruct %38441, %int16_35956, %int1_35957, %int128_35958 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38443 = torch.aten.view %38420, %38442 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38443, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35959 = torch.constant.int 4
%38444 = torch.aten.mul.int %int4_35959, %3095 : !torch.int, !torch.int -> !torch.int
%int16_35960 = torch.constant.int 16
%int1_35961 = torch.constant.int 1
%int128_35962 = torch.constant.int 128
%38445 = torch.prim.ListConstruct %38444, %int16_35960, %int1_35961, %int128_35962 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38446 = torch.aten.view %38422, %38445 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38446, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_35963 = torch.constant.int 4
%38447 = torch.aten.mul.int %int4_35963, %3095 : !torch.int, !torch.int -> !torch.int
%38448 = torch.prim.ListConstruct %38447 : (!torch.int) -> !torch.list<int>
%38449 = torch.aten.view %38399, %38448 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38449, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35964 = torch.constant.int 4
%38450 = torch.aten.mul.int %int4_35964, %3095 : !torch.int, !torch.int -> !torch.int
%38451 = torch.prim.ListConstruct %38450 : (!torch.int) -> !torch.list<int>
%38452 = torch.aten.view %38400, %38451 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38452, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35965 = torch.constant.int 4
%38453 = torch.aten.mul.int %int4_35965, %3095 : !torch.int, !torch.int -> !torch.int
%38454 = torch.prim.ListConstruct %38453 : (!torch.int) -> !torch.list<int>
%38455 = torch.aten.view %38401, %38454 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38455, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35966 = torch.constant.int 4
%38456 = torch.aten.mul.int %int4_35966, %3095 : !torch.int, !torch.int -> !torch.int
%38457 = torch.prim.ListConstruct %38456 : (!torch.int) -> !torch.list<int>
%38458 = torch.aten.view %38402, %38457 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38458, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35967 = torch.constant.int 4
%38459 = torch.aten.mul.int %int4_35967, %3095 : !torch.int, !torch.int -> !torch.int
%38460 = torch.prim.ListConstruct %38459 : (!torch.int) -> !torch.list<int>
%38461 = torch.aten.view %38403, %38460 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38461, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35968 = torch.constant.int 4
%38462 = torch.aten.mul.int %int4_35968, %3095 : !torch.int, !torch.int -> !torch.int
%38463 = torch.prim.ListConstruct %38462 : (!torch.int) -> !torch.list<int>
%38464 = torch.aten.view %38404, %38463 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38464, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35969 = torch.constant.int 4
%38465 = torch.aten.mul.int %int4_35969, %3095 : !torch.int, !torch.int -> !torch.int
%38466 = torch.prim.ListConstruct %38465 : (!torch.int) -> !torch.list<int>
%38467 = torch.aten.view %38405, %38466 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38467, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35970 = torch.constant.int 4
%38468 = torch.aten.mul.int %int4_35970, %3095 : !torch.int, !torch.int -> !torch.int
%38469 = torch.prim.ListConstruct %38468 : (!torch.int) -> !torch.list<int>
%38470 = torch.aten.view %38406, %38469 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38470, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_35971 = torch.constant.int 4
%int16_35972 = torch.constant.int 16
%int1_35973 = torch.constant.int 1
%int128_35974 = torch.constant.int 128
%38471 = torch.prim.ListConstruct %int4_35971, %3095, %int16_35972, %int1_35973, %int128_35974 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38472 = torch.aten.view %38060, %38471 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38472, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35975 = torch.constant.int 4
%int16_35976 = torch.constant.int 16
%int1_35977 = torch.constant.int 1
%int128_35978 = torch.constant.int 128
%38473 = torch.prim.ListConstruct %int4_35975, %3095, %int16_35976, %int1_35977, %int128_35978 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38474 = torch.aten.view %38062, %38473 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38474, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35979 = torch.constant.int 4
%int16_35980 = torch.constant.int 16
%int1_35981 = torch.constant.int 1
%int128_35982 = torch.constant.int 128
%38475 = torch.prim.ListConstruct %int4_35979, %3095, %int16_35980, %int1_35981, %int128_35982 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38476 = torch.aten.view %38064, %38475 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38476, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35983 = torch.constant.int 4
%int16_35984 = torch.constant.int 16
%int1_35985 = torch.constant.int 1
%int128_35986 = torch.constant.int 128
%38477 = torch.prim.ListConstruct %int4_35983, %3095, %int16_35984, %int1_35985, %int128_35986 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38478 = torch.aten.view %38066, %38477 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38478, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35987 = torch.constant.int 4
%int16_35988 = torch.constant.int 16
%int1_35989 = torch.constant.int 1
%int128_35990 = torch.constant.int 128
%38479 = torch.prim.ListConstruct %int4_35987, %3095, %int16_35988, %int1_35989, %int128_35990 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38480 = torch.aten.view %38068, %38479 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38480, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35991 = torch.constant.int 4
%int16_35992 = torch.constant.int 16
%int1_35993 = torch.constant.int 1
%int128_35994 = torch.constant.int 128
%38481 = torch.prim.ListConstruct %int4_35991, %3095, %int16_35992, %int1_35993, %int128_35994 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38482 = torch.aten.view %38070, %38481 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38482, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35995 = torch.constant.int 4
%int16_35996 = torch.constant.int 16
%int1_35997 = torch.constant.int 1
%int128_35998 = torch.constant.int 128
%38483 = torch.prim.ListConstruct %int4_35995, %3095, %int16_35996, %int1_35997, %int128_35998 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38484 = torch.aten.view %38072, %38483 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38484, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_35999 = torch.constant.int 4
%int16_36000 = torch.constant.int 16
%int1_36001 = torch.constant.int 1
%int128_36002 = torch.constant.int 128
%38485 = torch.prim.ListConstruct %int4_35999, %3095, %int16_36000, %int1_36001, %int128_36002 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38486 = torch.aten.view %38074, %38485 : !torch.vtensor<[4,?,1,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,16,1,128],f16>
torch.bind_symbolic_shape %38486, [%2336], affine_map<()[s0] -> (4, s0, 16, 1, 128)> : !torch.vtensor<[4,?,16,1,128],f16>
%int4_36003 = torch.constant.int 4
%38487 = torch.aten.mul.int %int4_36003, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36004 = torch.constant.int 16
%int1_36005 = torch.constant.int 1
%int128_36006 = torch.constant.int 128
%38488 = torch.prim.ListConstruct %38487, %int16_36004, %int1_36005, %int128_36006 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38489 = torch.aten.view %38472, %38488 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38489, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36007 = torch.constant.int 4
%38490 = torch.aten.mul.int %int4_36007, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36008 = torch.constant.int 16
%int1_36009 = torch.constant.int 1
%int128_36010 = torch.constant.int 128
%38491 = torch.prim.ListConstruct %38490, %int16_36008, %int1_36009, %int128_36010 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38492 = torch.aten.view %38474, %38491 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38492, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36011 = torch.constant.int 4
%38493 = torch.aten.mul.int %int4_36011, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36012 = torch.constant.int 16
%int1_36013 = torch.constant.int 1
%int128_36014 = torch.constant.int 128
%38494 = torch.prim.ListConstruct %38493, %int16_36012, %int1_36013, %int128_36014 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38495 = torch.aten.view %38476, %38494 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38495, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36015 = torch.constant.int 4
%38496 = torch.aten.mul.int %int4_36015, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36016 = torch.constant.int 16
%int1_36017 = torch.constant.int 1
%int128_36018 = torch.constant.int 128
%38497 = torch.prim.ListConstruct %38496, %int16_36016, %int1_36017, %int128_36018 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38498 = torch.aten.view %38478, %38497 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38498, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36019 = torch.constant.int 4
%38499 = torch.aten.mul.int %int4_36019, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36020 = torch.constant.int 16
%int1_36021 = torch.constant.int 1
%int128_36022 = torch.constant.int 128
%38500 = torch.prim.ListConstruct %38499, %int16_36020, %int1_36021, %int128_36022 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38501 = torch.aten.view %38480, %38500 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38501, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36023 = torch.constant.int 4
%38502 = torch.aten.mul.int %int4_36023, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36024 = torch.constant.int 16
%int1_36025 = torch.constant.int 1
%int128_36026 = torch.constant.int 128
%38503 = torch.prim.ListConstruct %38502, %int16_36024, %int1_36025, %int128_36026 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38504 = torch.aten.view %38482, %38503 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38504, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36027 = torch.constant.int 4
%38505 = torch.aten.mul.int %int4_36027, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36028 = torch.constant.int 16
%int1_36029 = torch.constant.int 1
%int128_36030 = torch.constant.int 128
%38506 = torch.prim.ListConstruct %38505, %int16_36028, %int1_36029, %int128_36030 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38507 = torch.aten.view %38484, %38506 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38507, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int4_36031 = torch.constant.int 4
%38508 = torch.aten.mul.int %int4_36031, %3095 : !torch.int, !torch.int -> !torch.int
%int16_36032 = torch.constant.int 16
%int1_36033 = torch.constant.int 1
%int128_36034 = torch.constant.int 128
%38509 = torch.prim.ListConstruct %38508, %int16_36032, %int1_36033, %int128_36034 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38510 = torch.aten.view %38486, %38509 : !torch.vtensor<[4,?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38510, [%2336], affine_map<()[s0] -> (s0 * 4, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int1_36035 = torch.constant.int 1
%int1_36036 = torch.constant.int 1
%38511 = torch.aten.add.Scalar %38399, %int1_36035, %int1_36036 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38511, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36037 = torch.constant.int 1
%int1_36038 = torch.constant.int 1
%38512 = torch.aten.add.Scalar %38400, %int1_36037, %int1_36038 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38512, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36039 = torch.constant.int 1
%int1_36040 = torch.constant.int 1
%38513 = torch.aten.add.Scalar %38401, %int1_36039, %int1_36040 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38513, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36041 = torch.constant.int 1
%int1_36042 = torch.constant.int 1
%38514 = torch.aten.add.Scalar %38402, %int1_36041, %int1_36042 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38514, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36043 = torch.constant.int 1
%int1_36044 = torch.constant.int 1
%38515 = torch.aten.add.Scalar %38403, %int1_36043, %int1_36044 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38515, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36045 = torch.constant.int 1
%int1_36046 = torch.constant.int 1
%38516 = torch.aten.add.Scalar %38404, %int1_36045, %int1_36046 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38516, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36047 = torch.constant.int 1
%int1_36048 = torch.constant.int 1
%38517 = torch.aten.add.Scalar %38405, %int1_36047, %int1_36048 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38517, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int1_36049 = torch.constant.int 1
%int1_36050 = torch.constant.int 1
%38518 = torch.aten.add.Scalar %38406, %int1_36049, %int1_36050 : !torch.vtensor<[4,?],si64>, !torch.int, !torch.int -> !torch.vtensor<[4,?],si64>
torch.bind_symbolic_shape %38518, [%2336], affine_map<()[s0] -> (4, s0)> : !torch.vtensor<[4,?],si64>
%int4_36051 = torch.constant.int 4
%38519 = torch.aten.mul.int %int4_36051, %3095 : !torch.int, !torch.int -> !torch.int
%38520 = torch.prim.ListConstruct %38519 : (!torch.int) -> !torch.list<int>
%38521 = torch.aten.view %38511, %38520 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38521, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36052 = torch.constant.int 4
%38522 = torch.aten.mul.int %int4_36052, %3095 : !torch.int, !torch.int -> !torch.int
%38523 = torch.prim.ListConstruct %38522 : (!torch.int) -> !torch.list<int>
%38524 = torch.aten.view %38512, %38523 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38524, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36053 = torch.constant.int 4
%38525 = torch.aten.mul.int %int4_36053, %3095 : !torch.int, !torch.int -> !torch.int
%38526 = torch.prim.ListConstruct %38525 : (!torch.int) -> !torch.list<int>
%38527 = torch.aten.view %38513, %38526 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38527, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36054 = torch.constant.int 4
%38528 = torch.aten.mul.int %int4_36054, %3095 : !torch.int, !torch.int -> !torch.int
%38529 = torch.prim.ListConstruct %38528 : (!torch.int) -> !torch.list<int>
%38530 = torch.aten.view %38514, %38529 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38530, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36055 = torch.constant.int 4
%38531 = torch.aten.mul.int %int4_36055, %3095 : !torch.int, !torch.int -> !torch.int
%38532 = torch.prim.ListConstruct %38531 : (!torch.int) -> !torch.list<int>
%38533 = torch.aten.view %38515, %38532 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38533, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36056 = torch.constant.int 4
%38534 = torch.aten.mul.int %int4_36056, %3095 : !torch.int, !torch.int -> !torch.int
%38535 = torch.prim.ListConstruct %38534 : (!torch.int) -> !torch.list<int>
%38536 = torch.aten.view %38516, %38535 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38536, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36057 = torch.constant.int 4
%38537 = torch.aten.mul.int %int4_36057, %3095 : !torch.int, !torch.int -> !torch.int
%38538 = torch.prim.ListConstruct %38537 : (!torch.int) -> !torch.list<int>
%38539 = torch.aten.view %38517, %38538 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38539, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%int4_36058 = torch.constant.int 4
%38540 = torch.aten.mul.int %int4_36058, %3095 : !torch.int, !torch.int -> !torch.int
%38541 = torch.prim.ListConstruct %38540 : (!torch.int) -> !torch.list<int>
%38542 = torch.aten.view %38518, %38541 : !torch.vtensor<[4,?],si64>, !torch.list<int> -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38542, [%2336], affine_map<()[s0] -> (s0 * 4)> : !torch.vtensor<[?],si64>
%38543 = torch.prim.ListConstruct %38449, %38521 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36059 = torch.constant.int 0
%38544 = torch.aten.cat %38543, %int0_36059 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38544, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38545 = torch.prim.ListConstruct %38452, %38524 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36060 = torch.constant.int 0
%38546 = torch.aten.cat %38545, %int0_36060 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38546, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38547 = torch.prim.ListConstruct %38455, %38527 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36061 = torch.constant.int 0
%38548 = torch.aten.cat %38547, %int0_36061 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38548, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38549 = torch.prim.ListConstruct %38458, %38530 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36062 = torch.constant.int 0
%38550 = torch.aten.cat %38549, %int0_36062 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38550, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38551 = torch.prim.ListConstruct %38461, %38533 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36063 = torch.constant.int 0
%38552 = torch.aten.cat %38551, %int0_36063 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38552, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38553 = torch.prim.ListConstruct %38464, %38536 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36064 = torch.constant.int 0
%38554 = torch.aten.cat %38553, %int0_36064 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38554, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38555 = torch.prim.ListConstruct %38467, %38539 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36065 = torch.constant.int 0
%38556 = torch.aten.cat %38555, %int0_36065 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38556, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38557 = torch.prim.ListConstruct %38470, %38542 : (!torch.vtensor<[?],si64>, !torch.vtensor<[?],si64>) -> !torch.list<vtensor>
%int0_36066 = torch.constant.int 0
%38558 = torch.aten.cat %38557, %int0_36066 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?],si64>
torch.bind_symbolic_shape %38558, [%2336], affine_map<()[s0] -> (s0 * 8)> : !torch.vtensor<[?],si64>
%38559 = torch.prim.ListConstruct %38425, %38489 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36067 = torch.constant.int 0
%38560 = torch.aten.cat %38559, %int0_36067 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38560, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38561 = torch.prim.ListConstruct %38428, %38492 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36068 = torch.constant.int 0
%38562 = torch.aten.cat %38561, %int0_36068 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38562, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38563 = torch.prim.ListConstruct %38431, %38495 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36069 = torch.constant.int 0
%38564 = torch.aten.cat %38563, %int0_36069 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38564, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38565 = torch.prim.ListConstruct %38434, %38498 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36070 = torch.constant.int 0
%38566 = torch.aten.cat %38565, %int0_36070 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38566, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38567 = torch.prim.ListConstruct %38437, %38501 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36071 = torch.constant.int 0
%38568 = torch.aten.cat %38567, %int0_36071 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38568, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38569 = torch.prim.ListConstruct %38440, %38504 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36072 = torch.constant.int 0
%38570 = torch.aten.cat %38569, %int0_36072 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38570, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38571 = torch.prim.ListConstruct %38443, %38507 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36073 = torch.constant.int 0
%38572 = torch.aten.cat %38571, %int0_36073 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38572, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38573 = torch.prim.ListConstruct %38446, %38510 : (!torch.vtensor<[?,16,1,128],f16>, !torch.vtensor<[?,16,1,128],f16>) -> !torch.list<vtensor>
%int0_36074 = torch.constant.int 0
%38574 = torch.aten.cat %38573, %int0_36074 : !torch.list<vtensor>, !torch.int -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38574, [%2336], affine_map<()[s0] -> (s0 * 8, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36075 = torch.constant.int 32
%int2_36076 = torch.constant.int 2
%int16_36077 = torch.constant.int 16
%int1_36078 = torch.constant.int 1
%int128_36079 = torch.constant.int 128
%38575 = torch.prim.ListConstruct %3023, %int32_36075, %int2_36076, %int16_36077, %int1_36078, %int128_36079 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38576 = torch.aten.view %36725, %38575 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38576, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36080 = torch.constant.int 32
%38577 = torch.aten.mul.int %3023, %int32_36080 : !torch.int, !torch.int -> !torch.int
%int2_36081 = torch.constant.int 2
%38578 = torch.aten.mul.int %38577, %int2_36081 : !torch.int, !torch.int -> !torch.int
%int16_36082 = torch.constant.int 16
%int1_36083 = torch.constant.int 1
%int128_36084 = torch.constant.int 128
%38579 = torch.prim.ListConstruct %38578, %int16_36082, %int1_36083, %int128_36084 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38580 = torch.aten.view %38576, %38579 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38580, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38581 = torch.prim.ListConstruct %38544 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36085 = torch.constant.bool false
%38582 = torch.aten.index_put %38580, %38581, %38560, %false_36085 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38582, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36086 = torch.constant.int 32
%int2_36087 = torch.constant.int 2
%int16_36088 = torch.constant.int 16
%int1_36089 = torch.constant.int 1
%int128_36090 = torch.constant.int 128
%38583 = torch.prim.ListConstruct %3023, %int32_36086, %int2_36087, %int16_36088, %int1_36089, %int128_36090 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38584 = torch.aten.view %38582, %38583 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38584, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36091 = torch.constant.int 131072
%38585 = torch.prim.ListConstruct %3023, %int131072_36091 : (!torch.int, !torch.int) -> !torch.list<int>
%38586 = torch.aten.view %38584, %38585 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38586, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36092 = torch.constant.int 32
%int2_36093 = torch.constant.int 2
%int16_36094 = torch.constant.int 16
%int1_36095 = torch.constant.int 1
%int128_36096 = torch.constant.int 128
%38587 = torch.prim.ListConstruct %3026, %int32_36092, %int2_36093, %int16_36094, %int1_36095, %int128_36096 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38588 = torch.aten.view %36737, %38587 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38588, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36097 = torch.constant.int 32
%38589 = torch.aten.mul.int %3026, %int32_36097 : !torch.int, !torch.int -> !torch.int
%int2_36098 = torch.constant.int 2
%38590 = torch.aten.mul.int %38589, %int2_36098 : !torch.int, !torch.int -> !torch.int
%int16_36099 = torch.constant.int 16
%int1_36100 = torch.constant.int 1
%int128_36101 = torch.constant.int 128
%38591 = torch.prim.ListConstruct %38590, %int16_36099, %int1_36100, %int128_36101 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38592 = torch.aten.view %38588, %38591 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38592, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38593 = torch.prim.ListConstruct %38546 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36102 = torch.constant.bool false
%38594 = torch.aten.index_put %38592, %38593, %38562, %false_36102 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38594, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36103 = torch.constant.int 32
%int2_36104 = torch.constant.int 2
%int16_36105 = torch.constant.int 16
%int1_36106 = torch.constant.int 1
%int128_36107 = torch.constant.int 128
%38595 = torch.prim.ListConstruct %3026, %int32_36103, %int2_36104, %int16_36105, %int1_36106, %int128_36107 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38596 = torch.aten.view %38594, %38595 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38596, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36108 = torch.constant.int 131072
%38597 = torch.prim.ListConstruct %3026, %int131072_36108 : (!torch.int, !torch.int) -> !torch.list<int>
%38598 = torch.aten.view %38596, %38597 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38598, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36109 = torch.constant.int 32
%int2_36110 = torch.constant.int 2
%int16_36111 = torch.constant.int 16
%int1_36112 = torch.constant.int 1
%int128_36113 = torch.constant.int 128
%38599 = torch.prim.ListConstruct %3029, %int32_36109, %int2_36110, %int16_36111, %int1_36112, %int128_36113 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38600 = torch.aten.view %36749, %38599 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38600, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36114 = torch.constant.int 32
%38601 = torch.aten.mul.int %3029, %int32_36114 : !torch.int, !torch.int -> !torch.int
%int2_36115 = torch.constant.int 2
%38602 = torch.aten.mul.int %38601, %int2_36115 : !torch.int, !torch.int -> !torch.int
%int16_36116 = torch.constant.int 16
%int1_36117 = torch.constant.int 1
%int128_36118 = torch.constant.int 128
%38603 = torch.prim.ListConstruct %38602, %int16_36116, %int1_36117, %int128_36118 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38604 = torch.aten.view %38600, %38603 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38604, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38605 = torch.prim.ListConstruct %38548 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36119 = torch.constant.bool false
%38606 = torch.aten.index_put %38604, %38605, %38564, %false_36119 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38606, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36120 = torch.constant.int 32
%int2_36121 = torch.constant.int 2
%int16_36122 = torch.constant.int 16
%int1_36123 = torch.constant.int 1
%int128_36124 = torch.constant.int 128
%38607 = torch.prim.ListConstruct %3029, %int32_36120, %int2_36121, %int16_36122, %int1_36123, %int128_36124 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38608 = torch.aten.view %38606, %38607 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38608, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36125 = torch.constant.int 131072
%38609 = torch.prim.ListConstruct %3029, %int131072_36125 : (!torch.int, !torch.int) -> !torch.list<int>
%38610 = torch.aten.view %38608, %38609 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38610, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36126 = torch.constant.int 32
%int2_36127 = torch.constant.int 2
%int16_36128 = torch.constant.int 16
%int1_36129 = torch.constant.int 1
%int128_36130 = torch.constant.int 128
%38611 = torch.prim.ListConstruct %3032, %int32_36126, %int2_36127, %int16_36128, %int1_36129, %int128_36130 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38612 = torch.aten.view %36761, %38611 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38612, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36131 = torch.constant.int 32
%38613 = torch.aten.mul.int %3032, %int32_36131 : !torch.int, !torch.int -> !torch.int
%int2_36132 = torch.constant.int 2
%38614 = torch.aten.mul.int %38613, %int2_36132 : !torch.int, !torch.int -> !torch.int
%int16_36133 = torch.constant.int 16
%int1_36134 = torch.constant.int 1
%int128_36135 = torch.constant.int 128
%38615 = torch.prim.ListConstruct %38614, %int16_36133, %int1_36134, %int128_36135 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38616 = torch.aten.view %38612, %38615 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38616, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38617 = torch.prim.ListConstruct %38550 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36136 = torch.constant.bool false
%38618 = torch.aten.index_put %38616, %38617, %38566, %false_36136 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38618, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36137 = torch.constant.int 32
%int2_36138 = torch.constant.int 2
%int16_36139 = torch.constant.int 16
%int1_36140 = torch.constant.int 1
%int128_36141 = torch.constant.int 128
%38619 = torch.prim.ListConstruct %3032, %int32_36137, %int2_36138, %int16_36139, %int1_36140, %int128_36141 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38620 = torch.aten.view %38618, %38619 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38620, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36142 = torch.constant.int 131072
%38621 = torch.prim.ListConstruct %3032, %int131072_36142 : (!torch.int, !torch.int) -> !torch.list<int>
%38622 = torch.aten.view %38620, %38621 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38622, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36143 = torch.constant.int 32
%int2_36144 = torch.constant.int 2
%int16_36145 = torch.constant.int 16
%int1_36146 = torch.constant.int 1
%int128_36147 = torch.constant.int 128
%38623 = torch.prim.ListConstruct %3035, %int32_36143, %int2_36144, %int16_36145, %int1_36146, %int128_36147 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38624 = torch.aten.view %36773, %38623 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38624, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36148 = torch.constant.int 32
%38625 = torch.aten.mul.int %3035, %int32_36148 : !torch.int, !torch.int -> !torch.int
%int2_36149 = torch.constant.int 2
%38626 = torch.aten.mul.int %38625, %int2_36149 : !torch.int, !torch.int -> !torch.int
%int16_36150 = torch.constant.int 16
%int1_36151 = torch.constant.int 1
%int128_36152 = torch.constant.int 128
%38627 = torch.prim.ListConstruct %38626, %int16_36150, %int1_36151, %int128_36152 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38628 = torch.aten.view %38624, %38627 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38628, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38629 = torch.prim.ListConstruct %38552 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36153 = torch.constant.bool false
%38630 = torch.aten.index_put %38628, %38629, %38568, %false_36153 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38630, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36154 = torch.constant.int 32
%int2_36155 = torch.constant.int 2
%int16_36156 = torch.constant.int 16
%int1_36157 = torch.constant.int 1
%int128_36158 = torch.constant.int 128
%38631 = torch.prim.ListConstruct %3035, %int32_36154, %int2_36155, %int16_36156, %int1_36157, %int128_36158 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38632 = torch.aten.view %38630, %38631 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38632, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36159 = torch.constant.int 131072
%38633 = torch.prim.ListConstruct %3035, %int131072_36159 : (!torch.int, !torch.int) -> !torch.list<int>
%38634 = torch.aten.view %38632, %38633 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38634, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36160 = torch.constant.int 32
%int2_36161 = torch.constant.int 2
%int16_36162 = torch.constant.int 16
%int1_36163 = torch.constant.int 1
%int128_36164 = torch.constant.int 128
%38635 = torch.prim.ListConstruct %3038, %int32_36160, %int2_36161, %int16_36162, %int1_36163, %int128_36164 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38636 = torch.aten.view %36785, %38635 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38636, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36165 = torch.constant.int 32
%38637 = torch.aten.mul.int %3038, %int32_36165 : !torch.int, !torch.int -> !torch.int
%int2_36166 = torch.constant.int 2
%38638 = torch.aten.mul.int %38637, %int2_36166 : !torch.int, !torch.int -> !torch.int
%int16_36167 = torch.constant.int 16
%int1_36168 = torch.constant.int 1
%int128_36169 = torch.constant.int 128
%38639 = torch.prim.ListConstruct %38638, %int16_36167, %int1_36168, %int128_36169 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38640 = torch.aten.view %38636, %38639 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38640, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38641 = torch.prim.ListConstruct %38554 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36170 = torch.constant.bool false
%38642 = torch.aten.index_put %38640, %38641, %38570, %false_36170 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38642, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36171 = torch.constant.int 32
%int2_36172 = torch.constant.int 2
%int16_36173 = torch.constant.int 16
%int1_36174 = torch.constant.int 1
%int128_36175 = torch.constant.int 128
%38643 = torch.prim.ListConstruct %3038, %int32_36171, %int2_36172, %int16_36173, %int1_36174, %int128_36175 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38644 = torch.aten.view %38642, %38643 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38644, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36176 = torch.constant.int 131072
%38645 = torch.prim.ListConstruct %3038, %int131072_36176 : (!torch.int, !torch.int) -> !torch.list<int>
%38646 = torch.aten.view %38644, %38645 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38646, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36177 = torch.constant.int 32
%int2_36178 = torch.constant.int 2
%int16_36179 = torch.constant.int 16
%int1_36180 = torch.constant.int 1
%int128_36181 = torch.constant.int 128
%38647 = torch.prim.ListConstruct %3041, %int32_36177, %int2_36178, %int16_36179, %int1_36180, %int128_36181 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38648 = torch.aten.view %36797, %38647 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38648, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36182 = torch.constant.int 32
%38649 = torch.aten.mul.int %3041, %int32_36182 : !torch.int, !torch.int -> !torch.int
%int2_36183 = torch.constant.int 2
%38650 = torch.aten.mul.int %38649, %int2_36183 : !torch.int, !torch.int -> !torch.int
%int16_36184 = torch.constant.int 16
%int1_36185 = torch.constant.int 1
%int128_36186 = torch.constant.int 128
%38651 = torch.prim.ListConstruct %38650, %int16_36184, %int1_36185, %int128_36186 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38652 = torch.aten.view %38648, %38651 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38652, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38653 = torch.prim.ListConstruct %38556 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36187 = torch.constant.bool false
%38654 = torch.aten.index_put %38652, %38653, %38572, %false_36187 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38654, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36188 = torch.constant.int 32
%int2_36189 = torch.constant.int 2
%int16_36190 = torch.constant.int 16
%int1_36191 = torch.constant.int 1
%int128_36192 = torch.constant.int 128
%38655 = torch.prim.ListConstruct %3041, %int32_36188, %int2_36189, %int16_36190, %int1_36191, %int128_36192 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38656 = torch.aten.view %38654, %38655 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38656, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36193 = torch.constant.int 131072
%38657 = torch.prim.ListConstruct %3041, %int131072_36193 : (!torch.int, !torch.int) -> !torch.list<int>
%38658 = torch.aten.view %38656, %38657 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38658, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int32_36194 = torch.constant.int 32
%int2_36195 = torch.constant.int 2
%int16_36196 = torch.constant.int 16
%int1_36197 = torch.constant.int 1
%int128_36198 = torch.constant.int 128
%38659 = torch.prim.ListConstruct %3044, %int32_36194, %int2_36195, %int16_36196, %int1_36197, %int128_36198 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38660 = torch.aten.view %36809, %38659 : !torch.vtensor<[?,131072],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38660, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int32_36199 = torch.constant.int 32
%38661 = torch.aten.mul.int %3044, %int32_36199 : !torch.int, !torch.int -> !torch.int
%int2_36200 = torch.constant.int 2
%38662 = torch.aten.mul.int %38661, %int2_36200 : !torch.int, !torch.int -> !torch.int
%int16_36201 = torch.constant.int 16
%int1_36202 = torch.constant.int 1
%int128_36203 = torch.constant.int 128
%38663 = torch.prim.ListConstruct %38662, %int16_36201, %int1_36202, %int128_36203 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38664 = torch.aten.view %38660, %38663 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38664, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%38665 = torch.prim.ListConstruct %38558 : (!torch.vtensor<[?],si64>) -> !torch.list<optional<vtensor>>
%false_36204 = torch.constant.bool false
%38666 = torch.aten.index_put %38664, %38665, %38574, %false_36204 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<optional<vtensor>>, !torch.vtensor<[?,16,1,128],f16>, !torch.bool -> !torch.vtensor<[?,16,1,128],f16>
torch.bind_symbolic_shape %38666, [%2337], affine_map<()[s0] -> (s0 * 64, 16, 1, 128)> : !torch.vtensor<[?,16,1,128],f16>
%int32_36205 = torch.constant.int 32
%int2_36206 = torch.constant.int 2
%int16_36207 = torch.constant.int 16
%int1_36208 = torch.constant.int 1
%int128_36209 = torch.constant.int 128
%38667 = torch.prim.ListConstruct %3044, %int32_36205, %int2_36206, %int16_36207, %int1_36208, %int128_36209 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38668 = torch.aten.view %38666, %38667 : !torch.vtensor<[?,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,32,2,16,1,128],f16>
torch.bind_symbolic_shape %38668, [%2337], affine_map<()[s0] -> (s0, 32, 2, 16, 1, 128)> : !torch.vtensor<[?,32,2,16,1,128],f16>
%int131072_36210 = torch.constant.int 131072
%38669 = torch.prim.ListConstruct %3044, %int131072_36210 : (!torch.int, !torch.int) -> !torch.list<int>
%38670 = torch.aten.view %38668, %38669 : !torch.vtensor<[?,32,2,16,1,128],f16>, !torch.list<int> -> !torch.vtensor<[?,131072],f16>
torch.bind_symbolic_shape %38670, [%2337], affine_map<()[s0] -> (s0, 131072)> : !torch.vtensor<[?,131072],f16>
%int-2_36211 = torch.constant.int -2
%38671 = torch.aten.unsqueeze %38285, %int-2_36211 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36212 = torch.constant.int -2
%38672 = torch.aten.unsqueeze %38300, %int-2_36212 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36213 = torch.constant.int -2
%38673 = torch.aten.unsqueeze %38315, %int-2_36213 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36214 = torch.constant.int -2
%38674 = torch.aten.unsqueeze %38330, %int-2_36214 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36215 = torch.constant.int -2
%38675 = torch.aten.unsqueeze %38345, %int-2_36215 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36216 = torch.constant.int -2
%38676 = torch.aten.unsqueeze %38360, %int-2_36216 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36217 = torch.constant.int -2
%38677 = torch.aten.unsqueeze %38375, %int-2_36217 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36218 = torch.constant.int -2
%38678 = torch.aten.unsqueeze %38390, %int-2_36218 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int4_36219 = torch.constant.int 4
%int1_36220 = torch.constant.int 1
%int4_36221 = torch.constant.int 4
%int128_36222 = torch.constant.int 128
%38679 = torch.prim.ListConstruct %int4_36219, %38271, %int1_36220, %int4_36221, %int128_36222 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36223 = torch.constant.bool false
%38680 = torch.aten.expand %38671, %38679, %false_36223 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36224 = torch.constant.int 4
%int1_36225 = torch.constant.int 1
%int4_36226 = torch.constant.int 4
%int128_36227 = torch.constant.int 128
%38681 = torch.prim.ListConstruct %int4_36224, %38271, %int1_36225, %int4_36226, %int128_36227 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36228 = torch.constant.bool false
%38682 = torch.aten.expand %38672, %38681, %false_36228 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36229 = torch.constant.int 4
%int1_36230 = torch.constant.int 1
%int4_36231 = torch.constant.int 4
%int128_36232 = torch.constant.int 128
%38683 = torch.prim.ListConstruct %int4_36229, %38271, %int1_36230, %int4_36231, %int128_36232 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36233 = torch.constant.bool false
%38684 = torch.aten.expand %38673, %38683, %false_36233 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36234 = torch.constant.int 4
%int1_36235 = torch.constant.int 1
%int4_36236 = torch.constant.int 4
%int128_36237 = torch.constant.int 128
%38685 = torch.prim.ListConstruct %int4_36234, %38271, %int1_36235, %int4_36236, %int128_36237 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36238 = torch.constant.bool false
%38686 = torch.aten.expand %38674, %38685, %false_36238 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36239 = torch.constant.int 4
%int1_36240 = torch.constant.int 1
%int4_36241 = torch.constant.int 4
%int128_36242 = torch.constant.int 128
%38687 = torch.prim.ListConstruct %int4_36239, %38271, %int1_36240, %int4_36241, %int128_36242 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36243 = torch.constant.bool false
%38688 = torch.aten.expand %38675, %38687, %false_36243 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36244 = torch.constant.int 4
%int1_36245 = torch.constant.int 1
%int4_36246 = torch.constant.int 4
%int128_36247 = torch.constant.int 128
%38689 = torch.prim.ListConstruct %int4_36244, %38271, %int1_36245, %int4_36246, %int128_36247 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36248 = torch.constant.bool false
%38690 = torch.aten.expand %38676, %38689, %false_36248 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36249 = torch.constant.int 4
%int1_36250 = torch.constant.int 1
%int4_36251 = torch.constant.int 4
%int128_36252 = torch.constant.int 128
%38691 = torch.prim.ListConstruct %int4_36249, %38271, %int1_36250, %int4_36251, %int128_36252 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36253 = torch.constant.bool false
%38692 = torch.aten.expand %38677, %38691, %false_36253 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36254 = torch.constant.int 4
%int1_36255 = torch.constant.int 1
%int4_36256 = torch.constant.int 4
%int128_36257 = torch.constant.int 128
%38693 = torch.prim.ListConstruct %int4_36254, %38271, %int1_36255, %int4_36256, %int128_36257 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36258 = torch.constant.bool false
%38694 = torch.aten.expand %38678, %38693, %false_36258 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36259 = torch.constant.int 4
%int4_36260 = torch.constant.int 4
%int128_36261 = torch.constant.int 128
%38695 = torch.prim.ListConstruct %int4_36259, %38271, %int4_36260, %int128_36261 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38696 = torch.aten.view %38680, %38695 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38696, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36262 = torch.constant.int 4
%int4_36263 = torch.constant.int 4
%int128_36264 = torch.constant.int 128
%38697 = torch.prim.ListConstruct %int4_36262, %38271, %int4_36263, %int128_36264 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38698 = torch.aten.view %38682, %38697 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38698, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36265 = torch.constant.int 4
%int4_36266 = torch.constant.int 4
%int128_36267 = torch.constant.int 128
%38699 = torch.prim.ListConstruct %int4_36265, %38271, %int4_36266, %int128_36267 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38700 = torch.aten.view %38684, %38699 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38700, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36268 = torch.constant.int 4
%int4_36269 = torch.constant.int 4
%int128_36270 = torch.constant.int 128
%38701 = torch.prim.ListConstruct %int4_36268, %38271, %int4_36269, %int128_36270 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38702 = torch.aten.view %38686, %38701 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38702, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36271 = torch.constant.int 4
%int4_36272 = torch.constant.int 4
%int128_36273 = torch.constant.int 128
%38703 = torch.prim.ListConstruct %int4_36271, %38271, %int4_36272, %int128_36273 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38704 = torch.aten.view %38688, %38703 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38704, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36274 = torch.constant.int 4
%int4_36275 = torch.constant.int 4
%int128_36276 = torch.constant.int 128
%38705 = torch.prim.ListConstruct %int4_36274, %38271, %int4_36275, %int128_36276 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38706 = torch.aten.view %38690, %38705 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38706, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36277 = torch.constant.int 4
%int4_36278 = torch.constant.int 4
%int128_36279 = torch.constant.int 128
%38707 = torch.prim.ListConstruct %int4_36277, %38271, %int4_36278, %int128_36279 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38708 = torch.aten.view %38692, %38707 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38708, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36280 = torch.constant.int 4
%int4_36281 = torch.constant.int 4
%int128_36282 = torch.constant.int 128
%38709 = torch.prim.ListConstruct %int4_36280, %38271, %int4_36281, %int128_36282 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38710 = torch.aten.view %38694, %38709 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38710, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int-2_36283 = torch.constant.int -2
%38711 = torch.aten.unsqueeze %38060, %int-2_36283 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38711, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36284 = torch.constant.int -2
%38712 = torch.aten.unsqueeze %38062, %int-2_36284 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38712, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36285 = torch.constant.int -2
%38713 = torch.aten.unsqueeze %38064, %int-2_36285 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38713, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36286 = torch.constant.int -2
%38714 = torch.aten.unsqueeze %38066, %int-2_36286 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38714, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36287 = torch.constant.int -2
%38715 = torch.aten.unsqueeze %38068, %int-2_36287 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38715, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36288 = torch.constant.int -2
%38716 = torch.aten.unsqueeze %38070, %int-2_36288 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38716, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36289 = torch.constant.int -2
%38717 = torch.aten.unsqueeze %38072, %int-2_36289 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int-2_36290 = torch.constant.int -2
%38718 = torch.aten.unsqueeze %38074, %int-2_36290 : !torch.vtensor<[4,?,1,128],f16>, !torch.int -> !torch.vtensor<[4,?,1,1,128],f16>
torch.bind_symbolic_shape %38718, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 1, 128)> : !torch.vtensor<[4,?,1,1,128],f16>
%int1_36291 = torch.constant.int 1
%38719 = torch.aten.size.int %37984, %int1_36291 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int4_36292 = torch.constant.int 4
%int1_36293 = torch.constant.int 1
%int4_36294 = torch.constant.int 4
%int128_36295 = torch.constant.int 128
%38720 = torch.prim.ListConstruct %int4_36292, %38719, %int1_36293, %int4_36294, %int128_36295 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36296 = torch.constant.bool false
%38721 = torch.aten.expand %38711, %38720, %false_36296 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38721, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36297 = torch.constant.int 4
%int1_36298 = torch.constant.int 1
%int4_36299 = torch.constant.int 4
%int128_36300 = torch.constant.int 128
%38722 = torch.prim.ListConstruct %int4_36297, %38719, %int1_36298, %int4_36299, %int128_36300 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36301 = torch.constant.bool false
%38723 = torch.aten.expand %38712, %38722, %false_36301 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36302 = torch.constant.int 4
%int1_36303 = torch.constant.int 1
%int4_36304 = torch.constant.int 4
%int128_36305 = torch.constant.int 128
%38724 = torch.prim.ListConstruct %int4_36302, %38719, %int1_36303, %int4_36304, %int128_36305 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36306 = torch.constant.bool false
%38725 = torch.aten.expand %38713, %38724, %false_36306 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38725, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36307 = torch.constant.int 4
%int1_36308 = torch.constant.int 1
%int4_36309 = torch.constant.int 4
%int128_36310 = torch.constant.int 128
%38726 = torch.prim.ListConstruct %int4_36307, %38719, %int1_36308, %int4_36309, %int128_36310 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36311 = torch.constant.bool false
%38727 = torch.aten.expand %38714, %38726, %false_36311 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38727, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36312 = torch.constant.int 4
%int1_36313 = torch.constant.int 1
%int4_36314 = torch.constant.int 4
%int128_36315 = torch.constant.int 128
%38728 = torch.prim.ListConstruct %int4_36312, %38719, %int1_36313, %int4_36314, %int128_36315 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36316 = torch.constant.bool false
%38729 = torch.aten.expand %38715, %38728, %false_36316 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36317 = torch.constant.int 4
%int1_36318 = torch.constant.int 1
%int4_36319 = torch.constant.int 4
%int128_36320 = torch.constant.int 128
%38730 = torch.prim.ListConstruct %int4_36317, %38719, %int1_36318, %int4_36319, %int128_36320 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36321 = torch.constant.bool false
%38731 = torch.aten.expand %38716, %38730, %false_36321 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38731, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36322 = torch.constant.int 4
%int1_36323 = torch.constant.int 1
%int4_36324 = torch.constant.int 4
%int128_36325 = torch.constant.int 128
%38732 = torch.prim.ListConstruct %int4_36322, %38719, %int1_36323, %int4_36324, %int128_36325 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36326 = torch.constant.bool false
%38733 = torch.aten.expand %38717, %38732, %false_36326 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38733, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36327 = torch.constant.int 4
%int1_36328 = torch.constant.int 1
%int4_36329 = torch.constant.int 4
%int128_36330 = torch.constant.int 128
%38734 = torch.prim.ListConstruct %int4_36327, %38719, %int1_36328, %int4_36329, %int128_36330 : (!torch.int, !torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%false_36331 = torch.constant.bool false
%38735 = torch.aten.expand %38718, %38734, %false_36331 : !torch.vtensor<[4,?,1,1,128],f16>, !torch.list<int>, !torch.bool -> !torch.vtensor<[4,?,1,4,128],f16>
torch.bind_symbolic_shape %38735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 4, 128)> : !torch.vtensor<[4,?,1,4,128],f16>
%int4_36332 = torch.constant.int 4
%int4_36333 = torch.constant.int 4
%int128_36334 = torch.constant.int 128
%38736 = torch.prim.ListConstruct %int4_36332, %38719, %int4_36333, %int128_36334 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38737 = torch.aten.view %38721, %38736 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38737, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36335 = torch.constant.int 4
%int4_36336 = torch.constant.int 4
%int128_36337 = torch.constant.int 128
%38738 = torch.prim.ListConstruct %int4_36335, %38719, %int4_36336, %int128_36337 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38739 = torch.aten.view %38723, %38738 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38739, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36338 = torch.constant.int 4
%int4_36339 = torch.constant.int 4
%int128_36340 = torch.constant.int 128
%38740 = torch.prim.ListConstruct %int4_36338, %38719, %int4_36339, %int128_36340 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38741 = torch.aten.view %38725, %38740 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36341 = torch.constant.int 4
%int4_36342 = torch.constant.int 4
%int128_36343 = torch.constant.int 128
%38742 = torch.prim.ListConstruct %int4_36341, %38719, %int4_36342, %int128_36343 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38743 = torch.aten.view %38727, %38742 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38743, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36344 = torch.constant.int 4
%int4_36345 = torch.constant.int 4
%int128_36346 = torch.constant.int 128
%38744 = torch.prim.ListConstruct %int4_36344, %38719, %int4_36345, %int128_36346 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38745 = torch.aten.view %38729, %38744 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38745, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36347 = torch.constant.int 4
%int4_36348 = torch.constant.int 4
%int128_36349 = torch.constant.int 128
%38746 = torch.prim.ListConstruct %int4_36347, %38719, %int4_36348, %int128_36349 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38747 = torch.aten.view %38731, %38746 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36350 = torch.constant.int 4
%int4_36351 = torch.constant.int 4
%int128_36352 = torch.constant.int 128
%38748 = torch.prim.ListConstruct %int4_36350, %38719, %int4_36351, %int128_36352 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38749 = torch.aten.view %38733, %38748 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38749, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36353 = torch.constant.int 4
%int4_36354 = torch.constant.int 4
%int128_36355 = torch.constant.int 128
%38750 = torch.prim.ListConstruct %int4_36353, %38719, %int4_36354, %int128_36355 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38751 = torch.aten.view %38735, %38750 : !torch.vtensor<[4,?,1,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38751, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36356 = torch.constant.int 1
%int2_36357 = torch.constant.int 2
%38752 = torch.aten.transpose.int %38127, %int1_36356, %int2_36357 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38752, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36358 = torch.constant.int 1
%int2_36359 = torch.constant.int 2
%38753 = torch.aten.transpose.int %38142, %int1_36358, %int2_36359 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38753, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36360 = torch.constant.int 1
%int2_36361 = torch.constant.int 2
%38754 = torch.aten.transpose.int %38157, %int1_36360, %int2_36361 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38754, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36362 = torch.constant.int 1
%int2_36363 = torch.constant.int 2
%38755 = torch.aten.transpose.int %38172, %int1_36362, %int2_36363 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38755, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36364 = torch.constant.int 1
%int2_36365 = torch.constant.int 2
%38756 = torch.aten.transpose.int %38187, %int1_36364, %int2_36365 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38756, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36366 = torch.constant.int 1
%int2_36367 = torch.constant.int 2
%38757 = torch.aten.transpose.int %38202, %int1_36366, %int2_36367 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38757, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36368 = torch.constant.int 1
%int2_36369 = torch.constant.int 2
%38758 = torch.aten.transpose.int %38217, %int1_36368, %int2_36369 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38758, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36370 = torch.constant.int 1
%int2_36371 = torch.constant.int 2
%38759 = torch.aten.transpose.int %38232, %int1_36370, %int2_36371 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38759, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36372 = torch.constant.int 1
%int2_36373 = torch.constant.int 2
%38760 = torch.aten.transpose.int %38696, %int1_36372, %int2_36373 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38760, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36374 = torch.constant.int 1
%int2_36375 = torch.constant.int 2
%38761 = torch.aten.transpose.int %38698, %int1_36374, %int2_36375 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38761, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36376 = torch.constant.int 1
%int2_36377 = torch.constant.int 2
%38762 = torch.aten.transpose.int %38700, %int1_36376, %int2_36377 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38762, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36378 = torch.constant.int 1
%int2_36379 = torch.constant.int 2
%38763 = torch.aten.transpose.int %38702, %int1_36378, %int2_36379 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38763, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36380 = torch.constant.int 1
%int2_36381 = torch.constant.int 2
%38764 = torch.aten.transpose.int %38704, %int1_36380, %int2_36381 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38764, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36382 = torch.constant.int 1
%int2_36383 = torch.constant.int 2
%38765 = torch.aten.transpose.int %38706, %int1_36382, %int2_36383 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38765, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36384 = torch.constant.int 1
%int2_36385 = torch.constant.int 2
%38766 = torch.aten.transpose.int %38708, %int1_36384, %int2_36385 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38766, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36386 = torch.constant.int 1
%int2_36387 = torch.constant.int 2
%38767 = torch.aten.transpose.int %38710, %int1_36386, %int2_36387 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38767, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36388 = torch.constant.int 1
%int2_36389 = torch.constant.int 2
%38768 = torch.aten.transpose.int %38737, %int1_36388, %int2_36389 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38768, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36390 = torch.constant.int 1
%int2_36391 = torch.constant.int 2
%38769 = torch.aten.transpose.int %38739, %int1_36390, %int2_36391 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38769, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36392 = torch.constant.int 1
%int2_36393 = torch.constant.int 2
%38770 = torch.aten.transpose.int %38741, %int1_36392, %int2_36393 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38770, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36394 = torch.constant.int 1
%int2_36395 = torch.constant.int 2
%38771 = torch.aten.transpose.int %38743, %int1_36394, %int2_36395 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38771, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36396 = torch.constant.int 1
%int2_36397 = torch.constant.int 2
%38772 = torch.aten.transpose.int %38745, %int1_36396, %int2_36397 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38772, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36398 = torch.constant.int 1
%int2_36399 = torch.constant.int 2
%38773 = torch.aten.transpose.int %38747, %int1_36398, %int2_36399 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38773, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36400 = torch.constant.int 1
%int2_36401 = torch.constant.int 2
%38774 = torch.aten.transpose.int %38749, %int1_36400, %int2_36401 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38774, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36402 = torch.constant.int 1
%int2_36403 = torch.constant.int 2
%38775 = torch.aten.transpose.int %38751, %int1_36402, %int2_36403 : !torch.vtensor<[4,?,4,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,4,?,128],f16>
torch.bind_symbolic_shape %38775, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36404 = torch.constant.float 0.000000e+00
%true_36405 = torch.constant.bool true
%none_36406 = torch.constant.none
%none_36407 = torch.constant.none
%38776:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38752, %38760, %38768, %float0.000000e00_36404, %true_36405, %none_36406, %none_36407) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38776#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36408 = torch.constant.float 0.000000e+00
%true_36409 = torch.constant.bool true
%none_36410 = torch.constant.none
%none_36411 = torch.constant.none
%38777:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38753, %38761, %38769, %float0.000000e00_36408, %true_36409, %none_36410, %none_36411) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38777#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36412 = torch.constant.float 0.000000e+00
%true_36413 = torch.constant.bool true
%none_36414 = torch.constant.none
%none_36415 = torch.constant.none
%38778:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38754, %38762, %38770, %float0.000000e00_36412, %true_36413, %none_36414, %none_36415) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38778#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36416 = torch.constant.float 0.000000e+00
%true_36417 = torch.constant.bool true
%none_36418 = torch.constant.none
%none_36419 = torch.constant.none
%38779:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38755, %38763, %38771, %float0.000000e00_36416, %true_36417, %none_36418, %none_36419) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38779#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36420 = torch.constant.float 0.000000e+00
%true_36421 = torch.constant.bool true
%none_36422 = torch.constant.none
%none_36423 = torch.constant.none
%38780:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38756, %38764, %38772, %float0.000000e00_36420, %true_36421, %none_36422, %none_36423) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38780#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36424 = torch.constant.float 0.000000e+00
%true_36425 = torch.constant.bool true
%none_36426 = torch.constant.none
%none_36427 = torch.constant.none
%38781:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38757, %38765, %38773, %float0.000000e00_36424, %true_36425, %none_36426, %none_36427) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38781#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36428 = torch.constant.float 0.000000e+00
%true_36429 = torch.constant.bool true
%none_36430 = torch.constant.none
%none_36431 = torch.constant.none
%38782:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38758, %38766, %38774, %float0.000000e00_36428, %true_36429, %none_36430, %none_36431) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38782#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%float0.000000e00_36432 = torch.constant.float 0.000000e+00
%true_36433 = torch.constant.bool true
%none_36434 = torch.constant.none
%none_36435 = torch.constant.none
%38783:2 = torch.operator "torch.aten._scaled_dot_product_flash_attention_for_cpu"(%38759, %38767, %38775, %float0.000000e00_36432, %true_36433, %none_36434, %none_36435) : (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?,128],f16>, !torch.float, !torch.bool, !torch.none, !torch.none) -> (!torch.vtensor<[4,4,?,128],f16>, !torch.vtensor<[4,4,?],f32>)
torch.bind_symbolic_shape %38783#0, [%2336], affine_map<()[s0] -> (4, 4, s0 * 16, 128)> : !torch.vtensor<[4,4,?,128],f16>
%int1_36436 = torch.constant.int 1
%int2_36437 = torch.constant.int 2
%38784 = torch.aten.transpose.int %38776#0, %int1_36436, %int2_36437 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38784, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36438 = torch.constant.int 1
%int2_36439 = torch.constant.int 2
%38785 = torch.aten.transpose.int %38777#0, %int1_36438, %int2_36439 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38785, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36440 = torch.constant.int 1
%int2_36441 = torch.constant.int 2
%38786 = torch.aten.transpose.int %38778#0, %int1_36440, %int2_36441 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38786, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36442 = torch.constant.int 1
%int2_36443 = torch.constant.int 2
%38787 = torch.aten.transpose.int %38779#0, %int1_36442, %int2_36443 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36444 = torch.constant.int 1
%int2_36445 = torch.constant.int 2
%38788 = torch.aten.transpose.int %38780#0, %int1_36444, %int2_36445 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38788, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36446 = torch.constant.int 1
%int2_36447 = torch.constant.int 2
%38789 = torch.aten.transpose.int %38781#0, %int1_36446, %int2_36447 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38789, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36448 = torch.constant.int 1
%int2_36449 = torch.constant.int 2
%38790 = torch.aten.transpose.int %38782#0, %int1_36448, %int2_36449 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38790, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_36450 = torch.constant.int 1
%int2_36451 = torch.constant.int 2
%38791 = torch.aten.transpose.int %38783#0, %int1_36450, %int2_36451 : !torch.vtensor<[4,4,?,128],f16>, !torch.int, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %38791, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_36452 = torch.constant.int 4
%int512_36453 = torch.constant.int 512
%38792 = torch.prim.ListConstruct %int4_36452, %38113, %int512_36453 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38793 = torch.aten.view %38784, %38792 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36454 = torch.constant.int 4
%int512_36455 = torch.constant.int 512
%38794 = torch.prim.ListConstruct %int4_36454, %38128, %int512_36455 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38795 = torch.aten.view %38785, %38794 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38795, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36456 = torch.constant.int 4
%int512_36457 = torch.constant.int 512
%38796 = torch.prim.ListConstruct %int4_36456, %38143, %int512_36457 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38797 = torch.aten.view %38786, %38796 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38797, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36458 = torch.constant.int 4
%int512_36459 = torch.constant.int 512
%38798 = torch.prim.ListConstruct %int4_36458, %38158, %int512_36459 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38799 = torch.aten.view %38787, %38798 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36460 = torch.constant.int 4
%int512_36461 = torch.constant.int 512
%38800 = torch.prim.ListConstruct %int4_36460, %38173, %int512_36461 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38801 = torch.aten.view %38788, %38800 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38801, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36462 = torch.constant.int 4
%int512_36463 = torch.constant.int 512
%38802 = torch.prim.ListConstruct %int4_36462, %38188, %int512_36463 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38803 = torch.aten.view %38789, %38802 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38803, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36464 = torch.constant.int 4
%int512_36465 = torch.constant.int 512
%38804 = torch.prim.ListConstruct %int4_36464, %38203, %int512_36465 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38805 = torch.aten.view %38790, %38804 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_36466 = torch.constant.int 4
%int512_36467 = torch.constant.int 512
%38806 = torch.prim.ListConstruct %int4_36466, %38218, %int512_36467 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38807 = torch.aten.view %38791, %38806 : !torch.vtensor<[4,?,4,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %38807, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_36468 = torch.constant.int 1
%int0_36469 = torch.constant.int 0
%38808 = torch.prim.ListConstruct %int1_36468, %int0_36469 : (!torch.int, !torch.int) -> !torch.list<int>
%38809 = torch.aten.permute %1408, %38808 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36470 = torch.constant.int 1
%int0_36471 = torch.constant.int 0
%38810 = torch.prim.ListConstruct %int1_36470, %int0_36471 : (!torch.int, !torch.int) -> !torch.list<int>
%38811 = torch.aten.permute %1409, %38810 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36472 = torch.constant.int 1
%int0_36473 = torch.constant.int 0
%38812 = torch.prim.ListConstruct %int1_36472, %int0_36473 : (!torch.int, !torch.int) -> !torch.list<int>
%38813 = torch.aten.permute %1410, %38812 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36474 = torch.constant.int 1
%int0_36475 = torch.constant.int 0
%38814 = torch.prim.ListConstruct %int1_36474, %int0_36475 : (!torch.int, !torch.int) -> !torch.list<int>
%38815 = torch.aten.permute %1411, %38814 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36476 = torch.constant.int 1
%int0_36477 = torch.constant.int 0
%38816 = torch.prim.ListConstruct %int1_36476, %int0_36477 : (!torch.int, !torch.int) -> !torch.list<int>
%38817 = torch.aten.permute %1412, %38816 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36478 = torch.constant.int 1
%int0_36479 = torch.constant.int 0
%38818 = torch.prim.ListConstruct %int1_36478, %int0_36479 : (!torch.int, !torch.int) -> !torch.list<int>
%38819 = torch.aten.permute %1413, %38818 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36480 = torch.constant.int 1
%int0_36481 = torch.constant.int 0
%38820 = torch.prim.ListConstruct %int1_36480, %int0_36481 : (!torch.int, !torch.int) -> !torch.list<int>
%38821 = torch.aten.permute %1414, %38820 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int1_36482 = torch.constant.int 1
%int0_36483 = torch.constant.int 0
%38822 = torch.prim.ListConstruct %int1_36482, %int0_36483 : (!torch.int, !torch.int) -> !torch.list<int>
%38823 = torch.aten.permute %1415, %38822 : !torch.vtensor<[4096,512],f16>, !torch.list<int> -> !torch.vtensor<[512,4096],f16>
%int4_36484 = torch.constant.int 4
%38824 = torch.aten.mul.int %int4_36484, %38113 : !torch.int, !torch.int -> !torch.int
%int512_36485 = torch.constant.int 512
%38825 = torch.prim.ListConstruct %38824, %int512_36485 : (!torch.int, !torch.int) -> !torch.list<int>
%38826 = torch.aten.view %38793, %38825 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38826, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38827 = torch.aten.mm %38826, %38809 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38827, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36486 = torch.constant.int 4
%int4096_36487 = torch.constant.int 4096
%38828 = torch.prim.ListConstruct %int4_36486, %38113, %int4096_36487 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38829 = torch.aten.view %38827, %38828 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38829, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36488 = torch.constant.int 4
%38830 = torch.aten.mul.int %int4_36488, %38128 : !torch.int, !torch.int -> !torch.int
%int512_36489 = torch.constant.int 512
%38831 = torch.prim.ListConstruct %38830, %int512_36489 : (!torch.int, !torch.int) -> !torch.list<int>
%38832 = torch.aten.view %38795, %38831 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38832, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38833 = torch.aten.mm %38832, %38811 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38833, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36490 = torch.constant.int 4
%int4096_36491 = torch.constant.int 4096
%38834 = torch.prim.ListConstruct %int4_36490, %38128, %int4096_36491 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38835 = torch.aten.view %38833, %38834 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38835, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36492 = torch.constant.int 4
%38836 = torch.aten.mul.int %int4_36492, %38143 : !torch.int, !torch.int -> !torch.int
%int512_36493 = torch.constant.int 512
%38837 = torch.prim.ListConstruct %38836, %int512_36493 : (!torch.int, !torch.int) -> !torch.list<int>
%38838 = torch.aten.view %38797, %38837 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38838, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38839 = torch.aten.mm %38838, %38813 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38839, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36494 = torch.constant.int 4
%int4096_36495 = torch.constant.int 4096
%38840 = torch.prim.ListConstruct %int4_36494, %38143, %int4096_36495 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38841 = torch.aten.view %38839, %38840 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38841, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36496 = torch.constant.int 4
%38842 = torch.aten.mul.int %int4_36496, %38158 : !torch.int, !torch.int -> !torch.int
%int512_36497 = torch.constant.int 512
%38843 = torch.prim.ListConstruct %38842, %int512_36497 : (!torch.int, !torch.int) -> !torch.list<int>
%38844 = torch.aten.view %38799, %38843 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38844, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38845 = torch.aten.mm %38844, %38815 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38845, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36498 = torch.constant.int 4
%int4096_36499 = torch.constant.int 4096
%38846 = torch.prim.ListConstruct %int4_36498, %38158, %int4096_36499 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38847 = torch.aten.view %38845, %38846 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38847, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36500 = torch.constant.int 4
%38848 = torch.aten.mul.int %int4_36500, %38173 : !torch.int, !torch.int -> !torch.int
%int512_36501 = torch.constant.int 512
%38849 = torch.prim.ListConstruct %38848, %int512_36501 : (!torch.int, !torch.int) -> !torch.list<int>
%38850 = torch.aten.view %38801, %38849 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38850, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38851 = torch.aten.mm %38850, %38817 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38851, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36502 = torch.constant.int 4
%int4096_36503 = torch.constant.int 4096
%38852 = torch.prim.ListConstruct %int4_36502, %38173, %int4096_36503 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38853 = torch.aten.view %38851, %38852 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38853, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36504 = torch.constant.int 4
%38854 = torch.aten.mul.int %int4_36504, %38188 : !torch.int, !torch.int -> !torch.int
%int512_36505 = torch.constant.int 512
%38855 = torch.prim.ListConstruct %38854, %int512_36505 : (!torch.int, !torch.int) -> !torch.list<int>
%38856 = torch.aten.view %38803, %38855 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38856, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38857 = torch.aten.mm %38856, %38819 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38857, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36506 = torch.constant.int 4
%int4096_36507 = torch.constant.int 4096
%38858 = torch.prim.ListConstruct %int4_36506, %38188, %int4096_36507 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38859 = torch.aten.view %38857, %38858 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38859, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36508 = torch.constant.int 4
%38860 = torch.aten.mul.int %int4_36508, %38203 : !torch.int, !torch.int -> !torch.int
%int512_36509 = torch.constant.int 512
%38861 = torch.prim.ListConstruct %38860, %int512_36509 : (!torch.int, !torch.int) -> !torch.list<int>
%38862 = torch.aten.view %38805, %38861 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38862, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38863 = torch.aten.mm %38862, %38821 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38863, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36510 = torch.constant.int 4
%int4096_36511 = torch.constant.int 4096
%38864 = torch.prim.ListConstruct %int4_36510, %38203, %int4096_36511 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38865 = torch.aten.view %38863, %38864 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38865, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int4_36512 = torch.constant.int 4
%38866 = torch.aten.mul.int %int4_36512, %38218 : !torch.int, !torch.int -> !torch.int
%int512_36513 = torch.constant.int 512
%38867 = torch.prim.ListConstruct %38866, %int512_36513 : (!torch.int, !torch.int) -> !torch.list<int>
%38868 = torch.aten.view %38807, %38867 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %38868, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%38869 = torch.aten.mm %38868, %38823 : !torch.vtensor<[?,512],f16>, !torch.vtensor<[512,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %38869, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36514 = torch.constant.int 4
%int4096_36515 = torch.constant.int 4096
%38870 = torch.prim.ListConstruct %int4_36514, %38218, %int4096_36515 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%38871 = torch.aten.view %38869, %38870 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38871, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38872 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36516 = arith.constant 1 : index
%dim_36517 = tensor.dim %38872, %c1_36516 : tensor<4x?x4096xf16>
%38873 = flow.tensor.transfer %38872 : tensor<4x?x4096xf16>{%dim_36517} to #hal.device.promise<@__device_0>
%38874 = torch_c.from_builtin_tensor %38873 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38874, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38875 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36518 = arith.constant 1 : index
%dim_36519 = tensor.dim %38875, %c1_36518 : tensor<4x?x4096xf16>
%38876 = flow.tensor.transfer %38875 : tensor<4x?x4096xf16>{%dim_36519} to #hal.device.promise<@__device_0>
%38877 = torch_c.from_builtin_tensor %38876 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38877, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38878 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36520 = arith.constant 1 : index
%dim_36521 = tensor.dim %38878, %c1_36520 : tensor<4x?x4096xf16>
%38879 = flow.tensor.transfer %38878 : tensor<4x?x4096xf16>{%dim_36521} to #hal.device.promise<@__device_0>
%38880 = torch_c.from_builtin_tensor %38879 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38880, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38881 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36522 = arith.constant 1 : index
%dim_36523 = tensor.dim %38881, %c1_36522 : tensor<4x?x4096xf16>
%38882 = flow.tensor.transfer %38881 : tensor<4x?x4096xf16>{%dim_36523} to #hal.device.promise<@__device_0>
%38883 = torch_c.from_builtin_tensor %38882 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38883, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38884 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36524 = arith.constant 1 : index
%dim_36525 = tensor.dim %38884, %c1_36524 : tensor<4x?x4096xf16>
%38885 = flow.tensor.transfer %38884 : tensor<4x?x4096xf16>{%dim_36525} to #hal.device.promise<@__device_0>
%38886 = torch_c.from_builtin_tensor %38885 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38886, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38887 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36526 = arith.constant 1 : index
%dim_36527 = tensor.dim %38887, %c1_36526 : tensor<4x?x4096xf16>
%38888 = flow.tensor.transfer %38887 : tensor<4x?x4096xf16>{%dim_36527} to #hal.device.promise<@__device_0>
%38889 = torch_c.from_builtin_tensor %38888 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38890 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36528 = arith.constant 1 : index
%dim_36529 = tensor.dim %38890, %c1_36528 : tensor<4x?x4096xf16>
%38891 = flow.tensor.transfer %38890 : tensor<4x?x4096xf16>{%dim_36529} to #hal.device.promise<@__device_0>
%38892 = torch_c.from_builtin_tensor %38891 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38892, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36530 = torch.constant.int 1
%38893 = torch.aten.add.Tensor %38829, %38874, %int1_36530 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36531 = torch.constant.int 1
%38894 = torch.aten.add.Tensor %38893, %38877, %int1_36531 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38894, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36532 = torch.constant.int 1
%38895 = torch.aten.add.Tensor %38894, %38880, %int1_36532 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36533 = torch.constant.int 1
%38896 = torch.aten.add.Tensor %38895, %38883, %int1_36533 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38896, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36534 = torch.constant.int 1
%38897 = torch.aten.add.Tensor %38896, %38886, %int1_36534 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36535 = torch.constant.int 1
%38898 = torch.aten.add.Tensor %38897, %38889, %int1_36535 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38898, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36536 = torch.constant.int 1
%38899 = torch.aten.add.Tensor %38898, %38892, %int1_36536 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38900 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36537 = arith.constant 1 : index
%dim_36538 = tensor.dim %38900, %c1_36537 : tensor<4x?x4096xf16>
%38901 = flow.tensor.transfer %38900 : tensor<4x?x4096xf16>{%dim_36538} to #hal.device.promise<@__device_1>
%38902 = torch_c.from_builtin_tensor %38901 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38902, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38903 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36539 = arith.constant 1 : index
%dim_36540 = tensor.dim %38903, %c1_36539 : tensor<4x?x4096xf16>
%38904 = flow.tensor.transfer %38903 : tensor<4x?x4096xf16>{%dim_36540} to #hal.device.promise<@__device_1>
%38905 = torch_c.from_builtin_tensor %38904 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38906 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36541 = arith.constant 1 : index
%dim_36542 = tensor.dim %38906, %c1_36541 : tensor<4x?x4096xf16>
%38907 = flow.tensor.transfer %38906 : tensor<4x?x4096xf16>{%dim_36542} to #hal.device.promise<@__device_1>
%38908 = torch_c.from_builtin_tensor %38907 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38908, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38909 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36543 = arith.constant 1 : index
%dim_36544 = tensor.dim %38909, %c1_36543 : tensor<4x?x4096xf16>
%38910 = flow.tensor.transfer %38909 : tensor<4x?x4096xf16>{%dim_36544} to #hal.device.promise<@__device_1>
%38911 = torch_c.from_builtin_tensor %38910 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38912 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36545 = arith.constant 1 : index
%dim_36546 = tensor.dim %38912, %c1_36545 : tensor<4x?x4096xf16>
%38913 = flow.tensor.transfer %38912 : tensor<4x?x4096xf16>{%dim_36546} to #hal.device.promise<@__device_1>
%38914 = torch_c.from_builtin_tensor %38913 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38914, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38915 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36547 = arith.constant 1 : index
%dim_36548 = tensor.dim %38915, %c1_36547 : tensor<4x?x4096xf16>
%38916 = flow.tensor.transfer %38915 : tensor<4x?x4096xf16>{%dim_36548} to #hal.device.promise<@__device_1>
%38917 = torch_c.from_builtin_tensor %38916 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38918 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36549 = arith.constant 1 : index
%dim_36550 = tensor.dim %38918, %c1_36549 : tensor<4x?x4096xf16>
%38919 = flow.tensor.transfer %38918 : tensor<4x?x4096xf16>{%dim_36550} to #hal.device.promise<@__device_1>
%38920 = torch_c.from_builtin_tensor %38919 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38920, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36551 = torch.constant.int 1
%38921 = torch.aten.add.Tensor %38902, %38835, %int1_36551 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36552 = torch.constant.int 1
%38922 = torch.aten.add.Tensor %38921, %38905, %int1_36552 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38922, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36553 = torch.constant.int 1
%38923 = torch.aten.add.Tensor %38922, %38908, %int1_36553 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36554 = torch.constant.int 1
%38924 = torch.aten.add.Tensor %38923, %38911, %int1_36554 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38924, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36555 = torch.constant.int 1
%38925 = torch.aten.add.Tensor %38924, %38914, %int1_36555 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36556 = torch.constant.int 1
%38926 = torch.aten.add.Tensor %38925, %38917, %int1_36556 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38926, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36557 = torch.constant.int 1
%38927 = torch.aten.add.Tensor %38926, %38920, %int1_36557 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38928 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36558 = arith.constant 1 : index
%dim_36559 = tensor.dim %38928, %c1_36558 : tensor<4x?x4096xf16>
%38929 = flow.tensor.transfer %38928 : tensor<4x?x4096xf16>{%dim_36559} to #hal.device.promise<@__device_2>
%38930 = torch_c.from_builtin_tensor %38929 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38930, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38931 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36560 = arith.constant 1 : index
%dim_36561 = tensor.dim %38931, %c1_36560 : tensor<4x?x4096xf16>
%38932 = flow.tensor.transfer %38931 : tensor<4x?x4096xf16>{%dim_36561} to #hal.device.promise<@__device_2>
%38933 = torch_c.from_builtin_tensor %38932 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38934 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36562 = arith.constant 1 : index
%dim_36563 = tensor.dim %38934, %c1_36562 : tensor<4x?x4096xf16>
%38935 = flow.tensor.transfer %38934 : tensor<4x?x4096xf16>{%dim_36563} to #hal.device.promise<@__device_2>
%38936 = torch_c.from_builtin_tensor %38935 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38936, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38937 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36564 = arith.constant 1 : index
%dim_36565 = tensor.dim %38937, %c1_36564 : tensor<4x?x4096xf16>
%38938 = flow.tensor.transfer %38937 : tensor<4x?x4096xf16>{%dim_36565} to #hal.device.promise<@__device_2>
%38939 = torch_c.from_builtin_tensor %38938 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38939, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38940 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36566 = arith.constant 1 : index
%dim_36567 = tensor.dim %38940, %c1_36566 : tensor<4x?x4096xf16>
%38941 = flow.tensor.transfer %38940 : tensor<4x?x4096xf16>{%dim_36567} to #hal.device.promise<@__device_2>
%38942 = torch_c.from_builtin_tensor %38941 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38942, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38943 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36568 = arith.constant 1 : index
%dim_36569 = tensor.dim %38943, %c1_36568 : tensor<4x?x4096xf16>
%38944 = flow.tensor.transfer %38943 : tensor<4x?x4096xf16>{%dim_36569} to #hal.device.promise<@__device_2>
%38945 = torch_c.from_builtin_tensor %38944 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38945, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38946 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36570 = arith.constant 1 : index
%dim_36571 = tensor.dim %38946, %c1_36570 : tensor<4x?x4096xf16>
%38947 = flow.tensor.transfer %38946 : tensor<4x?x4096xf16>{%dim_36571} to #hal.device.promise<@__device_2>
%38948 = torch_c.from_builtin_tensor %38947 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38948, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36572 = torch.constant.int 1
%38949 = torch.aten.add.Tensor %38930, %38933, %int1_36572 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38949, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36573 = torch.constant.int 1
%38950 = torch.aten.add.Tensor %38949, %38841, %int1_36573 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38950, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36574 = torch.constant.int 1
%38951 = torch.aten.add.Tensor %38950, %38936, %int1_36574 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38951, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36575 = torch.constant.int 1
%38952 = torch.aten.add.Tensor %38951, %38939, %int1_36575 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38952, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36576 = torch.constant.int 1
%38953 = torch.aten.add.Tensor %38952, %38942, %int1_36576 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38953, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36577 = torch.constant.int 1
%38954 = torch.aten.add.Tensor %38953, %38945, %int1_36577 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38954, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36578 = torch.constant.int 1
%38955 = torch.aten.add.Tensor %38954, %38948, %int1_36578 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38955, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38956 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36579 = arith.constant 1 : index
%dim_36580 = tensor.dim %38956, %c1_36579 : tensor<4x?x4096xf16>
%38957 = flow.tensor.transfer %38956 : tensor<4x?x4096xf16>{%dim_36580} to #hal.device.promise<@__device_3>
%38958 = torch_c.from_builtin_tensor %38957 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38958, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38959 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36581 = arith.constant 1 : index
%dim_36582 = tensor.dim %38959, %c1_36581 : tensor<4x?x4096xf16>
%38960 = flow.tensor.transfer %38959 : tensor<4x?x4096xf16>{%dim_36582} to #hal.device.promise<@__device_3>
%38961 = torch_c.from_builtin_tensor %38960 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38961, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38962 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36583 = arith.constant 1 : index
%dim_36584 = tensor.dim %38962, %c1_36583 : tensor<4x?x4096xf16>
%38963 = flow.tensor.transfer %38962 : tensor<4x?x4096xf16>{%dim_36584} to #hal.device.promise<@__device_3>
%38964 = torch_c.from_builtin_tensor %38963 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38964, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38965 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36585 = arith.constant 1 : index
%dim_36586 = tensor.dim %38965, %c1_36585 : tensor<4x?x4096xf16>
%38966 = flow.tensor.transfer %38965 : tensor<4x?x4096xf16>{%dim_36586} to #hal.device.promise<@__device_3>
%38967 = torch_c.from_builtin_tensor %38966 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38967, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38968 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36587 = arith.constant 1 : index
%dim_36588 = tensor.dim %38968, %c1_36587 : tensor<4x?x4096xf16>
%38969 = flow.tensor.transfer %38968 : tensor<4x?x4096xf16>{%dim_36588} to #hal.device.promise<@__device_3>
%38970 = torch_c.from_builtin_tensor %38969 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38970, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38971 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36589 = arith.constant 1 : index
%dim_36590 = tensor.dim %38971, %c1_36589 : tensor<4x?x4096xf16>
%38972 = flow.tensor.transfer %38971 : tensor<4x?x4096xf16>{%dim_36590} to #hal.device.promise<@__device_3>
%38973 = torch_c.from_builtin_tensor %38972 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38973, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38974 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36591 = arith.constant 1 : index
%dim_36592 = tensor.dim %38974, %c1_36591 : tensor<4x?x4096xf16>
%38975 = flow.tensor.transfer %38974 : tensor<4x?x4096xf16>{%dim_36592} to #hal.device.promise<@__device_3>
%38976 = torch_c.from_builtin_tensor %38975 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38976, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36593 = torch.constant.int 1
%38977 = torch.aten.add.Tensor %38958, %38961, %int1_36593 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38977, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36594 = torch.constant.int 1
%38978 = torch.aten.add.Tensor %38977, %38964, %int1_36594 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38978, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36595 = torch.constant.int 1
%38979 = torch.aten.add.Tensor %38978, %38847, %int1_36595 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38979, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36596 = torch.constant.int 1
%38980 = torch.aten.add.Tensor %38979, %38967, %int1_36596 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38980, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36597 = torch.constant.int 1
%38981 = torch.aten.add.Tensor %38980, %38970, %int1_36597 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38981, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36598 = torch.constant.int 1
%38982 = torch.aten.add.Tensor %38981, %38973, %int1_36598 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38982, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36599 = torch.constant.int 1
%38983 = torch.aten.add.Tensor %38982, %38976, %int1_36599 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38984 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36600 = arith.constant 1 : index
%dim_36601 = tensor.dim %38984, %c1_36600 : tensor<4x?x4096xf16>
%38985 = flow.tensor.transfer %38984 : tensor<4x?x4096xf16>{%dim_36601} to #hal.device.promise<@__device_4>
%38986 = torch_c.from_builtin_tensor %38985 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38986, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38987 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36602 = arith.constant 1 : index
%dim_36603 = tensor.dim %38987, %c1_36602 : tensor<4x?x4096xf16>
%38988 = flow.tensor.transfer %38987 : tensor<4x?x4096xf16>{%dim_36603} to #hal.device.promise<@__device_4>
%38989 = torch_c.from_builtin_tensor %38988 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38989, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38990 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36604 = arith.constant 1 : index
%dim_36605 = tensor.dim %38990, %c1_36604 : tensor<4x?x4096xf16>
%38991 = flow.tensor.transfer %38990 : tensor<4x?x4096xf16>{%dim_36605} to #hal.device.promise<@__device_4>
%38992 = torch_c.from_builtin_tensor %38991 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38992, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38993 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36606 = arith.constant 1 : index
%dim_36607 = tensor.dim %38993, %c1_36606 : tensor<4x?x4096xf16>
%38994 = flow.tensor.transfer %38993 : tensor<4x?x4096xf16>{%dim_36607} to #hal.device.promise<@__device_4>
%38995 = torch_c.from_builtin_tensor %38994 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38995, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38996 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36608 = arith.constant 1 : index
%dim_36609 = tensor.dim %38996, %c1_36608 : tensor<4x?x4096xf16>
%38997 = flow.tensor.transfer %38996 : tensor<4x?x4096xf16>{%dim_36609} to #hal.device.promise<@__device_4>
%38998 = torch_c.from_builtin_tensor %38997 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %38998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%38999 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36610 = arith.constant 1 : index
%dim_36611 = tensor.dim %38999, %c1_36610 : tensor<4x?x4096xf16>
%39000 = flow.tensor.transfer %38999 : tensor<4x?x4096xf16>{%dim_36611} to #hal.device.promise<@__device_4>
%39001 = torch_c.from_builtin_tensor %39000 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39001, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39002 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36612 = arith.constant 1 : index
%dim_36613 = tensor.dim %39002, %c1_36612 : tensor<4x?x4096xf16>
%39003 = flow.tensor.transfer %39002 : tensor<4x?x4096xf16>{%dim_36613} to #hal.device.promise<@__device_4>
%39004 = torch_c.from_builtin_tensor %39003 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39004, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36614 = torch.constant.int 1
%39005 = torch.aten.add.Tensor %38986, %38989, %int1_36614 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39005, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36615 = torch.constant.int 1
%39006 = torch.aten.add.Tensor %39005, %38992, %int1_36615 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39006, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36616 = torch.constant.int 1
%39007 = torch.aten.add.Tensor %39006, %38995, %int1_36616 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39007, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36617 = torch.constant.int 1
%39008 = torch.aten.add.Tensor %39007, %38853, %int1_36617 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39008, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36618 = torch.constant.int 1
%39009 = torch.aten.add.Tensor %39008, %38998, %int1_36618 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39009, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36619 = torch.constant.int 1
%39010 = torch.aten.add.Tensor %39009, %39001, %int1_36619 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39010, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36620 = torch.constant.int 1
%39011 = torch.aten.add.Tensor %39010, %39004, %int1_36620 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39011, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39012 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36621 = arith.constant 1 : index
%dim_36622 = tensor.dim %39012, %c1_36621 : tensor<4x?x4096xf16>
%39013 = flow.tensor.transfer %39012 : tensor<4x?x4096xf16>{%dim_36622} to #hal.device.promise<@__device_5>
%39014 = torch_c.from_builtin_tensor %39013 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39015 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36623 = arith.constant 1 : index
%dim_36624 = tensor.dim %39015, %c1_36623 : tensor<4x?x4096xf16>
%39016 = flow.tensor.transfer %39015 : tensor<4x?x4096xf16>{%dim_36624} to #hal.device.promise<@__device_5>
%39017 = torch_c.from_builtin_tensor %39016 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39018 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36625 = arith.constant 1 : index
%dim_36626 = tensor.dim %39018, %c1_36625 : tensor<4x?x4096xf16>
%39019 = flow.tensor.transfer %39018 : tensor<4x?x4096xf16>{%dim_36626} to #hal.device.promise<@__device_5>
%39020 = torch_c.from_builtin_tensor %39019 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39020, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39021 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36627 = arith.constant 1 : index
%dim_36628 = tensor.dim %39021, %c1_36627 : tensor<4x?x4096xf16>
%39022 = flow.tensor.transfer %39021 : tensor<4x?x4096xf16>{%dim_36628} to #hal.device.promise<@__device_5>
%39023 = torch_c.from_builtin_tensor %39022 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39023, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39024 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36629 = arith.constant 1 : index
%dim_36630 = tensor.dim %39024, %c1_36629 : tensor<4x?x4096xf16>
%39025 = flow.tensor.transfer %39024 : tensor<4x?x4096xf16>{%dim_36630} to #hal.device.promise<@__device_5>
%39026 = torch_c.from_builtin_tensor %39025 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39026, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39027 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36631 = arith.constant 1 : index
%dim_36632 = tensor.dim %39027, %c1_36631 : tensor<4x?x4096xf16>
%39028 = flow.tensor.transfer %39027 : tensor<4x?x4096xf16>{%dim_36632} to #hal.device.promise<@__device_5>
%39029 = torch_c.from_builtin_tensor %39028 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39030 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36633 = arith.constant 1 : index
%dim_36634 = tensor.dim %39030, %c1_36633 : tensor<4x?x4096xf16>
%39031 = flow.tensor.transfer %39030 : tensor<4x?x4096xf16>{%dim_36634} to #hal.device.promise<@__device_5>
%39032 = torch_c.from_builtin_tensor %39031 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36635 = torch.constant.int 1
%39033 = torch.aten.add.Tensor %39014, %39017, %int1_36635 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36636 = torch.constant.int 1
%39034 = torch.aten.add.Tensor %39033, %39020, %int1_36636 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39034, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36637 = torch.constant.int 1
%39035 = torch.aten.add.Tensor %39034, %39023, %int1_36637 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39035, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36638 = torch.constant.int 1
%39036 = torch.aten.add.Tensor %39035, %39026, %int1_36638 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39036, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36639 = torch.constant.int 1
%39037 = torch.aten.add.Tensor %39036, %38859, %int1_36639 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39037, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36640 = torch.constant.int 1
%39038 = torch.aten.add.Tensor %39037, %39029, %int1_36640 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39038, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36641 = torch.constant.int 1
%39039 = torch.aten.add.Tensor %39038, %39032, %int1_36641 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39039, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39040 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36642 = arith.constant 1 : index
%dim_36643 = tensor.dim %39040, %c1_36642 : tensor<4x?x4096xf16>
%39041 = flow.tensor.transfer %39040 : tensor<4x?x4096xf16>{%dim_36643} to #hal.device.promise<@__device_6>
%39042 = torch_c.from_builtin_tensor %39041 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39042, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39043 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36644 = arith.constant 1 : index
%dim_36645 = tensor.dim %39043, %c1_36644 : tensor<4x?x4096xf16>
%39044 = flow.tensor.transfer %39043 : tensor<4x?x4096xf16>{%dim_36645} to #hal.device.promise<@__device_6>
%39045 = torch_c.from_builtin_tensor %39044 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39045, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39046 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36646 = arith.constant 1 : index
%dim_36647 = tensor.dim %39046, %c1_36646 : tensor<4x?x4096xf16>
%39047 = flow.tensor.transfer %39046 : tensor<4x?x4096xf16>{%dim_36647} to #hal.device.promise<@__device_6>
%39048 = torch_c.from_builtin_tensor %39047 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39049 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36648 = arith.constant 1 : index
%dim_36649 = tensor.dim %39049, %c1_36648 : tensor<4x?x4096xf16>
%39050 = flow.tensor.transfer %39049 : tensor<4x?x4096xf16>{%dim_36649} to #hal.device.promise<@__device_6>
%39051 = torch_c.from_builtin_tensor %39050 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39051, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39052 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36650 = arith.constant 1 : index
%dim_36651 = tensor.dim %39052, %c1_36650 : tensor<4x?x4096xf16>
%39053 = flow.tensor.transfer %39052 : tensor<4x?x4096xf16>{%dim_36651} to #hal.device.promise<@__device_6>
%39054 = torch_c.from_builtin_tensor %39053 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39054, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39055 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36652 = arith.constant 1 : index
%dim_36653 = tensor.dim %39055, %c1_36652 : tensor<4x?x4096xf16>
%39056 = flow.tensor.transfer %39055 : tensor<4x?x4096xf16>{%dim_36653} to #hal.device.promise<@__device_6>
%39057 = torch_c.from_builtin_tensor %39056 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39057, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39058 = torch_c.to_builtin_tensor %38871 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36654 = arith.constant 1 : index
%dim_36655 = tensor.dim %39058, %c1_36654 : tensor<4x?x4096xf16>
%39059 = flow.tensor.transfer %39058 : tensor<4x?x4096xf16>{%dim_36655} to #hal.device.promise<@__device_6>
%39060 = torch_c.from_builtin_tensor %39059 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39060, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36656 = torch.constant.int 1
%39061 = torch.aten.add.Tensor %39042, %39045, %int1_36656 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39061, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36657 = torch.constant.int 1
%39062 = torch.aten.add.Tensor %39061, %39048, %int1_36657 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36658 = torch.constant.int 1
%39063 = torch.aten.add.Tensor %39062, %39051, %int1_36658 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36659 = torch.constant.int 1
%39064 = torch.aten.add.Tensor %39063, %39054, %int1_36659 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39064, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36660 = torch.constant.int 1
%39065 = torch.aten.add.Tensor %39064, %39057, %int1_36660 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39065, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36661 = torch.constant.int 1
%39066 = torch.aten.add.Tensor %39065, %38865, %int1_36661 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39066, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36662 = torch.constant.int 1
%39067 = torch.aten.add.Tensor %39066, %39060, %int1_36662 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39067, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39068 = torch_c.to_builtin_tensor %38829 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36663 = arith.constant 1 : index
%dim_36664 = tensor.dim %39068, %c1_36663 : tensor<4x?x4096xf16>
%39069 = flow.tensor.transfer %39068 : tensor<4x?x4096xf16>{%dim_36664} to #hal.device.promise<@__device_7>
%39070 = torch_c.from_builtin_tensor %39069 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39070, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39071 = torch_c.to_builtin_tensor %38835 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36665 = arith.constant 1 : index
%dim_36666 = tensor.dim %39071, %c1_36665 : tensor<4x?x4096xf16>
%39072 = flow.tensor.transfer %39071 : tensor<4x?x4096xf16>{%dim_36666} to #hal.device.promise<@__device_7>
%39073 = torch_c.from_builtin_tensor %39072 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39074 = torch_c.to_builtin_tensor %38841 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36667 = arith.constant 1 : index
%dim_36668 = tensor.dim %39074, %c1_36667 : tensor<4x?x4096xf16>
%39075 = flow.tensor.transfer %39074 : tensor<4x?x4096xf16>{%dim_36668} to #hal.device.promise<@__device_7>
%39076 = torch_c.from_builtin_tensor %39075 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39076, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39077 = torch_c.to_builtin_tensor %38847 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36669 = arith.constant 1 : index
%dim_36670 = tensor.dim %39077, %c1_36669 : tensor<4x?x4096xf16>
%39078 = flow.tensor.transfer %39077 : tensor<4x?x4096xf16>{%dim_36670} to #hal.device.promise<@__device_7>
%39079 = torch_c.from_builtin_tensor %39078 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39079, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39080 = torch_c.to_builtin_tensor %38853 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36671 = arith.constant 1 : index
%dim_36672 = tensor.dim %39080, %c1_36671 : tensor<4x?x4096xf16>
%39081 = flow.tensor.transfer %39080 : tensor<4x?x4096xf16>{%dim_36672} to #hal.device.promise<@__device_7>
%39082 = torch_c.from_builtin_tensor %39081 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39082, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39083 = torch_c.to_builtin_tensor %38859 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36673 = arith.constant 1 : index
%dim_36674 = tensor.dim %39083, %c1_36673 : tensor<4x?x4096xf16>
%39084 = flow.tensor.transfer %39083 : tensor<4x?x4096xf16>{%dim_36674} to #hal.device.promise<@__device_7>
%39085 = torch_c.from_builtin_tensor %39084 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39085, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39086 = torch_c.to_builtin_tensor %38865 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36675 = arith.constant 1 : index
%dim_36676 = tensor.dim %39086, %c1_36675 : tensor<4x?x4096xf16>
%39087 = flow.tensor.transfer %39086 : tensor<4x?x4096xf16>{%dim_36676} to #hal.device.promise<@__device_7>
%39088 = torch_c.from_builtin_tensor %39087 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36677 = torch.constant.int 1
%39089 = torch.aten.add.Tensor %39070, %39073, %int1_36677 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36678 = torch.constant.int 1
%39090 = torch.aten.add.Tensor %39089, %39076, %int1_36678 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39090, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36679 = torch.constant.int 1
%39091 = torch.aten.add.Tensor %39090, %39079, %int1_36679 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39091, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36680 = torch.constant.int 1
%39092 = torch.aten.add.Tensor %39091, %39082, %int1_36680 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36681 = torch.constant.int 1
%39093 = torch.aten.add.Tensor %39092, %39085, %int1_36681 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36682 = torch.constant.int 1
%39094 = torch.aten.add.Tensor %39093, %39088, %int1_36682 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39094, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36683 = torch.constant.int 1
%39095 = torch.aten.add.Tensor %39094, %38871, %int1_36683 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39095, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36684 = torch.constant.int 1
%39096 = torch.aten.add.Tensor %37755, %38899, %int1_36684 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39096, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36685 = torch.constant.int 1
%39097 = torch.aten.add.Tensor %37756, %38927, %int1_36685 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39097, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36686 = torch.constant.int 1
%39098 = torch.aten.add.Tensor %37757, %38955, %int1_36686 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39098, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36687 = torch.constant.int 1
%39099 = torch.aten.add.Tensor %37758, %38983, %int1_36687 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39099, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36688 = torch.constant.int 1
%39100 = torch.aten.add.Tensor %37759, %39011, %int1_36688 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39100, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36689 = torch.constant.int 1
%39101 = torch.aten.add.Tensor %37760, %39039, %int1_36689 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39101, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36690 = torch.constant.int 1
%39102 = torch.aten.add.Tensor %37761, %39067, %int1_36690 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39102, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36691 = torch.constant.int 1
%39103 = torch.aten.add.Tensor %37762, %39095, %int1_36691 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39103, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_36692 = torch.constant.int 6
%39104 = torch.prims.convert_element_type %39096, %int6_36692 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39104, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36693 = torch.constant.int 6
%39105 = torch.prims.convert_element_type %39097, %int6_36693 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39105, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36694 = torch.constant.int 6
%39106 = torch.prims.convert_element_type %39098, %int6_36694 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39106, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36695 = torch.constant.int 6
%39107 = torch.prims.convert_element_type %39099, %int6_36695 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39107, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36696 = torch.constant.int 6
%39108 = torch.prims.convert_element_type %39100, %int6_36696 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39108, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36697 = torch.constant.int 6
%39109 = torch.prims.convert_element_type %39101, %int6_36697 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39109, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36698 = torch.constant.int 6
%39110 = torch.prims.convert_element_type %39102, %int6_36698 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39110, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_36699 = torch.constant.int 6
%39111 = torch.prims.convert_element_type %39103, %int6_36699 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39111, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36700 = torch.constant.int 2
%39112 = torch.aten.pow.Tensor_Scalar %39104, %int2_36700 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39112, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36701 = torch.constant.int 2
%39113 = torch.aten.pow.Tensor_Scalar %39105, %int2_36701 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39113, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36702 = torch.constant.int 2
%39114 = torch.aten.pow.Tensor_Scalar %39106, %int2_36702 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39114, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36703 = torch.constant.int 2
%39115 = torch.aten.pow.Tensor_Scalar %39107, %int2_36703 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39115, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36704 = torch.constant.int 2
%39116 = torch.aten.pow.Tensor_Scalar %39108, %int2_36704 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39116, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36705 = torch.constant.int 2
%39117 = torch.aten.pow.Tensor_Scalar %39109, %int2_36705 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39117, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36706 = torch.constant.int 2
%39118 = torch.aten.pow.Tensor_Scalar %39110, %int2_36706 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39118, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_36707 = torch.constant.int 2
%39119 = torch.aten.pow.Tensor_Scalar %39111, %int2_36707 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39119, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_36708 = torch.constant.int -1
%39120 = torch.prim.ListConstruct %int-1_36708 : (!torch.int) -> !torch.list<int>
%true_36709 = torch.constant.bool true
%none_36710 = torch.constant.none
%39121 = torch.aten.mean.dim %39112, %39120, %true_36709, %none_36710 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39121, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36711 = torch.constant.int -1
%39122 = torch.prim.ListConstruct %int-1_36711 : (!torch.int) -> !torch.list<int>
%true_36712 = torch.constant.bool true
%none_36713 = torch.constant.none
%39123 = torch.aten.mean.dim %39113, %39122, %true_36712, %none_36713 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39123, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36714 = torch.constant.int -1
%39124 = torch.prim.ListConstruct %int-1_36714 : (!torch.int) -> !torch.list<int>
%true_36715 = torch.constant.bool true
%none_36716 = torch.constant.none
%39125 = torch.aten.mean.dim %39114, %39124, %true_36715, %none_36716 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39125, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36717 = torch.constant.int -1
%39126 = torch.prim.ListConstruct %int-1_36717 : (!torch.int) -> !torch.list<int>
%true_36718 = torch.constant.bool true
%none_36719 = torch.constant.none
%39127 = torch.aten.mean.dim %39115, %39126, %true_36718, %none_36719 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39127, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36720 = torch.constant.int -1
%39128 = torch.prim.ListConstruct %int-1_36720 : (!torch.int) -> !torch.list<int>
%true_36721 = torch.constant.bool true
%none_36722 = torch.constant.none
%39129 = torch.aten.mean.dim %39116, %39128, %true_36721, %none_36722 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39129, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36723 = torch.constant.int -1
%39130 = torch.prim.ListConstruct %int-1_36723 : (!torch.int) -> !torch.list<int>
%true_36724 = torch.constant.bool true
%none_36725 = torch.constant.none
%39131 = torch.aten.mean.dim %39117, %39130, %true_36724, %none_36725 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39131, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36726 = torch.constant.int -1
%39132 = torch.prim.ListConstruct %int-1_36726 : (!torch.int) -> !torch.list<int>
%true_36727 = torch.constant.bool true
%none_36728 = torch.constant.none
%39133 = torch.aten.mean.dim %39118, %39132, %true_36727, %none_36728 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39133, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_36729 = torch.constant.int -1
%39134 = torch.prim.ListConstruct %int-1_36729 : (!torch.int) -> !torch.list<int>
%true_36730 = torch.constant.bool true
%none_36731 = torch.constant.none
%39135 = torch.aten.mean.dim %39119, %39134, %true_36730, %none_36731 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39135, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36732 = torch.constant.float 9.9999997473787516E-6
%int1_36733 = torch.constant.int 1
%39136 = torch.aten.add.Scalar %39121, %float9.999990e-06_36732, %int1_36733 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39136, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36734 = torch.constant.float 9.9999997473787516E-6
%int1_36735 = torch.constant.int 1
%39137 = torch.aten.add.Scalar %39123, %float9.999990e-06_36734, %int1_36735 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39137, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36736 = torch.constant.float 9.9999997473787516E-6
%int1_36737 = torch.constant.int 1
%39138 = torch.aten.add.Scalar %39125, %float9.999990e-06_36736, %int1_36737 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39138, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36738 = torch.constant.float 9.9999997473787516E-6
%int1_36739 = torch.constant.int 1
%39139 = torch.aten.add.Scalar %39127, %float9.999990e-06_36738, %int1_36739 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39139, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36740 = torch.constant.float 9.9999997473787516E-6
%int1_36741 = torch.constant.int 1
%39140 = torch.aten.add.Scalar %39129, %float9.999990e-06_36740, %int1_36741 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39140, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36742 = torch.constant.float 9.9999997473787516E-6
%int1_36743 = torch.constant.int 1
%39141 = torch.aten.add.Scalar %39131, %float9.999990e-06_36742, %int1_36743 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36744 = torch.constant.float 9.9999997473787516E-6
%int1_36745 = torch.constant.int 1
%39142 = torch.aten.add.Scalar %39133, %float9.999990e-06_36744, %int1_36745 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_36746 = torch.constant.float 9.9999997473787516E-6
%int1_36747 = torch.constant.int 1
%39143 = torch.aten.add.Scalar %39135, %float9.999990e-06_36746, %int1_36747 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39143, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39144 = torch.aten.rsqrt %39136 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39144, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39145 = torch.aten.rsqrt %39137 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39146 = torch.aten.rsqrt %39138 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39147 = torch.aten.rsqrt %39139 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39147, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39148 = torch.aten.rsqrt %39140 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39148, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39149 = torch.aten.rsqrt %39141 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39149, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39150 = torch.aten.rsqrt %39142 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39150, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39151 = torch.aten.rsqrt %39143 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39151, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39152 = torch.aten.mul.Tensor %39104, %39144 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39152, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39153 = torch.aten.mul.Tensor %39105, %39145 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39153, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39154 = torch.aten.mul.Tensor %39106, %39146 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39154, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39155 = torch.aten.mul.Tensor %39107, %39147 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39155, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39156 = torch.aten.mul.Tensor %39108, %39148 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39156, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39157 = torch.aten.mul.Tensor %39109, %39149 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39157, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39158 = torch.aten.mul.Tensor %39110, %39150 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39158, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39159 = torch.aten.mul.Tensor %39111, %39151 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39159, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39160 = torch.aten.mul.Tensor %1416, %39152 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39160, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39161 = torch.aten.mul.Tensor %1417, %39153 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39161, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39162 = torch.aten.mul.Tensor %1418, %39154 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39162, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39163 = torch.aten.mul.Tensor %1419, %39155 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39163, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39164 = torch.aten.mul.Tensor %1420, %39156 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39164, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39165 = torch.aten.mul.Tensor %1421, %39157 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39165, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39166 = torch.aten.mul.Tensor %1422, %39158 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39166, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39167 = torch.aten.mul.Tensor %1423, %39159 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39167, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_36748 = torch.constant.int 5
%39168 = torch.prims.convert_element_type %39160, %int5_36748 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39168, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36749 = torch.constant.int 5
%39169 = torch.prims.convert_element_type %39161, %int5_36749 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39169, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36750 = torch.constant.int 5
%39170 = torch.prims.convert_element_type %39162, %int5_36750 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39170, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36751 = torch.constant.int 5
%39171 = torch.prims.convert_element_type %39163, %int5_36751 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39171, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36752 = torch.constant.int 5
%39172 = torch.prims.convert_element_type %39164, %int5_36752 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39172, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36753 = torch.constant.int 5
%39173 = torch.prims.convert_element_type %39165, %int5_36753 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39173, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36754 = torch.constant.int 5
%39174 = torch.prims.convert_element_type %39166, %int5_36754 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39174, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_36755 = torch.constant.int 5
%39175 = torch.prims.convert_element_type %39167, %int5_36755 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39175, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36756 = torch.constant.int 1
%int0_36757 = torch.constant.int 0
%39176 = torch.prim.ListConstruct %int1_36756, %int0_36757 : (!torch.int, !torch.int) -> !torch.list<int>
%39177 = torch.aten.permute %1424, %39176 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36758 = torch.constant.int 1
%int0_36759 = torch.constant.int 0
%39178 = torch.prim.ListConstruct %int1_36758, %int0_36759 : (!torch.int, !torch.int) -> !torch.list<int>
%39179 = torch.aten.permute %1425, %39178 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36760 = torch.constant.int 1
%int0_36761 = torch.constant.int 0
%39180 = torch.prim.ListConstruct %int1_36760, %int0_36761 : (!torch.int, !torch.int) -> !torch.list<int>
%39181 = torch.aten.permute %1426, %39180 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36762 = torch.constant.int 1
%int0_36763 = torch.constant.int 0
%39182 = torch.prim.ListConstruct %int1_36762, %int0_36763 : (!torch.int, !torch.int) -> !torch.list<int>
%39183 = torch.aten.permute %1427, %39182 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36764 = torch.constant.int 1
%int0_36765 = torch.constant.int 0
%39184 = torch.prim.ListConstruct %int1_36764, %int0_36765 : (!torch.int, !torch.int) -> !torch.list<int>
%39185 = torch.aten.permute %1428, %39184 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36766 = torch.constant.int 1
%int0_36767 = torch.constant.int 0
%39186 = torch.prim.ListConstruct %int1_36766, %int0_36767 : (!torch.int, !torch.int) -> !torch.list<int>
%39187 = torch.aten.permute %1429, %39186 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36768 = torch.constant.int 1
%int0_36769 = torch.constant.int 0
%39188 = torch.prim.ListConstruct %int1_36768, %int0_36769 : (!torch.int, !torch.int) -> !torch.list<int>
%39189 = torch.aten.permute %1430, %39188 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36770 = torch.constant.int 1
%int0_36771 = torch.constant.int 0
%39190 = torch.prim.ListConstruct %int1_36770, %int0_36771 : (!torch.int, !torch.int) -> !torch.list<int>
%39191 = torch.aten.permute %1431, %39190 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_36772 = torch.constant.int 4
%39192 = torch.aten.mul.int %int4_36772, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36773 = torch.constant.int 4096
%39193 = torch.prim.ListConstruct %39192, %int4096_36773 : (!torch.int, !torch.int) -> !torch.list<int>
%39194 = torch.aten.view %39168, %39193 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39194, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39195 = torch.aten.mm %39194, %39177 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39195, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36774 = torch.constant.int 4
%int1792_36775 = torch.constant.int 1792
%39196 = torch.prim.ListConstruct %int4_36774, %2482, %int1792_36775 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39197 = torch.aten.view %39195, %39196 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39197, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36776 = torch.constant.int 4
%39198 = torch.aten.mul.int %int4_36776, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36777 = torch.constant.int 4096
%39199 = torch.prim.ListConstruct %39198, %int4096_36777 : (!torch.int, !torch.int) -> !torch.list<int>
%39200 = torch.aten.view %39169, %39199 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39200, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39201 = torch.aten.mm %39200, %39179 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39201, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36778 = torch.constant.int 4
%int1792_36779 = torch.constant.int 1792
%39202 = torch.prim.ListConstruct %int4_36778, %2482, %int1792_36779 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39203 = torch.aten.view %39201, %39202 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39203, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36780 = torch.constant.int 4
%39204 = torch.aten.mul.int %int4_36780, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36781 = torch.constant.int 4096
%39205 = torch.prim.ListConstruct %39204, %int4096_36781 : (!torch.int, !torch.int) -> !torch.list<int>
%39206 = torch.aten.view %39170, %39205 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39206, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39207 = torch.aten.mm %39206, %39181 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39207, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36782 = torch.constant.int 4
%int1792_36783 = torch.constant.int 1792
%39208 = torch.prim.ListConstruct %int4_36782, %2482, %int1792_36783 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39209 = torch.aten.view %39207, %39208 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39209, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36784 = torch.constant.int 4
%39210 = torch.aten.mul.int %int4_36784, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36785 = torch.constant.int 4096
%39211 = torch.prim.ListConstruct %39210, %int4096_36785 : (!torch.int, !torch.int) -> !torch.list<int>
%39212 = torch.aten.view %39171, %39211 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39212, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39213 = torch.aten.mm %39212, %39183 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39213, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36786 = torch.constant.int 4
%int1792_36787 = torch.constant.int 1792
%39214 = torch.prim.ListConstruct %int4_36786, %2482, %int1792_36787 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39215 = torch.aten.view %39213, %39214 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39215, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36788 = torch.constant.int 4
%39216 = torch.aten.mul.int %int4_36788, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36789 = torch.constant.int 4096
%39217 = torch.prim.ListConstruct %39216, %int4096_36789 : (!torch.int, !torch.int) -> !torch.list<int>
%39218 = torch.aten.view %39172, %39217 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39218, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39219 = torch.aten.mm %39218, %39185 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39219, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36790 = torch.constant.int 4
%int1792_36791 = torch.constant.int 1792
%39220 = torch.prim.ListConstruct %int4_36790, %2482, %int1792_36791 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39221 = torch.aten.view %39219, %39220 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39221, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36792 = torch.constant.int 4
%39222 = torch.aten.mul.int %int4_36792, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36793 = torch.constant.int 4096
%39223 = torch.prim.ListConstruct %39222, %int4096_36793 : (!torch.int, !torch.int) -> !torch.list<int>
%39224 = torch.aten.view %39173, %39223 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39224, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39225 = torch.aten.mm %39224, %39187 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39225, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36794 = torch.constant.int 4
%int1792_36795 = torch.constant.int 1792
%39226 = torch.prim.ListConstruct %int4_36794, %2482, %int1792_36795 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39227 = torch.aten.view %39225, %39226 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39227, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36796 = torch.constant.int 4
%39228 = torch.aten.mul.int %int4_36796, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36797 = torch.constant.int 4096
%39229 = torch.prim.ListConstruct %39228, %int4096_36797 : (!torch.int, !torch.int) -> !torch.list<int>
%39230 = torch.aten.view %39174, %39229 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39230, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39231 = torch.aten.mm %39230, %39189 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39231, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36798 = torch.constant.int 4
%int1792_36799 = torch.constant.int 1792
%39232 = torch.prim.ListConstruct %int4_36798, %2482, %int1792_36799 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39233 = torch.aten.view %39231, %39232 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39233, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36800 = torch.constant.int 4
%39234 = torch.aten.mul.int %int4_36800, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36801 = torch.constant.int 4096
%39235 = torch.prim.ListConstruct %39234, %int4096_36801 : (!torch.int, !torch.int) -> !torch.list<int>
%39236 = torch.aten.view %39175, %39235 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39236, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39237 = torch.aten.mm %39236, %39191 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39237, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36802 = torch.constant.int 4
%int1792_36803 = torch.constant.int 1792
%39238 = torch.prim.ListConstruct %int4_36802, %2482, %int1792_36803 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39239 = torch.aten.view %39237, %39238 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39239, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39240 = torch.aten.silu %39197 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39240, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39241 = torch.aten.silu %39203 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39241, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39242 = torch.aten.silu %39209 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39242, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39243 = torch.aten.silu %39215 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39243, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39244 = torch.aten.silu %39221 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39244, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39245 = torch.aten.silu %39227 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39245, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39246 = torch.aten.silu %39233 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39246, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39247 = torch.aten.silu %39239 : !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39247, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_36804 = torch.constant.int 1
%int0_36805 = torch.constant.int 0
%39248 = torch.prim.ListConstruct %int1_36804, %int0_36805 : (!torch.int, !torch.int) -> !torch.list<int>
%39249 = torch.aten.permute %1432, %39248 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36806 = torch.constant.int 1
%int0_36807 = torch.constant.int 0
%39250 = torch.prim.ListConstruct %int1_36806, %int0_36807 : (!torch.int, !torch.int) -> !torch.list<int>
%39251 = torch.aten.permute %1433, %39250 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36808 = torch.constant.int 1
%int0_36809 = torch.constant.int 0
%39252 = torch.prim.ListConstruct %int1_36808, %int0_36809 : (!torch.int, !torch.int) -> !torch.list<int>
%39253 = torch.aten.permute %1434, %39252 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36810 = torch.constant.int 1
%int0_36811 = torch.constant.int 0
%39254 = torch.prim.ListConstruct %int1_36810, %int0_36811 : (!torch.int, !torch.int) -> !torch.list<int>
%39255 = torch.aten.permute %1435, %39254 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36812 = torch.constant.int 1
%int0_36813 = torch.constant.int 0
%39256 = torch.prim.ListConstruct %int1_36812, %int0_36813 : (!torch.int, !torch.int) -> !torch.list<int>
%39257 = torch.aten.permute %1436, %39256 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36814 = torch.constant.int 1
%int0_36815 = torch.constant.int 0
%39258 = torch.prim.ListConstruct %int1_36814, %int0_36815 : (!torch.int, !torch.int) -> !torch.list<int>
%39259 = torch.aten.permute %1437, %39258 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36816 = torch.constant.int 1
%int0_36817 = torch.constant.int 0
%39260 = torch.prim.ListConstruct %int1_36816, %int0_36817 : (!torch.int, !torch.int) -> !torch.list<int>
%39261 = torch.aten.permute %1438, %39260 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int1_36818 = torch.constant.int 1
%int0_36819 = torch.constant.int 0
%39262 = torch.prim.ListConstruct %int1_36818, %int0_36819 : (!torch.int, !torch.int) -> !torch.list<int>
%39263 = torch.aten.permute %1439, %39262 : !torch.vtensor<[1792,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,1792],f16>
%int4_36820 = torch.constant.int 4
%39264 = torch.aten.mul.int %int4_36820, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36821 = torch.constant.int 4096
%39265 = torch.prim.ListConstruct %39264, %int4096_36821 : (!torch.int, !torch.int) -> !torch.list<int>
%39266 = torch.aten.view %39168, %39265 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39266, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39267 = torch.aten.mm %39266, %39249 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39267, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36822 = torch.constant.int 4
%int1792_36823 = torch.constant.int 1792
%39268 = torch.prim.ListConstruct %int4_36822, %2482, %int1792_36823 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39269 = torch.aten.view %39267, %39268 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39269, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36824 = torch.constant.int 4
%39270 = torch.aten.mul.int %int4_36824, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36825 = torch.constant.int 4096
%39271 = torch.prim.ListConstruct %39270, %int4096_36825 : (!torch.int, !torch.int) -> !torch.list<int>
%39272 = torch.aten.view %39169, %39271 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39272, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39273 = torch.aten.mm %39272, %39251 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39273, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36826 = torch.constant.int 4
%int1792_36827 = torch.constant.int 1792
%39274 = torch.prim.ListConstruct %int4_36826, %2482, %int1792_36827 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39275 = torch.aten.view %39273, %39274 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39275, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36828 = torch.constant.int 4
%39276 = torch.aten.mul.int %int4_36828, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36829 = torch.constant.int 4096
%39277 = torch.prim.ListConstruct %39276, %int4096_36829 : (!torch.int, !torch.int) -> !torch.list<int>
%39278 = torch.aten.view %39170, %39277 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39278, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39279 = torch.aten.mm %39278, %39253 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39279, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36830 = torch.constant.int 4
%int1792_36831 = torch.constant.int 1792
%39280 = torch.prim.ListConstruct %int4_36830, %2482, %int1792_36831 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39281 = torch.aten.view %39279, %39280 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39281, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36832 = torch.constant.int 4
%39282 = torch.aten.mul.int %int4_36832, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36833 = torch.constant.int 4096
%39283 = torch.prim.ListConstruct %39282, %int4096_36833 : (!torch.int, !torch.int) -> !torch.list<int>
%39284 = torch.aten.view %39171, %39283 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39284, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39285 = torch.aten.mm %39284, %39255 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39285, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36834 = torch.constant.int 4
%int1792_36835 = torch.constant.int 1792
%39286 = torch.prim.ListConstruct %int4_36834, %2482, %int1792_36835 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39287 = torch.aten.view %39285, %39286 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39287, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36836 = torch.constant.int 4
%39288 = torch.aten.mul.int %int4_36836, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36837 = torch.constant.int 4096
%39289 = torch.prim.ListConstruct %39288, %int4096_36837 : (!torch.int, !torch.int) -> !torch.list<int>
%39290 = torch.aten.view %39172, %39289 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39290, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39291 = torch.aten.mm %39290, %39257 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39291, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36838 = torch.constant.int 4
%int1792_36839 = torch.constant.int 1792
%39292 = torch.prim.ListConstruct %int4_36838, %2482, %int1792_36839 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39293 = torch.aten.view %39291, %39292 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39293, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36840 = torch.constant.int 4
%39294 = torch.aten.mul.int %int4_36840, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36841 = torch.constant.int 4096
%39295 = torch.prim.ListConstruct %39294, %int4096_36841 : (!torch.int, !torch.int) -> !torch.list<int>
%39296 = torch.aten.view %39173, %39295 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39296, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39297 = torch.aten.mm %39296, %39259 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39297, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36842 = torch.constant.int 4
%int1792_36843 = torch.constant.int 1792
%39298 = torch.prim.ListConstruct %int4_36842, %2482, %int1792_36843 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39299 = torch.aten.view %39297, %39298 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39299, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36844 = torch.constant.int 4
%39300 = torch.aten.mul.int %int4_36844, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36845 = torch.constant.int 4096
%39301 = torch.prim.ListConstruct %39300, %int4096_36845 : (!torch.int, !torch.int) -> !torch.list<int>
%39302 = torch.aten.view %39174, %39301 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39302, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39303 = torch.aten.mm %39302, %39261 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39303, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36846 = torch.constant.int 4
%int1792_36847 = torch.constant.int 1792
%39304 = torch.prim.ListConstruct %int4_36846, %2482, %int1792_36847 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39305 = torch.aten.view %39303, %39304 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39305, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int4_36848 = torch.constant.int 4
%39306 = torch.aten.mul.int %int4_36848, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_36849 = torch.constant.int 4096
%39307 = torch.prim.ListConstruct %39306, %int4096_36849 : (!torch.int, !torch.int) -> !torch.list<int>
%39308 = torch.aten.view %39175, %39307 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39308, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39309 = torch.aten.mm %39308, %39263 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,1792],f16> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39309, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%int4_36850 = torch.constant.int 4
%int1792_36851 = torch.constant.int 1792
%39310 = torch.prim.ListConstruct %int4_36850, %2482, %int1792_36851 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39311 = torch.aten.view %39309, %39310 : !torch.vtensor<[?,1792],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39311, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39312 = torch.aten.mul.Tensor %39240, %39269 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39312, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39313 = torch.aten.mul.Tensor %39241, %39275 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39313, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39314 = torch.aten.mul.Tensor %39242, %39281 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39314, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39315 = torch.aten.mul.Tensor %39243, %39287 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39315, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39316 = torch.aten.mul.Tensor %39244, %39293 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39316, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39317 = torch.aten.mul.Tensor %39245, %39299 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39317, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39318 = torch.aten.mul.Tensor %39246, %39305 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39318, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%39319 = torch.aten.mul.Tensor %39247, %39311 : !torch.vtensor<[4,?,1792],f16>, !torch.vtensor<[4,?,1792],f16> -> !torch.vtensor<[4,?,1792],f16>
torch.bind_symbolic_shape %39319, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1792)> : !torch.vtensor<[4,?,1792],f16>
%int1_36852 = torch.constant.int 1
%int0_36853 = torch.constant.int 0
%39320 = torch.prim.ListConstruct %int1_36852, %int0_36853 : (!torch.int, !torch.int) -> !torch.list<int>
%39321 = torch.aten.permute %1440, %39320 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36854 = torch.constant.int 1
%int0_36855 = torch.constant.int 0
%39322 = torch.prim.ListConstruct %int1_36854, %int0_36855 : (!torch.int, !torch.int) -> !torch.list<int>
%39323 = torch.aten.permute %1441, %39322 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36856 = torch.constant.int 1
%int0_36857 = torch.constant.int 0
%39324 = torch.prim.ListConstruct %int1_36856, %int0_36857 : (!torch.int, !torch.int) -> !torch.list<int>
%39325 = torch.aten.permute %1442, %39324 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36858 = torch.constant.int 1
%int0_36859 = torch.constant.int 0
%39326 = torch.prim.ListConstruct %int1_36858, %int0_36859 : (!torch.int, !torch.int) -> !torch.list<int>
%39327 = torch.aten.permute %1443, %39326 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36860 = torch.constant.int 1
%int0_36861 = torch.constant.int 0
%39328 = torch.prim.ListConstruct %int1_36860, %int0_36861 : (!torch.int, !torch.int) -> !torch.list<int>
%39329 = torch.aten.permute %1444, %39328 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36862 = torch.constant.int 1
%int0_36863 = torch.constant.int 0
%39330 = torch.prim.ListConstruct %int1_36862, %int0_36863 : (!torch.int, !torch.int) -> !torch.list<int>
%39331 = torch.aten.permute %1445, %39330 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36864 = torch.constant.int 1
%int0_36865 = torch.constant.int 0
%39332 = torch.prim.ListConstruct %int1_36864, %int0_36865 : (!torch.int, !torch.int) -> !torch.list<int>
%39333 = torch.aten.permute %1446, %39332 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36866 = torch.constant.int 1
%int0_36867 = torch.constant.int 0
%39334 = torch.prim.ListConstruct %int1_36866, %int0_36867 : (!torch.int, !torch.int) -> !torch.list<int>
%39335 = torch.aten.permute %1447, %39334 : !torch.vtensor<[4096,1792],f16>, !torch.list<int> -> !torch.vtensor<[1792,4096],f16>
%int1_36868 = torch.constant.int 1
%39336 = torch.aten.size.int %39197, %int1_36868 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36869 = torch.constant.int 4
%39337 = torch.aten.mul.int %int4_36869, %39336 : !torch.int, !torch.int -> !torch.int
%int1792_36870 = torch.constant.int 1792
%39338 = torch.prim.ListConstruct %39337, %int1792_36870 : (!torch.int, !torch.int) -> !torch.list<int>
%39339 = torch.aten.view %39312, %39338 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39339, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39340 = torch.aten.mm %39339, %39321 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39340, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36871 = torch.constant.int 4
%int4096_36872 = torch.constant.int 4096
%39341 = torch.prim.ListConstruct %int4_36871, %39336, %int4096_36872 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39342 = torch.aten.view %39340, %39341 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39342, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36873 = torch.constant.int 1
%39343 = torch.aten.size.int %39203, %int1_36873 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36874 = torch.constant.int 4
%39344 = torch.aten.mul.int %int4_36874, %39343 : !torch.int, !torch.int -> !torch.int
%int1792_36875 = torch.constant.int 1792
%39345 = torch.prim.ListConstruct %39344, %int1792_36875 : (!torch.int, !torch.int) -> !torch.list<int>
%39346 = torch.aten.view %39313, %39345 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39346, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39347 = torch.aten.mm %39346, %39323 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39347, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36876 = torch.constant.int 4
%int4096_36877 = torch.constant.int 4096
%39348 = torch.prim.ListConstruct %int4_36876, %39343, %int4096_36877 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39349 = torch.aten.view %39347, %39348 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39349, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36878 = torch.constant.int 1
%39350 = torch.aten.size.int %39209, %int1_36878 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36879 = torch.constant.int 4
%39351 = torch.aten.mul.int %int4_36879, %39350 : !torch.int, !torch.int -> !torch.int
%int1792_36880 = torch.constant.int 1792
%39352 = torch.prim.ListConstruct %39351, %int1792_36880 : (!torch.int, !torch.int) -> !torch.list<int>
%39353 = torch.aten.view %39314, %39352 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39353, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39354 = torch.aten.mm %39353, %39325 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39354, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36881 = torch.constant.int 4
%int4096_36882 = torch.constant.int 4096
%39355 = torch.prim.ListConstruct %int4_36881, %39350, %int4096_36882 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39356 = torch.aten.view %39354, %39355 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39356, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36883 = torch.constant.int 1
%39357 = torch.aten.size.int %39215, %int1_36883 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36884 = torch.constant.int 4
%39358 = torch.aten.mul.int %int4_36884, %39357 : !torch.int, !torch.int -> !torch.int
%int1792_36885 = torch.constant.int 1792
%39359 = torch.prim.ListConstruct %39358, %int1792_36885 : (!torch.int, !torch.int) -> !torch.list<int>
%39360 = torch.aten.view %39315, %39359 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39360, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39361 = torch.aten.mm %39360, %39327 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39361, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36886 = torch.constant.int 4
%int4096_36887 = torch.constant.int 4096
%39362 = torch.prim.ListConstruct %int4_36886, %39357, %int4096_36887 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39363 = torch.aten.view %39361, %39362 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39363, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36888 = torch.constant.int 1
%39364 = torch.aten.size.int %39221, %int1_36888 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36889 = torch.constant.int 4
%39365 = torch.aten.mul.int %int4_36889, %39364 : !torch.int, !torch.int -> !torch.int
%int1792_36890 = torch.constant.int 1792
%39366 = torch.prim.ListConstruct %39365, %int1792_36890 : (!torch.int, !torch.int) -> !torch.list<int>
%39367 = torch.aten.view %39316, %39366 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39367, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39368 = torch.aten.mm %39367, %39329 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39368, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36891 = torch.constant.int 4
%int4096_36892 = torch.constant.int 4096
%39369 = torch.prim.ListConstruct %int4_36891, %39364, %int4096_36892 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39370 = torch.aten.view %39368, %39369 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39370, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36893 = torch.constant.int 1
%39371 = torch.aten.size.int %39227, %int1_36893 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36894 = torch.constant.int 4
%39372 = torch.aten.mul.int %int4_36894, %39371 : !torch.int, !torch.int -> !torch.int
%int1792_36895 = torch.constant.int 1792
%39373 = torch.prim.ListConstruct %39372, %int1792_36895 : (!torch.int, !torch.int) -> !torch.list<int>
%39374 = torch.aten.view %39317, %39373 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39374, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39375 = torch.aten.mm %39374, %39331 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39375, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36896 = torch.constant.int 4
%int4096_36897 = torch.constant.int 4096
%39376 = torch.prim.ListConstruct %int4_36896, %39371, %int4096_36897 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39377 = torch.aten.view %39375, %39376 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39377, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36898 = torch.constant.int 1
%39378 = torch.aten.size.int %39233, %int1_36898 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36899 = torch.constant.int 4
%39379 = torch.aten.mul.int %int4_36899, %39378 : !torch.int, !torch.int -> !torch.int
%int1792_36900 = torch.constant.int 1792
%39380 = torch.prim.ListConstruct %39379, %int1792_36900 : (!torch.int, !torch.int) -> !torch.list<int>
%39381 = torch.aten.view %39318, %39380 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39381, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39382 = torch.aten.mm %39381, %39333 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39382, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36901 = torch.constant.int 4
%int4096_36902 = torch.constant.int 4096
%39383 = torch.prim.ListConstruct %int4_36901, %39378, %int4096_36902 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39384 = torch.aten.view %39382, %39383 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39384, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36903 = torch.constant.int 1
%39385 = torch.aten.size.int %39239, %int1_36903 : !torch.vtensor<[4,?,1792],f16>, !torch.int -> !torch.int
%int4_36904 = torch.constant.int 4
%39386 = torch.aten.mul.int %int4_36904, %39385 : !torch.int, !torch.int -> !torch.int
%int1792_36905 = torch.constant.int 1792
%39387 = torch.prim.ListConstruct %39386, %int1792_36905 : (!torch.int, !torch.int) -> !torch.list<int>
%39388 = torch.aten.view %39319, %39387 : !torch.vtensor<[4,?,1792],f16>, !torch.list<int> -> !torch.vtensor<[?,1792],f16>
torch.bind_symbolic_shape %39388, [%2336], affine_map<()[s0] -> (s0 * 64, 1792)> : !torch.vtensor<[?,1792],f16>
%39389 = torch.aten.mm %39388, %39335 : !torch.vtensor<[?,1792],f16>, !torch.vtensor<[1792,4096],f16> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39389, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%int4_36906 = torch.constant.int 4
%int4096_36907 = torch.constant.int 4096
%39390 = torch.prim.ListConstruct %int4_36906, %39385, %int4096_36907 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39391 = torch.aten.view %39389, %39390 : !torch.vtensor<[?,4096],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39391, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39392 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36908 = arith.constant 1 : index
%dim_36909 = tensor.dim %39392, %c1_36908 : tensor<4x?x4096xf16>
%39393 = flow.tensor.transfer %39392 : tensor<4x?x4096xf16>{%dim_36909} to #hal.device.promise<@__device_0>
%39394 = torch_c.from_builtin_tensor %39393 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39394, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39395 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36910 = arith.constant 1 : index
%dim_36911 = tensor.dim %39395, %c1_36910 : tensor<4x?x4096xf16>
%39396 = flow.tensor.transfer %39395 : tensor<4x?x4096xf16>{%dim_36911} to #hal.device.promise<@__device_0>
%39397 = torch_c.from_builtin_tensor %39396 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39397, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39398 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36912 = arith.constant 1 : index
%dim_36913 = tensor.dim %39398, %c1_36912 : tensor<4x?x4096xf16>
%39399 = flow.tensor.transfer %39398 : tensor<4x?x4096xf16>{%dim_36913} to #hal.device.promise<@__device_0>
%39400 = torch_c.from_builtin_tensor %39399 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39400, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39401 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36914 = arith.constant 1 : index
%dim_36915 = tensor.dim %39401, %c1_36914 : tensor<4x?x4096xf16>
%39402 = flow.tensor.transfer %39401 : tensor<4x?x4096xf16>{%dim_36915} to #hal.device.promise<@__device_0>
%39403 = torch_c.from_builtin_tensor %39402 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39403, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39404 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36916 = arith.constant 1 : index
%dim_36917 = tensor.dim %39404, %c1_36916 : tensor<4x?x4096xf16>
%39405 = flow.tensor.transfer %39404 : tensor<4x?x4096xf16>{%dim_36917} to #hal.device.promise<@__device_0>
%39406 = torch_c.from_builtin_tensor %39405 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39406, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39407 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36918 = arith.constant 1 : index
%dim_36919 = tensor.dim %39407, %c1_36918 : tensor<4x?x4096xf16>
%39408 = flow.tensor.transfer %39407 : tensor<4x?x4096xf16>{%dim_36919} to #hal.device.promise<@__device_0>
%39409 = torch_c.from_builtin_tensor %39408 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39409, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39410 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36920 = arith.constant 1 : index
%dim_36921 = tensor.dim %39410, %c1_36920 : tensor<4x?x4096xf16>
%39411 = flow.tensor.transfer %39410 : tensor<4x?x4096xf16>{%dim_36921} to #hal.device.promise<@__device_0>
%39412 = torch_c.from_builtin_tensor %39411 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39412, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36922 = torch.constant.int 1
%39413 = torch.aten.add.Tensor %39342, %39394, %int1_36922 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39413, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36923 = torch.constant.int 1
%39414 = torch.aten.add.Tensor %39413, %39397, %int1_36923 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39414, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36924 = torch.constant.int 1
%39415 = torch.aten.add.Tensor %39414, %39400, %int1_36924 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39415, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36925 = torch.constant.int 1
%39416 = torch.aten.add.Tensor %39415, %39403, %int1_36925 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39416, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36926 = torch.constant.int 1
%39417 = torch.aten.add.Tensor %39416, %39406, %int1_36926 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39417, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36927 = torch.constant.int 1
%39418 = torch.aten.add.Tensor %39417, %39409, %int1_36927 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39418, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36928 = torch.constant.int 1
%39419 = torch.aten.add.Tensor %39418, %39412, %int1_36928 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39419, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39420 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36929 = arith.constant 1 : index
%dim_36930 = tensor.dim %39420, %c1_36929 : tensor<4x?x4096xf16>
%39421 = flow.tensor.transfer %39420 : tensor<4x?x4096xf16>{%dim_36930} to #hal.device.promise<@__device_1>
%39422 = torch_c.from_builtin_tensor %39421 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39422, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39423 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36931 = arith.constant 1 : index
%dim_36932 = tensor.dim %39423, %c1_36931 : tensor<4x?x4096xf16>
%39424 = flow.tensor.transfer %39423 : tensor<4x?x4096xf16>{%dim_36932} to #hal.device.promise<@__device_1>
%39425 = torch_c.from_builtin_tensor %39424 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39425, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39426 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36933 = arith.constant 1 : index
%dim_36934 = tensor.dim %39426, %c1_36933 : tensor<4x?x4096xf16>
%39427 = flow.tensor.transfer %39426 : tensor<4x?x4096xf16>{%dim_36934} to #hal.device.promise<@__device_1>
%39428 = torch_c.from_builtin_tensor %39427 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39428, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39429 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36935 = arith.constant 1 : index
%dim_36936 = tensor.dim %39429, %c1_36935 : tensor<4x?x4096xf16>
%39430 = flow.tensor.transfer %39429 : tensor<4x?x4096xf16>{%dim_36936} to #hal.device.promise<@__device_1>
%39431 = torch_c.from_builtin_tensor %39430 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39431, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39432 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36937 = arith.constant 1 : index
%dim_36938 = tensor.dim %39432, %c1_36937 : tensor<4x?x4096xf16>
%39433 = flow.tensor.transfer %39432 : tensor<4x?x4096xf16>{%dim_36938} to #hal.device.promise<@__device_1>
%39434 = torch_c.from_builtin_tensor %39433 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39434, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39435 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36939 = arith.constant 1 : index
%dim_36940 = tensor.dim %39435, %c1_36939 : tensor<4x?x4096xf16>
%39436 = flow.tensor.transfer %39435 : tensor<4x?x4096xf16>{%dim_36940} to #hal.device.promise<@__device_1>
%39437 = torch_c.from_builtin_tensor %39436 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39437, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39438 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36941 = arith.constant 1 : index
%dim_36942 = tensor.dim %39438, %c1_36941 : tensor<4x?x4096xf16>
%39439 = flow.tensor.transfer %39438 : tensor<4x?x4096xf16>{%dim_36942} to #hal.device.promise<@__device_1>
%39440 = torch_c.from_builtin_tensor %39439 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39440, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36943 = torch.constant.int 1
%39441 = torch.aten.add.Tensor %39422, %39349, %int1_36943 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39441, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36944 = torch.constant.int 1
%39442 = torch.aten.add.Tensor %39441, %39425, %int1_36944 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39442, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36945 = torch.constant.int 1
%39443 = torch.aten.add.Tensor %39442, %39428, %int1_36945 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39443, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36946 = torch.constant.int 1
%39444 = torch.aten.add.Tensor %39443, %39431, %int1_36946 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39444, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36947 = torch.constant.int 1
%39445 = torch.aten.add.Tensor %39444, %39434, %int1_36947 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39445, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36948 = torch.constant.int 1
%39446 = torch.aten.add.Tensor %39445, %39437, %int1_36948 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39446, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36949 = torch.constant.int 1
%39447 = torch.aten.add.Tensor %39446, %39440, %int1_36949 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39447, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39448 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36950 = arith.constant 1 : index
%dim_36951 = tensor.dim %39448, %c1_36950 : tensor<4x?x4096xf16>
%39449 = flow.tensor.transfer %39448 : tensor<4x?x4096xf16>{%dim_36951} to #hal.device.promise<@__device_2>
%39450 = torch_c.from_builtin_tensor %39449 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39450, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39451 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36952 = arith.constant 1 : index
%dim_36953 = tensor.dim %39451, %c1_36952 : tensor<4x?x4096xf16>
%39452 = flow.tensor.transfer %39451 : tensor<4x?x4096xf16>{%dim_36953} to #hal.device.promise<@__device_2>
%39453 = torch_c.from_builtin_tensor %39452 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39453, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39454 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36954 = arith.constant 1 : index
%dim_36955 = tensor.dim %39454, %c1_36954 : tensor<4x?x4096xf16>
%39455 = flow.tensor.transfer %39454 : tensor<4x?x4096xf16>{%dim_36955} to #hal.device.promise<@__device_2>
%39456 = torch_c.from_builtin_tensor %39455 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39456, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39457 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36956 = arith.constant 1 : index
%dim_36957 = tensor.dim %39457, %c1_36956 : tensor<4x?x4096xf16>
%39458 = flow.tensor.transfer %39457 : tensor<4x?x4096xf16>{%dim_36957} to #hal.device.promise<@__device_2>
%39459 = torch_c.from_builtin_tensor %39458 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39459, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39460 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36958 = arith.constant 1 : index
%dim_36959 = tensor.dim %39460, %c1_36958 : tensor<4x?x4096xf16>
%39461 = flow.tensor.transfer %39460 : tensor<4x?x4096xf16>{%dim_36959} to #hal.device.promise<@__device_2>
%39462 = torch_c.from_builtin_tensor %39461 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39462, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39463 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36960 = arith.constant 1 : index
%dim_36961 = tensor.dim %39463, %c1_36960 : tensor<4x?x4096xf16>
%39464 = flow.tensor.transfer %39463 : tensor<4x?x4096xf16>{%dim_36961} to #hal.device.promise<@__device_2>
%39465 = torch_c.from_builtin_tensor %39464 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39465, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39466 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36962 = arith.constant 1 : index
%dim_36963 = tensor.dim %39466, %c1_36962 : tensor<4x?x4096xf16>
%39467 = flow.tensor.transfer %39466 : tensor<4x?x4096xf16>{%dim_36963} to #hal.device.promise<@__device_2>
%39468 = torch_c.from_builtin_tensor %39467 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39468, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36964 = torch.constant.int 1
%39469 = torch.aten.add.Tensor %39450, %39453, %int1_36964 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39469, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36965 = torch.constant.int 1
%39470 = torch.aten.add.Tensor %39469, %39356, %int1_36965 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39470, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36966 = torch.constant.int 1
%39471 = torch.aten.add.Tensor %39470, %39456, %int1_36966 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39471, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36967 = torch.constant.int 1
%39472 = torch.aten.add.Tensor %39471, %39459, %int1_36967 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39472, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36968 = torch.constant.int 1
%39473 = torch.aten.add.Tensor %39472, %39462, %int1_36968 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39473, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36969 = torch.constant.int 1
%39474 = torch.aten.add.Tensor %39473, %39465, %int1_36969 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39474, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36970 = torch.constant.int 1
%39475 = torch.aten.add.Tensor %39474, %39468, %int1_36970 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39475, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39476 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36971 = arith.constant 1 : index
%dim_36972 = tensor.dim %39476, %c1_36971 : tensor<4x?x4096xf16>
%39477 = flow.tensor.transfer %39476 : tensor<4x?x4096xf16>{%dim_36972} to #hal.device.promise<@__device_3>
%39478 = torch_c.from_builtin_tensor %39477 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39478, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39479 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36973 = arith.constant 1 : index
%dim_36974 = tensor.dim %39479, %c1_36973 : tensor<4x?x4096xf16>
%39480 = flow.tensor.transfer %39479 : tensor<4x?x4096xf16>{%dim_36974} to #hal.device.promise<@__device_3>
%39481 = torch_c.from_builtin_tensor %39480 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39481, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39482 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36975 = arith.constant 1 : index
%dim_36976 = tensor.dim %39482, %c1_36975 : tensor<4x?x4096xf16>
%39483 = flow.tensor.transfer %39482 : tensor<4x?x4096xf16>{%dim_36976} to #hal.device.promise<@__device_3>
%39484 = torch_c.from_builtin_tensor %39483 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39484, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39485 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36977 = arith.constant 1 : index
%dim_36978 = tensor.dim %39485, %c1_36977 : tensor<4x?x4096xf16>
%39486 = flow.tensor.transfer %39485 : tensor<4x?x4096xf16>{%dim_36978} to #hal.device.promise<@__device_3>
%39487 = torch_c.from_builtin_tensor %39486 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39487, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39488 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36979 = arith.constant 1 : index
%dim_36980 = tensor.dim %39488, %c1_36979 : tensor<4x?x4096xf16>
%39489 = flow.tensor.transfer %39488 : tensor<4x?x4096xf16>{%dim_36980} to #hal.device.promise<@__device_3>
%39490 = torch_c.from_builtin_tensor %39489 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39490, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39491 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36981 = arith.constant 1 : index
%dim_36982 = tensor.dim %39491, %c1_36981 : tensor<4x?x4096xf16>
%39492 = flow.tensor.transfer %39491 : tensor<4x?x4096xf16>{%dim_36982} to #hal.device.promise<@__device_3>
%39493 = torch_c.from_builtin_tensor %39492 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39493, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39494 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36983 = arith.constant 1 : index
%dim_36984 = tensor.dim %39494, %c1_36983 : tensor<4x?x4096xf16>
%39495 = flow.tensor.transfer %39494 : tensor<4x?x4096xf16>{%dim_36984} to #hal.device.promise<@__device_3>
%39496 = torch_c.from_builtin_tensor %39495 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39496, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36985 = torch.constant.int 1
%39497 = torch.aten.add.Tensor %39478, %39481, %int1_36985 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39497, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36986 = torch.constant.int 1
%39498 = torch.aten.add.Tensor %39497, %39484, %int1_36986 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39498, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36987 = torch.constant.int 1
%39499 = torch.aten.add.Tensor %39498, %39363, %int1_36987 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39499, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36988 = torch.constant.int 1
%39500 = torch.aten.add.Tensor %39499, %39487, %int1_36988 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39500, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36989 = torch.constant.int 1
%39501 = torch.aten.add.Tensor %39500, %39490, %int1_36989 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39501, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36990 = torch.constant.int 1
%39502 = torch.aten.add.Tensor %39501, %39493, %int1_36990 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39502, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_36991 = torch.constant.int 1
%39503 = torch.aten.add.Tensor %39502, %39496, %int1_36991 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39503, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39504 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36992 = arith.constant 1 : index
%dim_36993 = tensor.dim %39504, %c1_36992 : tensor<4x?x4096xf16>
%39505 = flow.tensor.transfer %39504 : tensor<4x?x4096xf16>{%dim_36993} to #hal.device.promise<@__device_4>
%39506 = torch_c.from_builtin_tensor %39505 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39506, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39507 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36994 = arith.constant 1 : index
%dim_36995 = tensor.dim %39507, %c1_36994 : tensor<4x?x4096xf16>
%39508 = flow.tensor.transfer %39507 : tensor<4x?x4096xf16>{%dim_36995} to #hal.device.promise<@__device_4>
%39509 = torch_c.from_builtin_tensor %39508 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39509, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39510 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36996 = arith.constant 1 : index
%dim_36997 = tensor.dim %39510, %c1_36996 : tensor<4x?x4096xf16>
%39511 = flow.tensor.transfer %39510 : tensor<4x?x4096xf16>{%dim_36997} to #hal.device.promise<@__device_4>
%39512 = torch_c.from_builtin_tensor %39511 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39512, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39513 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_36998 = arith.constant 1 : index
%dim_36999 = tensor.dim %39513, %c1_36998 : tensor<4x?x4096xf16>
%39514 = flow.tensor.transfer %39513 : tensor<4x?x4096xf16>{%dim_36999} to #hal.device.promise<@__device_4>
%39515 = torch_c.from_builtin_tensor %39514 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39515, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39516 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37000 = arith.constant 1 : index
%dim_37001 = tensor.dim %39516, %c1_37000 : tensor<4x?x4096xf16>
%39517 = flow.tensor.transfer %39516 : tensor<4x?x4096xf16>{%dim_37001} to #hal.device.promise<@__device_4>
%39518 = torch_c.from_builtin_tensor %39517 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39518, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39519 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37002 = arith.constant 1 : index
%dim_37003 = tensor.dim %39519, %c1_37002 : tensor<4x?x4096xf16>
%39520 = flow.tensor.transfer %39519 : tensor<4x?x4096xf16>{%dim_37003} to #hal.device.promise<@__device_4>
%39521 = torch_c.from_builtin_tensor %39520 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39521, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39522 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37004 = arith.constant 1 : index
%dim_37005 = tensor.dim %39522, %c1_37004 : tensor<4x?x4096xf16>
%39523 = flow.tensor.transfer %39522 : tensor<4x?x4096xf16>{%dim_37005} to #hal.device.promise<@__device_4>
%39524 = torch_c.from_builtin_tensor %39523 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39524, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37006 = torch.constant.int 1
%39525 = torch.aten.add.Tensor %39506, %39509, %int1_37006 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39525, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37007 = torch.constant.int 1
%39526 = torch.aten.add.Tensor %39525, %39512, %int1_37007 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39526, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37008 = torch.constant.int 1
%39527 = torch.aten.add.Tensor %39526, %39515, %int1_37008 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39527, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37009 = torch.constant.int 1
%39528 = torch.aten.add.Tensor %39527, %39370, %int1_37009 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39528, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37010 = torch.constant.int 1
%39529 = torch.aten.add.Tensor %39528, %39518, %int1_37010 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39529, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37011 = torch.constant.int 1
%39530 = torch.aten.add.Tensor %39529, %39521, %int1_37011 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39530, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37012 = torch.constant.int 1
%39531 = torch.aten.add.Tensor %39530, %39524, %int1_37012 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39531, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39532 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37013 = arith.constant 1 : index
%dim_37014 = tensor.dim %39532, %c1_37013 : tensor<4x?x4096xf16>
%39533 = flow.tensor.transfer %39532 : tensor<4x?x4096xf16>{%dim_37014} to #hal.device.promise<@__device_5>
%39534 = torch_c.from_builtin_tensor %39533 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39534, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39535 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37015 = arith.constant 1 : index
%dim_37016 = tensor.dim %39535, %c1_37015 : tensor<4x?x4096xf16>
%39536 = flow.tensor.transfer %39535 : tensor<4x?x4096xf16>{%dim_37016} to #hal.device.promise<@__device_5>
%39537 = torch_c.from_builtin_tensor %39536 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39537, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39538 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37017 = arith.constant 1 : index
%dim_37018 = tensor.dim %39538, %c1_37017 : tensor<4x?x4096xf16>
%39539 = flow.tensor.transfer %39538 : tensor<4x?x4096xf16>{%dim_37018} to #hal.device.promise<@__device_5>
%39540 = torch_c.from_builtin_tensor %39539 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39540, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39541 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37019 = arith.constant 1 : index
%dim_37020 = tensor.dim %39541, %c1_37019 : tensor<4x?x4096xf16>
%39542 = flow.tensor.transfer %39541 : tensor<4x?x4096xf16>{%dim_37020} to #hal.device.promise<@__device_5>
%39543 = torch_c.from_builtin_tensor %39542 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39543, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39544 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37021 = arith.constant 1 : index
%dim_37022 = tensor.dim %39544, %c1_37021 : tensor<4x?x4096xf16>
%39545 = flow.tensor.transfer %39544 : tensor<4x?x4096xf16>{%dim_37022} to #hal.device.promise<@__device_5>
%39546 = torch_c.from_builtin_tensor %39545 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39546, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39547 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37023 = arith.constant 1 : index
%dim_37024 = tensor.dim %39547, %c1_37023 : tensor<4x?x4096xf16>
%39548 = flow.tensor.transfer %39547 : tensor<4x?x4096xf16>{%dim_37024} to #hal.device.promise<@__device_5>
%39549 = torch_c.from_builtin_tensor %39548 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39549, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39550 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37025 = arith.constant 1 : index
%dim_37026 = tensor.dim %39550, %c1_37025 : tensor<4x?x4096xf16>
%39551 = flow.tensor.transfer %39550 : tensor<4x?x4096xf16>{%dim_37026} to #hal.device.promise<@__device_5>
%39552 = torch_c.from_builtin_tensor %39551 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39552, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37027 = torch.constant.int 1
%39553 = torch.aten.add.Tensor %39534, %39537, %int1_37027 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39553, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37028 = torch.constant.int 1
%39554 = torch.aten.add.Tensor %39553, %39540, %int1_37028 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39554, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37029 = torch.constant.int 1
%39555 = torch.aten.add.Tensor %39554, %39543, %int1_37029 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39555, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37030 = torch.constant.int 1
%39556 = torch.aten.add.Tensor %39555, %39546, %int1_37030 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39556, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37031 = torch.constant.int 1
%39557 = torch.aten.add.Tensor %39556, %39377, %int1_37031 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39557, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37032 = torch.constant.int 1
%39558 = torch.aten.add.Tensor %39557, %39549, %int1_37032 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39558, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37033 = torch.constant.int 1
%39559 = torch.aten.add.Tensor %39558, %39552, %int1_37033 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39559, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39560 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37034 = arith.constant 1 : index
%dim_37035 = tensor.dim %39560, %c1_37034 : tensor<4x?x4096xf16>
%39561 = flow.tensor.transfer %39560 : tensor<4x?x4096xf16>{%dim_37035} to #hal.device.promise<@__device_6>
%39562 = torch_c.from_builtin_tensor %39561 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39562, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39563 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37036 = arith.constant 1 : index
%dim_37037 = tensor.dim %39563, %c1_37036 : tensor<4x?x4096xf16>
%39564 = flow.tensor.transfer %39563 : tensor<4x?x4096xf16>{%dim_37037} to #hal.device.promise<@__device_6>
%39565 = torch_c.from_builtin_tensor %39564 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39565, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39566 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37038 = arith.constant 1 : index
%dim_37039 = tensor.dim %39566, %c1_37038 : tensor<4x?x4096xf16>
%39567 = flow.tensor.transfer %39566 : tensor<4x?x4096xf16>{%dim_37039} to #hal.device.promise<@__device_6>
%39568 = torch_c.from_builtin_tensor %39567 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39568, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39569 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37040 = arith.constant 1 : index
%dim_37041 = tensor.dim %39569, %c1_37040 : tensor<4x?x4096xf16>
%39570 = flow.tensor.transfer %39569 : tensor<4x?x4096xf16>{%dim_37041} to #hal.device.promise<@__device_6>
%39571 = torch_c.from_builtin_tensor %39570 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39571, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39572 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37042 = arith.constant 1 : index
%dim_37043 = tensor.dim %39572, %c1_37042 : tensor<4x?x4096xf16>
%39573 = flow.tensor.transfer %39572 : tensor<4x?x4096xf16>{%dim_37043} to #hal.device.promise<@__device_6>
%39574 = torch_c.from_builtin_tensor %39573 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39574, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39575 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37044 = arith.constant 1 : index
%dim_37045 = tensor.dim %39575, %c1_37044 : tensor<4x?x4096xf16>
%39576 = flow.tensor.transfer %39575 : tensor<4x?x4096xf16>{%dim_37045} to #hal.device.promise<@__device_6>
%39577 = torch_c.from_builtin_tensor %39576 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39577, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39578 = torch_c.to_builtin_tensor %39391 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37046 = arith.constant 1 : index
%dim_37047 = tensor.dim %39578, %c1_37046 : tensor<4x?x4096xf16>
%39579 = flow.tensor.transfer %39578 : tensor<4x?x4096xf16>{%dim_37047} to #hal.device.promise<@__device_6>
%39580 = torch_c.from_builtin_tensor %39579 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39580, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37048 = torch.constant.int 1
%39581 = torch.aten.add.Tensor %39562, %39565, %int1_37048 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39581, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37049 = torch.constant.int 1
%39582 = torch.aten.add.Tensor %39581, %39568, %int1_37049 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39582, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37050 = torch.constant.int 1
%39583 = torch.aten.add.Tensor %39582, %39571, %int1_37050 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39583, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37051 = torch.constant.int 1
%39584 = torch.aten.add.Tensor %39583, %39574, %int1_37051 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39584, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37052 = torch.constant.int 1
%39585 = torch.aten.add.Tensor %39584, %39577, %int1_37052 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39585, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37053 = torch.constant.int 1
%39586 = torch.aten.add.Tensor %39585, %39384, %int1_37053 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39586, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37054 = torch.constant.int 1
%39587 = torch.aten.add.Tensor %39586, %39580, %int1_37054 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39587, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39588 = torch_c.to_builtin_tensor %39342 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37055 = arith.constant 1 : index
%dim_37056 = tensor.dim %39588, %c1_37055 : tensor<4x?x4096xf16>
%39589 = flow.tensor.transfer %39588 : tensor<4x?x4096xf16>{%dim_37056} to #hal.device.promise<@__device_7>
%39590 = torch_c.from_builtin_tensor %39589 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39590, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39591 = torch_c.to_builtin_tensor %39349 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37057 = arith.constant 1 : index
%dim_37058 = tensor.dim %39591, %c1_37057 : tensor<4x?x4096xf16>
%39592 = flow.tensor.transfer %39591 : tensor<4x?x4096xf16>{%dim_37058} to #hal.device.promise<@__device_7>
%39593 = torch_c.from_builtin_tensor %39592 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39593, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39594 = torch_c.to_builtin_tensor %39356 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37059 = arith.constant 1 : index
%dim_37060 = tensor.dim %39594, %c1_37059 : tensor<4x?x4096xf16>
%39595 = flow.tensor.transfer %39594 : tensor<4x?x4096xf16>{%dim_37060} to #hal.device.promise<@__device_7>
%39596 = torch_c.from_builtin_tensor %39595 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39596, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39597 = torch_c.to_builtin_tensor %39363 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37061 = arith.constant 1 : index
%dim_37062 = tensor.dim %39597, %c1_37061 : tensor<4x?x4096xf16>
%39598 = flow.tensor.transfer %39597 : tensor<4x?x4096xf16>{%dim_37062} to #hal.device.promise<@__device_7>
%39599 = torch_c.from_builtin_tensor %39598 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39599, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39600 = torch_c.to_builtin_tensor %39370 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37063 = arith.constant 1 : index
%dim_37064 = tensor.dim %39600, %c1_37063 : tensor<4x?x4096xf16>
%39601 = flow.tensor.transfer %39600 : tensor<4x?x4096xf16>{%dim_37064} to #hal.device.promise<@__device_7>
%39602 = torch_c.from_builtin_tensor %39601 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39602, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39603 = torch_c.to_builtin_tensor %39377 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37065 = arith.constant 1 : index
%dim_37066 = tensor.dim %39603, %c1_37065 : tensor<4x?x4096xf16>
%39604 = flow.tensor.transfer %39603 : tensor<4x?x4096xf16>{%dim_37066} to #hal.device.promise<@__device_7>
%39605 = torch_c.from_builtin_tensor %39604 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39605, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%39606 = torch_c.to_builtin_tensor %39384 : !torch.vtensor<[4,?,4096],f16> -> tensor<4x?x4096xf16>
%c1_37067 = arith.constant 1 : index
%dim_37068 = tensor.dim %39606, %c1_37067 : tensor<4x?x4096xf16>
%39607 = flow.tensor.transfer %39606 : tensor<4x?x4096xf16>{%dim_37068} to #hal.device.promise<@__device_7>
%39608 = torch_c.from_builtin_tensor %39607 : tensor<4x?x4096xf16> -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39608, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37069 = torch.constant.int 1
%39609 = torch.aten.add.Tensor %39590, %39593, %int1_37069 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39609, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37070 = torch.constant.int 1
%39610 = torch.aten.add.Tensor %39609, %39596, %int1_37070 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39610, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37071 = torch.constant.int 1
%39611 = torch.aten.add.Tensor %39610, %39599, %int1_37071 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39611, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37072 = torch.constant.int 1
%39612 = torch.aten.add.Tensor %39611, %39602, %int1_37072 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39612, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37073 = torch.constant.int 1
%39613 = torch.aten.add.Tensor %39612, %39605, %int1_37073 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39613, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37074 = torch.constant.int 1
%39614 = torch.aten.add.Tensor %39613, %39608, %int1_37074 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39614, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37075 = torch.constant.int 1
%39615 = torch.aten.add.Tensor %39614, %39391, %int1_37075 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39615, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37076 = torch.constant.int 1
%39616 = torch.aten.add.Tensor %39096, %39419, %int1_37076 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39616, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37077 = torch.constant.int 1
%39617 = torch.aten.add.Tensor %39097, %39447, %int1_37077 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39617, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37078 = torch.constant.int 1
%39618 = torch.aten.add.Tensor %39098, %39475, %int1_37078 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39618, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37079 = torch.constant.int 1
%39619 = torch.aten.add.Tensor %39099, %39503, %int1_37079 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39619, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37080 = torch.constant.int 1
%39620 = torch.aten.add.Tensor %39100, %39531, %int1_37080 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39620, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37081 = torch.constant.int 1
%39621 = torch.aten.add.Tensor %39101, %39559, %int1_37081 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39621, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37082 = torch.constant.int 1
%39622 = torch.aten.add.Tensor %39102, %39587, %int1_37082 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39622, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37083 = torch.constant.int 1
%39623 = torch.aten.add.Tensor %39103, %39615, %int1_37083 : !torch.vtensor<[4,?,4096],f16>, !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39623, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int6_37084 = torch.constant.int 6
%39624 = torch.prims.convert_element_type %39616, %int6_37084 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39624, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37085 = torch.constant.int 6
%39625 = torch.prims.convert_element_type %39617, %int6_37085 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39625, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37086 = torch.constant.int 6
%39626 = torch.prims.convert_element_type %39618, %int6_37086 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39626, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37087 = torch.constant.int 6
%39627 = torch.prims.convert_element_type %39619, %int6_37087 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39627, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37088 = torch.constant.int 6
%39628 = torch.prims.convert_element_type %39620, %int6_37088 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39628, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37089 = torch.constant.int 6
%39629 = torch.prims.convert_element_type %39621, %int6_37089 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39629, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37090 = torch.constant.int 6
%39630 = torch.prims.convert_element_type %39622, %int6_37090 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39630, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int6_37091 = torch.constant.int 6
%39631 = torch.prims.convert_element_type %39623, %int6_37091 : !torch.vtensor<[4,?,4096],f16>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39631, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37092 = torch.constant.int 2
%39632 = torch.aten.pow.Tensor_Scalar %39624, %int2_37092 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39632, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37093 = torch.constant.int 2
%39633 = torch.aten.pow.Tensor_Scalar %39625, %int2_37093 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39633, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37094 = torch.constant.int 2
%39634 = torch.aten.pow.Tensor_Scalar %39626, %int2_37094 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39634, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37095 = torch.constant.int 2
%39635 = torch.aten.pow.Tensor_Scalar %39627, %int2_37095 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39635, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37096 = torch.constant.int 2
%39636 = torch.aten.pow.Tensor_Scalar %39628, %int2_37096 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39636, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37097 = torch.constant.int 2
%39637 = torch.aten.pow.Tensor_Scalar %39629, %int2_37097 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39637, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37098 = torch.constant.int 2
%39638 = torch.aten.pow.Tensor_Scalar %39630, %int2_37098 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39638, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int2_37099 = torch.constant.int 2
%39639 = torch.aten.pow.Tensor_Scalar %39631, %int2_37099 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39639, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int-1_37100 = torch.constant.int -1
%39640 = torch.prim.ListConstruct %int-1_37100 : (!torch.int) -> !torch.list<int>
%true_37101 = torch.constant.bool true
%none_37102 = torch.constant.none
%39641 = torch.aten.mean.dim %39632, %39640, %true_37101, %none_37102 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39641, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37103 = torch.constant.int -1
%39642 = torch.prim.ListConstruct %int-1_37103 : (!torch.int) -> !torch.list<int>
%true_37104 = torch.constant.bool true
%none_37105 = torch.constant.none
%39643 = torch.aten.mean.dim %39633, %39642, %true_37104, %none_37105 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39643, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37106 = torch.constant.int -1
%39644 = torch.prim.ListConstruct %int-1_37106 : (!torch.int) -> !torch.list<int>
%true_37107 = torch.constant.bool true
%none_37108 = torch.constant.none
%39645 = torch.aten.mean.dim %39634, %39644, %true_37107, %none_37108 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39645, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37109 = torch.constant.int -1
%39646 = torch.prim.ListConstruct %int-1_37109 : (!torch.int) -> !torch.list<int>
%true_37110 = torch.constant.bool true
%none_37111 = torch.constant.none
%39647 = torch.aten.mean.dim %39635, %39646, %true_37110, %none_37111 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39647, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37112 = torch.constant.int -1
%39648 = torch.prim.ListConstruct %int-1_37112 : (!torch.int) -> !torch.list<int>
%true_37113 = torch.constant.bool true
%none_37114 = torch.constant.none
%39649 = torch.aten.mean.dim %39636, %39648, %true_37113, %none_37114 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39649, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37115 = torch.constant.int -1
%39650 = torch.prim.ListConstruct %int-1_37115 : (!torch.int) -> !torch.list<int>
%true_37116 = torch.constant.bool true
%none_37117 = torch.constant.none
%39651 = torch.aten.mean.dim %39637, %39650, %true_37116, %none_37117 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39651, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37118 = torch.constant.int -1
%39652 = torch.prim.ListConstruct %int-1_37118 : (!torch.int) -> !torch.list<int>
%true_37119 = torch.constant.bool true
%none_37120 = torch.constant.none
%39653 = torch.aten.mean.dim %39638, %39652, %true_37119, %none_37120 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39653, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%int-1_37121 = torch.constant.int -1
%39654 = torch.prim.ListConstruct %int-1_37121 : (!torch.int) -> !torch.list<int>
%true_37122 = torch.constant.bool true
%none_37123 = torch.constant.none
%39655 = torch.aten.mean.dim %39639, %39654, %true_37122, %none_37123 : !torch.vtensor<[4,?,4096],f32>, !torch.list<int>, !torch.bool, !torch.none -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39655, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37124 = torch.constant.float 9.9999997473787516E-6
%int1_37125 = torch.constant.int 1
%39656 = torch.aten.add.Scalar %39641, %float9.999990e-06_37124, %int1_37125 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39656, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37126 = torch.constant.float 9.9999997473787516E-6
%int1_37127 = torch.constant.int 1
%39657 = torch.aten.add.Scalar %39643, %float9.999990e-06_37126, %int1_37127 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39657, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37128 = torch.constant.float 9.9999997473787516E-6
%int1_37129 = torch.constant.int 1
%39658 = torch.aten.add.Scalar %39645, %float9.999990e-06_37128, %int1_37129 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39658, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37130 = torch.constant.float 9.9999997473787516E-6
%int1_37131 = torch.constant.int 1
%39659 = torch.aten.add.Scalar %39647, %float9.999990e-06_37130, %int1_37131 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39659, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37132 = torch.constant.float 9.9999997473787516E-6
%int1_37133 = torch.constant.int 1
%39660 = torch.aten.add.Scalar %39649, %float9.999990e-06_37132, %int1_37133 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39660, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37134 = torch.constant.float 9.9999997473787516E-6
%int1_37135 = torch.constant.int 1
%39661 = torch.aten.add.Scalar %39651, %float9.999990e-06_37134, %int1_37135 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39661, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37136 = torch.constant.float 9.9999997473787516E-6
%int1_37137 = torch.constant.int 1
%39662 = torch.aten.add.Scalar %39653, %float9.999990e-06_37136, %int1_37137 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39662, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%float9.999990e-06_37138 = torch.constant.float 9.9999997473787516E-6
%int1_37139 = torch.constant.int 1
%39663 = torch.aten.add.Scalar %39655, %float9.999990e-06_37138, %int1_37139 : !torch.vtensor<[4,?,1],f32>, !torch.float, !torch.int -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39663, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39664 = torch.aten.rsqrt %39656 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39664, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39665 = torch.aten.rsqrt %39657 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39665, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39666 = torch.aten.rsqrt %39658 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39666, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39667 = torch.aten.rsqrt %39659 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39667, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39668 = torch.aten.rsqrt %39660 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39668, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39669 = torch.aten.rsqrt %39661 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39669, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39670 = torch.aten.rsqrt %39662 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39670, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39671 = torch.aten.rsqrt %39663 : !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,1],f32>
torch.bind_symbolic_shape %39671, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1)> : !torch.vtensor<[4,?,1],f32>
%39672 = torch.aten.mul.Tensor %39624, %39664 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39672, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39673 = torch.aten.mul.Tensor %39625, %39665 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39673, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39674 = torch.aten.mul.Tensor %39626, %39666 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39674, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39675 = torch.aten.mul.Tensor %39627, %39667 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39675, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39676 = torch.aten.mul.Tensor %39628, %39668 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39676, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39677 = torch.aten.mul.Tensor %39629, %39669 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39677, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39678 = torch.aten.mul.Tensor %39630, %39670 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39678, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39679 = torch.aten.mul.Tensor %39631, %39671 : !torch.vtensor<[4,?,4096],f32>, !torch.vtensor<[4,?,1],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39679, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39680 = torch.aten.mul.Tensor %1448, %39672 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39680, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39681 = torch.aten.mul.Tensor %1449, %39673 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39681, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39682 = torch.aten.mul.Tensor %1450, %39674 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39682, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39683 = torch.aten.mul.Tensor %1451, %39675 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39683, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39684 = torch.aten.mul.Tensor %1452, %39676 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39684, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39685 = torch.aten.mul.Tensor %1453, %39677 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39685, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39686 = torch.aten.mul.Tensor %1454, %39678 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39686, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%39687 = torch.aten.mul.Tensor %1455, %39679 : !torch.vtensor<[4096],f32>, !torch.vtensor<[4,?,4096],f32> -> !torch.vtensor<[4,?,4096],f32>
torch.bind_symbolic_shape %39687, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f32>
%int5_37140 = torch.constant.int 5
%39688 = torch.prims.convert_element_type %39680, %int5_37140 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39688, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37141 = torch.constant.int 5
%39689 = torch.prims.convert_element_type %39681, %int5_37141 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39689, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37142 = torch.constant.int 5
%39690 = torch.prims.convert_element_type %39682, %int5_37142 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39690, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37143 = torch.constant.int 5
%39691 = torch.prims.convert_element_type %39683, %int5_37143 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39691, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37144 = torch.constant.int 5
%39692 = torch.prims.convert_element_type %39684, %int5_37144 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39692, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37145 = torch.constant.int 5
%39693 = torch.prims.convert_element_type %39685, %int5_37145 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39693, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37146 = torch.constant.int 5
%39694 = torch.prims.convert_element_type %39686, %int5_37146 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39694, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int5_37147 = torch.constant.int 5
%39695 = torch.prims.convert_element_type %39687, %int5_37147 : !torch.vtensor<[4,?,4096],f32>, !torch.int -> !torch.vtensor<[4,?,4096],f16>
torch.bind_symbolic_shape %39695, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4096)> : !torch.vtensor<[4,?,4096],f16>
%int1_37148 = torch.constant.int 1
%int0_37149 = torch.constant.int 0
%39696 = torch.prim.ListConstruct %int1_37148, %int0_37149 : (!torch.int, !torch.int) -> !torch.list<int>
%39697 = torch.aten.permute %1456, %39696 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37150 = torch.constant.int 1
%int0_37151 = torch.constant.int 0
%39698 = torch.prim.ListConstruct %int1_37150, %int0_37151 : (!torch.int, !torch.int) -> !torch.list<int>
%39699 = torch.aten.permute %1457, %39698 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37152 = torch.constant.int 1
%int0_37153 = torch.constant.int 0
%39700 = torch.prim.ListConstruct %int1_37152, %int0_37153 : (!torch.int, !torch.int) -> !torch.list<int>
%39701 = torch.aten.permute %1458, %39700 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37154 = torch.constant.int 1
%int0_37155 = torch.constant.int 0
%39702 = torch.prim.ListConstruct %int1_37154, %int0_37155 : (!torch.int, !torch.int) -> !torch.list<int>
%39703 = torch.aten.permute %1459, %39702 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37156 = torch.constant.int 1
%int0_37157 = torch.constant.int 0
%39704 = torch.prim.ListConstruct %int1_37156, %int0_37157 : (!torch.int, !torch.int) -> !torch.list<int>
%39705 = torch.aten.permute %1460, %39704 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37158 = torch.constant.int 1
%int0_37159 = torch.constant.int 0
%39706 = torch.prim.ListConstruct %int1_37158, %int0_37159 : (!torch.int, !torch.int) -> !torch.list<int>
%39707 = torch.aten.permute %1461, %39706 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37160 = torch.constant.int 1
%int0_37161 = torch.constant.int 0
%39708 = torch.prim.ListConstruct %int1_37160, %int0_37161 : (!torch.int, !torch.int) -> !torch.list<int>
%39709 = torch.aten.permute %1462, %39708 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int1_37162 = torch.constant.int 1
%int0_37163 = torch.constant.int 0
%39710 = torch.prim.ListConstruct %int1_37162, %int0_37163 : (!torch.int, !torch.int) -> !torch.list<int>
%39711 = torch.aten.permute %1463, %39710 : !torch.vtensor<[512,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,512],f16>
%int4_37164 = torch.constant.int 4
%39712 = torch.aten.mul.int %int4_37164, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37165 = torch.constant.int 4096
%39713 = torch.prim.ListConstruct %39712, %int4096_37165 : (!torch.int, !torch.int) -> !torch.list<int>
%39714 = torch.aten.view %39688, %39713 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39714, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39715 = torch.aten.mm %39714, %39697 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39715, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37166 = torch.constant.int 4
%int512_37167 = torch.constant.int 512
%39716 = torch.prim.ListConstruct %int4_37166, %2482, %int512_37167 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39717 = torch.aten.view %39715, %39716 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39717, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37168 = torch.constant.int 4
%39718 = torch.aten.mul.int %int4_37168, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37169 = torch.constant.int 4096
%39719 = torch.prim.ListConstruct %39718, %int4096_37169 : (!torch.int, !torch.int) -> !torch.list<int>
%39720 = torch.aten.view %39689, %39719 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39720, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39721 = torch.aten.mm %39720, %39699 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39721, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37170 = torch.constant.int 4
%int512_37171 = torch.constant.int 512
%39722 = torch.prim.ListConstruct %int4_37170, %2482, %int512_37171 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39723 = torch.aten.view %39721, %39722 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39723, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37172 = torch.constant.int 4
%39724 = torch.aten.mul.int %int4_37172, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37173 = torch.constant.int 4096
%39725 = torch.prim.ListConstruct %39724, %int4096_37173 : (!torch.int, !torch.int) -> !torch.list<int>
%39726 = torch.aten.view %39690, %39725 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39726, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39727 = torch.aten.mm %39726, %39701 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39727, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37174 = torch.constant.int 4
%int512_37175 = torch.constant.int 512
%39728 = torch.prim.ListConstruct %int4_37174, %2482, %int512_37175 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39729 = torch.aten.view %39727, %39728 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39729, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37176 = torch.constant.int 4
%39730 = torch.aten.mul.int %int4_37176, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37177 = torch.constant.int 4096
%39731 = torch.prim.ListConstruct %39730, %int4096_37177 : (!torch.int, !torch.int) -> !torch.list<int>
%39732 = torch.aten.view %39691, %39731 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39732, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39733 = torch.aten.mm %39732, %39703 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39733, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37178 = torch.constant.int 4
%int512_37179 = torch.constant.int 512
%39734 = torch.prim.ListConstruct %int4_37178, %2482, %int512_37179 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39735 = torch.aten.view %39733, %39734 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39735, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37180 = torch.constant.int 4
%39736 = torch.aten.mul.int %int4_37180, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37181 = torch.constant.int 4096
%39737 = torch.prim.ListConstruct %39736, %int4096_37181 : (!torch.int, !torch.int) -> !torch.list<int>
%39738 = torch.aten.view %39692, %39737 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39738, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39739 = torch.aten.mm %39738, %39705 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39739, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37182 = torch.constant.int 4
%int512_37183 = torch.constant.int 512
%39740 = torch.prim.ListConstruct %int4_37182, %2482, %int512_37183 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39741 = torch.aten.view %39739, %39740 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39741, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37184 = torch.constant.int 4
%39742 = torch.aten.mul.int %int4_37184, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37185 = torch.constant.int 4096
%39743 = torch.prim.ListConstruct %39742, %int4096_37185 : (!torch.int, !torch.int) -> !torch.list<int>
%39744 = torch.aten.view %39693, %39743 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39744, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39745 = torch.aten.mm %39744, %39707 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39745, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37186 = torch.constant.int 4
%int512_37187 = torch.constant.int 512
%39746 = torch.prim.ListConstruct %int4_37186, %2482, %int512_37187 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39747 = torch.aten.view %39745, %39746 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39747, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37188 = torch.constant.int 4
%39748 = torch.aten.mul.int %int4_37188, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37189 = torch.constant.int 4096
%39749 = torch.prim.ListConstruct %39748, %int4096_37189 : (!torch.int, !torch.int) -> !torch.list<int>
%39750 = torch.aten.view %39694, %39749 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39750, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39751 = torch.aten.mm %39750, %39709 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39751, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37190 = torch.constant.int 4
%int512_37191 = torch.constant.int 512
%39752 = torch.prim.ListConstruct %int4_37190, %2482, %int512_37191 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39753 = torch.aten.view %39751, %39752 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39753, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int4_37192 = torch.constant.int 4
%39754 = torch.aten.mul.int %int4_37192, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37193 = torch.constant.int 4096
%39755 = torch.prim.ListConstruct %39754, %int4096_37193 : (!torch.int, !torch.int) -> !torch.list<int>
%39756 = torch.aten.view %39695, %39755 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39756, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39757 = torch.aten.mm %39756, %39711 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,512],f16> -> !torch.vtensor<[?,512],f16>
torch.bind_symbolic_shape %39757, [%2336], affine_map<()[s0] -> (s0 * 64, 512)> : !torch.vtensor<[?,512],f16>
%int4_37194 = torch.constant.int 4
%int512_37195 = torch.constant.int 512
%39758 = torch.prim.ListConstruct %int4_37194, %2482, %int512_37195 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39759 = torch.aten.view %39757, %39758 : !torch.vtensor<[?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,512],f16>
torch.bind_symbolic_shape %39759, [%2336], affine_map<()[s0] -> (4, s0 * 16, 512)> : !torch.vtensor<[4,?,512],f16>
%int1_37196 = torch.constant.int 1
%int0_37197 = torch.constant.int 0
%39760 = torch.prim.ListConstruct %int1_37196, %int0_37197 : (!torch.int, !torch.int) -> !torch.list<int>
%39761 = torch.aten.permute %1464, %39760 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37198 = torch.constant.int 1
%int0_37199 = torch.constant.int 0
%39762 = torch.prim.ListConstruct %int1_37198, %int0_37199 : (!torch.int, !torch.int) -> !torch.list<int>
%39763 = torch.aten.permute %1465, %39762 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37200 = torch.constant.int 1
%int0_37201 = torch.constant.int 0
%39764 = torch.prim.ListConstruct %int1_37200, %int0_37201 : (!torch.int, !torch.int) -> !torch.list<int>
%39765 = torch.aten.permute %1466, %39764 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37202 = torch.constant.int 1
%int0_37203 = torch.constant.int 0
%39766 = torch.prim.ListConstruct %int1_37202, %int0_37203 : (!torch.int, !torch.int) -> !torch.list<int>
%39767 = torch.aten.permute %1467, %39766 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37204 = torch.constant.int 1
%int0_37205 = torch.constant.int 0
%39768 = torch.prim.ListConstruct %int1_37204, %int0_37205 : (!torch.int, !torch.int) -> !torch.list<int>
%39769 = torch.aten.permute %1468, %39768 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37206 = torch.constant.int 1
%int0_37207 = torch.constant.int 0
%39770 = torch.prim.ListConstruct %int1_37206, %int0_37207 : (!torch.int, !torch.int) -> !torch.list<int>
%39771 = torch.aten.permute %1469, %39770 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37208 = torch.constant.int 1
%int0_37209 = torch.constant.int 0
%39772 = torch.prim.ListConstruct %int1_37208, %int0_37209 : (!torch.int, !torch.int) -> !torch.list<int>
%39773 = torch.aten.permute %1470, %39772 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37210 = torch.constant.int 1
%int0_37211 = torch.constant.int 0
%39774 = torch.prim.ListConstruct %int1_37210, %int0_37211 : (!torch.int, !torch.int) -> !torch.list<int>
%39775 = torch.aten.permute %1471, %39774 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_37212 = torch.constant.int 4
%39776 = torch.aten.mul.int %int4_37212, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37213 = torch.constant.int 4096
%39777 = torch.prim.ListConstruct %39776, %int4096_37213 : (!torch.int, !torch.int) -> !torch.list<int>
%39778 = torch.aten.view %39688, %39777 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39778, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39779 = torch.aten.mm %39778, %39761 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39779, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37214 = torch.constant.int 4
%int128_37215 = torch.constant.int 128
%39780 = torch.prim.ListConstruct %int4_37214, %2482, %int128_37215 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39781 = torch.aten.view %39779, %39780 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39781, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37216 = torch.constant.int 4
%39782 = torch.aten.mul.int %int4_37216, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37217 = torch.constant.int 4096
%39783 = torch.prim.ListConstruct %39782, %int4096_37217 : (!torch.int, !torch.int) -> !torch.list<int>
%39784 = torch.aten.view %39689, %39783 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39784, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39785 = torch.aten.mm %39784, %39763 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39785, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37218 = torch.constant.int 4
%int128_37219 = torch.constant.int 128
%39786 = torch.prim.ListConstruct %int4_37218, %2482, %int128_37219 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39787 = torch.aten.view %39785, %39786 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39787, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37220 = torch.constant.int 4
%39788 = torch.aten.mul.int %int4_37220, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37221 = torch.constant.int 4096
%39789 = torch.prim.ListConstruct %39788, %int4096_37221 : (!torch.int, !torch.int) -> !torch.list<int>
%39790 = torch.aten.view %39690, %39789 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39790, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39791 = torch.aten.mm %39790, %39765 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39791, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37222 = torch.constant.int 4
%int128_37223 = torch.constant.int 128
%39792 = torch.prim.ListConstruct %int4_37222, %2482, %int128_37223 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39793 = torch.aten.view %39791, %39792 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39793, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37224 = torch.constant.int 4
%39794 = torch.aten.mul.int %int4_37224, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37225 = torch.constant.int 4096
%39795 = torch.prim.ListConstruct %39794, %int4096_37225 : (!torch.int, !torch.int) -> !torch.list<int>
%39796 = torch.aten.view %39691, %39795 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39796, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39797 = torch.aten.mm %39796, %39767 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39797, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37226 = torch.constant.int 4
%int128_37227 = torch.constant.int 128
%39798 = torch.prim.ListConstruct %int4_37226, %2482, %int128_37227 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39799 = torch.aten.view %39797, %39798 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39799, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37228 = torch.constant.int 4
%39800 = torch.aten.mul.int %int4_37228, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37229 = torch.constant.int 4096
%39801 = torch.prim.ListConstruct %39800, %int4096_37229 : (!torch.int, !torch.int) -> !torch.list<int>
%39802 = torch.aten.view %39692, %39801 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39802, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39803 = torch.aten.mm %39802, %39769 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39803, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37230 = torch.constant.int 4
%int128_37231 = torch.constant.int 128
%39804 = torch.prim.ListConstruct %int4_37230, %2482, %int128_37231 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39805 = torch.aten.view %39803, %39804 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39805, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37232 = torch.constant.int 4
%39806 = torch.aten.mul.int %int4_37232, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37233 = torch.constant.int 4096
%39807 = torch.prim.ListConstruct %39806, %int4096_37233 : (!torch.int, !torch.int) -> !torch.list<int>
%39808 = torch.aten.view %39693, %39807 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39808, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39809 = torch.aten.mm %39808, %39771 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39809, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37234 = torch.constant.int 4
%int128_37235 = torch.constant.int 128
%39810 = torch.prim.ListConstruct %int4_37234, %2482, %int128_37235 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39811 = torch.aten.view %39809, %39810 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39811, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37236 = torch.constant.int 4
%39812 = torch.aten.mul.int %int4_37236, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37237 = torch.constant.int 4096
%39813 = torch.prim.ListConstruct %39812, %int4096_37237 : (!torch.int, !torch.int) -> !torch.list<int>
%39814 = torch.aten.view %39694, %39813 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39814, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39815 = torch.aten.mm %39814, %39773 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39815, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37238 = torch.constant.int 4
%int128_37239 = torch.constant.int 128
%39816 = torch.prim.ListConstruct %int4_37238, %2482, %int128_37239 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39817 = torch.aten.view %39815, %39816 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39817, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37240 = torch.constant.int 4
%39818 = torch.aten.mul.int %int4_37240, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37241 = torch.constant.int 4096
%39819 = torch.prim.ListConstruct %39818, %int4096_37241 : (!torch.int, !torch.int) -> !torch.list<int>
%39820 = torch.aten.view %39695, %39819 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39820, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39821 = torch.aten.mm %39820, %39775 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39821, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37242 = torch.constant.int 4
%int128_37243 = torch.constant.int 128
%39822 = torch.prim.ListConstruct %int4_37242, %2482, %int128_37243 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39823 = torch.aten.view %39821, %39822 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39823, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int1_37244 = torch.constant.int 1
%int0_37245 = torch.constant.int 0
%39824 = torch.prim.ListConstruct %int1_37244, %int0_37245 : (!torch.int, !torch.int) -> !torch.list<int>
%39825 = torch.aten.permute %1472, %39824 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37246 = torch.constant.int 1
%int0_37247 = torch.constant.int 0
%39826 = torch.prim.ListConstruct %int1_37246, %int0_37247 : (!torch.int, !torch.int) -> !torch.list<int>
%39827 = torch.aten.permute %1473, %39826 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37248 = torch.constant.int 1
%int0_37249 = torch.constant.int 0
%39828 = torch.prim.ListConstruct %int1_37248, %int0_37249 : (!torch.int, !torch.int) -> !torch.list<int>
%39829 = torch.aten.permute %1474, %39828 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37250 = torch.constant.int 1
%int0_37251 = torch.constant.int 0
%39830 = torch.prim.ListConstruct %int1_37250, %int0_37251 : (!torch.int, !torch.int) -> !torch.list<int>
%39831 = torch.aten.permute %1475, %39830 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37252 = torch.constant.int 1
%int0_37253 = torch.constant.int 0
%39832 = torch.prim.ListConstruct %int1_37252, %int0_37253 : (!torch.int, !torch.int) -> !torch.list<int>
%39833 = torch.aten.permute %1476, %39832 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37254 = torch.constant.int 1
%int0_37255 = torch.constant.int 0
%39834 = torch.prim.ListConstruct %int1_37254, %int0_37255 : (!torch.int, !torch.int) -> !torch.list<int>
%39835 = torch.aten.permute %1477, %39834 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37256 = torch.constant.int 1
%int0_37257 = torch.constant.int 0
%39836 = torch.prim.ListConstruct %int1_37256, %int0_37257 : (!torch.int, !torch.int) -> !torch.list<int>
%39837 = torch.aten.permute %1478, %39836 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int1_37258 = torch.constant.int 1
%int0_37259 = torch.constant.int 0
%39838 = torch.prim.ListConstruct %int1_37258, %int0_37259 : (!torch.int, !torch.int) -> !torch.list<int>
%39839 = torch.aten.permute %1479, %39838 : !torch.vtensor<[128,4096],f16>, !torch.list<int> -> !torch.vtensor<[4096,128],f16>
%int4_37260 = torch.constant.int 4
%39840 = torch.aten.mul.int %int4_37260, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37261 = torch.constant.int 4096
%39841 = torch.prim.ListConstruct %39840, %int4096_37261 : (!torch.int, !torch.int) -> !torch.list<int>
%39842 = torch.aten.view %39688, %39841 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39842, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39843 = torch.aten.mm %39842, %39825 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39843, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37262 = torch.constant.int 4
%int128_37263 = torch.constant.int 128
%39844 = torch.prim.ListConstruct %int4_37262, %2482, %int128_37263 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39845 = torch.aten.view %39843, %39844 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39845, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37264 = torch.constant.int 4
%39846 = torch.aten.mul.int %int4_37264, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37265 = torch.constant.int 4096
%39847 = torch.prim.ListConstruct %39846, %int4096_37265 : (!torch.int, !torch.int) -> !torch.list<int>
%39848 = torch.aten.view %39689, %39847 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39848, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39849 = torch.aten.mm %39848, %39827 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39849, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37266 = torch.constant.int 4
%int128_37267 = torch.constant.int 128
%39850 = torch.prim.ListConstruct %int4_37266, %2482, %int128_37267 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39851 = torch.aten.view %39849, %39850 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39851, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37268 = torch.constant.int 4
%39852 = torch.aten.mul.int %int4_37268, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37269 = torch.constant.int 4096
%39853 = torch.prim.ListConstruct %39852, %int4096_37269 : (!torch.int, !torch.int) -> !torch.list<int>
%39854 = torch.aten.view %39690, %39853 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39854, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39855 = torch.aten.mm %39854, %39829 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39855, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37270 = torch.constant.int 4
%int128_37271 = torch.constant.int 128
%39856 = torch.prim.ListConstruct %int4_37270, %2482, %int128_37271 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39857 = torch.aten.view %39855, %39856 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39857, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37272 = torch.constant.int 4
%39858 = torch.aten.mul.int %int4_37272, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37273 = torch.constant.int 4096
%39859 = torch.prim.ListConstruct %39858, %int4096_37273 : (!torch.int, !torch.int) -> !torch.list<int>
%39860 = torch.aten.view %39691, %39859 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39860, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39861 = torch.aten.mm %39860, %39831 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39861, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37274 = torch.constant.int 4
%int128_37275 = torch.constant.int 128
%39862 = torch.prim.ListConstruct %int4_37274, %2482, %int128_37275 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39863 = torch.aten.view %39861, %39862 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39863, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37276 = torch.constant.int 4
%39864 = torch.aten.mul.int %int4_37276, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37277 = torch.constant.int 4096
%39865 = torch.prim.ListConstruct %39864, %int4096_37277 : (!torch.int, !torch.int) -> !torch.list<int>
%39866 = torch.aten.view %39692, %39865 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39866, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39867 = torch.aten.mm %39866, %39833 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39867, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37278 = torch.constant.int 4
%int128_37279 = torch.constant.int 128
%39868 = torch.prim.ListConstruct %int4_37278, %2482, %int128_37279 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39869 = torch.aten.view %39867, %39868 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39869, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37280 = torch.constant.int 4
%39870 = torch.aten.mul.int %int4_37280, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37281 = torch.constant.int 4096
%39871 = torch.prim.ListConstruct %39870, %int4096_37281 : (!torch.int, !torch.int) -> !torch.list<int>
%39872 = torch.aten.view %39693, %39871 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39872, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39873 = torch.aten.mm %39872, %39835 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39873, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37282 = torch.constant.int 4
%int128_37283 = torch.constant.int 128
%39874 = torch.prim.ListConstruct %int4_37282, %2482, %int128_37283 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39875 = torch.aten.view %39873, %39874 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39875, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37284 = torch.constant.int 4
%39876 = torch.aten.mul.int %int4_37284, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37285 = torch.constant.int 4096
%39877 = torch.prim.ListConstruct %39876, %int4096_37285 : (!torch.int, !torch.int) -> !torch.list<int>
%39878 = torch.aten.view %39694, %39877 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39878, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39879 = torch.aten.mm %39878, %39837 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39879, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37286 = torch.constant.int 4
%int128_37287 = torch.constant.int 128
%39880 = torch.prim.ListConstruct %int4_37286, %2482, %int128_37287 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39881 = torch.aten.view %39879, %39880 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39881, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37288 = torch.constant.int 4
%39882 = torch.aten.mul.int %int4_37288, %2482 : !torch.int, !torch.int -> !torch.int
%int4096_37289 = torch.constant.int 4096
%39883 = torch.prim.ListConstruct %39882, %int4096_37289 : (!torch.int, !torch.int) -> !torch.list<int>
%39884 = torch.aten.view %39695, %39883 : !torch.vtensor<[4,?,4096],f16>, !torch.list<int> -> !torch.vtensor<[?,4096],f16>
torch.bind_symbolic_shape %39884, [%2336], affine_map<()[s0] -> (s0 * 64, 4096)> : !torch.vtensor<[?,4096],f16>
%39885 = torch.aten.mm %39884, %39839 : !torch.vtensor<[?,4096],f16>, !torch.vtensor<[4096,128],f16> -> !torch.vtensor<[?,128],f16>
torch.bind_symbolic_shape %39885, [%2336], affine_map<()[s0] -> (s0 * 64, 128)> : !torch.vtensor<[?,128],f16>
%int4_37290 = torch.constant.int 4
%int128_37291 = torch.constant.int 128
%39886 = torch.prim.ListConstruct %int4_37290, %2482, %int128_37291 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39887 = torch.aten.view %39885, %39886 : !torch.vtensor<[?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,128],f16>
torch.bind_symbolic_shape %39887, [%2336], affine_map<()[s0] -> (4, s0 * 16, 128)> : !torch.vtensor<[4,?,128],f16>
%int4_37292 = torch.constant.int 4
%int4_37293 = torch.constant.int 4
%int128_37294 = torch.constant.int 128
%39888 = torch.prim.ListConstruct %int4_37292, %2482, %int4_37293, %int128_37294 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39889 = torch.aten.view %39717, %39888 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39889, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37295 = torch.constant.int 4
%int4_37296 = torch.constant.int 4
%int128_37297 = torch.constant.int 128
%39890 = torch.prim.ListConstruct %int4_37295, %2482, %int4_37296, %int128_37297 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39891 = torch.aten.view %39723, %39890 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39891, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37298 = torch.constant.int 4
%int4_37299 = torch.constant.int 4
%int128_37300 = torch.constant.int 128
%39892 = torch.prim.ListConstruct %int4_37298, %2482, %int4_37299, %int128_37300 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39893 = torch.aten.view %39729, %39892 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39893, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37301 = torch.constant.int 4
%int4_37302 = torch.constant.int 4
%int128_37303 = torch.constant.int 128
%39894 = torch.prim.ListConstruct %int4_37301, %2482, %int4_37302, %int128_37303 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39895 = torch.aten.view %39735, %39894 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39895, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37304 = torch.constant.int 4
%int4_37305 = torch.constant.int 4
%int128_37306 = torch.constant.int 128
%39896 = torch.prim.ListConstruct %int4_37304, %2482, %int4_37305, %int128_37306 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39897 = torch.aten.view %39741, %39896 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39897, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37307 = torch.constant.int 4
%int4_37308 = torch.constant.int 4
%int128_37309 = torch.constant.int 128
%39898 = torch.prim.ListConstruct %int4_37307, %2482, %int4_37308, %int128_37309 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39899 = torch.aten.view %39747, %39898 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39899, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37310 = torch.constant.int 4
%int4_37311 = torch.constant.int 4
%int128_37312 = torch.constant.int 128
%39900 = torch.prim.ListConstruct %int4_37310, %2482, %int4_37311, %int128_37312 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39901 = torch.aten.view %39753, %39900 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39901, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37313 = torch.constant.int 4
%int4_37314 = torch.constant.int 4
%int128_37315 = torch.constant.int 128
%39902 = torch.prim.ListConstruct %int4_37313, %2482, %int4_37314, %int128_37315 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39903 = torch.aten.view %39759, %39902 : !torch.vtensor<[4,?,512],f16>, !torch.list<int> -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39903, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int4_37316 = torch.constant.int 4
%int1_37317 = torch.constant.int 1
%int128_37318 = torch.constant.int 128
%39904 = torch.prim.ListConstruct %int4_37316, %2482, %int1_37317, %int128_37318 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39905 = torch.aten.view %39781, %39904 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39905, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37319 = torch.constant.int 4
%int1_37320 = torch.constant.int 1
%int128_37321 = torch.constant.int 128
%39906 = torch.prim.ListConstruct %int4_37319, %2482, %int1_37320, %int128_37321 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39907 = torch.aten.view %39787, %39906 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39907, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37322 = torch.constant.int 4
%int1_37323 = torch.constant.int 1
%int128_37324 = torch.constant.int 128
%39908 = torch.prim.ListConstruct %int4_37322, %2482, %int1_37323, %int128_37324 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39909 = torch.aten.view %39793, %39908 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39909, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37325 = torch.constant.int 4
%int1_37326 = torch.constant.int 1
%int128_37327 = torch.constant.int 128
%39910 = torch.prim.ListConstruct %int4_37325, %2482, %int1_37326, %int128_37327 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39911 = torch.aten.view %39799, %39910 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39911, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37328 = torch.constant.int 4
%int1_37329 = torch.constant.int 1
%int128_37330 = torch.constant.int 128
%39912 = torch.prim.ListConstruct %int4_37328, %2482, %int1_37329, %int128_37330 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39913 = torch.aten.view %39805, %39912 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39913, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37331 = torch.constant.int 4
%int1_37332 = torch.constant.int 1
%int128_37333 = torch.constant.int 128
%39914 = torch.prim.ListConstruct %int4_37331, %2482, %int1_37332, %int128_37333 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39915 = torch.aten.view %39811, %39914 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39915, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37334 = torch.constant.int 4
%int1_37335 = torch.constant.int 1
%int128_37336 = torch.constant.int 128
%39916 = torch.prim.ListConstruct %int4_37334, %2482, %int1_37335, %int128_37336 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39917 = torch.aten.view %39817, %39916 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39917, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37337 = torch.constant.int 4
%int1_37338 = torch.constant.int 1
%int128_37339 = torch.constant.int 128
%39918 = torch.prim.ListConstruct %int4_37337, %2482, %int1_37338, %int128_37339 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39919 = torch.aten.view %39823, %39918 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39919, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37340 = torch.constant.int 4
%int1_37341 = torch.constant.int 1
%int128_37342 = torch.constant.int 128
%39920 = torch.prim.ListConstruct %int4_37340, %2482, %int1_37341, %int128_37342 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39921 = torch.aten.view %39845, %39920 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39921, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37343 = torch.constant.int 4
%int1_37344 = torch.constant.int 1
%int128_37345 = torch.constant.int 128
%39922 = torch.prim.ListConstruct %int4_37343, %2482, %int1_37344, %int128_37345 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39923 = torch.aten.view %39851, %39922 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39923, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37346 = torch.constant.int 4
%int1_37347 = torch.constant.int 1
%int128_37348 = torch.constant.int 128
%39924 = torch.prim.ListConstruct %int4_37346, %2482, %int1_37347, %int128_37348 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39925 = torch.aten.view %39857, %39924 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39925, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37349 = torch.constant.int 4
%int1_37350 = torch.constant.int 1
%int128_37351 = torch.constant.int 128
%39926 = torch.prim.ListConstruct %int4_37349, %2482, %int1_37350, %int128_37351 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39927 = torch.aten.view %39863, %39926 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39927, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37352 = torch.constant.int 4
%int1_37353 = torch.constant.int 1
%int128_37354 = torch.constant.int 128
%39928 = torch.prim.ListConstruct %int4_37352, %2482, %int1_37353, %int128_37354 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39929 = torch.aten.view %39869, %39928 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39929, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37355 = torch.constant.int 4
%int1_37356 = torch.constant.int 1
%int128_37357 = torch.constant.int 128
%39930 = torch.prim.ListConstruct %int4_37355, %2482, %int1_37356, %int128_37357 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39931 = torch.aten.view %39875, %39930 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39931, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37358 = torch.constant.int 4
%int1_37359 = torch.constant.int 1
%int128_37360 = torch.constant.int 128
%39932 = torch.prim.ListConstruct %int4_37358, %2482, %int1_37359, %int128_37360 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39933 = torch.aten.view %39881, %39932 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39933, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int4_37361 = torch.constant.int 4
%int1_37362 = torch.constant.int 1
%int128_37363 = torch.constant.int 128
%39934 = torch.prim.ListConstruct %int4_37361, %2482, %int1_37362, %int128_37363 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>
%39935 = torch.aten.view %39887, %39934 : !torch.vtensor<[4,?,128],f16>, !torch.list<int> -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %39935, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int131072_37364 = torch.constant.int 131072
%none_37365 = torch.constant.none
%none_37366 = torch.constant.none
%cpu_37367 = torch.constant.device "cpu"
%false_37368 = torch.constant.bool false
%39936 = torch.aten.arange %int131072_37364, %none_37365, %none_37366, %cpu_37367, %false_37368 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_37369 = torch.constant.int 0
%int128_37370 = torch.constant.int 128
%int2_37371 = torch.constant.int 2
%none_37372 = torch.constant.none
%none_37373 = torch.constant.none
%cpu_37374 = torch.constant.device "cpu"
%false_37375 = torch.constant.bool false
%39937 = torch.aten.arange.start_step %int0_37369, %int128_37370, %int2_37371, %none_37372, %none_37373, %cpu_37374, %false_37375 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_37376 = torch.constant.int 0
%int0_37377 = torch.constant.int 0
%int64_37378 = torch.constant.int 64
%int1_37379 = torch.constant.int 1
%39938 = torch.aten.slice.Tensor %39937, %int0_37376, %int0_37377, %int64_37378, %int1_37379 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_37380 = torch.constant.int 6
%39939 = torch.prims.convert_element_type %39938, %int6_37380 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_37381 = torch.constant.int 128
%39940 = torch.aten.div.Scalar %39939, %int128_37381 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_37382 = torch.constant.float 5.000000e+05
%39941 = torch.aten.pow.Scalar %float5.000000e05_37382, %39940 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%39942 = torch.aten.reciprocal %39941 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_37383 = torch.constant.float 1.000000e+00
%39943 = torch.aten.mul.Scalar %39942, %float1.000000e00_37383 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_37384 = torch.constant.int 131072
%int1_37385 = torch.constant.int 1
%39944 = torch.prim.ListConstruct %int131072_37384, %int1_37385 : (!torch.int, !torch.int) -> !torch.list<int>
%39945 = torch.aten.view %39936, %39944 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%39946 = torch.aten.mul.Tensor %39945, %39943 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%39947 = torch.aten.cos %39946 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%39948 = torch.aten.sin %39946 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%39949 = torch.aten.complex %39947, %39948 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%39950 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39951 = flow.tensor.transfer %39950 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%39952 = torch_c.from_builtin_tensor %39951 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39953 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39954 = flow.tensor.transfer %39953 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%39955 = torch_c.from_builtin_tensor %39954 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39956 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39957 = flow.tensor.transfer %39956 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%39958 = torch_c.from_builtin_tensor %39957 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39959 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39960 = flow.tensor.transfer %39959 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%39961 = torch_c.from_builtin_tensor %39960 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39962 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39963 = flow.tensor.transfer %39962 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%39964 = torch_c.from_builtin_tensor %39963 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39965 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39966 = flow.tensor.transfer %39965 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%39967 = torch_c.from_builtin_tensor %39966 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39968 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39969 = flow.tensor.transfer %39968 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%39970 = torch_c.from_builtin_tensor %39969 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%39971 = torch_c.to_builtin_tensor %39949 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%39972 = flow.tensor.transfer %39971 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%39973 = torch_c.from_builtin_tensor %39972 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_37386 = torch.constant.int 1
%39974 = torch.aten.size.int %39717, %int1_37386 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37387 = torch.constant.int 0
%39975 = torch.aten.add.int %int0_37387, %39974 : !torch.int, !torch.int -> !torch.int
%int0_37388 = torch.constant.int 0
%int0_37389 = torch.constant.int 0
%int1_37390 = torch.constant.int 1
%39976 = torch.aten.slice.Tensor %39952, %int0_37388, %int0_37389, %39975, %int1_37390 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %39976, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37391 = torch.constant.int 1
%int0_37392 = torch.constant.int 0
%int9223372036854775807_37393 = torch.constant.int 9223372036854775807
%int1_37394 = torch.constant.int 1
%39977 = torch.aten.slice.Tensor %39976, %int1_37391, %int0_37392, %int9223372036854775807_37393, %int1_37394 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %39977, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37395 = torch.constant.int 0
%39978 = torch.aten.unsqueeze %39977, %int0_37395 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %39978, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37396 = torch.constant.int 2
%39979 = torch.aten.unsqueeze %39978, %int2_37396 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %39979, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37397 = torch.constant.int 3
%int0_37398 = torch.constant.int 0
%int9223372036854775807_37399 = torch.constant.int 9223372036854775807
%int1_37400 = torch.constant.int 1
%39980 = torch.aten.slice.Tensor %39979, %int3_37397, %int0_37398, %int9223372036854775807_37399, %int1_37400 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %39980, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%39981 = torch_c.to_builtin_tensor %39889 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37401 = arith.constant 1 : index
%dim_37402 = tensor.dim %39981, %c1_37401 : tensor<4x?x4x128xf16>
%39982 = flow.tensor.bitcast %39981 : tensor<4x?x4x128xf16>{%dim_37402} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37402}
%39983 = torch_c.from_builtin_tensor %39982 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %39983, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%39984 = torch.aten.mul.Tensor %39983, %39980 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %39984, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%39985 = torch_c.to_builtin_tensor %39984 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37403 = arith.constant 1 : index
%dim_37404 = tensor.dim %39985, %c1_37403 : tensor<4x?x4x64xcomplex<f32>>
%39986 = flow.tensor.bitcast %39985 : tensor<4x?x4x64xcomplex<f32>>{%dim_37404} -> tensor<4x?x4x128xf32>{%dim_37404}
%39987 = torch_c.from_builtin_tensor %39986 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %39987, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37405 = torch.constant.int 5
%39988 = torch.prims.convert_element_type %39987, %int5_37405 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %39988, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37406 = torch.constant.int 1
%39989 = torch.aten.size.int %39723, %int1_37406 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37407 = torch.constant.int 0
%39990 = torch.aten.add.int %int0_37407, %39989 : !torch.int, !torch.int -> !torch.int
%int0_37408 = torch.constant.int 0
%int0_37409 = torch.constant.int 0
%int1_37410 = torch.constant.int 1
%39991 = torch.aten.slice.Tensor %39955, %int0_37408, %int0_37409, %39990, %int1_37410 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %39991, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37411 = torch.constant.int 1
%int0_37412 = torch.constant.int 0
%int9223372036854775807_37413 = torch.constant.int 9223372036854775807
%int1_37414 = torch.constant.int 1
%39992 = torch.aten.slice.Tensor %39991, %int1_37411, %int0_37412, %int9223372036854775807_37413, %int1_37414 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %39992, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37415 = torch.constant.int 0
%39993 = torch.aten.unsqueeze %39992, %int0_37415 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %39993, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37416 = torch.constant.int 2
%39994 = torch.aten.unsqueeze %39993, %int2_37416 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %39994, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37417 = torch.constant.int 3
%int0_37418 = torch.constant.int 0
%int9223372036854775807_37419 = torch.constant.int 9223372036854775807
%int1_37420 = torch.constant.int 1
%39995 = torch.aten.slice.Tensor %39994, %int3_37417, %int0_37418, %int9223372036854775807_37419, %int1_37420 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %39995, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%39996 = torch_c.to_builtin_tensor %39891 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37421 = arith.constant 1 : index
%dim_37422 = tensor.dim %39996, %c1_37421 : tensor<4x?x4x128xf16>
%39997 = flow.tensor.bitcast %39996 : tensor<4x?x4x128xf16>{%dim_37422} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37422}
%39998 = torch_c.from_builtin_tensor %39997 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %39998, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%39999 = torch.aten.mul.Tensor %39998, %39995 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %39999, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40000 = torch_c.to_builtin_tensor %39999 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37423 = arith.constant 1 : index
%dim_37424 = tensor.dim %40000, %c1_37423 : tensor<4x?x4x64xcomplex<f32>>
%40001 = flow.tensor.bitcast %40000 : tensor<4x?x4x64xcomplex<f32>>{%dim_37424} -> tensor<4x?x4x128xf32>{%dim_37424}
%40002 = torch_c.from_builtin_tensor %40001 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40002, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37425 = torch.constant.int 5
%40003 = torch.prims.convert_element_type %40002, %int5_37425 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40003, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37426 = torch.constant.int 1
%40004 = torch.aten.size.int %39729, %int1_37426 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37427 = torch.constant.int 0
%40005 = torch.aten.add.int %int0_37427, %40004 : !torch.int, !torch.int -> !torch.int
%int0_37428 = torch.constant.int 0
%int0_37429 = torch.constant.int 0
%int1_37430 = torch.constant.int 1
%40006 = torch.aten.slice.Tensor %39958, %int0_37428, %int0_37429, %40005, %int1_37430 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40006, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37431 = torch.constant.int 1
%int0_37432 = torch.constant.int 0
%int9223372036854775807_37433 = torch.constant.int 9223372036854775807
%int1_37434 = torch.constant.int 1
%40007 = torch.aten.slice.Tensor %40006, %int1_37431, %int0_37432, %int9223372036854775807_37433, %int1_37434 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40007, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37435 = torch.constant.int 0
%40008 = torch.aten.unsqueeze %40007, %int0_37435 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40008, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37436 = torch.constant.int 2
%40009 = torch.aten.unsqueeze %40008, %int2_37436 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40009, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37437 = torch.constant.int 3
%int0_37438 = torch.constant.int 0
%int9223372036854775807_37439 = torch.constant.int 9223372036854775807
%int1_37440 = torch.constant.int 1
%40010 = torch.aten.slice.Tensor %40009, %int3_37437, %int0_37438, %int9223372036854775807_37439, %int1_37440 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40010, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40011 = torch_c.to_builtin_tensor %39893 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37441 = arith.constant 1 : index
%dim_37442 = tensor.dim %40011, %c1_37441 : tensor<4x?x4x128xf16>
%40012 = flow.tensor.bitcast %40011 : tensor<4x?x4x128xf16>{%dim_37442} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37442}
%40013 = torch_c.from_builtin_tensor %40012 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %40013, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%40014 = torch.aten.mul.Tensor %40013, %40010 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %40014, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40015 = torch_c.to_builtin_tensor %40014 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37443 = arith.constant 1 : index
%dim_37444 = tensor.dim %40015, %c1_37443 : tensor<4x?x4x64xcomplex<f32>>
%40016 = flow.tensor.bitcast %40015 : tensor<4x?x4x64xcomplex<f32>>{%dim_37444} -> tensor<4x?x4x128xf32>{%dim_37444}
%40017 = torch_c.from_builtin_tensor %40016 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40017, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37445 = torch.constant.int 5
%40018 = torch.prims.convert_element_type %40017, %int5_37445 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40018, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37446 = torch.constant.int 1
%40019 = torch.aten.size.int %39735, %int1_37446 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37447 = torch.constant.int 0
%40020 = torch.aten.add.int %int0_37447, %40019 : !torch.int, !torch.int -> !torch.int
%int0_37448 = torch.constant.int 0
%int0_37449 = torch.constant.int 0
%int1_37450 = torch.constant.int 1
%40021 = torch.aten.slice.Tensor %39961, %int0_37448, %int0_37449, %40020, %int1_37450 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40021, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37451 = torch.constant.int 1
%int0_37452 = torch.constant.int 0
%int9223372036854775807_37453 = torch.constant.int 9223372036854775807
%int1_37454 = torch.constant.int 1
%40022 = torch.aten.slice.Tensor %40021, %int1_37451, %int0_37452, %int9223372036854775807_37453, %int1_37454 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40022, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37455 = torch.constant.int 0
%40023 = torch.aten.unsqueeze %40022, %int0_37455 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40023, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37456 = torch.constant.int 2
%40024 = torch.aten.unsqueeze %40023, %int2_37456 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40024, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37457 = torch.constant.int 3
%int0_37458 = torch.constant.int 0
%int9223372036854775807_37459 = torch.constant.int 9223372036854775807
%int1_37460 = torch.constant.int 1
%40025 = torch.aten.slice.Tensor %40024, %int3_37457, %int0_37458, %int9223372036854775807_37459, %int1_37460 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40025, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40026 = torch_c.to_builtin_tensor %39895 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37461 = arith.constant 1 : index
%dim_37462 = tensor.dim %40026, %c1_37461 : tensor<4x?x4x128xf16>
%40027 = flow.tensor.bitcast %40026 : tensor<4x?x4x128xf16>{%dim_37462} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37462}
%40028 = torch_c.from_builtin_tensor %40027 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %40028, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%40029 = torch.aten.mul.Tensor %40028, %40025 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %40029, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40030 = torch_c.to_builtin_tensor %40029 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37463 = arith.constant 1 : index
%dim_37464 = tensor.dim %40030, %c1_37463 : tensor<4x?x4x64xcomplex<f32>>
%40031 = flow.tensor.bitcast %40030 : tensor<4x?x4x64xcomplex<f32>>{%dim_37464} -> tensor<4x?x4x128xf32>{%dim_37464}
%40032 = torch_c.from_builtin_tensor %40031 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40032, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37465 = torch.constant.int 5
%40033 = torch.prims.convert_element_type %40032, %int5_37465 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40033, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37466 = torch.constant.int 1
%40034 = torch.aten.size.int %39741, %int1_37466 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37467 = torch.constant.int 0
%40035 = torch.aten.add.int %int0_37467, %40034 : !torch.int, !torch.int -> !torch.int
%int0_37468 = torch.constant.int 0
%int0_37469 = torch.constant.int 0
%int1_37470 = torch.constant.int 1
%40036 = torch.aten.slice.Tensor %39964, %int0_37468, %int0_37469, %40035, %int1_37470 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40036, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37471 = torch.constant.int 1
%int0_37472 = torch.constant.int 0
%int9223372036854775807_37473 = torch.constant.int 9223372036854775807
%int1_37474 = torch.constant.int 1
%40037 = torch.aten.slice.Tensor %40036, %int1_37471, %int0_37472, %int9223372036854775807_37473, %int1_37474 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40037, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37475 = torch.constant.int 0
%40038 = torch.aten.unsqueeze %40037, %int0_37475 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40038, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37476 = torch.constant.int 2
%40039 = torch.aten.unsqueeze %40038, %int2_37476 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40039, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37477 = torch.constant.int 3
%int0_37478 = torch.constant.int 0
%int9223372036854775807_37479 = torch.constant.int 9223372036854775807
%int1_37480 = torch.constant.int 1
%40040 = torch.aten.slice.Tensor %40039, %int3_37477, %int0_37478, %int9223372036854775807_37479, %int1_37480 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40040, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40041 = torch_c.to_builtin_tensor %39897 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37481 = arith.constant 1 : index
%dim_37482 = tensor.dim %40041, %c1_37481 : tensor<4x?x4x128xf16>
%40042 = flow.tensor.bitcast %40041 : tensor<4x?x4x128xf16>{%dim_37482} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37482}
%40043 = torch_c.from_builtin_tensor %40042 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %40043, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%40044 = torch.aten.mul.Tensor %40043, %40040 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %40044, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40045 = torch_c.to_builtin_tensor %40044 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37483 = arith.constant 1 : index
%dim_37484 = tensor.dim %40045, %c1_37483 : tensor<4x?x4x64xcomplex<f32>>
%40046 = flow.tensor.bitcast %40045 : tensor<4x?x4x64xcomplex<f32>>{%dim_37484} -> tensor<4x?x4x128xf32>{%dim_37484}
%40047 = torch_c.from_builtin_tensor %40046 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40047, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37485 = torch.constant.int 5
%40048 = torch.prims.convert_element_type %40047, %int5_37485 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40048, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37486 = torch.constant.int 1
%40049 = torch.aten.size.int %39747, %int1_37486 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37487 = torch.constant.int 0
%40050 = torch.aten.add.int %int0_37487, %40049 : !torch.int, !torch.int -> !torch.int
%int0_37488 = torch.constant.int 0
%int0_37489 = torch.constant.int 0
%int1_37490 = torch.constant.int 1
%40051 = torch.aten.slice.Tensor %39967, %int0_37488, %int0_37489, %40050, %int1_37490 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40051, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37491 = torch.constant.int 1
%int0_37492 = torch.constant.int 0
%int9223372036854775807_37493 = torch.constant.int 9223372036854775807
%int1_37494 = torch.constant.int 1
%40052 = torch.aten.slice.Tensor %40051, %int1_37491, %int0_37492, %int9223372036854775807_37493, %int1_37494 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40052, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37495 = torch.constant.int 0
%40053 = torch.aten.unsqueeze %40052, %int0_37495 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40053, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37496 = torch.constant.int 2
%40054 = torch.aten.unsqueeze %40053, %int2_37496 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40054, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37497 = torch.constant.int 3
%int0_37498 = torch.constant.int 0
%int9223372036854775807_37499 = torch.constant.int 9223372036854775807
%int1_37500 = torch.constant.int 1
%40055 = torch.aten.slice.Tensor %40054, %int3_37497, %int0_37498, %int9223372036854775807_37499, %int1_37500 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40055, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40056 = torch_c.to_builtin_tensor %39899 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37501 = arith.constant 1 : index
%dim_37502 = tensor.dim %40056, %c1_37501 : tensor<4x?x4x128xf16>
%40057 = flow.tensor.bitcast %40056 : tensor<4x?x4x128xf16>{%dim_37502} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37502}
%40058 = torch_c.from_builtin_tensor %40057 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %40058, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%40059 = torch.aten.mul.Tensor %40058, %40055 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %40059, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40060 = torch_c.to_builtin_tensor %40059 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37503 = arith.constant 1 : index
%dim_37504 = tensor.dim %40060, %c1_37503 : tensor<4x?x4x64xcomplex<f32>>
%40061 = flow.tensor.bitcast %40060 : tensor<4x?x4x64xcomplex<f32>>{%dim_37504} -> tensor<4x?x4x128xf32>{%dim_37504}
%40062 = torch_c.from_builtin_tensor %40061 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40062, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37505 = torch.constant.int 5
%40063 = torch.prims.convert_element_type %40062, %int5_37505 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40063, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37506 = torch.constant.int 1
%40064 = torch.aten.size.int %39753, %int1_37506 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37507 = torch.constant.int 0
%40065 = torch.aten.add.int %int0_37507, %40064 : !torch.int, !torch.int -> !torch.int
%int0_37508 = torch.constant.int 0
%int0_37509 = torch.constant.int 0
%int1_37510 = torch.constant.int 1
%40066 = torch.aten.slice.Tensor %39970, %int0_37508, %int0_37509, %40065, %int1_37510 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40066, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37511 = torch.constant.int 1
%int0_37512 = torch.constant.int 0
%int9223372036854775807_37513 = torch.constant.int 9223372036854775807
%int1_37514 = torch.constant.int 1
%40067 = torch.aten.slice.Tensor %40066, %int1_37511, %int0_37512, %int9223372036854775807_37513, %int1_37514 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40067, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37515 = torch.constant.int 0
%40068 = torch.aten.unsqueeze %40067, %int0_37515 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40068, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37516 = torch.constant.int 2
%40069 = torch.aten.unsqueeze %40068, %int2_37516 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40069, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37517 = torch.constant.int 3
%int0_37518 = torch.constant.int 0
%int9223372036854775807_37519 = torch.constant.int 9223372036854775807
%int1_37520 = torch.constant.int 1
%40070 = torch.aten.slice.Tensor %40069, %int3_37517, %int0_37518, %int9223372036854775807_37519, %int1_37520 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40070, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40071 = torch_c.to_builtin_tensor %39901 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37521 = arith.constant 1 : index
%dim_37522 = tensor.dim %40071, %c1_37521 : tensor<4x?x4x128xf16>
%40072 = flow.tensor.bitcast %40071 : tensor<4x?x4x128xf16>{%dim_37522} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37522}
%40073 = torch_c.from_builtin_tensor %40072 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %40073, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%40074 = torch.aten.mul.Tensor %40073, %40070 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %40074, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40075 = torch_c.to_builtin_tensor %40074 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37523 = arith.constant 1 : index
%dim_37524 = tensor.dim %40075, %c1_37523 : tensor<4x?x4x64xcomplex<f32>>
%40076 = flow.tensor.bitcast %40075 : tensor<4x?x4x64xcomplex<f32>>{%dim_37524} -> tensor<4x?x4x128xf32>{%dim_37524}
%40077 = torch_c.from_builtin_tensor %40076 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40077, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37525 = torch.constant.int 5
%40078 = torch.prims.convert_element_type %40077, %int5_37525 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40078, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int1_37526 = torch.constant.int 1
%40079 = torch.aten.size.int %39759, %int1_37526 : !torch.vtensor<[4,?,512],f16>, !torch.int -> !torch.int
%int0_37527 = torch.constant.int 0
%40080 = torch.aten.add.int %int0_37527, %40079 : !torch.int, !torch.int -> !torch.int
%int0_37528 = torch.constant.int 0
%int0_37529 = torch.constant.int 0
%int1_37530 = torch.constant.int 1
%40081 = torch.aten.slice.Tensor %39973, %int0_37528, %int0_37529, %40080, %int1_37530 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40081, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37531 = torch.constant.int 1
%int0_37532 = torch.constant.int 0
%int9223372036854775807_37533 = torch.constant.int 9223372036854775807
%int1_37534 = torch.constant.int 1
%40082 = torch.aten.slice.Tensor %40081, %int1_37531, %int0_37532, %int9223372036854775807_37533, %int1_37534 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40082, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37535 = torch.constant.int 0
%40083 = torch.aten.unsqueeze %40082, %int0_37535 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40083, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37536 = torch.constant.int 2
%40084 = torch.aten.unsqueeze %40083, %int2_37536 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40084, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37537 = torch.constant.int 3
%int0_37538 = torch.constant.int 0
%int9223372036854775807_37539 = torch.constant.int 9223372036854775807
%int1_37540 = torch.constant.int 1
%40085 = torch.aten.slice.Tensor %40084, %int3_37537, %int0_37538, %int9223372036854775807_37539, %int1_37540 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40085, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40086 = torch_c.to_builtin_tensor %39903 : !torch.vtensor<[4,?,4,128],f16> -> tensor<4x?x4x128xf16>
%c1_37541 = arith.constant 1 : index
%dim_37542 = tensor.dim %40086, %c1_37541 : tensor<4x?x4x128xf16>
%40087 = flow.tensor.bitcast %40086 : tensor<4x?x4x128xf16>{%dim_37542} -> tensor<4x?x4x64xcomplex<f16>>{%dim_37542}
%40088 = torch_c.from_builtin_tensor %40087 : tensor<4x?x4x64xcomplex<f16>> -> !torch.vtensor<[4,?,4,64],complex<f16>>
torch.bind_symbolic_shape %40088, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f16>>
%40089 = torch.aten.mul.Tensor %40088, %40085 : !torch.vtensor<[4,?,4,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,4,64],complex<f32>>
torch.bind_symbolic_shape %40089, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 64)> : !torch.vtensor<[4,?,4,64],complex<f32>>
%40090 = torch_c.to_builtin_tensor %40089 : !torch.vtensor<[4,?,4,64],complex<f32>> -> tensor<4x?x4x64xcomplex<f32>>
%c1_37543 = arith.constant 1 : index
%dim_37544 = tensor.dim %40090, %c1_37543 : tensor<4x?x4x64xcomplex<f32>>
%40091 = flow.tensor.bitcast %40090 : tensor<4x?x4x64xcomplex<f32>>{%dim_37544} -> tensor<4x?x4x128xf32>{%dim_37544}
%40092 = torch_c.from_builtin_tensor %40091 : tensor<4x?x4x128xf32> -> !torch.vtensor<[4,?,4,128],f32>
torch.bind_symbolic_shape %40092, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f32>
%int5_37545 = torch.constant.int 5
%40093 = torch.prims.convert_element_type %40092, %int5_37545 : !torch.vtensor<[4,?,4,128],f32>, !torch.int -> !torch.vtensor<[4,?,4,128],f16>
torch.bind_symbolic_shape %40093, [%2336], affine_map<()[s0] -> (4, s0 * 16, 4, 128)> : !torch.vtensor<[4,?,4,128],f16>
%int131072_37546 = torch.constant.int 131072
%none_37547 = torch.constant.none
%none_37548 = torch.constant.none
%cpu_37549 = torch.constant.device "cpu"
%false_37550 = torch.constant.bool false
%40094 = torch.aten.arange %int131072_37546, %none_37547, %none_37548, %cpu_37549, %false_37550 : !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[131072],si64>
%int0_37551 = torch.constant.int 0
%int128_37552 = torch.constant.int 128
%int2_37553 = torch.constant.int 2
%none_37554 = torch.constant.none
%none_37555 = torch.constant.none
%cpu_37556 = torch.constant.device "cpu"
%false_37557 = torch.constant.bool false
%40095 = torch.aten.arange.start_step %int0_37551, %int128_37552, %int2_37553, %none_37554, %none_37555, %cpu_37556, %false_37557 : !torch.int, !torch.int, !torch.int, !torch.none, !torch.none, !torch.Device, !torch.bool -> !torch.vtensor<[64],si64>
%int0_37558 = torch.constant.int 0
%int0_37559 = torch.constant.int 0
%int64_37560 = torch.constant.int 64
%int1_37561 = torch.constant.int 1
%40096 = torch.aten.slice.Tensor %40095, %int0_37558, %int0_37559, %int64_37560, %int1_37561 : !torch.vtensor<[64],si64>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[64],si64>
%int6_37562 = torch.constant.int 6
%40097 = torch.prims.convert_element_type %40096, %int6_37562 : !torch.vtensor<[64],si64>, !torch.int -> !torch.vtensor<[64],f32>
%int128_37563 = torch.constant.int 128
%40098 = torch.aten.div.Scalar %40097, %int128_37563 : !torch.vtensor<[64],f32>, !torch.int -> !torch.vtensor<[64],f32>
%float5.000000e05_37564 = torch.constant.float 5.000000e+05
%40099 = torch.aten.pow.Scalar %float5.000000e05_37564, %40098 : !torch.float, !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%40100 = torch.aten.reciprocal %40099 : !torch.vtensor<[64],f32> -> !torch.vtensor<[64],f32>
%float1.000000e00_37565 = torch.constant.float 1.000000e+00
%40101 = torch.aten.mul.Scalar %40100, %float1.000000e00_37565 : !torch.vtensor<[64],f32>, !torch.float -> !torch.vtensor<[64],f32>
%int131072_37566 = torch.constant.int 131072
%int1_37567 = torch.constant.int 1
%40102 = torch.prim.ListConstruct %int131072_37566, %int1_37567 : (!torch.int, !torch.int) -> !torch.list<int>
%40103 = torch.aten.view %40094, %40102 : !torch.vtensor<[131072],si64>, !torch.list<int> -> !torch.vtensor<[131072,1],si64>
%40104 = torch.aten.mul.Tensor %40103, %40101 : !torch.vtensor<[131072,1],si64>, !torch.vtensor<[64],f32> -> !torch.vtensor<[131072,64],f32>
%40105 = torch.aten.cos %40104 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%40106 = torch.aten.sin %40104 : !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],f32>
%40107 = torch.aten.complex %40105, %40106 : !torch.vtensor<[131072,64],f32>, !torch.vtensor<[131072,64],f32> -> !torch.vtensor<[131072,64],complex<f32>>
%40108 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40109 = flow.tensor.transfer %40108 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_0>
%40110 = torch_c.from_builtin_tensor %40109 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40111 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40112 = flow.tensor.transfer %40111 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_1>
%40113 = torch_c.from_builtin_tensor %40112 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40114 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40115 = flow.tensor.transfer %40114 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_2>
%40116 = torch_c.from_builtin_tensor %40115 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40117 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40118 = flow.tensor.transfer %40117 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_3>
%40119 = torch_c.from_builtin_tensor %40118 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40120 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40121 = flow.tensor.transfer %40120 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_4>
%40122 = torch_c.from_builtin_tensor %40121 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40123 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40124 = flow.tensor.transfer %40123 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_5>
%40125 = torch_c.from_builtin_tensor %40124 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40126 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40127 = flow.tensor.transfer %40126 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_6>
%40128 = torch_c.from_builtin_tensor %40127 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%40129 = torch_c.to_builtin_tensor %40107 : !torch.vtensor<[131072,64],complex<f32>> -> tensor<131072x64xcomplex<f32>>
%40130 = flow.tensor.transfer %40129 : tensor<131072x64xcomplex<f32>> to #hal.device.promise<@__device_7>
%40131 = torch_c.from_builtin_tensor %40130 : tensor<131072x64xcomplex<f32>> -> !torch.vtensor<[131072,64],complex<f32>>
%int1_37568 = torch.constant.int 1
%40132 = torch.aten.size.int %39781, %int1_37568 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_37569 = torch.constant.int 0
%40133 = torch.aten.add.int %int0_37569, %40132 : !torch.int, !torch.int -> !torch.int
%int0_37570 = torch.constant.int 0
%int0_37571 = torch.constant.int 0
%int1_37572 = torch.constant.int 1
%40134 = torch.aten.slice.Tensor %40110, %int0_37570, %int0_37571, %40133, %int1_37572 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40134, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37573 = torch.constant.int 1
%int0_37574 = torch.constant.int 0
%int9223372036854775807_37575 = torch.constant.int 9223372036854775807
%int1_37576 = torch.constant.int 1
%40135 = torch.aten.slice.Tensor %40134, %int1_37573, %int0_37574, %int9223372036854775807_37575, %int1_37576 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40135, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37577 = torch.constant.int 0
%40136 = torch.aten.unsqueeze %40135, %int0_37577 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40136, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37578 = torch.constant.int 2
%40137 = torch.aten.unsqueeze %40136, %int2_37578 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40137, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%int3_37579 = torch.constant.int 3
%int0_37580 = torch.constant.int 0
%int9223372036854775807_37581 = torch.constant.int 9223372036854775807
%int1_37582 = torch.constant.int 1
%40138 = torch.aten.slice.Tensor %40137, %int3_37579, %int0_37580, %int9223372036854775807_37581, %int1_37582 : !torch.vtensor<[1,?,1,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40138, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>
%40139 = torch_c.to_builtin_tensor %39905 : !torch.vtensor<[4,?,1,128],f16> -> tensor<4x?x1x128xf16>
%c1_37583 = arith.constant 1 : index
%dim_37584 = tensor.dim %40139, %c1_37583 : tensor<4x?x1x128xf16>
%40140 = flow.tensor.bitcast %40139 : tensor<4x?x1x128xf16>{%dim_37584} -> tensor<4x?x1x64xcomplex<f16>>{%dim_37584}
%40141 = torch_c.from_builtin_tensor %40140 : tensor<4x?x1x64xcomplex<f16>> -> !torch.vtensor<[4,?,1,64],complex<f16>>
torch.bind_symbolic_shape %40141, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f16>>
%40142 = torch.aten.mul.Tensor %40141, %40138 : !torch.vtensor<[4,?,1,64],complex<f16>>, !torch.vtensor<[1,?,1,64],complex<f32>> -> !torch.vtensor<[4,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40142, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 64)> : !torch.vtensor<[4,?,1,64],complex<f32>>
%40143 = torch_c.to_builtin_tensor %40142 : !torch.vtensor<[4,?,1,64],complex<f32>> -> tensor<4x?x1x64xcomplex<f32>>
%c1_37585 = arith.constant 1 : index
%dim_37586 = tensor.dim %40143, %c1_37585 : tensor<4x?x1x64xcomplex<f32>>
%40144 = flow.tensor.bitcast %40143 : tensor<4x?x1x64xcomplex<f32>>{%dim_37586} -> tensor<4x?x1x128xf32>{%dim_37586}
%40145 = torch_c.from_builtin_tensor %40144 : tensor<4x?x1x128xf32> -> !torch.vtensor<[4,?,1,128],f32>
torch.bind_symbolic_shape %40145, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f32>
%int5_37587 = torch.constant.int 5
%40146 = torch.prims.convert_element_type %40145, %int5_37587 : !torch.vtensor<[4,?,1,128],f32>, !torch.int -> !torch.vtensor<[4,?,1,128],f16>
torch.bind_symbolic_shape %40146, [%2336], affine_map<()[s0] -> (4, s0 * 16, 1, 128)> : !torch.vtensor<[4,?,1,128],f16>
%int1_37588 = torch.constant.int 1
%40147 = torch.aten.size.int %39787, %int1_37588 : !torch.vtensor<[4,?,128],f16>, !torch.int -> !torch.int
%int0_37589 = torch.constant.int 0
%40148 = torch.aten.add.int %int0_37589, %40147 : !torch.int, !torch.int -> !torch.int
%int0_37590 = torch.constant.int 0
%int0_37591 = torch.constant.int 0
%int1_37592 = torch.constant.int 1
%40149 = torch.aten.slice.Tensor %40113, %int0_37590, %int0_37591, %40148, %int1_37592 : !torch.vtensor<[131072,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40149, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int1_37593 = torch.constant.int 1
%int0_37594 = torch.constant.int 0
%int9223372036854775807_37595 = torch.constant.int 9223372036854775807
%int1_37596 = torch.constant.int 1
%40150 = torch.aten.slice.Tensor %40149, %int1_37593, %int0_37594, %int9223372036854775807_37595, %int1_37596 : !torch.vtensor<[?,64],complex<f32>>, !torch.int, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[?,64],complex<f32>>
torch.bind_symbolic_shape %40150, [%2336], affine_map<()[s0] -> (s0 * 16, 64)> : !torch.vtensor<[?,64],complex<f32>>
%int0_37597 = torch.constant.int 0
%40151 = torch.aten.unsqueeze %40150, %int0_37597 : !torch.vtensor<[?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,64],complex<f32>>
torch.bind_symbolic_shape %40151, [%2336], affine_map<()[s0] -> (1, s0 * 16, 64)> : !torch.vtensor<[1,?,64],complex<f32>>
%int2_37598 = torch.constant.int 2
%40152 = torch.aten.unsqueeze %40151, %int2_37598 : !torch.vtensor<[1,?,64],complex<f32>>, !torch.int -> !torch.vtensor<[1,?,1,64],complex<f32>>
torch.bind_symbolic_shape %40152, [%2336], affine_map<()[s0] -> (1, s0 * 16, 1, 64)> : !torch.vtensor<[1,?,1,64],complex<f32>>